Skip to Content

Rule

Occupational Exposure to Respirable Crystalline Silica

This document was corrected by an document published on 05/18/2016. View Correction

Document Details

Information about this document as published in the Federal Register.

Published Document

This document has been published in the Federal Register. Use the PDF linked in the document sidebar for the official electronic format.

Start Preamble Start Printed Page 16286

AGENCY:

Occupational Safety and Health Administration (OSHA), Department of Labor.

ACTION:

Final rule.

SUMMARY:

The Occupational Safety and Health Administration (OSHA) is amending its existing standards for occupational exposure to respirable crystalline silica. OSHA has determined that employees exposed to respirable crystalline silica at the previous permissible exposure limits face a significant risk of material impairment to their health. The evidence in the record for this rulemaking indicates that workers exposed to respirable crystalline silica are at increased risk of developing silicosis and other non-malignant respiratory diseases, lung cancer, and kidney disease. This final rule establishes a new permissible exposure limit of 50 micrograms of respirable crystalline silica per cubic meter of air (50 μg/m3) as an 8-hour time-weighted average in all industries covered by the rule. It also includes other provisions to protect employees, such as requirements for exposure assessment, methods for controlling exposure, respiratory protection, medical surveillance, hazard communication, and recordkeeping.

OSHA is issuing two separate standards—one for general industry and maritime, and the other for construction—in order to tailor requirements to the circumstances found in these sectors.

DATES:

The final rule is effective on June 23, 2016. Start-up dates for specific provisions are set in § 1910.1053(l) for general industry and maritime and in § 1926.1153(k) for construction.

Collections of Information

There are a number of collections of information contained in this final rule (see Section VIII, Paperwork Reduction Act). Notwithstanding the general date of applicability that applies to all other requirements contained in the final rule, affected parties do not have to comply with the collections of information until the Department of Labor publishes a separate notice in the Federal Register announcing the Office of Management and Budget has approved them under the Paperwork Reduction Act.

ADDRESSES:

In accordance with 28 U.S.C. 2112(a), the Agency designates Ann Rosenthal, Associate Solicitor of Labor for Occupational Safety and Health, Office of the Solicitor of Labor, Room S-4004, U.S. Department of Labor, 200 Constitution Avenue NW., Washington, DC 20210, to receive petitions for review of the final rule.

Start Further Info

FOR FURTHER INFORMATION CONTACT:

For general information and press inquiries, contact Frank Meilinger, Director, Office of Communications, Room N-3647, OSHA, U.S. Department of Labor, 200 Constitution Avenue NW., Washington, DC 20210; telephone (202) 693-1999; email meilinger.francis2@dol.gov.

For technical inquiries, contact William Perry or David O'Connor, Directorate of Standards and Guidance, Room N-3718, OSHA, U.S. Department of Labor, 200 Constitution Avenue NW., Washington, DC 20210; telephone (202) 693-1950.

End Further Info End Preamble Start Supplemental Information

SUPPLEMENTARY INFORMATION:

The preamble to the rule on occupational exposure to respirable crystalline silica follows this outline:

I. Executive Summary

II. Pertinent Legal Authority

III. Events Leading to the Final Standards

IV. Chemical Properties and Industrial Uses

V. Health Effects

VI. Final Quantitative Risk Assessment and Significance of Risk

VII. Summary of the Final Economic Analysis and Final Regulatory Flexibility Analysis

VIII. Paperwork Reduction Act

IX. Federalism

X. State-Plan States

XI. Unfunded Mandates

XII. Protecting Children From Environmental Health and Safety Risks

XIII. Consultation and Coordination With Indian Tribal Governments

XIV. Environmental Impacts

XV. Summary and Explanation of the Standards

Scope

Definitions

Specified Exposure Control Methods

Alternative Exposure Control Methods

Permissible Exposure Limit

Exposure Assessment

Regulated Areas

Methods of Compliance

Respiratory Protection

Housekeeping

Written Exposure Control Plan

Medical Surveillance

Communication of Respirable Crystalline Silica Hazards to Employees

Recordkeeping

Dates

Authority and Signature

Citation Method

In the docket for the respirable crystalline silica rulemaking, found at http://www.regulations.gov, every submission was assigned a document identification (ID) number that consists of the docket number (OSHA-2010-0034) followed by an additional four-digit number. For example, the document ID number for OSHA's Preliminary Economic Analysis and Initial Regulatory Flexibility Analysis is OSHA-2010-0034-1720. Some document ID numbers include one or more attachments, such as the National Institute for Occupational Safety and Health (NIOSH) prehearing submission (see Document ID OSHA 2010-0034-2177).

When citing exhibits in the docket, OSHA includes the term “Document ID” followed by the last four digits of the document ID number, the attachment number or other attachment identifier, if applicable, page numbers (designated “p.” or “Tr.” for pages from a hearing transcript), and in a limited number of cases a footnote number (designated “Fn”). In a citation that contains two or more document ID numbers, the document ID numbers are separated by semi-colons. For example, a citation referring to the NIOSH prehearing comments and NIOSH testimony obtained from the hearing transcript would be indicated as follows: (Document ID 2177, Attachment B, pp. 2-3; 3579, Tr. 132). In some sections, such as Section V, Health Effects, author names and year of study publication are included before the document ID number in a citation, for example: (Hughes et al., 2001, Document ID 1060; McDonald et al., 2001, 1091; McDonald et al., 2005, 1092; Rando et al., 2001, 0415).

I. Executive Summary

This final rule establishes a permissible exposure limit (PEL) for respirable crystalline silica of 50 μg/m3 as an 8-hour time-weighted average (TWA) in all industries covered by the rule. In addition to the PEL, the rule includes provisions to protect employees such as requirements for exposure assessment, methods for controlling exposure, respiratory protection, medical surveillance, hazard communication, and recordkeeping. OSHA is issuing two separate standards—one for general industry and maritime, and the other for construction—in order to tailor requirements to the circumstances found in these sectors. There are, however, numerous common elements in the two standards.Start Printed Page 16287

The final rule is based on the requirements of the Occupational Safety and Health Act (OSH Act) and court interpretations of the Act. For health standards issued under section 6(b)(5) of the OSH Act, OSHA is required to promulgate a standard that reduces significant risk to the extent that it is technologically and economically feasible to do so. See Section II, Pertinent Legal Authority, for a full discussion of OSH Act legal requirements.

OSHA has conducted an extensive review of the literature on adverse health effects associated with exposure to respirable crystalline silica. OSHA has also developed estimates of the risk of silica-related diseases, assuming exposure over a working lifetime, at the preceding PELs as well as at the revised PEL and action level. Comments received on OSHA's preliminary analysis, and the Agency's final findings, are discussed in Section V, Health Effects, and Section VI, Final Quantitative Risk Assessment and Significance of Risk. OSHA finds that employees exposed to respirable crystalline silica at the preceding PELs are at an increased risk of lung cancer mortality and silicosis mortality and morbidity. Occupational exposures to respirable crystalline silica also result in increased risk of death from other nonmalignant respiratory diseases including chronic obstructive pulmonary disease (COPD), and from kidney disease. OSHA further concludes that exposure to respirable crystalline silica constitutes a significant risk of material impairment to health and that the final rule will substantially lower that risk. The Agency considers the level of risk remaining at the new PEL to be significant. However, based on the evidence evaluated during the rulemaking process, OSHA has determined a PEL of 50 μg/m3 is appropriate because it is the lowest level feasible for all affected industries.

OSHA's examination of the technological and economic feasibility of the rule is presented in the Final Economic Analysis and Final Regulatory Flexibility Analysis (FEA), and is summarized in Section VII of this preamble. OSHA concludes that the PEL of 50 μg/m3 is technologically feasible for most operations in all affected industries, although it will be a technological challenge for several affected sectors and will require the use of respirators for a limited number of job categories and tasks.

OSHA developed quantitative estimates of the compliance costs of the rule for each of the affected industry sectors. The estimated compliance costs were compared with industry revenues and profits to provide a screening analysis of the economic feasibility of complying with the rule and an evaluation of the economic impacts. Industries with unusually high costs as a percentage of revenues or profits were further analyzed for possible economic feasibility issues. After performing these analyses, OSHA finds that compliance with the requirements of the rule is economically feasible in every affected industry sector.

The final rule includes several major changes from the proposed rule as a result of OSHA's analysis of comments and evidence received during the comment periods and public hearings. The major changes are summarized below and are fully discussed in Section XV, Summary and Explanation of the Standards.

Scope. As proposed, the standards covered all occupational exposures to respirable crystalline silica with the exception of agricultural operations covered under 29 CFR part 1928. OSHA has made a final determination to exclude exposures in general industry and maritime where the employer has objective data demonstrating that employee exposure to respirable crystalline silica will remain below 25 μg/m3 as an 8-hour TWA under any foreseeable conditions. OSHA is also excluding exposures in construction where employee exposure to respirable crystalline silica will remain below 25 μg/m3 as an 8-hour TWA under any foreseeable conditions. In addition, OSHA is excluding exposures that result from the processing of sorptive clays from the scope of the rule. The standard for general industry and maritime also allows employers to comply with the standard for construction in certain circumstances.

Specified Exposure Control Methods. OSHA has revised the structure of the standard for construction to emphasize the specified exposure control methods for construction tasks that are presented in Table 1 of the standard. Unlike in the proposed rule, employers who fully and properly implement the controls listed on Table 1 are not separately required to comply with the PEL, and are not subject to provisions for exposure assessment and methods of compliance. The entries on Table 1 have also been revised extensively.

Protective Clothing. The proposed rule would have required use of protective clothing in certain limited situations. The final rule does not include requirements for use of protective clothing to address exposure to respirable crystalline silica.

Housekeeping. The proposed rule would have prohibited use of compressed air, dry sweeping, and dry brushing to clean clothing or surfaces contaminated with crystalline silica where such activities could contribute to employee exposure to respirable crystalline silica that exceeds the PEL. The final rule allows for use of compressed air, dry sweeping, and dry brushing in certain limited situations.

Written Exposure Control Plan. OSHA did not propose a requirement for employers to develop a written exposure control plan. The final rule includes a requirement for employers covered by the rule to develop a written exposure control plan, and the standard for construction includes a provision for a competent person (i.e., a designated individual who is capable of identifying crystalline silica hazards in the workplace and who possesses the authority to take corrective measures to address them) to implement the written exposure control plan.

Regulated Areas. OSHA proposed to provide employers covered by the rule with the alternative of either establishing a regulated area or an access control plan to limit access to areas where exposure to respirable crystalline silica exceeds the PEL. The final standard for general industry and maritime requires employers to establish a regulated area in such circumstances. The final standard for construction does not include a provision for regulated areas, but includes a requirement that the written exposure control plan include procedures used to restrict access to work areas, when necessary, to minimize the numbers of employees exposed to respirable crystalline silica and their level of exposure. The access control plan alternative is not included in the final rule.

Medical Surveillance. The proposed rule would have required employers to make medical surveillance available to employees exposed to respirable crystalline silica above the PEL for 30 or more days per year. The final standard for general industry and maritime requires that medical surveillance be made available to employees exposed to respirable crystalline silica at or above the action level of 25 μg/m3 as an 8-hour TWA for 30 or more days per year. The final standard for construction requires that medical surveillance be made available to employees who are required by the standard to use respirators for 30 or more days per year.

The rule requires the employer to obtain a written medical opinion from physicians or other licensed health care professionals (PLHCPs) for medical Start Printed Page 16288examinations provided under the rule but limits the information provided to the employer to the date of the examination, a statement that the examination has met the requirements of the standard, and any recommended limitations on the employee's use of respirators. The proposed rule would have required that such opinions contain additional information, without requiring employee authorization, such as any recommended limitations upon the employee's exposure to respirable crystalline silica, and any referral to a specialist. In the final rule, the written opinion provided to the employer will only include recommended limitations on the employee's exposure to respirable crystalline silica and referral to a specialist if the employee provides written authorization. The final rule requires a separate written medical report provided to the employee to include this additional information, as well as detailed information related to the employee's health.

Dates. OSHA proposed identical requirements for both standards: an effective date 60 days after publication of the rule; a date for compliance with all provisions except engineering controls and laboratory requirements of 180 days after the effective date; a date for compliance with engineering controls requirements, which was one year after the effective date; and a date for compliance with laboratory requirements of two years after the effective date.

OSHA has revised the proposed compliance dates in both standards. The final rule is effective 90 days after publication. For general industry and maritime, all obligations for compliance commence two years after the effective date, with two exceptions: The obligation for engineering controls commences five years after the effective date for hydraulic fracturing operations in the oil and gas industry; and the obligation for employers in general industry and maritime to offer medical surveillance commences two years after the effective date for employees exposed above the PEL, and four years after the effective date for employees exposed at or above the action level. For construction, all obligations for compliance commence one year after the effective date, with the exception that certain requirements for laboratory analysis commence two years after the effective date.

Under the OSH Act's legal standard directing OSHA to set health standards based on findings of significant risk of material impairment and technological and economic feasibility, OSHA does not use cost-benefit analysis to determine the PEL or other aspects of the rule. It does, however, determine and analyze costs and benefits for its own informational purposes and to meet certain Executive Order requirements, as discussed in Section VII. Summary of the Final Economic Analysis and Final Regulatory Flexibility Analysis and in the FEA. Table I-1—which is derived from material presented in Section VII of this preamble—provides a summary of OSHA's best estimate of the costs and benefits of the rule using a discount rate of 3 percent. As shown, the rule is estimated to prevent 642 fatalities and 918 moderate-to-severe silicosis cases annually once it is fully effective, and the estimated cost of the rule is $1,030 million annually. Also as shown in Table I-1, the discounted monetized benefits of the rule are estimated to be $8.7 billion annually, and the rule is estimated to generate net benefits of approximately $7.7 billion annually.

Start Printed Page 16289

II. Pertinent Legal Authority

The purpose of the Occupational Safety and Health Act (29 U.S.C. 651 et seq.) (“the Act” or “the OSH Act”), is “to assure so far as possible every working man and woman in the Nation safe and healthful working conditions and to preserve our human resources” (29 U.S.C. 651(b)). To achieve this goal Congress authorized the Secretary of Labor (“the Secretary”) “to set mandatory occupational safety and health standards applicable to businesses affecting interstate commerce” (29 U.S.C. 651(b)(3); see 29 U.S.C. 654(a) (requiring employers to comply with OSHA standards), 655(a) (authorizing summary adoption of existing consensus and federal standards within two years of the Act's enactment), and 655(b) (authorizing promulgation, modification or revocation of standards pursuant to notice and comment)). The primary statutory provision relied upon by the Agency in promulgating health standards is section 6(b)(5) of the Act; other sections of the OSH Act, however, authorize the Occupational Safety and Health Administration (OSHA) to require labeling and other appropriate forms of warning, exposure assessment, medical examinations, and recordkeeping in its standards (29 U.S.C. 655(b)(5), 655(b)(7), 657(c)).

The Act provides that in promulgating standards dealing with toxic materials or harmful physical agents, such as respirable crystalline silica, the Secretary shall set the standard which “most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health . . . even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life” (29 U.S.C. 655(b)(5)). Thus, “[w]hen Congress passed the Occupational Safety and Health Act in 1970, it chose to place pre-eminent value on assuring employees a safe and healthful working environment, limited only by the feasibility of achieving such an environment” (American Textile Mfrs. Institute, Inc. v. Donovan, 452 US 490, 541 (1981) (“Cotton Dust”)).

OSHA proposed this new standard for respirable crystalline silica and conducted its rulemaking pursuant to Start Printed Page 16290section 6(b)(5) of the Act ((29 U.S.C. 655(b)(5)). The preceding silica standard, however, was adopted under the Secretary's authority in section 6(a) of the OSH Act (29 U.S.C. 655(a)), to adopt national consensus and established Federal standards within two years of the Act's enactment (see 29 CFR 1910.1000 Table Z-1). Any rule that “differs substantially from an existing national consensus standard” must “better effectuate the purposes of this Act than the national consensus standard” (29 U.S.C. 655(b)(8)). Several additional legal requirements arise from the statutory language in sections 3(8) and 6(b)(5) of the Act (29 U.S.C. 652(8), 655(b)(5)). The remainder of this section discusses these requirements, which OSHA must consider and meet before it may promulgate this occupational health standard regulating exposure to respirable crystalline silica.

Material Impairment of Health

Subject to the limitations discussed below, when setting standards regulating exposure to toxic materials or harmful physical agents, the Secretary is required to set health standards that ensure that “no employee will suffer material impairment of health or functional capacity . . .” (29 U.S.C. 655(b)(5)). OSHA has, under this section, considered medical conditions such as irritation of the skin, eyes, and respiratory system, asthma, and cancer to be material impairments of health. What constitutes material impairment in any given case is a policy determination on which OSHA is given substantial leeway. “OSHA is not required to state with scientific certainty or precision the exact point at which each type of [harm] becomes a material impairment” (AFL-CIO v. OSHA, 965 F.2d 962, 975 (11th Cir. 1992)). Courts have also noted that OSHA should consider all forms and degrees of material impairment—not just death or serious physical harm (AFL-CIO, 965 F.2d at 975). Thus the Agency has taken the position that “subclinical” health effects, which may be precursors to more serious disease, can be material impairments of health that OSHA should address when feasible (43 FR 52952, 52954 (11/14/78) (Preamble to the Lead Standard)).

Significant Risk

Section 3(8) of the Act requires that workplace safety and health standards be “reasonably necessary or appropriate to provide safe or healthful employment” (29 U.S.C. 652(8)). The Supreme Court, in its decision on OSHA's benzene standard, interpreted section 3(8) to mean that “before promulgating any standard, the Secretary must make a finding that the workplaces in question are not safe” (Indus. Union Dep't, AFL-CIO v. Am. Petroleum Inst., 448 U.S. 607, 642 (1980) (plurality opinion) (“Benzene”)). The Court further described OSHA's obligation as requiring it to evaluate “whether significant risks are present and can be eliminated or lessened by a change in practices” (Benzene, 448 U.S. at 642). The Court's holding is consistent with evidence in the legislative record, with regard to section 6(b)(5) of the Act (29 U.S.C. 655(b)(5)), that Congress intended the Agency to regulate unacceptably severe occupational hazards, and not “to establish a utopia free from any hazards” or to address risks comparable to those that exist in virtually any occupation or workplace (116 Cong. Rec. 37614 (1970), Leg. Hist. 480-82). It is also consistent with Section 6(g) of the OSH Act, which states that, in determining regulatory priorities, “the Secretary shall give due regard to the urgency of the need for mandatory safety and health standards for particular industries, trades, crafts, occupations, businesses, workplaces or work environments” (29 U.S.C. 655(g)).

The Supreme Court in Benzene clarified that OSHA has considerable latitude in defining significant risk and in determining the significance of any particular risk. The Court did not specify a means to distinguish significant from insignificant risks, but rather instructed OSHA to develop a reasonable approach to making its significant risk determination. The Court stated that “[i]t is the Agency's responsibility to determine, in the first instance, what it considers to be a `significant' risk” (Benzene, 448 U.S. at 655), and it did not “express any opinion on the . . . difficult question of what factual determinations would warrant a conclusion that significant risks are present which make promulgation of a new standard reasonably necessary or appropriate” (Benzene, 448 U.S. at 659). The Court stated, however, that the section 6(f) (29 U.S.C. 655(b)(f)) substantial evidence standard applicable to OSHA's significant risk determination does not require the Agency “to support its finding that a significant risk exists with anything approaching scientific certainty” (Benzene, 448 U.S. at 656). Rather, OSHA may rely on “a body of reputable scientific thought” to which “conservative assumptions in interpreting the data . . . ” may be applied, “risking error on the side of overprotection” (Benzene, 448 U.S. at 656; see also United Steelworkers of Am., AFL-CIO-CLC v. Marshall, 647 F.2d 1189, 1248 (D.C. Cir. 1980) (“Lead I”) (noting the Benzene Court's application of this principle to carcinogens and applying it to the lead standard, which was not based on carcinogenic effects)). OSHA may thus act with a “pronounced bias towards worker safety” in making its risk determinations (Bldg & Constr. Trades Dep't v. Brock, 838 F.2d 1258, 1266 (D.C. Cir. 1988) (“Asbestos II”).

The Supreme Court further recognized that what constitutes “significant risk” is “not a mathematical straitjacket” (Benzene, 448 U.S. at 655) and will be “based largely on policy considerations” (Benzene, 448 U.S. at 655 n.62). The Court gave the following example:

If . . . the odds are one in a billion that a person will die from cancer by taking a drink of chlorinated water, the risk clearly could not be considered significant. On the other hand, if the odds are one in a thousand that regular inhalation of gasoline vapors that are 2% benzene will be fatal, a reasonable person might well consider the risk significant . . . (Benzene, 448 U.S. at 655).

Following Benzene, OSHA has, in many of its health standards, considered the one-in-a-thousand metric when determining whether a significant risk exists. Moreover, as “a prerequisite to more stringent regulation” in all subsequent health standards, OSHA has, consistent with the Benzene plurality decision, based each standard on a finding of significant risk at the “then prevailing standard” of exposure to the relevant hazardous substance (Asbestos II, 838 F.2d at 1263). Once a significant risk of material impairment of health is demonstrated, it is of no import that the incidence of the illness may be declining (see Nat'l Min. Assoc. v. Sec'y, U.S. Dep't of Labor, Nos. 14-11942, 14-12163, slip op. at 80 (11th Cir. Jan. 25, 2016) (interpreting the Mine Act, 30 U.S.C. 811(a)(6)(A), which contains the same language as section 6(b)(5) of the OSH Act requiring the Secretary to set standards that assure no employee will suffer material impairment of health)).

The Agency's final risk assessment is derived from existing scientific and enforcement data and its final conclusions are made only after considering all evidence in the rulemaking record. Courts reviewing the validity of these standards have uniformly held the Secretary to the significant risk standard first articulated by the Benzene plurality and have generally upheld the Secretary's significant risk determinations as supported by substantial evidence and “a reasoned explanation for his policy Start Printed Page 16291assumptions and conclusions” (Asbestos II, 838 F.2d at 1266).

Once OSHA makes its significant risk finding, the “more stringent regulation” (Asbestos II, 838 F.2d at 1263) it promulgates must be “reasonably necessary or appropriate” to reduce or eliminate that risk, within the meaning of section 3(8) of the Act (29 U.S.C. 652(8)) and Benzene (448 U.S. at 642) (see Asbestos II, 838 F.2d at 1269). The courts have interpreted section 6(b)(5) of the OSH Act as requiring OSHA to set the standard that eliminates or reduces risk to the lowest feasible level; as discussed below, the limits of technological and economic feasibility usually determine where the new standard is set (see UAW v. Pendergrass, 878 F.2d 389, 390 (D.C. Cir. 1989)). In choosing among regulatory alternatives, however, “[t]he determination that [one standard] is appropriate, as opposed to a marginally [more or less protective] standard, is a technical decision entrusted to the expertise of the agency. . . ” (Nat'l Mining Ass'n v. Mine Safety and Health Admin., 116 F.3d 520, 528 (D.C. Cir. 1997)) (analyzing a Mine Safety and Health Administration (“MSHA”) standard under the Benzene significant risk standard). In making its choice, OSHA may incorporate a margin of safety even if it theoretically regulates below the lower limit of significant risk (Nat'l Mining Ass'n, 116 F.3d at 528 (citing American Petroleum Inst. v. Costle, 665 F.2d 1176, 1186 (D.C. Cir. 1982))).

Working Life Assumption

The OSH Act requires OSHA to set the standard that most adequately protects employees against harmful workplace exposures for the period of their “working life” (29 U.S.C. 655(b)(5)). OSHA's longstanding policy is to define “working life” as constituting 45 years; thus, it assumes 45 years of exposure when evaluating the risk of material impairment to health caused by a toxic or hazardous substance. This policy is not based on empirical data that most employees are exposed to a particular hazard for 45 years. Instead, OSHA has adopted the practice to be consistent with the statutory directive that “no employee” suffer material impairment of health “even if” such employee is exposed to the hazard for the period of his or her working life (see 74 FR 44796 (8/31/09)). OSHA's policy was given judicial approval in a challenge to an OSHA standard that lowered the permissible exposure limit (PEL) for asbestos (Asbestos II, 838 F.2d at 1264-1265). In that case, the petitioners claimed that the median duration of employment in the affected industry sectors was only five years. Therefore, according to petitioners, OSHA erred in assuming a 45-year working life in calculating the risk of health effects caused by asbestos exposure. The D.C. Circuit disagreed, stating,

Even if it is only the rare worker who stays with asbestos-related tasks for 45 years, that worker would face a 64/1000 excess risk of contracting cancer; Congress clearly authorized OSHA to protect such a worker (Asbestos II, 838 F.2d at 1264-1265).

OSHA might calculate the health risks of exposure, and the related benefits of lowering the exposure limit, based on an assumption of a shorter working life, such as 25 years, but such estimates are for informational purposes only.

Best Available Evidence

Section 6(b)(5) of the Act requires OSHA to set standards “on the basis of the best available evidence” and to consider the “latest available scientific data in the field” (29 U.S.C. 655(b)(5)). As noted above, the Supreme Court, in its Benzene decision, explained that OSHA must look to “a body of reputable scientific thought” in making its material harm and significant risk determinations, while noting that a reviewing court must “give OSHA some leeway where its findings must be made on the frontiers of scientific knowledge” (Benzene, 448 U.S. at 656). The courts of appeals have afforded OSHA similar latitude to issue health standards in the face of scientific uncertainty. The Second Circuit, in upholding the vinyl chloride standard, stated:

. . . the ultimate facts here in dispute are `on the frontiers of scientific knowledge', and, though the factual finger points, it does not conclude. Under the command of OSHA, it remains the duty of the Secretary to act to protect the workingman, and to act even in circumstances where existing methodology or research is deficient (Society of the Plastics Industry, Inc. v. OSHA, 509 F.2d 1301, 1308 (2d Cir. 1975) (quoting Indus. Union Dep't, AFL-CIO v. Hodgson, 499 F.2d 467, 474 (D.C. Cir. 1974) (“Asbestos I”))).

The D.C. Circuit, in upholding the cotton dust standard, stated: “OSHA's mandate necessarily requires it to act even if information is incomplete when the best available evidence indicates a serious threat to the health of workers” (Am. Fed'n of Labor & Cong. of Indus. Orgs. v. Marshall, 617 F.2d 636, 651 (D.C. Cir. 1979), aff'd in part and vacated in part on other grounds, American Textile Mfrs. Inst., Inc. v. Donovan, 452 U.S. 490 (1981)).

When there is disputed scientific evidence, OSHA must review the evidence on both sides and “reasonably resolve” the dispute (Pub. Citizen Health Research Grp. v. Tyson, 796 F.2d 1479, 1500 (D.C. Cir. 1986)). In Public Citizen, there was disputed scientific evidence regarding whether there was a threshold exposure level for the health effects of ethylene oxide. The Court noted that, where “OSHA has the expertise we lack and it has exercised that expertise by carefully reviewing the scientific data,” a dispute within the scientific community is not occasion for it to take sides about which view is correct (Pub. Citizen Health Research Grp., 796 F.2d at 1500). “Indeed, Congress did `not [intend] that the Secretary be paralyzed by debate surrounding diverse medical opinions' ” (Pub. Citizen Health Research Grp., 796 F.2d at 1497 (quoting H.R.Rep. No. 91-1291, 91st Cong., 2d Sess. 18 (1970), reprinted in Legislative History of the Occupational Safety and Health Act of 1970 at 848 (1971))).

A recent decision by the Eleventh Circuit Court of Appeals upholding a coal dust standard promulgated by MSHA emphasized that courts should give “an extreme degree of deference to the agency when it is evaluating scientific data within its technical expertise” (Nat'l Min. Assoc. v. Sec'y, U.S. Dep't of Labor, Nos. 14-11942, 14-12163, slip op. at 43 (11th Cir. Jan. 25, 2016) (quoting Kennecott Greens Creek Min. Co. v. MSHA, 476 F.3d 946, 954-955 (D.C. Cir. 2007) (internal quotation marks omitted)). The Court emphasized that because the Mine Act, like the OSH Act, “evinces a clear bias in favor of [ ] health and safety,” the agency's responsibility to use the best evidence and consider feasibility should not be used as a counterweight to the agency's duty to protect the lives and health of workers (Nat'l Min. Assoc., Nos. 14-11942, 14-12163, slip op. at 43 (11th Cir. Jan. 25, 2016)).

Feasibility

The OSH Act requires that, in setting a standard, OSHA must eliminate the risk of material health impairment “to the extent feasible” (29 U.S.C. 655(b)(5)). The statutory mandate to consider the feasibility of the standard encompasses both technological and economic feasibility; these analyses have been done primarily on an industry-by-industry basis (Lead I, 647 F.2d at 1264, 1301) in general industry. The Agency has also used application groups, defined by common tasks, as the structure for its feasibility analyses in construction (Pub. Citizen Health Research Grp. v. OSHA, 557 F.3d 165, 177-179 (3d Cir. 2009) (“Chromium (VI)”). The Supreme Court has broadly defined feasible as “capable of being Start Printed Page 16292done” (Cotton Dust, 452 U.S. at 509-510).

Although OSHA must set the most protective PEL that the Agency finds to be technologically and economically feasible, it retains discretion to set a uniform PEL even when the evidence demonstrates that certain industries or operations could reasonably be expected to meet a lower PEL. OSHA health standards generally set a single PEL for all affected employers; OSHA exercised this discretion most recently in its final rule on occupational exposure to chromium (VI) (71 FR 10100, 10337-10338 (2/28/2006); see also 62 FR 1494, 1575 (1/10/97) (methylene chloride)). In its decision upholding the chromium (VI) standard, including the uniform PEL, the Court of Appeals for the Third Circuit addressed this issue as one of deference, stating “OSHA's decision to select a uniform exposure limit is a legislative policy decision that we will uphold as long as it was reasonably drawn from the record” (Chromium (VI), 557 F.3d at 183 (3d Cir. 2009)); see also Am. Iron & Steel Inst. v. OSHA, 577 F.2d 825, 833 (3d Cir. 1978)). OSHA's reasons for choosing one chromium (VI) PEL, rather than imposing different PELs on different application groups or industries, included: Multiple PELs would create enforcement and compliance problems because many workplaces, and even workers, were affected by multiple categories of chromium (VI) exposure; discerning individual PELs for different groups of establishments would impose a huge evidentiary burden on the Agency and unnecessarily delay implementation of the standard; and a uniform PEL would, by eliminating confusion and simplifying compliance, enhance worker protection (Chromium (VI), 557 F.3d at 173, 183-184). The Court held that OSHA's rationale for choosing a uniform PEL, despite evidence that some application groups or industries could meet a lower PEL, was reasonably drawn from the record and that the Agency's decision was within its discretion and supported by past practice (Chromium (VI), 557 F.3d at 183-184).

Technological Feasibility

A standard is technologically feasible if the protective measures it requires already exist, can be brought into existence with available technology, or can be created with technology that can reasonably be expected to be developed (Lead I, 647 F.2d at 1272; Amer. Iron & Steel Inst. v. OSHA, 939 F.2d 975, 980 (D.C. Cir. 1991) (“Lead II”)). While the test for technological feasibility is normally articulated in terms of the ability of employers to decrease exposures to the PEL, provisions such as exposure measurement requirements must also be technologically feasible (Forging Indus. Ass'n v. Sec'y of Labor, 773 F.2d 1436, 1453 (4th Cir. 1985)).

OSHA's standards may be “technology forcing,” i.e., where the Agency gives an industry a reasonable amount of time to develop new technologies, OSHA is not bound by the “technological status quo” (Lead I, 647 F.2d at 1264); see also Kennecott Greens Creek Min. Co. v. MSHA, 476 F.3d 946, 957 (D.C. Cir. 2007) (MSHA standards, like OSHA standards, may be technology-forcing); Nat'l Petrochemical & Refiners Ass'n v. EPA, 287 F.3d 1130, 1136 (D.C. Cir. 2002) (agency is “not obliged to provide detailed solutions to every engineering problem,” but only to “identify the major steps for improvement and give plausible reasons for its belief that the industry will be able to solve those problems in the time remaining.”).

In its Lead decisions, the D.C. Circuit described OSHA's obligation to demonstrate the technological feasibility of reducing occupational exposure to a hazardous substance.

[W]ithin the limits of the best available evidence . . . OSHA must prove a reasonable possibility that the typical firm will be able to develop and install engineering and work practice controls that can meet the PEL in most of its operations . . . The effect of such proof is to establish a presumption that industry can meet the PEL without relying on respirators . . . Insufficient proof of technological feasibility for a few isolated operations within an industry, or even OSHA's concession that respirators will be necessary in a few such operations, will not undermine this general presumption in favor of feasibility. Rather, in such operations firms will remain responsible for installing engineering and work practice controls to the extent feasible, and for using them to reduce . . . exposure as far as these controls can do so (Lead I, 647 F.2d at 1272).

Additionally, the D.C. Circuit explained that “[f]easibility of compliance turns on whether exposure levels at or below [the PEL] can be met in most operations most of the time . . .” (Lead II, 939 F.2d at 990).

Courts have given OSHA significant deference in reviewing its technological feasibility findings.

So long as we require OSHA to show that any required means of compliance, even if it carries no guarantee of meeting the PEL, will substantially lower . . . exposure, we can uphold OSHA's determination that every firm must exploit all possible means to meet the standard (Lead I, 647 F.2d at 1273).

Even in the face of significant uncertainty about technological feasibility in a given industry, OSHA has been granted broad discretion in making its findings (Lead I, 647 F.2d at 1285).

OSHA cannot let workers suffer while it awaits . . . scientific certainty. It can and must make reasonable [technological feasibility] predictions on the basis of `credible sources of information,' whether data from existing plants or expert testimony (Lead I, 647 F.2d at 1266 (quoting Am. Fed'n of Labor & Cong. of Indus. Orgs., 617 F.2d at 658)).

For example, in Lead I, the D.C. Circuit allowed OSHA to use, as best available evidence, information about new and expensive industrial smelting processes that had not yet been adopted in the U.S. and would require the rebuilding of plants (Lead I, 647 F.2d at 1283-1284). Even under circumstances where OSHA's feasibility findings were less certain and the Agency was relying on its “legitimate policy of technology forcing,” the D.C. Circuit approved of OSHA's feasibility findings when the Agency granted lengthy phase-in periods to allow particular industries time to comply (Lead I, 647 F.2d at 1279-1281, 1285).

OSHA is permitted to adopt a standard that some employers will not be able to meet some of the time, with employers limited to challenging feasibility at the enforcement stage (Lead I, 647 F.2d at 1273 & n. 125; Asbestos II, 838 F.2d at 1268). Even when the Agency recognized that it might have to balance its general feasibility findings with flexible enforcement of the standard in individual cases, the courts of appeals have generally upheld OSHA's technological feasibility findings (Lead II, 939 F.2d at 980; see Lead I, 647 F.2d at 1266-1273; Asbestos II, 838 F.2d at 1268). Flexible enforcement policies have been approved where there is variability in measurement of the regulated hazardous substance or where exposures can fluctuate uncontrollably (Asbestos II, 838 F.2d at 1267-1268; Lead II, 939 F.2d at 991). A common means of dealing with the measurement variability inherent in sampling and analysis is for the Agency to add the standard sampling error to its exposure measurements before determining whether to issue a citation (e.g., 51 FR 22612, 22654 (06/20/86) (Preamble to the Asbestos Standard)).

Economic Feasibility

In addition to technological feasibility, OSHA is required to demonstrate that its standards are economically feasible. A reviewing court will examine the cost of compliance with an OSHA standard “in relation to the financial health and Start Printed Page 16293profitability of the industry and the likely effect of such costs on unit consumer prices . . .” (Lead I, 647 F.2d at 1265 (omitting citation)). As articulated by the D.C. Circuit in Lead I,

OSHA must construct a reasonable estimate of compliance costs and demonstrate a reasonable likelihood that these costs will not threaten the existence or competitive structure of an industry, even if it does portend disaster for some marginal firms (Lead I, 647 F.2d at 1272).

A reasonable estimate entails assessing “the likely range of costs and the likely effects of those costs on the industry” (Lead I, 647 F.2d at 1266). As with OSHA's consideration of scientific data and control technology, however, the estimates need not be precise (Cotton Dust, 452 U.S. at 528-29 & n.54) as long as they are adequately explained. Thus, as the D.C. Circuit further explained:

Standards may be economically feasible even though, from the standpoint of employers, they are financially burdensome and affect profit margins adversely. Nor does the concept of economic feasibility necessarily guarantee the continued existence of individual employers. It would appear to be consistent with the purposes of the Act to envisage the economic demise of an employer who has lagged behind the rest of the industry in protecting the health and safety of employees and is consequently financially unable to comply with new standards as quickly as other employers. As the effect becomes more widespread within an industry, the problem of economic feasibility becomes more pressing (Asbestos I, 499 F.2d. at 478).

OSHA standards therefore satisfy the economic feasibility criterion even if they impose significant costs on regulated industries so long as they do not cause massive economic dislocations within a particular industry or imperil the very existence of the industry (Lead II, 939 F.2d at 980; Lead I, 647 F.2d at 1272; Asbestos I, 499 F.2d. at 478). As with its other legal findings, OSHA “is not required to prove economic feasibility with certainty, but is required to use the best available evidence and to support its conclusions with substantial evidence” (Lead II, 939 F.2d at 980-981) (citing Lead I, 647 F.2d at 1267)). Granting industries additional time to comply with new PELs may enhance the economic, as well as technological, feasibility of a standard (Lead I, 647 F.2d at 1265).

Because section 6(b)(5) of the Act explicitly imposes the “to the extent feasible” limitation on the setting of health standards, OSHA is not permitted to use cost-benefit analysis to make its standards-setting decisions (29 U.S.C. 655(b)(5)).

Congress itself defined the basic relationship between costs and benefits, by placing the “benefit” of worker health above all other considerations save those making attainment of this “benefit” unachievable. Any standard based on a balancing of costs and benefits by the Secretary that strikes a different balance than that struck by Congress would be inconsistent with the command set forth in § 6(b)(5) (Cotton Dust, 452 U.S. at 509).

Thus, while OSHA estimates the costs and benefits of its proposed and final rules, these calculations do not form the basis for the Agency's regulatory decisions; rather, they are performed in acknowledgement of requirements such as those in Executive Orders 12866 and 13563.

Structure of OSHA Health Standards

OSHA's health standards traditionally incorporate a comprehensive approach to reducing occupational disease. OSHA substance-specific health standards generally include the “hierarchy of controls,” which, as a matter of OSHA's preferred policy, mandates that employers install and implement all feasible engineering and work practice controls before respirators may be used. The Agency's adherence to the hierarchy of controls has been upheld by the courts (ASARCO, Inc. v. OSHA, 746 F.2d 483, 496-498 (9th Cir. 1984); Am. Iron & Steel Inst. v. OSHA, 182 F.3d 1261, 1271 (11th Cir. 1999)). In fact, courts view the legal standard for proving technological feasibility as incorporating the hierarchy:

OSHA must prove a reasonable possibility that the typical firm will be able to develop and install engineering and work practice controls that can meet the PEL in most of its operations. . . . The effect of such proof is to establish a presumption that industry can meet the PEL without relying on respirators (Lead I, 647 F.2d at 1272).

The hierarchy of controls focuses on removing harmful materials at their source. OSHA allows employers to rely on respiratory protection to protect their employees only when engineering and work practice controls are insufficient or infeasible. In fact, in the control of “those occupational diseases caused by breathing air contaminated with harmful dusts, fogs, fumes, mists, gases, smokes, sprays, or vapors,” the employers' primary objective “shall be to prevent atmospheric contamination. This shall be accomplished as far as feasible by accepted engineering control measures (for example, enclosure or confinement of the operation, general and local ventilation, and substitution of less toxic materials). When effective engineering controls are not feasible, or while they are being instituted, appropriate respirators shall be used pursuant to this section” (29 CFR 1910.134).

The reasons supporting OSHA's continued reliance on the hierarchy of controls, as well as its reasons for limiting the use of respirators, are numerous and grounded in good industrial hygiene principles (see Section XV, Summary and Explanation of the Standards, Methods of Compliance). Courts have upheld OSHA's emphasis on engineering and work practice controls over personal protective equipment in challenges to previous health standards, such as chromium (VI): “Nothing in . . . any case reviewing an airborne toxin standard, can be read to support a technological feasibility rule that would effectively encourage the routine and widespread use of respirators to comply with a PEL” (Chromium (VI), 557 F.3d at 179; see Am. Fed'n of Labor & Cong. of Indus. Orgs. v. Marshall, 617 F.2d 636, 653 (D.C. Cir. 1979) cert. granted, judgment vacated sub nom. Cotton Warehouse Ass'n v. Marshall, 449 U.S. 809 (1980) and aff'd in part, vacated in part sub nom. Am. Textile Mfrs. Inst., Inc. v. Donovan, 452 U.S. 490 (1981) (finding “uncontradicted testimony in the record that respirators can cause severe physical discomfort and create safety problems of their own”)).

In health standards such as this one, the hierarchy of controls is augmented by ancillary provisions. These provisions work with the hierarchy of controls and personal protective equipment requirements to provide comprehensive protection to employees in affected workplaces. Such provisions typically include exposure assessment, medical surveillance, hazard communication, and recordkeeping. This approach is recognized as effective in dealing with air contaminants such as respirable crystalline silica; for example, the industry standards for respirable crystalline silica, ASTM E 1132-06, Standard Practice for Health Requirements Relating to Occupational Exposure to Respirable Crystalline Silica, and ASTM E 2626-09, Standard Practice for Controlling Occupational Exposure to Respirable Crystalline Silica for Construction and Demolition Activities, take a similar comprehensive approach (Document ID 1466; 1504).

The OSH Act compels OSHA to require all feasible measures for reducing significant health risks (29 U.S.C. 655(b)(5); Pub. Citizen Health Research Grp., 796 F.2d at 1505 (“if in fact a STEL [short-term exposure limit] would further reduce a significant Start Printed Page 16294health risk and is feasible to implement, then the OSH Act compels the agency to adopt it (barring alternative avenues to the same result)”). When there is significant risk below the PEL, as is the case with respirable crystalline silica, the DC Circuit indicated that OSHA should use its regulatory authority to impose additional requirements on employers when those requirements will result in a greater than de minimis incremental benefit to workers' health (Asbestos II, 838 F.2d at 1274). The Supreme Court alluded to a similar issue in Benzene, pointing out that “in setting a permissible exposure level in reliance on less-than-perfect methods, OSHA would have the benefit of a backstop in the form of monitoring and medical testing” (Benzene, 448 U.S. at 657). OSHA believes that the ancillary provisions in this final standard provide significant benefits to worker health by providing additional layers and types of protection to employees exposed to respirable crystalline silica.

Finally, while OSHA is bound by evidence in the rulemaking record, and generally looks to its prior standards for guidance on how to structure and specify requirements in a new standard, it is not limited to past approaches to regulation. In promulgating health standards, “[w]henever practicable, the standard promulgated shall be expressed in terms of objective criteria and of the performance desired” (29 U.S.C. 655(b)(5)). In cases of industries or tasks presenting unique challenges in terms of assessing and controlling exposures, it may be more practicable and provide greater certainty to require specific controls with a demonstrated track record of efficacy in reducing exposures and, therefore, risk (especially when supplemented by appropriate respirator usage). Such an approach could more effectively protect workers than the traditional exposure assessment-and-control approach when exposures may vary because of factors such as changing environmental conditions or materials, and an assessment may not reflect typical exposures associated with a task or operation. As discussed at length in Section XV, Summary and Explanation of the Standards, the specified exposure control measures option in the construction standard (i.e., Table 1, in paragraph (c)(1)) for respirable crystalline silica represents the type of innovative, objective approach available to the Secretary when fashioning a rule under these circumstances.

III. Events Leading to the Final Standards

The Occupational Safety and Health Administration's (OSHA's) previous standards for workplace exposure to respirable crystalline silica were adopted in 1971, pursuant to section 6(a) of the Occupational Safety and Health Act (29 U.S.C. 651 et seq.) (“the Act” or “the OSH Act”) (36 FR 10466 (5/29/71)). Section 6(a) (29 U.S.C. 655(a)) authorized OSHA, in the first two years after the effective date of the Act, to promulgate “start-up” standards, on an expedited basis and without public hearing or comment, based on national consensus or established Federal standards that improved employee safety or health. Pursuant to that authority, OSHA in 1971 promulgated approximately 425 permissible exposure limits (PELs) for air contaminants, including crystalline silica, which were derived principally from Federal standards applicable to government contractors under the Walsh-Healey Public Contracts Act, 41 U.S.C. 35, and the Contract Work Hours and Safety Standards Act (commonly known as the Construction Safety Act), 40 U.S.C. 333. The Walsh-Healey Act and Construction Safety Act standards had been adopted primarily from recommendations of the American Conference of Governmental Industrial Hygienists (ACGIH).

For general industry (see 29 CFR 1910.1000, Table Z-3), the PEL for crystalline silica in the form of respirable quartz was based on two alternative formulas: (1) A particle-count formula, PELmppcf=250/(% quartz + 5) as respirable dust; and (2) a mass formula proposed by ACGIH in 1968, PEL=(10 mg/m3)/(% quartz + 2) as respirable dust. The general industry PELs for crystalline silica in the form of cristobalite and tridymite were one-half of the value calculated from either of the above two formulas for quartz. For construction (see 29 CFR 1926.55, Appendix A) and shipyards (see 29 CFR 1915.1000, Table Z), the formula for the PEL for crystalline silica in the form of quartz (PELmppcf=250/(% quartz + 5) as respirable dust), which requires particle counting, was derived from the 1970 ACGIH threshold limit value (TLV).[1] Based on the formulas, the PELs for quartz, expressed as time-weighted averages (TWAs), were approximately equivalent to 100 μg/m3 for general industry and 250 μg/m3 for construction and shipyards. The PELs were not supplemented by additional protective provisions—such as medical surveillance requirements—as are included in other OSHA standards. OSHA believes that the formula based on particle-counting technology used in the general industry, construction, and shipyard PELs has been rendered obsolete by respirable mass (gravimetric) sampling.

In 1974, the National Institute for Occupational Safety and Health (NIOSH), an agency within the Department of Health and Human Services created by the OSH Act and designed to carry out research and recommend standards for occupational safety and health hazards, evaluated crystalline silica as a workplace hazard and issued criteria for a recommended standard (29 U.S.C. 669, 671; Document ID 0388). NIOSH recommended that occupational exposure to crystalline silica be controlled so that no worker is exposed to a TWA of free (respirable crystalline) silica greater than 50 μg/m3 as determined by a full-shift sample for up to a 10-hour workday over a 40-hour workweek. The document also recommended a number of ancillary provisions for a standard, such as exposure monitoring and medical surveillance.

In December 1974, OSHA published an Advance Notice of Proposed Rulemaking (ANPRM) based on the recommendations in the NIOSH criteria document (39 FR 44771 (12/27/74)). In the ANPRM, OSHA solicited “public participation on the issues of whether a new standard for crystalline silica should be issued on the basis of the [NIOSH] criteria or any other information, and, if so, what should be the contents of a proposed standard for crystalline silica” (39 FR at 44771). OSHA also set forth the particular issues of concern on which comments were requested. The Agency did not issue a proposed rule or pursue a final rule for crystalline silica at that time.

As information on the health effects of silica exposure developed during the 1980s and 1990s, national and international classification organizations came to recognize crystalline silica as a human carcinogen. In June 1986, the International Agency for Research on Cancer (IARC), which is the specialized cancer agency within the World Health Organization, evaluated the available evidence regarding crystalline silica carcinogenicity and concluded, in 1987, that crystalline silica is probably carcinogenic to Start Printed Page 16295humans (http://monographs.iarc.fr/​ENG/​Monographs/​suppl7/​Suppl7.pdf). An IARC working group met again in October 1996 to evaluate the complete body of research, including research that had been conducted since the initial 1986 evaluation. IARC concluded, more decisively this time, that “crystalline silica inhaled in the form of quartz or cristobalite from occupational sources is carcinogenic to humans” (Document ID 2258, Attachment 8, p. 211). In 2012, IARC reaffirmed that “Crystalline silica in the form of quartz or cristobalite dust is carcinogenic to humans” (Document ID 1473, p. 396).

In 1991, in the Sixth Annual Report on Carcinogens, the U.S. National Toxicology Program (NTP), within the U.S. Department of Health and Human Services, concluded that respirable crystalline silica was “reasonably anticipated to be a human carcinogen” (as referenced in Document ID 1417, p. 1). NTP reevaluated the available evidence and concluded, in the Ninth Report on Carcinogens, that “respirable crystalline silica (RCS), primarily quartz dust occurring in industrial and occupational settings, is known to be a human carcinogen, based on sufficient evidence of carcinogenicity from studies in humans indicating a causal relationship between exposure to RCS and increased lung cancer rates in workers exposed to crystalline silica dust” (Document ID 1417, p. 1). ACGIH listed respirable crystalline silica (in the form of quartz) as a suspected human carcinogen in 2000, while lowering the TLV to 0.05 mg/m3 (50 μg/m3) (Document ID 1503, p. 15). ACGIH subsequently lowered the TLV for crystalline silica to 0.025 mg/m3 (25 μg/m3) in 2006, which is ACGIH's current recommended exposure limit (Document ID 1503, pp. 1, 15).

In 1989, OSHA established 8-hour TWA PELs of 0.1 mg/m3 (100 μg/m3) for quartz and 0.05 mg/m3 (50 μg/m3) for cristobalite and tridymite, as part of the Air Contaminants final rule for general industry (54 FR 2332 (1/19/89)). OSHA stated that these limits presented no substantial change from the Agency's former formula limits, but would simplify sampling procedures. In providing comments on the proposed rule, NIOSH recommended that crystalline silica be considered a potential carcinogen.

In 1992, OSHA, as part of the Air Contaminants proposed rule for maritime, construction, and agriculture, proposed the same PELs as for general industry, to make the PELs consistent across all the OSHA-regulated sectors (57 FR 26002 (6/12/92)). However, the U.S. Court of Appeals for the Eleventh Circuit vacated the 1989 Air Contaminants final rule for general industry (Am. Fed'n of Labor and Cong. of Indus. Orgs. v. OSHA, 965 F.2d 962 (1992)), and also mooted the proposed rule for maritime, construction, and agriculture. The Court's decision to vacate the rule forced the Agency to return to the original 1971 PELs for all compounds, including silica, adopted as section 6(a) standards.

In 1994, OSHA initiated a process to determine which safety and health hazards in the U.S. needed the most attention. A priority planning committee included safety and health experts from OSHA, NIOSH, and the Mine Safety and Health Administration (MSHA). The committee reviewed available information on occupational deaths, injuries, and illnesses and communicated extensively with representatives of labor, industry, professional and academic organizations, the States, voluntary standards organizations, and the public. The OSHA National Advisory Committee on Occupational Safety and Health and the Advisory Committee on Construction Safety and Health (ACCSH) also made recommendations. Rulemaking for crystalline silica exposure was one of the priorities designated by this process. OSHA indicated that crystalline silica would be added to the Agency's regulatory agenda as other standards were completed and resources became available.

In 1996, OSHA instituted a Special Emphasis Program (SEP) to step up enforcement of the crystalline silica standards. The SEP was intended to reduce worker silica dust exposures that can cause silicosis and lung cancer. It included extensive outreach designed to educate and train employers and employees about the hazards of silica and how to control them, as well as inspections to enforce the standards. Among the outreach materials available were slides presenting information on hazard recognition and crystalline silica control technology, a video on crystalline silica and silicosis, and informational cards for workers explaining crystalline silica, health effects related to exposure, and methods of control. The SEP provided guidance for targeting inspections of worksites that had employees at risk of developing silicosis. The inspections resulted in the collection of exposure data from the various worksites visited by OSHA's compliance officers.

As a follow-up to the SEP, OSHA undertook numerous non-regulatory actions to address silica exposures. For example, in October of 1996, OSHA launched a joint silicosis prevention effort with MSHA, NIOSH, and the American Lung Association (see https://www.osha.gov/​pls/​oshaweb/​owadisp.show_​document?​p_​table=​NEWS_​RELEASES&​p_​id=​14110). This public education campaign involved distribution of materials on how to prevent silicosis, including a guide for working safely with silica and stickers for hard hats to remind workers of crystalline silica hazards. Spanish language versions of these materials were also made available. OSHA and MSHA inspectors distributed materials at mines, construction sites, and other affected workplaces. The joint silicosis prevention effort included a National Conference to Eliminate Silicosis in Washington, DC, in March of 1997, which brought together approximately 650 participants from labor, business, government, and the health and safety professions to exchange ideas and share solutions regarding the goal of eliminating silicosis (see https://industrydocuments.library.ucsf.edu/​documentstore/​s/​h/​d/​p/​/shdp0052/​shdp0052.pdf).

In 1997, OSHA announced in its Unified Agenda under Long-Term Actions that it planned to publish a proposed rule on crystalline silica

. . . because the agency has concluded that there will be no significant progress in the prevention of silica-related diseases without the adoption of a full and comprehensive silica standard, including provisions for product substitution, engineering controls, training and education, respiratory protection and medical screening and surveillance. A full standard will improve worker protection, ensure adequate prevention programs, and further reduce silica-related diseases (62 FR 57755, 57758 (10/29/97)).

In November 1998, OSHA moved “Occupational Exposure to Crystalline Silica” to the pre-rule stage in the Regulatory Plan (63 FR 61284, 61303-61304 (11/9/98)). OSHA held a series of stakeholder meetings in 1999 and 2000 to get input on the rulemaking. Stakeholder meetings for all industry sectors were held in Washington, Chicago, and San Francisco. A separate stakeholder meeting for the construction sector was held in Atlanta.

OSHA initiated Small Business Regulatory Enforcement Fairness Act (SBREFA) proceedings in 2003, seeking the advice of small business representatives on the proposed rule (68 FR 30583, 30584 (5/27/03)). The SBREFA panel, including representatives from OSHA, the Small Business Administration's Office of Advocacy, and the Office of Management and Budget (OMB), was Start Printed Page 16296convened on October 20, 2003. The panel conferred with small entity representatives (SERs) from general industry, maritime, and construction on November 10 and 12, 2003, and delivered its final report, which included comments from the SERs and recommendations to OSHA for the proposed rule, to OSHA's Assistant Secretary on December 19, 2003 (Document ID 0937).

In 2003, OSHA examined enforcement data for the years 1997 to 2002 and identified high rates of noncompliance with the OSHA respirable crystalline silica PELs, particularly in construction. This period covers the first five years of the SEP. These enforcement data, presented in Table III-1, indicate that 24 percent of silica samples from the construction industry and 13 percent from general industry were at least three times the then-existing OSHA PELs. The data indicate that 66 percent of the silica samples obtained during inspections in general industry were in compliance with the PEL, while only 58 percent of the samples collected in construction were in compliance.

In an effort to expand the 1996 SEP, on January 24, 2008, OSHA implemented a National Emphasis Program (NEP) to identify and reduce or eliminate the health hazards associated with occupational exposure to crystalline silica (CPL-03-007 (1/24/08)). The NEP targeted worksites with elevated exposures to crystalline silica and included new program evaluation procedures designed to ensure that the goals of the NEP were measured as accurately as possible, detailed procedures for conducting inspections, updated information for selecting sites for inspection, development of outreach programs by each Regional and Area Office emphasizing the formation of voluntary partnerships to share information, and guidance on calculating PELs in construction and shipyards. In each OSHA Region, at least two percent of inspections every year are silica-related inspections. Additionally, the silica-related inspections are conducted at a range of facilities reasonably representing the distribution of general industry and construction work sites in that region.

A more recent analysis of OSHA enforcement data from January 2003 to December 2009 (covering the period of continued implementation of the SEP and the first two years of the NEP) shows that considerable noncompliance with the then-existing PELs continued to occur. These enforcement data, presented in Table III-2, indicate that 14 percent of silica samples from the construction industry and 19 percent for general industry were at least three times the OSHA PEL during this period. The data indicate that 70 percent of the silica samples obtained during inspections in general industry were in compliance with the PEL, and 75 percent of the samples collected in construction were in compliance.

Start Printed Page 16297

Both industry and worker groups have recognized that a comprehensive standard is needed to protect workers exposed to respirable crystalline silica. For example, ASTM International (originally known as the American Society for Testing and Materials) has published voluntary consensus standards for addressing the hazards of crystalline silica, and the Building and Construction Trades Department, AFL-CIO also has recommended a comprehensive program standard. These recommended standards include provisions for methods of compliance, exposure monitoring, training, and medical surveillance. The National Industrial Sand Association has also developed an occupational exposure program for crystalline silica that addresses exposure assessment and medical surveillance.

Throughout the crystalline silica rulemaking process, OSHA has presented information to, and consulted with, ACCSH and the Maritime Advisory Committee on Occupational Safety and Health. In December of 2009, OSHA representatives met with ACCSH to discuss the rulemaking and receive their comments and recommendations. On December 11, 2009, ACCSH passed motions supporting the concept of Table 1 in the draft proposed construction rule, recognizing that the controls listed in Table 1 are effective. As discussed with regard to paragraph (f) of the proposed standard for construction (paragraph (c) of the final standard for construction), Table 1 presents specified control measures for selected construction tasks. ACCSH also recommended that OSHA maintain the protective clothing provision found in the SBREFA panel draft regulatory text and restore the “competent person” requirement and responsibilities to the proposed rule. Additionally, the group recommended that OSHA move forward expeditiously with the rulemaking process.

In January 2010, OSHA completed a peer review of the draft Health Effects Analysis and Preliminary Quantitative Risk Assessment following procedures set forth by OMB in the Final Information Quality Bulletin for Peer Review, published on the OMB Web site on December 16, 2004 (see 70 FR 2664 (1/14/05)). Each peer reviewer submitted a written report to OSHA. The Agency revised its draft documents as appropriate and made the revised documents available to the public as part of its Notice of Proposed Rulemaking (NPRM). OSHA also made the written charge to the peer reviewers, the peer reviewers' names, the peer reviewers' reports, and the Agency's response to the peer reviewers' reports publicly available with publication of the proposed rule (Document ID 1711; 1716). Five of the seven original peer reviewers submitted post-hearing reports, commenting on OSHA's disposition of their original peer review comments in the proposed rule, as well as commenting on written and oral testimony presented at the silica hearing (Document ID 3574).

On August 23, 2013, OSHA posted its NPRM for respirable crystalline silica on its Web site and requested comments on the proposed rule. On September 12, 2013, OSHA published the NPRM in the Federal Register (78 FR 56273 (9/12/13)). In the NPRM, the Agency made a preliminary determination that employees exposed to respirable crystalline silica at the current PELs face a significant risk to their health and that promulgating the proposed standards would substantially reduce that risk. The NPRM required commenters to submit their comments by December 11, 2013. In response to stakeholder requests, OSHA extended the comment period until January 27, 2014 (78 FR 65242 (10/31/13)). On January 14, 2014, OSHA held a web chat to provide small businesses and other stakeholders an additional opportunity to obtain information from the Agency about the proposed rule. Subsequently, OSHA further extended the comment period to February 11, 2014 (79 FR 4641 (1/29/14)).

As part of the instructions for submitting comments, OSHA requested (but did not require) that parties submitting technical or scientific studies or research results and those submitting comments or testimony on the Agency's analyses disclose the nature of financial relationships with (e.g., consulting agreement), and extent of review by, parties interested in or Start Printed Page 16298affected by the rulemaking (78 FR 56274). Parties submitting studies or research results were also asked to disclose sources of funding and sponsorship for their research. OSHA intended for the disclosure of such information to promote the transparency and scientific integrity of evidence submitted to the record and stated that the request was consistent with Executive Order 13563.

The Agency received several comments related to this request. For example, an industrial hygiene engineer supported the disclosure of potential conflict of interest information (Document ID 2278, p. 5). Other commenters, such as congressional representatives and industry associations, opposed the request, asserting that it could lead to prejudgment or questioning of integrity, in addition to dissuading participation in the rulemaking; some also questioned the legality of such a request or OSHA's interpretation of Executive Order 13563 (e.g., Document ID 1811, p. 2; 2101, pp. 2-3). A number of stakeholders from academia and industry submitted information related to the request for funding, sponsorships, and review by interested parties (e.g., Document ID 1766, p. 1; 2004, p. 2; 2211, p. 2; 2195, p. 17). OSHA emphasizes that it reviewed and considered all evidence submitted to the record.

An informal public hearing on the proposed standards was held in Washington, DC from March 18 through April 4, 2014. Administrative Law Judges Daniel F. Solomon and Stephen L. Purcell presided over the hearing. The Agency heard testimony from over 200 stakeholders representing more than 70 organizations, such as public health groups, trade associations, and labor unions. Chief Administrative Law Judge Stephen L. Purcell closed the public hearing on April 4, 2014, allowing 45 days—until May 19, 2014—for participants who filed a notice of intention to appear at the hearings to submit additional evidence and data, and an additional 45 days—until July 3, 2014—to submit final briefs, arguments, and summations (Document ID 3589, Tr. 4415-4416). After the hearing concluded, OSHA extended the deadline to give those participants who filed a notice of intention to appear at the hearings until June 3, 2014 to submit additional information and data to the record, and until July 18, 2014 to submit final briefs and arguments (Document ID 3569). Based upon requests from stakeholders, the second deadline was extended, and parties who filed a notice of intention to appear at the hearing were given until August 18, 2014, to submit their final briefs and arguments (Document ID 4192).

OSHA provided the public with multiple opportunities to participate in the rulemaking process, including stakeholder meetings, the SBREFA panel, two comment periods (pre- and post-hearing), and a 14-day public hearing. Commenters were provided more than five months to comment on the rule before the hearing, and nearly as long to submit additional information, final briefs, and arguments after the hearing. OSHA received more than 2,000 comments on the silica NPRM during the entire pre-and post-hearing public participation period. In OSHA's view, therefore, the public was given sufficient opportunities and ample time to fully participate in this rulemaking.

The final rule on occupational exposure to respirable crystalline silica is based on consideration of the entire record of this rulemaking proceeding, including materials discussed or relied upon in the proposal, the record of the hearing, and all written comments and exhibits timely received. Thus, in promulgating this final rule, OSHA considered all comments in the record, including those that suggested that OSHA withdraw its proposal and merely enforce the existing silica standards, as well as those that argued the proposed rule was not protective enough. Based on this comprehensive record, OSHA concludes that employees exposed to respirable crystalline silica are at significant risk of developing silicosis and other non-malignant respiratory disease, lung cancer, kidney effects, and immune system effects. The Agency concludes that the PEL of 50 μg/m3 reduces the significant risks of material impairments of health posed to workers by occupational exposure to respirable crystalline silica to the maximum extent that is technologically and economically feasible. OSHA's substantive determinations with regard to the comments, testimony, and other information in the record, the legal standards governing the decision-making process, and the Agency's analysis of the data resulting in its assessments of risks, benefits, technological and economic feasibility, and compliance costs are discussed elsewhere in this preamble.

IV. Chemical Properties and Industrial Uses

Silica is a compound composed of the elements silicon and oxygen (chemical formula SiO2). Silica has a molecular weight of 60.08, and exists in crystalline and amorphous states, both in the natural environment and as produced during manufacturing or other processes. These substances are odorless solids, have no vapor pressure, and create non-explosive dusts when particles are suspended in air (Document ID 3637, pp. 1-3).

Silica is classified as part of the “silicate” class of minerals, which includes compounds that are composed of silicon and oxygen and which may also be bonded to metal ions or their oxides. The basic structural units of silicates are silicon tetrahedrons (SiO4), pyramidal structures with four triangular sides where a silicon atom is located in the center of the structure and an oxygen atom is located at each of the four corners. When silica tetrahedrons bond exclusively with other silica tetrahedrons, each oxygen atom is bonded to the silicon atom of its original ion, as well as to the silicon atom from another silica ion. This results in a ratio of one atom of silicon to two atoms of oxygen, expressed as SiO2. The silicon-oxygen bonds within the tetrahedrons use only one-half of each oxygen's total bonding energy. This leaves negatively charged oxygen ions available to bond with available positively charged ions. When they bond with metal and metal oxides, commonly of iron, magnesium, aluminum, sodium, potassium, and calcium, they form the silicate minerals commonly found in nature (Document ID 1334, p. 7).

In crystalline silica, the silicon and oxygen atoms are arranged in a three-dimensional repeating pattern. Silica is said to be polymorphic, as different forms are created when the silica tetrahedrons combine in different crystalline structures. The primary forms of crystalline silica are quartz, cristobalite, and tridymite. In an amorphous state, silicon and oxygen atoms are present in the same proportions but are not organized in a repeating pattern. Amorphous silica includes natural and manufactured glasses (vitreous and fused silica, quartz glass), biogenic silica, and opals, which are amorphous silica hydrates (Document ID 2258, Attachment 8, pp. 45-50).

Quartz is the most common form of crystalline silica and accounts for almost 12% by volume of the earth's crust. Alpha quartz, the quartz form that is stable below 573 °C, is the most prevalent form of crystalline silica found in the workplace. It accounts for the overwhelming majority of naturally found silica and is present in varying amounts in almost every type of mineral. Alpha quartz is found in igneous, sedimentary, and metamorphic rock, and all soils contain at least a trace amount of quartz (Document ID 1334, p. Start Printed Page 162999). Alpha quartz is used in many products throughout various industries and is a common component of building materials (Document ID 1334, pp. 11-15). Common trade names for commercially available quartz include: CSQZ, DQ 12, Min-U-Sil, Sil-Co-Sil, Snowit, Sykron F300, and Sykron F600 (Document ID 2258, Attachment 8, p. 43).

Cristobalite is a form of crystalline silica that is formed at high temperatures (>1470 °C). Although naturally occurring cristobalite is relatively rare, volcanic eruptions, such as Mount St. Helens, can release cristobalite dust into the air. Cristobalite can also be created during some processes conducted in the workplace. For example, flux-calcined diatomaceous earth is a material used as a filtering aid and as a filler in other products (Document ID 2258, Attachment 8, p. 44). It is produced when diatomaceous earth (diatomite), a geological product of decayed unicellular organisms called diatoms, is heated with flux. The finished product can contain between 40 and 60 percent cristobalite. Also, high temperature furnaces are often lined with bricks that contain quartz. When subjected to prolonged high temperatures, this quartz can convert to cristobalite.

Tridymite is another material formed at high temperatures (>870 °C) that is associated with volcanic activity. The creation of tridymite requires the presence of a flux such as sodium oxide. Tridymite is rarely found in nature and rarely reported in the workplace (Document ID 1424 pp. 5, 14).

When heated or cooled sufficiently, crystalline silica can transition between the polymorphic forms, with specific transitions occurring at different temperatures. At higher temperatures the linkages between the silica tetrahedrons break and reform, resulting in new crystalline structures. Quartz converts to cristobalite at 1470 °C, and at 1723 °C cristobalite loses its crystalline structure and becomes amorphous fused silica. These high temperature transitions reverse themselves at extremely slow rates, with different forms co-existing for a long time after the crystal cools (Document ID 2258, Attachment 8, p. 47).

Other types of transitions occur at lower temperatures when the silica-oxygen bonds in the silica tetrahedron rotate or stretch, resulting in a new crystalline structure. These low-temperature, or alpha to beta, transitions are readily and rapidly reversed as the crystal cools. At temperatures encountered by workers, only the alpha form of crystalline silica exists (Document ID 2258, Attachment 8, pp. 46-48).

Crystalline silica minerals produce distinct X-ray diffraction patterns, specific to their crystalline structure. The patterns can be used to distinguish the crystalline polymorphs from each other and from amorphous silica (Document ID 2258, Attachment 8, p. 45).

The specific gravity and melting point of silica vary between polymorphs. Silica is insoluble in water at 20 °C and in most acids, but its solubility increases with higher temperatures and pH, and it dissolves readily in hydrofluoric acid. Solubility is also affected by the presence of trace metals and by particle size. Under humid conditions water vapor in the air reacts with the surface of silica particles to form an external layer of silinols (SiOH). When these silinols are present the crystalline silica becomes more hydrophilic. Heating or acid washing reduces the amount of silinols on the surface area of crystalline silica particles. There is an external amorphous layer found in aged quartz, called the Beilby layer, which is not found on freshly cut quartz. This amorphous layer is more water soluble than the underlying crystalline core. Etching with hydrofluoric acid removes the Beilby layer as well as the principal metal impurities on quartz (Document ID 2258, Attachment 8, pp. 44-49).

Crystalline silica has limited chemical reactivity. It reacts with alkaline aqueous solutions, but does not readily react with most acids, with the exception of hydrofluoric acid. In contrast, amorphous silica and most silicates react with most mineral acids and alkaline solutions. Analytical chemists relied on this difference in acid reactivity to develop the silica point count analytical method that was widely used prior to the current X-ray diffraction and infrared methods (Document ID 2258, Attachment 8, pp. 48-51; 1355, p. 994).

Crystalline silica is used in industry in a wide variety of applications. Sand and gravel are used in road building and concrete construction. Sand with greater than 98% silica is used in the manufacture of glass and ceramics. Silica sand is used to form molds for metal castings in foundries, and in abrasive blasting operations. Silica is also used as a filler in plastics, rubber, and paint, and as an abrasive in soaps and scouring cleansers. Silica sand is used to filter impurities from municipal water and sewage treatment plants, and in hydraulic fracturing for oil and gas recovery (Document ID 1334, p. 11). Silica is also used to manufacture artificial stone products used as bathroom and kitchen countertops, and the silica content in those products can exceed 85 percent (Document ID 1477, pp. 3 and 11; 2178, Attachment 5, p. 420).

There are over 30 major industries and operations where exposures to crystalline silica can occur. They include such diverse workplaces as foundries, dental laboratories, concrete products and paint and coating manufacture, as well as construction activities including masonry cutting, drilling, grinding and tuckpointing, and use of heavy equipment during demolition activities involving silica-containing materials. A more detailed discussion of the industries affected by the proposed standard is presented in Section VII, Summary of the Final Economic Analysis and Final Regulatory Flexibility Analysis. Crystalline silica exposures can also occur in mining (which is under the jurisdiction of the Mine Safety and Health Administration), and in agriculture during plowing and harvesting.

V. Health Effects

A. Introduction

As discussed more thoroughly in Section II of this preamble, Pertinent Legal Authority, section 6(b)(5) of the Occupational Safety and Health Act (OSH Act or Act) requires the Secretary of Labor, in promulgating standards dealing with toxic materials or harmful physical agents, to “set the standard which most adequately assures, to the extent feasible, on the basis of the best available evidence, that no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life” (29 U.S.C. 655). Thus, in order to set a new health standard, the Secretary must determine that there is a significant risk of material impairment of health at the existing PEL and that issuance of a new standard will significantly reduce or eliminate that risk.

The Secretary's significant risk and material impairment determinations must be made “on the basis of the best available evidence” (29 U.S.C. 655(b)(5)). Although the Supreme Court, in its decision on OSHA's Benzene standard, explained that OSHA must look to “a body of reputable scientific thought” in making its material harm and significant risk determinations, the Court added that a reviewing court must “give OSHA some leeway where its findings must be made on the frontiers Start Printed Page 16300of scientific knowledge” (Indus. Union Dep't, AFL-CIO v. Am. Petroleum Inst., 448 U.S. 607, 656 (1980) (plurality opinion) (“Benzene”)). Thus, while OSHA's significant risk determination must be supported by substantial evidence, the Agency “is not required to support the finding that a significant risk exists with anything approaching scientific certainty” (Benzene, 448 U.S. at 656).

This section provides an overview of OSHA's material harm and significant risk determinations: (1) Summarizing OSHA's preliminary methods and findings from the proposal; (2) addressing public comments dealing with OSHA's evaluation of the scientific literature and methods used to estimate quantitative risk; and (3) presenting OSHA's final conclusions, with consideration of the rulemaking record, on the health effects and quantitative risk estimates associated with worker exposure to respirable crystalline silica. The quantitative risk estimates and significance of those risks are then discussed in detail in Section VI, Final Quantitative Risk Assessment and Significance of Risk.

B. Summary of Health and Risk Findings

As discussed in detail throughout this section and in Section VI, Final Quantitative Risk Assessment and Significance of Risk, OSHA finds, based upon the best available evidence in the published, peer-reviewed scientific literature, that exposure to respirable crystalline silica increases the risk of silicosis, lung cancer, other non-malignant respiratory disease (NMRD), and renal and autoimmune effects. In its Preliminary Quantitative Risk Assessment (QRA), OSHA used the best available exposure-response data from epidemiological studies to estimate quantitative risks. After carefully reviewing stakeholder comments on the Preliminary QRA and new information provided to the rulemaking record, OSHA finds there to be a clearly significant risk at the previous PELs for respirable crystalline silica (equivalent to approximately 100 μg/m3 for general industry and between 250 and 500 μg/m3 for construction/shipyards), with excess lifetime risk estimates for lung cancer mortality, silicosis mortality, and NMRD mortality each being much greater than 1 death per 1,000 workers exposed for a working life of 45 years. Cumulative risk estimates for silicosis morbidity are also well above 1 case per 1,000 workers exposed at the previous PELs. At the revised PEL of 50 μg/m3 respirable crystalline silica, these estimated risks are substantially reduced. Thus, OSHA concludes that the new PEL of 50 μg/m3 provides a large reduction in the lifetime and cumulative risk posed to workers exposed to respirable crystalline silica.

These findings and conclusions are consistent with those of the World Health Organization's International Agency for Research on Cancer (IARC), the U.S. Department of Health and Human Services' (HHS) National Toxicology Program (NTP), the National Institute for Occupational Safety and Health (NIOSH), and many other organizations and individuals, as evidenced in the rulemaking record and further discussed below. Many other scientific organizations and governments have recognized the strong body of scientific evidence pointing to the health risks of respirable crystalline silica and have deemed it necessary to take action to reduce those risks. As far back as 1974, NIOSH recommended that the exposure limit for crystalline silica be reduced to 50 μg/m3 (Document ID 2177b, p. 2). In 2000, the American Conference of Governmental Industrial Hygienists (ACGIH), a professional society that has recommended workplace exposure limits for six decades, revised their Threshold Limit Value (TLV) for respirable crystalline silica to 50 μg/m3 and has since further lowered its TLV for respirable crystalline silica to 25 μg/m3. OSHA is setting its revised PEL at 50 μg/m3 based on consideration of the body of evidence describing the health risks of crystalline silica as well as on technological feasibility considerations, as discussed in Section VII of this preamble and Chapter IV of the Final Economic Analysis and Final Regulatory Flexibility Analysis (FEA).

To reach these conclusions, OSHA performed an extensive search and review of the peer-reviewed scientific literature on the health effects of inhalation exposure to crystalline silica, particularly silicosis, lung cancer, other NMRD, and renal and autoimmune effects (Document ID 1711, pp. 7-265). Based upon this review, OSHA preliminarily determined that there was substantial evidence that exposure to respirable crystalline silica increases the risk of silicosis, lung cancer, NMRD, and renal and autoimmune effects (Document ID 1711, pp. 164, 181-208, 229). OSHA also found there to be suitable exposure-response data from many well-conducted epidemiological studies that permitted the Agency to estimate quantitative risks for lung cancer mortality, silicosis and NMRD mortality, renal disease mortality, and silicosis morbidity (Document ID 1711, p. 266).

As part of the preliminary quantitative risk assessment, OSHA calculated estimates of the risk of silica-related diseases assuming exposure over a working life (45 years) to 25, 50, 100, 250, and 500 μg/m3 respirable crystalline silica (corresponding to cumulative exposures over 45 years to 1.125, 2.25, 4.5, 11.25, and 22.5 mg/m3-yrs) (see Bldg & Constr. Trades Dep't v. Brock, 838 F.2d 1258, 1264-65 (D.C. Cir. 1988) approving OSHA's policy of using 45 years for the working life of an employee in setting a toxic substance standard). To estimate lifetime excess mortality risks at these exposure levels, OSHA used, for each key study, the exposure-response risk model(s) and regression coefficient from the model(s) in a life table analysis that accounted for competing causes of death due to background causes and cumulated risk through age 85 (Document ID 1711, pp. 360-378). For these analyses, OSHA used lung cancer, NMRD, or renal disease mortality and all-cause mortality rates to account for background risks and competing risks (U.S. 2006 data for lung cancer and NMRD mortality in all males, 1998 data for renal disease mortality, obtained from cause-specific death rate tables published by the National Center for Health Statistics (2009, Document ID 1104)). The mortality risk estimates were presented in terms of lifetime excess risk per 1,000 workers for exposure over an 8-hour working day, 250 days per year, and a 45-year working lifetime. For silicosis morbidity, OSHA based its risk estimates on the cumulative risk model(s) used in each study to develop quantitative exposure-response relationships. These models characterized the risk of developing silicosis, as detected by chest radiography, up to the time that cohort members, including both active and retired workers, were last examined (78 FR 56273, 56312 (9/12/13)).

OSHA then combined its review of the health effects literature and preliminary quantitative risk assessment into a draft document, entitled “Occupational Exposure to Respirable Crystalline Silica—Review of Health Effects Literature and Preliminary Quantitative Risk Assessment,” and submitted it to a panel of scientific experts [2] for independent peer review, Start Printed Page 16301in accordance with the Office of Management and Budget's (OMB) “Final Information Quality Bulletin for Peer Review” (Document ID 1336). The peer reviewers reviewed OSHA's draft Review of Health Effects Literature and Preliminary QRA. The peer-review panel responded to nearly 20 charge questions from OSHA and commented on various aspects of OSHA's analysis (Document ID 1716).

Overall, the peer reviewers found that OSHA was very thorough in its review of the literature and was reasonable in its interpretation of the studies with regards to the various endpoints examined, such that the Agency's conclusions on health effects were generally well founded (Document ID 1711, p. 381). The reviewers had various comments on OSHA's draft Preliminary QRA (Document ID 1716, pp. 107-218). OSHA provided a response to each comment in the Review of Health Effects Literature and Preliminary QRA and, where appropriate, made revisions (Document ID 1711, pp. 381-399). The Agency then placed the Review of Health Effects Literature and Preliminary QRA into the rulemaking docket as a background document (Document ID 1711). With the publication of the Notice of Proposed Rulemaking (78 FR 56723 on 9/12/13), all aspects of the Review of Health Effects Literature and Preliminary QRA were open for public comment.

Following the publication of the proposed rule (78 FR 56273 (9/12/13)) and accompanying revised Review of Health Effects Literature and Preliminary QRA (Document ID 1711), the peer reviewers were invited to review the revised analysis, examine the written comments in the docket, and attend the public hearing to listen to oral testimony as it applied to the health effects and quantitative risk assessment. Five peer reviewers were available and attended. In their final comments, provided to OSHA following the hearings, all five peer reviewers indicated that OSHA had adequately addressed their original comments (Document ID 3574). The peer reviewers also offered additional comments on concerns raised during the hearing. Many of the reviewers commented on the difficulty of evaluating exposure-response thresholds, and responded to public comments regarding causation and other specific issues (Document ID 3574). OSHA has incorporated many of the peer reviewers' additional comments into its risk assessment discussion in the preamble. Thus, OSHA believes that the external, independent peer-review process supports and lends legitimacy to its risk assessment methods and findings.

OSHA also received substantial public comment and testimony from a wide variety of stakeholders supporting its Review of Health Effects Literature and Preliminary QRA. In general, supportive comments and testimony were received from NIOSH (Document ID 2177; 3998; 4233), the public health and medical community, labor unions, affected workers, private citizens, and others.

Regarding health effects, NIOSH commented that the adverse health effects of exposure to respirable crystalline silica are “well-known, long lasting, and preventable” (Document ID 2177b, p. 2). Darius Sivin, Ph.D., of the UAW, commented, “[o]ccupational exposure to silica has been recognized for centuries as a serious workplace hazard” (Document ID 2282, Attachment 3, p. 4). Similarly, David Goldsmith, Ph.D., testified:

There have been literally thousands of research studies on exposure to crystalline silica in the past 30 years. Almost every study tells the occupational research community that workers need better protection to prevent severe chronic respiratory diseases, including lung cancer and other diseases in the future. What OSHA is proposing to do in revising the workplace standard for silica seems to be a rational response to the accumulation of published evidence (Document ID 3577, Tr. 865-866).

Franklin Mirer, Ph.D., CIH, Professor of Environmental and Occupational Health at CUNY School of Public Health, on behalf of the American Federation of Labor and Congress of Industrial Organizations (AFL-CIO), reiterated that silica “is a clear and present danger to workers health at exposure levels prevailing now in a large number of industries. Workers are at significant risk for mortality and illnesses including lung cancer and non-malignant respiratory disease including COPD, and silicosis” (Document ID 2256, Attachment 3, p. 3). The AFL-CIO also noted that there is “overwhelming evidence in the record that exposure to respirable crystalline silica poses a significant health risk to workers” (Document ID 4204, p. 11). The Building and Construction Trades Department, AFL-CIO, further commented that the rulemaking record “clearly supports OSHA's risk determination” (Document ID 4223, p. 2). Likewise, the Sorptive Minerals Institute, a national trade association, commented, “It is beyond dispute that OSHA has correctly determined that industrial exposure to certain types of silica can cause extremely serious, sometimes even fatal disease. In the massive rulemaking docket being compiled by the Agency, credible claims to the contrary are sparse to non-existent” (Document ID 4230, p. 8). OSHA also received numerous comments supportive of the revised standard from affected workers and citizens (e.g., Document ID 1724, 1726, 1731, 1752, 1756, 1759, 1762, 1764, 1787, 1798, 1800, 1802).

Regarding OSHA's literature review for its quantitative risk assessment, the American Public Health Association (APHA) and the National Consumers League (NCL) commented, “OSHA has thoroughly reviewed and evaluated the peer-reviewed literature on the health effects associated with exposure to respirable crystalline silica. OSHA's quantitative risk assessment is sound. The agency has relied on the best available evidence and acted appropriately in giving greater weight to those studies with the most robust designs and statistical analyses” (Document ID 2178, Attachment 1, p. 1; 2373, p. 1).

Dr. Mirer, who has served on several National Academy of Sciences committees setting risk assessment guidelines, further commented that OSHA's risk analysis is “scientifically correct, and consistent with the latest thinking on risk assessment,” (Document ID 2256, Attachment 3, p. 3), citing the National Academies' National Research Council's Science and Decisions: Advancing Risk Assessment (Document ID 4052), which makes technical recommendations on risk assessment and risk-based decision making (Document ID 3578, Tr. 935-936). In post-hearing comments expanding on this testimony, the AFL-CIO also noted that OSHA's risk assessment methodologies are transparent and consistent with practices recommended by the National Research Council in its publication, Risk Assessment in the Federal Government: Managing the Process, and with the Environmental Protection Agency's Guidelines for Carcinogenic Risk Assessment (Document ID 4204, p. 20). Similarly, Kyle Steenland, Ph.D., Professor in the Department of Environmental Health at Rollins School of Public Health, Emory University, one of the researchers on whose studies OSHA relied, testified that “OSHA has Start Printed Page 16302done a very capable job in conducting the summary of the literature and doing its own risk assessment” (Document ID 3580, Tr. 1235). Collectively, these comments and testimony support OSHA's use of the best available evidence and methods to estimate quantitative risks of lung cancer mortality, silicosis and NMRD mortality, renal disease mortality, and silicosis morbidity from exposure to respirable crystalline silica.

Based on OSHA's Preliminary QRA, many commenters recognized that reducing the permissible exposure limit is necessary to reduce significant risks presented by exposure to respirable crystalline silica (Document ID 4204, pp. 11-12; 2080, p. 1; 2339, p. 2). For example, the AFL-CIO stated that “OSHA based its proposal on more than adequate evidence, but more recent publications have described further the risk posed by silica exposure, and further justify the need for new silica standards” (Document ID 4204, pp. 11-12). Similarly, the American Society of Safety Engineers (ASSE) remarked that “[w]hile some may debate the science underlying the findings set forth in the proposed rule, overexposure to crystalline silica has been linked to occupational illness since the time of the ancient Greeks, and reduction of the current permissible exposure limit (PEL) to that recommended for years by the National Institute for Occupational Safety and Health (NIOSH) is long overdue” (Document ID 2339, p. 2).

Not every commenter agreed, however, as OSHA also received critical comments and testimony from various employers and their representatives, as well as some organizations representing affected industries. In general, these comments were critical of the underlying studies on which OSHA relied for its quantitative risk assessment, or with the methods used by OSHA to estimate quantitative risks. Some commenters also presented additional studies for OSHA to consider. OSHA thoroughly reviewed these and did not find them adequate to alter OSHA's overall conclusions of health risk, as discussed in great detail in the sections that follow.

After considering the evidence and testimony in the record, as discussed below, OSHA affirms its approach to quantify health risks related to exposure to respirable crystalline silica and the Agency's preliminary conclusions. In the final risk assessment that is now presented as part of this final rule in Section VI, Final Quantitative Risk Assessment and Significance of Risk, OSHA concludes that there is a clearly significant risk at the previous PELs for respirable crystalline silica, with excess lifetime risk estimates for lung cancer mortality, silicosis mortality, and NMRD mortality each being much greater than 1 death per 1,000 workers as a result of exposure for 45 working years (see Section VI, Final Quantitative Risk Assessment and Significance of Risk). At the revised PEL of 50 µg/m3 respirable crystalline silica, OSHA finds the estimated risks to be substantially reduced. Cumulative risk estimates for silicosis morbidity are also well above 1 case per 1,000 workers at the previous PELs, with a substantial reduction at the revised PEL (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1).

The health effects associated with silica exposure are well-established and supported by the record. Based on the record evidence, OSHA concludes that exposure to respirable crystalline silica causes silicosis and is the only known cause of silicosis. This causal relationship has long been accepted in the scientific and medical communities. In fact, the Department of Labor produced a video in 1938 featuring then Secretary of Labor Frances Perkins discussing the occurrence of silicosis among workers exposed to silica (see https://www.osha.gov/​silica/​index.html). Silicosis is a progressive disease induced by the inflammatory effects of respirable crystalline silica in the lung, which leads to lung damage and scarring and, in some cases, progresses to complications resulting in disability and death (see Section VI, Final Quantitative Risk Assessment and Significance of Risk). OSHA used a weight-of-evidence approach to evaluate the scientific studies in the literature to determine their overall quality and whether there is substantial evidence that exposure to respirable crystalline silica increases the risk of a particular health effect.

For lung cancer, OSHA reviewed the published, peer-reviewed scientific literature, including 60 epidemiological studies covering more than 30 occupational groups in over a dozen industrial sectors (see Document ID 1711, pp. 77-170). Based on this comprehensive review, and after considering the rulemaking record as a whole, OSHA concludes that the data provide ample evidence that exposure to respirable crystalline silica increases the risk of lung cancer among workers (see Document ID 1711, p. 164). OSHA's conclusion is consistent with that of IARC, which is the specialized cancer agency that is part of the World Health Organization and utilizes interdisciplinary (e.g., biostatistics, epidemiology, and laboratory sciences) experts to comprehensively identify the causes of cancer. In 1997, IARC classified respirable crystalline silica dust, in the form of quartz or cristobalite, as Group 1, i.e., “carcinogenic to humans,” following a thorough expert committee review of the peer-reviewed scientific literature (Document ID 2258, Attachment 8, p. 211). OSHA notes that IARC classifications and accompanying monographs are well recognized in the scientific community, having been described as “the most comprehensive and respected collection of systematically evaluated agents in the field of cancer epidemiology” (Demetriou et al., 2012, Document ID 4131, p. 1273). For silica, IARC's overall finding was based on studies of nine occupational cohorts that it considered to be the least influenced by confounding factors (see Document ID 1711, p. 76). OSHA included these studies in its review, in addition to several other studies (Document ID 1711, pp. 77-170).

Since IARC's 1997 determination that respirable crystalline silica is a Group 1 carcinogen, the scientific community has reaffirmed the soundness of this finding. In March of 2009, 27 scientists from eight countries participated in an additional IARC review of the scientific literature and reaffirmed that respirable crystalline silica dust is a Group 1 human carcinogen (Document ID 1473, p. 396). Additionally, in 2000, the NTP, which is a widely-respected interagency program under HHS that evaluates chemicals for possible toxic effects on public health, also concluded that respirable crystalline silica is a known human carcinogen (Document ID 1164, p. 1).

For NMRD other than silicosis, based on its review of several studies and all subsequent record evidence, OSHA concludes that exposure to respirable crystalline silica increases the risk of emphysema, chronic bronchitis, and pulmonary function impairment (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, pp. 181-208). For renal disease, OSHA reviewed the epidemiological literature and finds that a number of epidemiological studies reported statistically significant associations between occupational exposure to silica dust and chronic renal disease, subclinical renal changes, end-stage renal disease morbidity, chronic renal disease mortality, and granulomatosis with polyangitis (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 228). For autoimmune effects, OSHA reviewed Start Printed Page 16303epidemiological information in the record suggesting an association between respirable crystalline silica exposure and increased risk of systemic autoimmune diseases, including scleroderma, rheumatoid arthritis, and systemic lupus erythematosus (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 229). Therefore, OSHA concludes that there is substantial evidence that silica exposure increases the risks of renal and of autoimmune disease (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 229).

OSHA also finds there to be suitable exposure-response data from many well-conducted studies that permit the Agency to estimate quantitative risks for lung cancer mortality, silicosis and NMRD mortality, renal disease mortality, and silicosis morbidity (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 266). OSHA believes the exposure-response data in these studies collectively represent the best available evidence for use in estimating the quantitative risks related to silica exposure. For lung cancer mortality, OSHA relies upon a number of published studies that analyzed exposure-response relationships between respirable crystalline silica and lung cancer. These included studies of cohorts from several industry sectors: Diatomaceous earth workers (Rice et al., 2001, Document ID 1118), Vermont granite workers (Attfield and Costello, 2004, Document ID 0285), North American industrial sand workers (Hughes et al., 2001, Document ID 1060), and British coal miners (Miller and MacCalman, 2009, Document ID 1306). These studies are scientifically sound due to their sufficient size and adequate years of follow-up, sufficient quantitative exposure data, lack of serious confounding by exposure to other occupational carcinogens, consideration (for the most part) of potential confounding by smoking, and absence of any apparent selection bias (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 165). They all demonstrated positive, statistically significant exposure-response relationships between exposure to crystalline silica and lung cancer mortality. Also compelling was a pooled analysis (Steenland et al., 2001a, Document ID 0452) of 10 occupational cohorts (with a total of 65,980 workers and 1,072 lung cancer deaths), which was also used as a basis for IARC's 2009 reaffirmation of respirable crystalline silica as a human carcinogen. This analysis by Steenland et al. found an overall positive exposure-response relationship between cumulative exposure to crystalline silica and lung cancer mortality (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, pp. 269-292). Based on these studies, OSHA estimates that the lifetime lung cancer mortality excess risk associated with 45 years of exposure to respirable crystalline silica ranges from 11 to 54 deaths per 1,000 workers at the previous general industry PEL of 100 µg/m3 respirable crystalline silica, and 5 to 23 deaths per 1,000 workers at the revised PEL of 50 µg/m3 respirable crystalline silica (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1). These estimates exceed by a substantial margin the one in a thousand benchmark that OSHA has generally applied to its health standards following the Supreme Court's Benzene decision (448 U.S. 607, 655 (1980)).

For silicosis and NMRD mortality, OSHA relies upon two published, peer-reviewed studies: A pooled analysis of silicosis mortality data from six epidemiological studies (Mannetje et al., 2002b, Document ID 1089), and an exposure-response analysis of NMRD mortality among diatomaceous earth workers (Park et al, 2002, Document ID 0405) (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 292). The pooled analysis had a total of 18,634 subjects, 150 silicosis deaths, and 20 deaths from unspecified pneumoconiosis, and demonstrated an increasing mortality rate with silica exposure (Mannetje et al., 2002b, Document ID 1089; see also 1711, pp. 292-295). To estimate the risks of silicosis mortality, OSHA used the model described by Mannetje et al. but used rate ratios that were estimated from a sensitivity analysis conducted by ToxaChemica, Inc. that was expected to better control for age and exposure measurement uncertainty (2004, Document ID 0469; 1711, p. 295). OSHA's estimate of lifetime silicosis mortality risk is 11 deaths per 1,000 workers at the previous general industry PEL, and 7 deaths per 1,000 workers at the revised PEL (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1).

The NMRD analysis by Park et al. (2002, Document 0405) included pneumoconiosis (including silicosis), chronic bronchitis, and emphysema, since silicosis is a cause of death that is often misclassified as emphysema or chronic bronchitis (see Document ID 1711, p. 295). Positive exposure-response relationships were found between exposure to crystalline silica and excess risk for NMRD mortality (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, pp. 204-206, 295-297). OSHA's estimate of excess lifetime NMRD mortality risk, calculated using the results from Park et al., is 85 deaths per 1,000 workers at the previous general industry PEL of 100 µg/m[3] respirable crystalline silica, and 44 deaths per 1,000 workers at the revised PEL (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1).[3]

For renal disease mortality, Steenland et al. (2002a, Document ID 0448) conducted a pooled analysis of three cohorts (with a total of 13,382 workers) that found a positive exposure-response relationship for both multiple-cause mortality (i.e., any mention of renal disease on the death certificate) and underlying cause mortality. OSHA used the Steenland et al. (2002a, Document ID 0448) pooled analysis to estimate risks, given its large number of workers from cohorts with sufficient exposure data (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, pp. 314-315). OSHA's analysis for renal disease mortality shows estimated lifetime excess risk of 39 deaths per 1,000 workers at the previous general industry PEL of 100 µg/m3 respirable crystalline silica, and 32 deaths per 1,000 workers exposed at the revised PEL of 50 µg/m3 (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1). OSHA acknowledges, however, that there are considerably less data for renal disease mortality, and thus the findings based on them are less robust than those for silicosis, lung cancer, and NMRD mortality (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 229). For autoimmune disease, there were no quantitative exposure-response data available for a quantitative risk assessment (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 229).Start Printed Page 16304

For silicosis morbidity, OSHA reviewed the principal studies available in the scientific literature that have characterized the risk to exposed workers of acquiring silicosis, as detected by the appearance of opacities on chest radiographs (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, p. 357). The most reliable estimates of silicosis morbidity came from five studies that evaluated radiographs over time, including after workers left employment: The U.S. gold miner cohort studied by Steenland and Brown (1995b, Document ID 0451); the Scottish coal miner cohort studied by Buchanan et al. (2003, Document ID 0306); the Chinese tin mining cohort studied by Chen et al. (2001, Document ID 0332); the Chinese tin, tungsten, and pottery worker cohorts studied by Chen et al. (2005, Document ID 0985); and the South African gold miner cohort studied by Hnizdo and Sluis-Cremer (1993, Document ID 1052) (see Section VI, Final Quantitative Risk Assessment and Significance of Risk; Document ID 1711, pp. 316-343). These studies demonstrated positive exposure-response relationships between exposure to crystalline silica and silicosis risk. Based on the results of these studies, OSHA estimates a cumulative risk for silicosis morbidity of between 60 and 773 cases per 1,000 workers for a 45-year exposure to the previous general industry PEL of 100 µg/m3 respirable crystalline silica depending upon the study used, and between 20 and 170 cases per 1,000 workers exposed at the new PEL of 50 µg/m3 depending upon the study used (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1). Thus, like OSHA's risk estimates for other health endpoints, the risk is substantially lower, though still significant, at the revised PEL.

In conclusion, OSHA finds, based on the best available evidence and methods to estimate quantitative risks of disease resulting from exposure to respirable crystalline silica, that there are significant risks of material health impairment at the former PELs for respirable crystalline silica, which would be substantially reduced (but not entirely eliminated) at the new PEL of 50 μg/m3. In meeting its legal burden to estimate the health risks posed by respirable crystalline silica, OSHA has used the best available evidence and methods to estimate quantitative risks of disease resulting from exposure to respirable crystalline silica. As a result, the Agency finds that the lifetime excess mortality risks (for lung cancer, NMRD and silicosis, and renal disease) and cumulative risk (silicosis morbidity) posed to workers exposed to respirable crystalline silica over a working life represent significant risks that warrant mitigation, and that these risks will be substantially reduced at the revised PEL of 50 μg/m3 respirable crystalline silica.

C. Summary of the Review of Health Effects Literature and Preliminary QRA

As noted above, a wide variety of stakeholders offered comments and testimony in this rulemaking on issues related to health and risk. Many of these comments were submitted in response to OSHA's preliminary risk and material impairment determinations, which were presented in two background documents, entitled “Occupational Exposure to Respirable Crystalline Silica—Review of Health Effects Literature and Preliminary Quantitative Risk Assessment” (Document ID 1711) and “Supplemental Literature Review of Epidemiological Studies on Lung Cancer Associated with Exposure to Respirable Crystalline Silica” (Document ID 1711, Attachment 1), and summarized in the proposal in Section V, Health Effects Summary, and Section VI, Summary of OSHA's Preliminary Quantitative Risk Assessment.

In this subsection, OSHA summarizes the major findings of the two background documents. The Agency intends for this subsection to provide the detailed background necessary to fully understand stakeholders' comments and OSHA's responses.

1. Background

As noted above, OSHA's Review and Supplemental Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (Document ID 1711; 1711, Attachment 1) were the result of the Agency's extensive search and review of the peer-reviewed scientific literature on the health effects of inhalation exposure to crystalline silica, particularly silicosis, lung cancer and cancer at other sites, non-malignant respiratory diseases (NMRD) other than silicosis, and renal and autoimmune effects. The purposes of this detailed search and scientific review were to determine the nature of the hazards presented by exposure to respirable crystalline silica, and to evaluate whether there was an adequate basis, with suitable data availability, for quantitative risk assessment.

Much of the scientific evidence that describes the health effects and risks associated with exposure to crystalline silica consisted of epidemiological studies of worker populations; OSHA also reviewed animal and in vitro studies. OSHA used a weight-of-evidence approach in evaluating this evidence. Under this approach, OSHA evaluated the relevant studies to determine their overall quality. Factors considered in assessing the quality of studies included: (1) The size of the cohort studied and the power of the study to detect a sufficiently low level of disease risk; (2) the duration of follow-up of the study population; (3) the potential for study bias (e.g., selection bias in case-control studies or survivor effects in cross-sectional studies); and (4) the adequacy of underlying exposure information for examining exposure-response relationships. Studies were deemed suitable for inclusion in OSHA's Preliminary Quantitative Risk Assessment (QRA) where there was adequate quantitative information on exposure and disease risks and the study was judged to be sufficiently high quality according to these criteria.

Based upon this weight-of-evidence approach, OSHA preliminarily determined that there is substantial evidence in the peer-reviewed scientific literature that exposure to respirable crystalline silica increases the risk of silicosis, lung cancer, other NMRD, and renal and autoimmune effects. The Preliminary QRA indicated that, for silicosis and NMRD mortality, lung cancer mortality, and renal disease mortality, there is a significant risk at the previous PELs for respirable crystalline silica, with excess lifetime risk estimates substantially greater than 1 death per 1,000 workers as a result of exposure over a working life (45 years, from age 20 to age 65). At the revised PEL of 50 μg/m3 respirable crystalline silica, OSHA estimated that these risks would be substantially reduced. Cumulative risk estimates for silicosis morbidity were also well above 1 case per 1,000 workers at the previous PELs, with a substantial reduction at the revised PEL.

2. Summary of the Review of Health Effects Literature

In its Review of Health Effects Literature, OSHA identified the adverse health effects associated with the inhalation of respirable crystalline silica (Document ID 1711). OSHA covered the following topics: Silicosis (including relevant data from U.S. disease surveillance efforts), lung cancer and cancer at other sites, non-malignant respiratory diseases (NMRD) other than silicosis, renal and autoimmune effects, and physical factors affecting the toxicity of crystalline silica. Most of the evidence that described the health risks associated with exposure to silica Start Printed Page 16305consisted of epidemiological studies of worker populations; animal and in vitro studies on mode of action and molecular toxicology were also described. OSHA focused solely on those studies associated with airborne exposure to respirable crystalline silica due to the lack of evidence of health hazards from dermal or oral exposure. The review was further confined to issues related to the inhalation of respirable dust, which is generally defined as particles that are capable of reaching the pulmonary region of the lung (i.e., particles less than 10 microns (μm) in aerodynamic diameter), in the form of either quartz or cristobalite, the two forms of crystalline silica most often encountered in the workplace.

a. Silicosis

i. Types

Silicosis is an irreversible, progressive disease induced by the inflammatory effects of respirable crystalline silica in the lung, leading to lung damage and scarring and, in some cases, progressing to complications resulting in disability and death. Exposure to respirable crystalline silica is the only known cause of silicosis. Three types of silicosis have been described: An acute form following intense exposure to respirable dust of high crystalline silica content for a relatively short period (i.e., a few months or years); an accelerated form, resulting from about 5 to 15 years of heavy exposure to respirable dusts of high crystalline silica content; and, most commonly, a chronic form that typically follows less intense exposure of more than 20 years (Becklake, 1994, Document ID 0294; Balaan and Banks, 1992, 0289). In both the accelerated and chronic forms of the disease, lung inflammation leads to the formation of excess connective tissue, or fibrosis, in the lung. The hallmark of the chronic form of silicosis is the silicotic islet or nodule, one of the few agent-specific lesions in pathology (Balaan and Banks, 1992, Document ID 0289). As the disease progresses, these nodules, or fibrotic lesions, increase in density and can develop into large fibrotic masses, resulting in progressive massive fibrosis (PMF). Once established, the fibrotic process of chronic silicosis is thought to be irreversible (Becklake, 1994, Document ID 0294). There is no specific treatment for silicosis (Davis, 1996, Document ID 0998; Banks, 2005, 0291).

Chronic silicosis is the most frequently observed type of silicosis in the U.S. today. Affected workers may have a dry chronic cough, sputum production, shortness of breath, and reduced pulmonary function. These symptoms result from airway restriction and/or obstruction caused by the development of fibrotic scarring in the alveolar sacs and lower region of the lung. Prospective studies that follow the exposed cohort over a long period of time with periodic examinations can provide the best information on factors affecting the development and progression of silicosis, which has a latency period (the interval between beginning of exposure to silica and the onset of disease) from 10 to 30 years after first exposure (Weissman and Wagner, 2005; Document ID 0481).

ii. Diagnosis

The scarring caused by silicosis can be detected by chest x-ray or computerized tomography (CT) when the lesions become large enough to appear as visible opacities. The clinical diagnosis of silicosis has three requirements: Recognition by the physician that exposure to crystalline silica has occurred; the presence of chest radiographic abnormalities consistent with silicosis; the absence of other illnesses that could resemble silicosis on a chest radiograph (e.g., pulmonary fungal infection or tuberculosis) (Balaan and Banks, 1992, Document ID 0289; Banks, 2005, 0291). A standardized system to classify opacities seen in chest radiographs was developed by the International Labour Organization (ILO) to describe the presence and severity of silicosis on the basis of size, shape, and density of opacities, which together indicate the severity and extent of lung involvement (ILO, 1980, Document ID 1063; ILO, 2002, 1064; ILO, 2011, 1475; Merchant and Schwartz, 1998, 1096; NIOSH, 2011, 1513). The density of opacities seen on chest radiographs is classified on a 4-point category scale (0, 1, 2, or 3), with each category divided into three, giving a 12-subcategory scale between 0/0 and 3/+. For each subcategory, the top number indicates the major category that the profusion most closely resembles, and the bottom number indicates the major category that was given secondary consideration. Category 0 indicates the absence of visible opacities and categories 1 to 3 reflect increasing profusion of opacities and a concomitant increase in severity of disease. The bottom number can deviate from the top number by 1. At the extremes of the scale, a designation of 0/− or 3/+ may be used. Subcategory 0/− represents a radiograph that is obviously absent of small opacities. Subcategory 3/+ represents a radiograph that shows much greater profusion than depicted on a standard 3/3 radiograph.

To address the low sensitivity of chest x-rays for detecting silicosis, Hnizdo et al. (1993, Document ID 1050) recommended that radiographs consistent with an ILO category of 0/1 or greater be considered indicative of silicosis among workers exposed to a high concentration of silica-containing dust. In like manner, to maintain high specificity, chest x-rays classified as category 1/0 or 1/1 should be considered as a positive diagnosis of silicosis. A biopsy is not necessary to make a diagnosis and a diagnosis does not require that chest x-ray films or digital radiographic images be rated using the ILO system (NIOSH, 2002, Document ID 1110).

iii. Review of Occupation-Based Epidemiological Studies

The causal relationship between exposure to crystalline silica and silicosis has long been accepted in the scientific and medical communities. OSHA reviewed a large number of cross-sectional and retrospective studies conducted to estimate the quantitative relationship between exposure to crystalline silica and the development of silicosis (e.g., Kreiss and Zhen, 1996, Document ID 1080; Love et al., 1999, 0369; Ng and Chan, 1994, 0382; Rosenman et al., 1996, 0423; Churchyard et al., 2003, 1295; Churchyard et al., 2004, 0986; Hughes et al., 1998, 1059; Muir et al., 1989a, 1102; Muir et al., 1989b, 1101; Park et al., 2002, 0405; Chen et al., 2001, 0332; Chen et al., 2005, 0985; Hnizdo and Sluis-Cremer, 1993, 1052; Miller et al., 1998, 0374; Buchanan et al., 2003, 0306; Steenland and Brown, 1995b, 0451). In general, these studies, particularly those that included retirees, found a risk of radiological silicosis (usually defined as x-ray films classified as ILO major category 1 or greater) among workers exposed near the range of cumulative exposures permitted by current exposure limits. The studies' methods and findings are presented in detail in the Preliminary QRA (Document ID 1711, pp. 316-340); those studies on which OSHA relied for its risk estimates are also discussed in the Summary of the Preliminary QRA, below.

OSHA's review of the silicosis literature also focused on specific issues associated with the factors that affect the progression of the disease and the relationship between the appearance of radiological abnormalities indicative of silicosis and pulmonary function decline. From its review of the health literature, OSHA made a number of preliminary findings. First, the size of opacities apparent on initial x-ray films is a determinant of future disease Start Printed Page 16306progression, with subjects exhibiting large opacities more likely to experience progression than those having smaller opacities (Hughes et al., 1982, Document ID 0362; Lee et al., 2001, 1086; Ogawa et al., 2003, 0398). Second, continued exposure to respirable crystalline silica following diagnosis of radiological silicosis increases the probability of disease progression compared to those who are not further exposed (Hessel et al., 1988, Document ID 1042), although there remains a likelihood of progression even absent continued exposure (Hessel et al., 1988, Document ID 1042; Miller et al., 1998, 0374; Ogawa et al., 2003, 0398; Yang et al., 2006, 1134).

With respect to the relationship between radiological silicosis and pulmonary function declines, literature findings are mixed. A number of studies have reported pulmonary function declines among workers exhibiting a degree of small-opacity profusion consistent with ILO categories 2 and 3 (e.g., Ng and Chan, 1992, Document ID 1107). However, although some studies have not found pulmonary function declines associated with silicosis scored as ILO category 1, a number of other studies have documented declines in pulmonary function in persons exposed to silica and whose radiograph readings are in the major ILO category 1 (i.e., 1/0, 1/1, 1/2), or even before changes were seen on chest x-ray (Cowie, 1998, 0993; Cowie and Mabena, 1991, 0342; Ng et al., 1987(a), 1108; Wang et al., 1997, 0478). Thus, OSHA preliminarily concluded that at least some individuals will develop pulmonary function declines absent radiological changes indicative of silicosis. The Agency posited that this may reflect the relatively poor sensitivity of x-ray films in detecting silicosis or may be due to pulmonary function declines related to silica-induced chronic obstructive pulmonary disease (see Document ID 1711, pp. 49-75).

iv. Surveillance

Unlike most occupational diseases, surveillance statistics are available on silicosis mortality and morbidity in the U.S. The most comprehensive and current source of surveillance data in the U.S. related to occupational lung diseases, including silicosis, is the National Institute for Occupational Safety and Health (NIOSH) Work-Related Lung Disease (WoRLD) Surveillance System (NIOSH, 2008c, Document ID 1308). Other sources are detailed in the Review of Health Effects Literature (Document ID 1711). Mortality data are compiled from death certificates reported to state vital statistics offices, which are collected by the National Center for Health Statistics (NCHS), an agency within the Centers for Disease Control and Prevention (e.g., CDC, 2005, Document ID 0319).

Silicosis-related mortality has declined in the U.S. over the time period for which these data have been collected. From 1968 to 2005, the annual number of silicosis deaths decreased from 1,157 to 161 (NIOSH, 2008c, Document ID 1308; http://wwwn.cdc.gov/​eworld). The CDC cited two main factors that were likely responsible for the declining trend in silicosis mortality since 1968 (CDC, 2005, Document ID 0319). First, many deaths during the early part of the study period were among workers whose main exposure to respirable crystalline silica probably occurred before introduction of national silica standards established by OSHA and the Mine Safety and Health Administration (MSHA) (i.e., permissible exposure limits (PELs)); these standards likely led to reduced silica dust exposure beginning in the 1970s. Second, employment has declined in heavy industries (e.g., foundries) where silica exposure was prevalent (CDC, 2005, Document ID 0319).

Despite this decline, silicosis deaths among workers of all ages result in significant premature mortality; between 1996 and 2005, a total of 1,746 deaths resulted in a total of 20,234 years of life lost from life expectancy, with an average of 11.6 years of life lost. For the same period, among 307 decedents who died before age 65 (the end of a working life), there were 3,045 years of life lost up to age 65, with an average of 9.9 years of life lost from a working life (NIOSH, 2008c, Document ID 1308).

Surveillance data on silicosis morbidity, primarily from hospital discharge records, are available only from the few states that have administered disease surveillance programs for silicosis. For the reporting period 1993-2002, these states recorded 879 cases of silicosis (NIOSH 2008c, Document ID 1308). Nationwide hospital discharge data compiled by NIOSH (2008c, Document ID 1308) and the Council of State and Territorial Epidemiologists (CSTE, 2005, Document ID 0996) indicate that, for the years 1970 to 2004, there were at least 1,000 hospitalizations that were coded for silicosis each year, except one.

Relying exclusively on such passive case-based disease surveillance systems that depend on the health care community to generate records is likely to understate the prevalence of diseases associated with respirable crystalline silica (Froines et al., 1989, Document ID 0385). In order to diagnose occupational diseases, health care professionals must have information about occupational histories and must be able to recognize occupational diseases (Goldman and Peters, 1981, Document ID 1027; Rutstein et al., 1983, 0425). The first criterion to be met in diagnosing silicosis is knowing a patient's history of exposure to crystalline silica. In addition to the lack of information about exposure histories, difficulty in recognizing occupational illnesses like silicosis, that manifest themselves long after initial exposure, contributes to under-recognition and underreporting by health care providers. Based on an analysis of data from Michigan's silicosis surveillance activities, Rosenman et al. (2003, Document ID 0420) estimated that silicosis mortality and morbidity were understated by a factor of between 2.5 and 5, and estimated that between 3,600 and 7,300 new cases of silicosis likely occurred in the U.S. annually between 1987 and 1996.

b. Lung Cancer

i. International Agency for Research on Cancer (IARC) Classification

In 1997, the IARC determined that there was sufficient evidence to regard crystalline silica as a human carcinogen (IARC, 1997, Document ID 1062). This finding was based largely on nine studies of cohorts in four industry sectors that IARC considered to be the least influenced by confounding factors (sectors included quarries and granite works, gold mining, ceramic/pottery/refractory brick industries, and the diatomaceous earth industry). NIOSH also determined that crystalline silica is a human carcinogen after evaluating updated literature (2002, Document ID 1110).

ii. Review of Occupation-Based Epidemiological Studies

OSHA conducted an independent review of the epidemiological literature on exposure to respirable crystalline silica and lung cancer, covering more than 30 occupational groups in over a dozen industrial sectors. OSHA's review included approximately 60 primary epidemiological studies. Based on this review, OSHA preliminarily concluded that the human data provides ample evidence that exposure to respirable crystalline silica increases the risk of lung cancer among workers.

The strongest evidence for carcinogenicity came from studies in five industry sectors:

  • Diatomaceous Earth Workers (Checkoway et al., 1993, Document ID 0324; Checkoway et al., 1996, 0325; Checkoway et al., 1997, 0326; Start Printed Page 16307Checkoway et al., 1999, 0327; Seixas et al., 1997, 0431);
  • British Pottery Workers (Cherry et al., 1998, Document ID 0335; McDonald et al., 1995, 0371);
  • Vermont Granite Workers (Attfield and Costello, 2004, Document ID 0285; Graham et al., 2004, 1031; Costello and Graham, 1988, 0991; Davis et al., 1983, 0999);
  • North American Industrial Sand Workers (Hughes et al., 2001, Document ID 1060; McDonald et al., 2001, 1091; McDonald et al., 2005, 1092; Rando et al., 2001, 0415; Sanderson et al., 2000, 0429; Steenland and Sanderson, 2001, 0455); and
  • British Coal Miners (Miller et al., 2007, Document ID 1305; Miller and MacCalman, 2009, 1306).

OSHA considered these studies as providing the strongest evidence for several reasons. They were all retrospective cohort or case-control studies that demonstrated positive, statistically significant exposure-response relationships between exposure to crystalline silica and lung cancer mortality. Except for the British pottery studies, where exposure-response trends were noted for average exposure only, lung cancer risk was found to be related to cumulative exposure. In general, these studies were of sufficient size and had adequate years of follow up, and had sufficient quantitative exposure data to reliably estimate exposures of cohort members. As part of their analyses, the authors of these studies also found positive exposure-response relationships for silicosis, indicating that underlying estimates of worker exposures were not likely to be substantially misclassified. Furthermore, the authors of these studies addressed potential confounding due to other carcinogenic exposures through study design or data analysis.

In the diatomaceous earth industry, Checkoway et al. developed a “semi-quantitative” cumulative exposure estimate that demonstrated a statistically significant positive exposure-response trend between duration of employment or cumulative exposure and lung cancer mortality (1993, Document ID 0324). The quartile analysis with a 15-year lag showed an increasing trend in relative risks (RR) of lung cancer mortality, with the highest exposure quartile having a RR of 2.74 for lung cancer mortality. Checkoway et al. conducted a re-analysis to address criticisms of potential confounding due to asbestos and again demonstrated a positive exposure-response risk gradient when controlling for asbestos exposure and other variables (1996, Document ID 0325). Rice et al. (2001, Document ID 1118) conducted a re-analysis and quantitative risk assessment of the Checkoway et al. (1997, Document ID 0326) study, finding that exposure to crystalline silica was a significant predictor of lung cancer mortality. OSHA included this re-analysis in its Preliminary QRA (Document ID 1711).

In the British pottery industry, excess lung cancer risk was found to be associated with crystalline silica exposure among workers in a proportionate mortality ratio (PMR) study [4] (McDonald et al., 1995, Document ID 0371) and in a cohort and nested case-control study [5] (Cherry et al., 1998, Document ID 0335). In the former, elevated PMRs for lung cancer were found after adjusting for potential confounding by asbestos exposure. In the study by Cherry et al., odds ratios for lung cancer mortality were statistically significantly elevated after adjusting for smoking. Odds ratios were related to average, but not cumulative, exposure to crystalline silica.

In the Vermont granite cohort, Costello and Graham (1988, Document ID 0991) and Graham et al. (2004, Document ID 1031) in a follow-up study found that workers employed prior to 1930 had an excess risk of lung cancer. Lung cancer mortality among granite workers hired after 1940 (post-implementation of controls), however, was not elevated in the Costello and Graham study and was only somewhat elevated (not statistically significant) in the Graham et al. study. Graham et al. (2004, Document ID 1031) concluded that their results did not support a causal relationship between granite dust exposure and lung cancer mortality.

Looking at the same population, Attfield and Costello (2004, Document ID 0285) developed a quantitative estimate of cumulative exposure (8 exposure categories) adapted from a job exposure matrix developed by Davis et al. (1983, Document ID 0999). They found a statistically significant trend between lung cancer mortality and log-transformed cumulative exposure to crystalline silica. Lung cancer mortality rose reasonably consistently through the first seven increasing exposure groups, but fell in the highest cumulative exposure group. With the highest exposure group omitted, a strong positive dose-response trend was found for both untransformed and log-transformed cumulative exposures. The authors explained that the highest exposure group would have included the most unreliable exposure estimates being reconstructed from exposures 20 years prior to study initiation when exposure estimation was less precise. OSHA expressed its belief that the study by Attfield and Costello (2004, Document ID 0285) was of superior design in that it used quantitative estimates of exposure and evaluated lung cancer mortality rates by exposure group. In contrast, the findings by Graham et al. (2004, Document ID 1031) were based on a dichotomous comparison of risk among high- versus low-exposure groups, where date-of-hire before and after implementation of ventilation controls was used as a surrogate for exposure. Consequently, OSHA used the Attfield and Costello study in its Preliminary QRA (Document ID 1711). In its Supplemental Literature Review of Epidemiological Studies on Lung Cancer Associated with Exposure to Respirable Crystalline Silica, OSHA also discussed a more recent study of Vermont granite workers by Vacek et al. (2011, Document ID 1486) that did not find an association between silica exposure and lung cancer mortality (Document ID 1711, Attachment 1, pp. 2-5). (OSHA examines this study in great length in Section V.F, Comments and Responses Concerning Lung Cancer Mortality.)

In the North American industrial sand industry, studies of two overlapping cohorts found a statistically significant increased risk of lung cancer mortality with increased cumulative exposure in both categorical and continuous analyses (Hughes et al., 2001, Document ID 1060; McDonald et al., 2001, 1091; McDonald et al., 2005, 1092; Rando et al., 2001, 0415; Sanderson et al., 2000, 0429; Steenland and Sanderson, 2001, 0455). McDonald et al. (2001, Document ID 1091) examined a cohort that entered the workforce, on average, a decade earlier than the cohorts that Steenland and Sanderson (2001, Document ID 0455) examined. The McDonald cohort, drawn from eight plants, had more years of exposure in the industry (19 versus 8.8 years). The Steenland and Sanderson (2001, Document ID 0455) cohort worked in 16 plants, 7 of which overlapped with the McDonald, et al. Start Printed Page 16308(2001, Document ID 1091) cohort. McDonald et al. (2001, Document ID 1091), Hughes et al. (2001, Document ID 1060), and Rando et al. (2001, Document ID 0415) had access to smoking histories, plant records, and exposure measurements that allowed for historical reconstruction and the development of a job exposure matrix. The McDonald et al. (2005, Document ID 1092) study was a later update, with follow-up through 2000, of both the cohort and nested case-control studies. Steenland and Sanderson (2001, Document ID 0455) had limited access to plant facilities, less detailed historic exposure data, and used MSHA enforcement records for estimates of recent exposure. These studies (Hughes et al., 2001, Document ID 1060; McDonald et al., 2005, 1092; Steenland and Sanderson, 2001, 0455) showed very similar exposure-response patterns of increased lung cancer mortality with increased exposure. OSHA included the quantitative exposure-response analysis from the Hughes et al. (2001, Document ID 1060) study in its Preliminary QRA, as it allowed for individual job, exposure, and smoking histories to be taken into account.

OSHA noted that Brown and Rushton (2005a, Document ID 0303; 2005b, 0304) found no association between risk of lung cancer mortality and exposure to respirable crystalline silica among British industrial sand workers. However, a large portion of the cohort had relatively short service times in the industry, with over one-half the cohort deaths and almost three-fourths of the lung cancer mortalities having had less than 10 years of service. Considering the apparent high turnover in this industry and the absence of prior occupational histories, exposures from work experience other than in the industrial sand industry could be a significant confounder (Document ID 1711, p. 131). Additionally, as Steenland noted in a letter review (2005a, Document ID 1313), the cumulative exposures of workers in the Brown and Ruston (2005b, Document ID 0304) study were over 10 times lower than the cumulative exposures experienced by the cohorts in the pooled analysis that Steenland et al. (2001a, Document ID 0452) performed. The low exposures experienced by this cohort would have made detecting a positive association with lung cancer mortality even more difficult.

In British coal miners, excess lung cancer mortality was reported in a large cohort study, which examined the mortality experience of 17,800 miners through the end of 2005 (Miller et al., 2007, Document ID 1305; Miller and MacCalman, 2009, 1306). By that time, the cohort had accumulated 516,431 person years of observation (an average of 29 years per miner), with 10,698 deaths from all causes. Overall lung cancer mortality was elevated (SMR = 115.7, 95% C.I. 104.8-127.7), and a positive exposure-response relationship with crystalline silica exposure was determined from Cox regression after adjusting for smoking history. Three of the strengths of this study were the detailed time-exposure measurements of both quartz and total mine dust, detailed individual work histories, and individual smoking histories. For lung cancer, analyses based on Cox regression provided strong evidence that, for these coal miners, although quartz exposures were associated with increased lung cancer risk, simultaneous exposures to coal dust did not cause increased lung cancer risk. Because of these strengths, OSHA included this study in its Preliminary QRA (Document ID 1711).

In addition to the studies in these cohorts, OSHA also reviewed studies of lung cancer mortality in metal ore mining populations. Many of these mining studies, which showed mixed results, were subject to confounding due to exposure to other potential carcinogens such as radon and arsenic. IARC noted that only a few ore mining studies accounted for confounding from other occupational carcinogens and that, when confounding was absent or accounted for, an association between silica exposure and lung cancer was absent (1997, Document ID 1062). Many of the studies conducted since IARC's review, however, more strongly implicate crystalline silica as a human carcinogen (1997, Document ID 1062). Pelucchi et al. (2006, Document ID 0408), in a meta-analysis of studies conducted since IARC's (1997, Document ID 1062) review, reported statistically significantly elevated relative risks of lung cancer mortality in underground and surface miners in three cohort and four case-control studies. Cassidy et al., in a pooled case-control analysis, showed a statistically significant increased risk of lung cancer mortality among miners (OR = 1.48), and demonstrated a linear trend of increasing odds ratios with increasing exposures (2007, Document ID 0313).

OSHA also preliminarily determined that the results of the studies conducted in three industry sectors (foundry, silicon carbide, and construction sectors) were confounded by the presence of exposures to other carcinogens. Exposure data from these studies were not sufficient to distinguish between exposure to silica dust and exposure to other occupational carcinogens. IARC previously made a similar determination in reference to the foundry industry. However, with respect to the construction industry, Cassidy et al. (2007, Document ID 0313), in a large European community-based case-control study, reported finding a clear linear trend of increasing odds ratios with increasing cumulative exposure to crystalline silica (estimated semi-quantitatively) after adjusting for smoking and exposure to insulation and wood dusts.

In addition, an analysis of 4.8 million death certificates from 27 states within the U.S. for the years 1982 to 1995 showed statistically significant excesses in lung cancer mortality, silicosis mortality, tuberculosis, and NMRD among persons with occupations involving medium and high exposure to respirable crystalline silica (Calvert et al., 2003, Document ID 0309). A national records and death certificate study was also conducted in Finland by Pukkala et al., who found a statistically significant excess of lung cancer incidence among men and women with estimated medium and heavy exposures (2005, Document ID 0412).

One of the more compelling studies OSHA evaluated and used in the Preliminary QRA (Document ID 1711) was Steenland et al.'s (2001a, Document ID 0452) pooled analysis of 10 occupational cohorts (5 mines and 5 industrial facilities), which demonstrated an overall positive exposure-response relationship between cumulative exposure to crystalline silica and lung cancer mortality. These 10 cohorts included 65,980 workers and 1,072 lung cancer deaths, and were selected because of the availability of raw data on exposure to crystalline silica and health outcomes. The investigators found lung cancer risk increased with increasing cumulative exposure, log cumulative exposure, and average exposure. Exposure-response trends were similar between mining and non-mining cohorts.

iii. Confounding

Smoking is known to be a major risk factor for lung cancer. However, OSHA maintained in the Preliminary QRA that it is unlikely that smoking explained the observed exposure-response trends in the studies described above (Document ID 1711). Studies by Hnizdo et al. (1997, Document ID 1049), McLaughlin et al. (1992, Document ID 0372), Hughes et al. (2001, Document ID 1060), McDonald et al. (2001, Document ID 1091; 2005, 1092), Miller and MacCalman (2009, Document ID 1306), and Cassidy et al. (2007, Document ID 0313) had detailed smoking histories with sufficiently large Start Printed Page 16309populations and a sufficient number of years of follow-up time to quantify the interaction between crystalline silica exposure and cigarette smoking. In a cohort of white South African gold miners (Hnizdo and Sluis-Cremer, 1991, Document ID 1051) and in the follow-up nested case-control study (Hnizdo et al., 1997, Document ID 1049), the combined effect of exposure to respirable crystalline silica and smoking was greater than additive, suggesting a multiplicative effect. This effect appeared to be greatest for miners with greater than 35 pack-years of smoking and higher cumulative exposure to silica. In the Chinese nested case-control studies (McLaughlin et al., 1992, Document ID 0372), cigarette smoking was associated with lung cancer, but control for smoking did not influence the association between silica and lung cancer in the mining and pottery cohorts studied. The studies of industrial sand workers (Hughes et al., 2001, Document ID 1060) and British coal workers (Miller and MacCalman, 2009, Document ID 1306) found positive exposure-response trends after adjusting for smoking histories, as did Cassidy et al. (2007, Document ID 0313) in their community-based case-control study of exposed European workers.

Given these findings of investigators who have accounted for the impact of smoking, OSHA preliminarily determined that the weight of the evidence reviewed identified respirable crystalline silica as an independent risk factor for lung cancer mortality. OSHA also determined that its finding was further supported by animal studies demonstrating that exposure to silica alone can cause lung cancer (e.g., Muhle et al., 1995, Document ID 0378).

iv. Lung Cancer and Silicosis

Animal and in vitro studies have demonstrated that the early steps in the proposed mechanistic pathways that lead to silicosis and lung cancer seem to share some common features (see Document ID 1711, pp. 171-172). This has led some researchers to suggest that silicosis is a prerequisite to lung cancer. Some have suggested that any increased lung cancer risk associated with silica may be a consequence of inflammation (and concomitant oxidative stress) and increased epithelial cell proliferation associated with the development of silicosis. However, other researchers have noted additional genotoxic and non-genotoxic mechanisms that may also be involved in carcinogenesis induced by silica (see Section V.H, Mechanisms of Silica-Induced Adverse Health Effects, and Document ID 1711, pp. 230-239). IARC also noted that a direct genotoxic mechanism from silica to induce a carcinogenic effect cannot be ruled out (2012, Document ID 1473). Thus, OSHA preliminarily concluded that available animal and in vitro studies do not support the hypothesis that development of silicosis is necessary for silica exposure to cause lung cancer.

In general, studies of workers with silicosis, as well as meta-analyses that include these studies, have shown that workers with radiologic evidence of silicosis have higher lung cancer risk than those without radiologic abnormalities or mixed cohorts. Three meta-analyses attempted to look at the association of increasing ILO radiographic categories of silicosis with increasing lung cancer mortality. Two of these analyses (Kurihara and Wada, 2004, Document ID 1084; Tsuda et al., 1997, 1127) showed no association with increasing lung cancer mortality, while Lacasse et al. (2005, Document ID 0365) demonstrated a positive dose-response for lung cancer with increasing ILO radiographic category. A number of other studies found increased lung cancer risk among exposed workers absent radiological evidence of silicosis (Cassidy et al., 2007, Document ID 0313; Checkoway et al., 1999, 0327; Cherry et al., 1998, 0335; Hnizdo et al., 1997, 1049; McLaughlin et al., 1992, 0372). For example, the diatomaceous earth study by Checkoway et al. showed a statistically significant exposure-response relationship for lung cancer among persons without silicosis (1999, Document ID 0327). Checkoway and Franzblau, reviewing the international literature, found that all epidemiological studies conducted to that date were insufficient to conclusively determine the role of silicosis in the etiology of lung cancer (2000, Document ID 0323). OSHA preliminarily concluded that the more recent pooled and meta-analyses do not provide compelling evidence that silicosis is a necessary precursor to lung cancer.

c. Non-Malignant Respiratory Diseases (Other Than Silicosis)

In addition to causing silicosis, exposure to crystalline silica has been associated with increased risks of other non-malignant respiratory diseases (NMRD), primarily chronic obstructive pulmonary disease (COPD), chronic bronchitis, and emphysema. COPD is a disease state characterized by airflow limitation that is usually progressive and not fully reversible. In patients with COPD, either chronic bronchitis or emphysema may be present or both conditions may be present together.

As detailed in the Review of Health Effects Literature, OSHA reviewed several studies of NMRD morbidity and preliminarily concluded that exposure to respirable crystalline silica may increase the risk of emphysema, chronic bronchitis, and pulmonary function impairment, regardless of whether signs of silicosis are present (Document ID 1711). Smokers may be at an increased risk relative to nonsmokers.

OSHA also reviewed studies of NMRD mortality that focused on causes of death other than silicosis. Wyndham et al. found a significant excess mortality for chronic respiratory diseases in a cohort of white South African gold miners (1986, Document ID 0490). A case-referent analysis found that, although the major risk factor for chronic respiratory disease was smoking, there was a statistically significant additional effect of cumulative exposure to silica-containing dust. A multiplicative effect of smoking and cumulative dust exposure on mortality from COPD was found in another study of white South African gold miners (Hnizdo, 1990, Document ID 1045). Analysis of various combinations of dust exposure and smoking found a trend in odds ratios that indicated this synergism. There was a statistically significant increasing trend for dust particle-years and for cigarette-years of smoking.

Park et al. (2002, Document ID 0405) analyzed the California diatomaceous earth cohort data originally studied by Checkoway et al. (1997, Document ID 0326), consisting of 2,570 diatomaceous earth workers employed for 12 months or more from 1942 to 1994, to quantify the relationship between exposure to cristobalite and mortality from chronic lung disease other than cancer (LDOC). Diseases in this category included pneumoconiosis (which included silicosis), chronic bronchitis, and emphysema, but excluded pneumonia and other infectious diseases. Smoking information was available for about 50 percent of the cohort and for 22 of the 67 LDOC deaths available for analysis, permitting at least partial adjustment for smoking. Using the exposure estimates developed for the cohort by Rice et al. (2001, Document ID 1118) in their exposure-response study of lung cancer risks, Park et al. (2002, Document 0405) evaluated the quantitative exposure-response relationship for LDOC mortality and found a strong positive relationship with exposure to respirable crystalline silica. OSHA found this study particularly compelling because of the strengths of the study design and availability of smoking history data on part of the cohort, as well as the high-Start Printed Page 16310quality exposure and job history data. The study authors noted:

Data on smoking, collected since the 1960s in the company's radiographic screening programme, were available for 1171 of the subjects (50%). However, smoking habits were unknown for 45 of the 67 workers that died from LDOC (67%). Our Poisson regression analyses for LDOC, stratified on smoking, have partially rectified the confounding by smoking issue. Furthermore, analyses performed without control for smoking produced slightly smaller and less precise estimates of the effects of silica, suggesting that smoking is a negative confounder. In their analysis of this cohort, Checkoway et al. applied the method of Axelson concluding that it was very unlikely that cigarette smoking could account for the association found between mortality from LDOC and cumulative exposure to silica (Document ID 0405, p. 41).

Consequently, OSHA used this study in its Preliminary QRA (Document ID 1711, pp. 295-298).

Based on this evidence, and the other studies discussed in the Review of Health Effects Literature, OSHA preliminarily concluded that respirable crystalline silica increases the risk for mortality from non-malignant respiratory disease (not including silicosis) in an exposure-related manner. The Agency also preliminarily concluded that the risk is strongly influenced by smoking, and opined that the effects of smoking and silica exposure may be synergistic.

d. Renal Disease and Autoimmune Diseases

In its Review of Health Effects Literature, OSHA described the available experimental and epidemiological data evaluating respirable crystalline silica exposure and renal and/or autoimmune effects (Document ID 1711). In addition to a number of case reports, epidemiological studies have found statistically significant associations between occupational exposure to silica dust and chronic renal disease (Calvert et al., 1997, Document ID 0976), subclinical renal changes (Ng et al., 1992c, Document ID 0386), end-stage renal disease morbidity (Steenland et al., 1990, Document ID 1125), chronic renal disease mortality (Steenland et al., 2001b, Document ID 0456; 2002a, 0448), and granulomatosis with polyangitis, a condition that can affect the kidneys (Nuyts et al., 1995, Document ID 0397). In other findings, silica-exposed individuals, both with and without silicosis, had an increased prevalence of abnormal renal function (Hotz et al., 1995, Document ID 0361), and renal effects have been reported to persist after cessation of silica exposure (Ng et al., 1992c, Document ID 0386). Possible mechanisms suggested for silica-induced renal disease include a direct toxic effect on the kidney, deposition of immune complexes (IgA) in the kidney following silica related pulmonary inflammation, and an autoimmune mechanism (Calvert et al., 1997, Document ID 0976; Gregorini et al., 1993, 1032).

In a pooled cohort analysis, Steenland et al. (2002a, Document ID 0448) combined the industrial sand cohort from Steenland et al. (2001b, Document ID 0456), the gold mining cohort from Steenland and Brown (1995a, Document ID 0450), and the Vermont granite cohort studies by Costello and Graham (1988, Document ID 0991). In all, the combined cohort consisted of 13,382 workers with exposure information available for 12,783. The analysis demonstrated statistically significant exposure-response trends for acute and chronic renal disease mortality with quartiles of cumulative exposure to respirable crystalline silica. In a nested case-control study design, a positive exposure-response relationship was found across the three cohorts for both multiple-cause mortality (i.e., any mention of renal disease on the death certificate) and underlying cause mortality. Renal disease risk was most prevalent among workers with cumulative exposures of 500 µg/m3 or more (Steenland et al., 2002a, Document ID 0448).

OSHA noted that other studies failed to find an excess renal disease risk among silica-exposed workers. Davis et al. (1983, Document ID 0999) found elevated, but not statistically significant, mortality from diseases of the genitourinary system among Vermont granite shed workers. There was no observed relationship between mortality from this cause and cumulative exposure. A similar finding was reported by Koskela et al. (1987, Document ID 0363) among Finnish granite workers, where there were 4 deaths due to urinary tract disease compared to 1.8 expected. Both Carta et al. (1994, Document ID 0312) and Cocco et al. (1994, Document ID 0988) reported finding no increased mortality from urinary tract disease among workers in an Italian lead mine and zinc mine. However, Cocco et al. (1994, Document ID 0988) commented that exposures to respirable crystalline silica were low, averaging 7 and 90 µg/m3 in the two mines, respectively, and that their study in particular had low statistical power to detect excess mortality.

OSHA expressed its belief that there is substantial evidence, particularly the 3-cohort pooled analysis conducted by Steenland et al. (2002a, Document ID 0448), on which to base a finding that exposure to respirable crystalline silica increases the risk of renal disease mortality and morbidity. The pooled analysis by Steenland et al. involved a large number of workers from three cohorts with well-documented, validated job-exposure matrices; it found a positive, monotonic increase in renal disease risk with increasing exposure for both underlying and multiple cause data (2002a, Document ID 0448). However, there are considerably less data available for renal disease than there are for silicosis mortality and lung cancer mortality. The findings based on these data are, therefore, less robust. Nevertheless, OSHA preliminarily concluded that the underlying data are sufficient to provide useful estimates of risk and included the Steenland et al. (2002a, Document ID 0448) analysis in its Preliminary QRA.

For autoimmune effects, OSHA reviewed epidemiological information suggesting an association between respirable silica exposure and autoimmune diseases, including scleroderma (Sluis-Cremer et al., 1985, Document ID 0439), rheumatoid arthritis (Klockars et al., 1987, Document ID 1075; Rosenman and Zhu, 1995, 0424), and systemic lupus erythematosus (Brown et al., 1997, Document ID 0974). However, there were no quantitative exposure-response data available on which to base a quantitative risk assessment for autoimmune diseases.

e. Physical Factors Affecting Toxicity of Crystalline Silica

OSHA also examined evidence on the comparative toxicity of the silica polymorphs (quartz, cristobalite, and tridymite). A number of animal studies appear to suggest that cristobalite and tridymite are more toxic to the lung than quartz and more tumorigenic (e.g., King et al., 1953, Document ID 1072; Wagner et al., 1980, 0476). However, in contrast to these findings, several authors have reviewed the studies done in this area and concluded that cristobalite and tridymite are not more toxic than quartz (e.g., Bolsaitis and Wallace, 1996, Document ID 0298; Guthrie and Heaney, 1995, 1035). Furthermore, a difference in toxicity between cristobalite and quartz has not been observed in epidemiological studies (tridymite has not been studied) (NIOSH, 2002, Document ID 1110). In an analysis of exposure-response for lung cancer, Steenland et al. found similar exposure-response trends between cristobalite-exposed workers and other cohorts Start Printed Page 16311exposed to quartz (2001a, Document ID 0452).

OSHA also discussed other physical factors that may influence the toxicologic potency of crystalline silica. A number of animal studies compared the toxicity of freshly fractured silica to that of aged silica (Porter et al., 2002, Document ID 1114; Shoemaker et al., 1995, 0437; Vallyathan et al., 1995, 1128). These studies have demonstrated that although freshly fractured silica is more toxic than aged silica, aged silica still retains significant toxicity. There have been no studies comparing workers exposed to freshly fractured silica to those exposed to aged silica. However, similarities between the results of animal and human studies involving freshly fractured silica suggest that the animal studies involving aged silica may also apply to humans. For example, studies of workers exposed to freshly fractured silica have demonstrated that these workers exhibit the same cellular effects as seen in animals exposed to freshly fractured silica (Castranova et al., 1998, Document ID 1294; Goodman et al., 1992, 1029). Animal studies also suggest that pulmonary reactions of rats to short-duration exposure to freshly fractured silica mimic those seen in acute silicosis in humans (Vallyathan et al., 1995, Document ID 1128).

Surface impurities, particularly metals, have been shown to alter silica toxicity. Iron, depending on its state and quantity, has been shown to either increase or decrease toxicity (see Document ID 1711, pp. 247-258). Aluminum has been shown to decrease toxicity (Castranova et al., 1997, Document ID 0978; Donaldson and Borm, 1998, 1004; Fubini, 1998, 1016). Silica coated with aluminosilicate clay exhibits lower toxicity, possibly as a result of reduced bioavailability of the silica particle surface (Donaldson and Borm, 1998, Document ID 1004; Fubini, 1998, 1016). Aluminum as well as other metal ions are thought to modify silanol groups on the silica surface, thus decreasing the membranolytic and cytotoxic potency and resulting in enhanced particle clearance from the lung before damage can take place (Fubini, 1998, Document ID 1016). An epidemiological study found that the risk of silicosis was less in pottery workers than in tin and tungsten miners (Chen et al., 2005, Document ID 0985; Harrison et al., 2005, 1036), possibly reflecting that pottery workers were exposed to silica particles having less biologically-available, non-clay-occluded surface area than was the case for miners.

Although it is evident that a number of factors can act to mediate the toxicological potency of crystalline silica, it is not clear how such considerations should be taken into account to evaluate lung cancer and silicosis risks to exposed workers. After evaluating many in vitro studies that investigated the surface characteristics of crystalline silica particles and their influence on fibrogenic activity, NIOSH concluded that further research is needed to associate specific surface characteristics that can affect toxicity with specific occupational exposure situations and consequent health risks to workers (2002, Document ID 1110). Thus, OSHA preliminarily concluded that while there was considerable evidence that several environmental influences can modify surface activity to either enhance or diminish the toxicity of silica, the available information was insufficient to determine in any quantitative way how these influences may affect disease risk to workers in any particular workplace setting.

3. Summary of the Preliminary QRA

OSHA presented in the Preliminary QRA estimates of the risk of silica-related diseases assuming exposure over a working life (45 years, from age 20 to age 65) to the revised 8-hour time-weighted average (TWA) PEL of 50 µg/m3 respirable crystalline silica, the new action level of 25 µg/m3, and the previous PELs. OSHA's previous general industry PEL for respirable quartz was expressed both in terms of a particle count formula and a gravimetric concentration formula; the previous construction and shipyard employment PELs for respirable quartz were only expressed in terms of a particle count formula. For general industry, as the quartz content increases, the gravimetric PEL approached a limit of 100 µg/m3 respirable quartz. For construction and shipyard employment, OSHA's previous PELs used a formula that limits exposure to respirable dust, depending upon the quartz content, expressed as a respirable particle count concentration. There was no single mass concentration equivalent for the construction and shipyard employment PELs; OSHA reviewed several studies that suggest that the previous construction/shipyard PEL likely was between 250 and 500 µg/m3 respirable quartz. In general industry, for both the gravimetric and particle count PELs, OSHA's previous PELs for cristobalite and tridymite were half the value for quartz. Based upon these previous PELs and the new action level, OSHA presented risk estimates associated with exposure over a working life to 25, 50, 100, 250, and 500 µg/m3 respirable silica (corresponding to cumulative exposures over 45 years to 1.125, 2.25, 4.5, 11.25, and 22.5 mg/m3-yrs).

To estimate lifetime excess mortality risks at these exposure levels, OSHA implemented each of the risk models in a life table analysis that accounted for competing causes of death due to background causes and cumulated risk through age 85. For these analyses, OSHA used lung cancer, NMRD, or renal disease mortality and all-cause mortality rates to account for background risks and competing risks (U.S. 2006 data for lung cancer and NMRD mortality in all males, 1998 data for renal disease mortality, obtained from cause-specific death rate tables published by the National Center for Health Statistics (2009, Document ID 1104)). OSHA calculated these risk estimates assuming occupational exposure from age 20 to age 65. The mortality risk estimates were presented in terms of lifetime excess risk per 1,000 workers for exposure over an 8-hour working day, 250 days per year, and a 45-year working life.

For silicosis morbidity, OSHA based its risk estimates on cumulative risk models used by various investigators to develop quantitative exposure-response relationships. These models characterized the risk of developing silicosis (as detected by chest radiography) up to the time that cohort members (including both active and retired workers) were last examined. Thus, risk estimates derived from these studies represented less-than-lifetime risks of developing radiographic silicosis. OSHA did not attempt to estimate lifetime risk (i.e., up to age 85) for silicosis morbidity because the relationships between age, time, and disease onset post-exposure have not been well characterized.

a. Silicosis and NMRD Mortality

i. Exposure-Response Studies

In the Preliminary QRA, OSHA relied upon two published quantitative risk studies of silicosis and NMRD mortality (Document ID 1711). The first, Mannetje et al. (2002b, Document ID 1089) conducted a pooled analysis of silicosis mortality in which there were 18,634 subjects, 150 silicosis deaths, and 20 deaths from unspecified pneumoconiosis. Rates for silicosis adjusted for age, calendar time, and study were estimated by Poisson regression and increased nearly monotonically with deciles of cumulative exposure, from a mortality rate of 5/100,000 person-years in the lowest exposure category (0-0.99 Start Printed Page 16312mg/m3-yrs) to 299/100,000 person-years in the highest category (>28.10 mg/m3-yrs).

As previously discussed, the second, Park et al. (2002, Document ID 0405) analyzed the California diatomaceous earth cohort data from Checkoway et al. (1997, Document ID 0326), and examined mortality from chronic lung disease other than cancer (LDOC; also known as non-malignant respiratory disease (NMRD)). Smoking information was available for about 50 percent of the cohort and for 22 of the 67 LDOC deaths available for analysis, permitting Park et al. (2002, Document ID 0405) to partially adjust for smoking. Estimates of LDOC mortality risks were derived via Poisson and Cox proportional hazards models; a variety of relative rate model forms were fit to the data, with a linear relative rate model selected for estimating risks.

ii. Risk Estimates

As silicosis is only caused by exposure to respirable crystalline silica (i.e., there is no background rate of silicosis in the unexposed population), absolute risks of silicosis mortality rather than excess risks were calculated for the Mannetje et al. pooled analysis (2002b, Document ID 1089). These risk estimates were derived from the rate ratios incorporating simulated measurement error reported by ToxaChemica (Document ID 0469). OSHA's estimate of lifetime risk of silicosis mortality, for 45 years of exposure to the previous general industry PEL, was 11 deaths per 1,000 workers for the pooled analysis (Document ID 1711). At the revised PEL, the risk estimate was 7 deaths per 1,000.

OSHA also calculated preliminary risk estimates for NMRD mortality. These estimates were derived from Park et al. (2002, Document ID 0405). For 45 years of exposure to the previous general industry PEL, OSHA preliminarily estimated lifetime excess risk at 83 deaths per 1,000 workers. At the revised PEL, OSHA estimated 43 deaths per 1,000 workers.

OSHA noted that, for exposures up to 250 µg/m3, the mortality risk estimates based on Park et al. (2002, Document ID 0405) are about 5 to 11 times as great as those calculated for the pooled analysis of silicosis mortality (Mannetje et al., 2002b, Document ID 1089). These two sets of risk estimates, however, are not directly comparable, as the endpoint for the Park et al. (2002, Document ID 0405) analysis was death from all non-cancer lung diseases, including pneumoconiosis, emphysema, and chronic bronchitis, whereas the pooled analysis by Mannetje et al. (2002b, Document ID 1089) included only deaths coded as silicosis or other pneumoconiosis. Less than 25 percent of the LDOC deaths in the Park et al. analysis were coded as silicosis or other pneumoconiosis (15 of 67), suggesting that silicosis as a cause of death may be misclassified as emphysema or chronic bronchitis. Thus, Mannetje et al.'s (2002b, Document ID 1089) selection of deaths may tend to underestimate the true risk of silicosis mortality, and Park et al.'s (2002, Document ID 0405) analysis may more completely capture the total respiratory mortality risk from all non-malignant causes.

Since the time of OSHA's analysis, NCHS has released updated all-cause mortality and NMRD mortality background rates from 2011 (http://wonder.cdc.gov/​ucd-icd10.html); OSHA's final risk estimates for NMRD mortality, which incorporate these updated rates (ICD10 codes J40-J47, chronic lower respiratory diseases; J60-J66, J68, pneumoconiosis and chemical effects), are available in Section VI, Final Quantitative Risk Assessment and Significance of Risk.

b. Lung Cancer Mortality

i. Exposure-Response Studies

In 1997, when IARC determined that there was sufficient evidence to regard crystalline silica as a human carcinogen, it also noted that some epidemiological studies did not demonstrate an excess risk of lung cancer and that exposure-response trends were not always consistent among studies that were able to describe such trends (Document ID 1062). These findings led Steenland et al. (2001a, Document ID 0452) to conduct a comprehensive exposure-response analysis—the IARC multi-center study—of the risk of lung cancer associated with exposure to crystalline silica. This study relied on all available cohort data from previously-published epidemiological studies for which there were adequate quantitative data on worker silica exposures to derive pooled estimates of disease risk. In addition, as discussed previously, OSHA identified four more recent studies suitable for quantitative risk assessment: (1) An exposure-response analysis by Rice et al. (2001, Document ID 1118) of a cohort of diatomaceous earth workers primarily exposed to cristobalite; (2) an analysis by Attfield and Costello (2004, Document ID 0285) of U.S. granite workers; (3) an exposure-response analysis by Hughes et al. (2001, Document ID 1060) of U.S. industrial sand workers; and (4) a risk analysis by Miller et al. (2007, Document ID 1305) and Miller and MacCalman (2009, Document ID 1306) of British coal miners. OSHA thoroughly described each of these studies in its Preliminary QRA (Document ID 1711); a brief summary of the exposure-response models used in each study is provided here.

The Steenland et al. pooled exposure-response analysis was based on data obtained from ten cohorts of silica-exposed workers (65,980 workers, 1,072 lung cancer deaths) (2001a, Document ID 0452). The pooled analysis cohorts included U.S. gold miners (Steenland and Brown, 1995a, Document ID 0450), U.S. diatomaceous earth workers (Checkoway et al., 1997, Document ID 0326), Australian gold miners (de Klerk and Musk, 1998, Document ID 0345), Finnish granite workers (Koskela et al., 1994, Document ID 1078), U.S. industrial sand employees (Steenland and Sanderson, 2001, Document ID 0455), Vermont granite workers (Costello and Graham, 1988, Document ID 0991), South African gold miners (Hnizdo and Sluis-Cremer, 1991, Document ID 1051; Hnizdo et al., 1997, 1049), and Chinese pottery workers, tin miners, and tungsten miners (Chen et al., 1992, Document ID 0329).

Steenland et al. (2001a, Document ID 0452) performed a nested case-control analysis via Cox regression. There were 100 controls chosen for each case randomly from among cohort members who survived past the age at which the case died; controls were matched on age (the time variable in Cox regression), study, race/ethnicity, sex, and date of birth within 5 years. Steenland et al. found that the use of any of the following continuous exposure variables in a log linear relative risk model resulted in positive statistically significant (p ≤ 0.05) exposure-response coefficients: (1) Cumulative exposure with a 15-year lag; (2) the log of cumulative exposure with a 15-year lag; and (3) average exposure (2001a, Document ID 0452). The models that provided the best fit to the data used cumulative exposure and log-transformed cumulative exposure. Models that used log-transformed cumulative exposure also showed no statistically significant heterogeneity among cohorts (p = 0.36), possibly because they are less influenced by very high exposures. At OSHA's request, Steenland (2010, Document ID 1312) also conducted a categorical analysis of the pooled data set and additional analyses using linear relative risk models (with and without the log transformation of cumulative exposure) as well as a two-piece spline model (see Document ID 1711, pp. 276-278).Start Printed Page 16313

Rice et al. (2001, Document ID 1118) applied a variety of exposure-response models to the California diatomaceous earth cohort data originally studied by Checkoway et al. (1993, Document ID 0324; 1996, 0325; 1997, 0326) and included in the Steenland et al. (2001a, Document ID 0452) pooled analysis. The cohort consisted of 2,342 white males employed for at least one year between 1942 and 1987 in a California diatomaceous earth mining and processing plant. The cohort was followed until 1994, and included 77 lung cancer deaths. Rice et al. reported that exposure to crystalline silica was a significant predictor of lung cancer mortality for nearly all of the models employed, with the linear relative risk model providing the best fit to the data in the Poisson regression analysis (2001, Document ID 1118).

Attfield and Costello (2004, Document ID 0285) analyzed the U.S. granite cohort originally studied by Costello and Graham (1988, Document ID 0991) and Davis et al. (1983, Document ID 0999) and included in the Steenland et al. (2001a, Document ID 0452) pooled analysis. The cohort consisted of 5,414 male granite workers who were employed in the Vermont granite industry between 1950 and 1982 and who had received at least one chest x-ray from the surveillance program of the Vermont Department of Industrial Hygiene. The 2004 report by Attfield and Costello extended follow-up from 1982 to 1994, and found 201 deaths (Document ID 0285). Using Poisson regression models, the results of a categorical analysis showed a generally increasing trend of lung cancer rate ratios with increasing cumulative exposure.

As mentioned previously, however, the rate ratio for the highest exposure group in the Attfield and Costello analysis (cumulative exposures of 6.0 mg/m3-yrs or higher) was substantially lower than that for other exposure groups (2004, Document ID 0285). The authors reported that the best-fitting model had a 15-year lag, untransformed cumulative exposure, and the omission of this highest exposure group. The authors argued that it was appropriate to omit the highest exposure group for several reasons, including that the exposure estimates for the highest exposure group were less reliable, and there was a greater likelihood of cohort selection effects, competing causes of death, and misdiagnosis (Document ID 0285, p. 136).

McDonald et al. (2001, Document ID 1091), Hughes et al. (2001, Document ID 1060) and McDonald et al. (2005, Document ID 1092) followed up on a cohort study of North American industrial sand workers included in the Steenland et al. (2001a, Document ID 0452) pooled analysis. The McDonald et al. cohort included 2,670 men employed before 1980 for three years or more in one of nine North American (8 U.S. and 1 Canadian) sand-producing plants, including 1 large associated office complex (2001, Document ID 1091). A nested case-control study based on 90 lung cancer deaths (through 1994) from this cohort was conducted by Hughes et al. (2001, Document ID 1060). A subsequent update (through 2000, 105 lung cancer deaths) eliminated the Canadian plant, following 2,452 men from the eight U.S. plants (McDonald et al., 2005, Document ID 1092). These nested case-control studies, Hughes et al. (2001, Document ID 1060) and McDonald et al. (2005, Document ID 1092), allowed for individual job, exposure, and smoking histories to be taken into account in the exposure-response analysis. Hughes et al. (2001, Document ID 1060) found statistically significant positive exposure-response trends for lung cancer for both cumulative exposure (lagged 15 years) and average exposure concentration, but not for duration of employment. With exposure lagged 15 years and after adjusting for smoking, increasing quartiles of cumulative silica exposure were also associated with lung cancer mortality (p-value for trend = 0.04). McDonald et al. (2005, Document ID 1092) found very similar results, with increasing quartiles of cumulative silica exposure (lagged 15 years) associated with lung cancer mortality (p-value for trend = 0.006). Because McDonald et al. (2005, Document ID 1092) did not report the medians of the exposure categories, and given the similar results of both case-control studies, OSHA chose to base its risk estimates on the Hughes et al. (2001, Document ID 1060) study.

Miller et al. (2007, Document ID 1305) and Miller and MacCalman (2009, Document ID 1306) continued a follow-up mortality study, begun in 1970, of coal miners from 10 British coal mines initially followed through the end of 1992 (Miller et al., 1997, Document ID 1304) and extended it to 2005. In the analysis using internal controls and Cox regression methods, the relative risk of lung cancer mortality, adjusted for concurrent dust exposure and smoking status, at a cumulative quartz exposure (lagged 15 years) equivalent of approximately 55 μg/m3 for 45 years was 1.14 (95% C.I., 1.04 to 1.25).

ii. Risk Estimates

In the Preliminary QRA, OSHA presented estimates of excess lung cancer mortality risk from occupational exposure to crystalline silica, based on data from the five epidemiology studies discussed above (Document ID 1711). In its preliminary analysis, OSHA used background all-cause mortality and lung cancer mortality rates from 2006, as reported by the National Center for Health Statistics (NCHS) (Document ID 1104). These rates were used in life table analyses to estimate lifetime risks at the exposure levels of interest, ranging from 25 to 500 μg/m3 respirable crystalline silica.

OSHA's preliminary estimates of lifetime excess lung cancer risk associated with 45 years of exposure to crystalline silica at 100 μg/m3 (approximately the previous general industry PEL) ranged between 13 and 60 deaths per 1,000 workers, depending upon the study used. For exposure to the revised PEL of 50 μg/m3, the lifetime risk estimates were in the range of between 6 and 26 deaths per 1,000 workers, depending upon the study used. For a 45 year exposure at the new action level of 25 μg/m3, OSHA estimated the risk to range between 3 and 23 deaths per 1,000 workers. The Agency found that the results from these preliminary assessments were reasonably consistent despite the use of data from different cohorts and the reliance on different analytical techniques for evaluating dose-response relationships.

OSHA also estimated the lung cancer risk associated with 45 years of exposure to the previous construction/shipyard PEL (in the range of 250 μg/m3 to 500 μg/m3) to range between 37 and 653 deaths per 1,000 workers, depending upon the study used. OSHA acknowledges that the 653 deaths is the upper limit for 45 years of exposure to 500 μg/m3, and recognizes that actual risk, to the extent that workers are exposed for less than 45 years or intermittently, is likely to be lower. In addition, exposure to 250 or 500 μg/m3 over 45 years represents cumulative exposures of 11.25 and 22.5 mg/m3-yrs, respectively. This range of cumulative exposure is well above the median cumulative exposure for most of the cohorts used in the preliminary risk assessment. Thus, OSHA explained that estimating lung cancer excess risks over this higher range of cumulative exposures of interest to OSHA required some degree of upward extrapolation of the exposure-response function to model these high exposures, thus adding uncertainty to the estimates.

Since the time of that original analysis, NCHS has released updated all-cause mortality and lung cancer mortality background rates from 2011. Start Printed Page 16314OSHA's final risk estimates, which incorporate these updated rates, are available in this preamble at Section VI, Final Quantitative Risk Assessment and Significance of Risk.

c. Uncertainty Analysis of Pooled Studies of Lung Cancer Mortality and Silicosis Mortality

In the Preliminary QRA, OSHA recognized that risk estimates can be inherently uncertain and can be affected by confounding, selection bias, and measurement error (Document ID 1711). OSHA presented several reasons as to why it does not believe that confounding or selection bias had a substantial impact on the risk estimates for lung cancer or silicosis mortality (Document ID 1711, pp. 299-302). However, because it was more difficult to assess the importance of exposure measurement error, OSHA's contractor, ToxaChemica, Inc., commissioned Drs. Kyle Steenland and Scott Bartell to perform an uncertainty analysis to examine the effect of uncertainty due to measurement error in the pooled studies (Steenland et al., 2001a, Document ID 0452; Mannetje 2002b, 1089) on the lung cancer and silicosis mortality risk estimates (ToxaChemica, Inc., 2004, Document ID 0469).

There are two main sources of error in the silica exposure measurements. The first arises from the assignment of individual workers' exposures based on either exposure measurements for a sample of workers in the same job or estimated exposure levels for specific jobs in the past when no measurements were available, via a job-exposure matrix (JEM) (Mannetje et al., 2002a, Document ID 1090). The second arises from the conversion of historically-available dust measurements, typically particle count concentrations, to gravimetric respirable silica concentrations. ToxaChemica, Inc. conducted an uncertainty analysis using the raw data from the IARC multi-centric study to address these sources of error (2004, Document ID 0469).

i. Lung Cancer Mortality

To examine the effect of error in the assignment of individual exposure values in the cohorts studied by Steenland et al. (2001a, Document ID 0452), ToxaChemica, Inc. used a Monte Carlo analysis (a type of simulation analysis that varies the values of an uncertain input to an analysis—in this case, exposure estimates—to explore the effects of different values on the outcome of the analysis) to randomly sample new values for each worker's job-specific exposure levels from a distribution that they believed characterized the variability in exposures of individual workers in each job (see Document ID 1711, pp. 303-305). That is, ToxaChemica created a distribution of values for each member of each cohort where the mean exposure for each member was equal to the original exposure value and the distribution of exposure values was based on a log-normal distribution having a standard deviation that was based on the exposure variation observed in industrial sand plants observed by Steenland and Sanderson (2001, Document ID 0455). From this distribution, new sets of exposure values from each cohort member were randomly drawn for 50 trials. This simulation was designed to test whether sets of exposure values that were plausibly different from the original estimates would lead to substantially different results of the exposure-response analysis. Except for the simulated exposure values and the correction of a few minor errors in the original data sets, the simulation analysis used the same data as the original analyses conducted by Steenland et al. (2001a, Document ID 0452).

When an entire set of cumulative exposure values was assembled for all workers based on these randomly sampled values, the set was used in a conditional logistic regression to fit a new exposure-response model. The extent to which altering the exposure values led to changes in the results indicated how sensitive the previously presented risk estimates may have been to error in the exposure estimates. Among the individual cohorts, most of the mean regression coefficients resulting from the simulation analysis were consistent with the coefficients from the exposure-response analyses reported in Steenland et al. (2001, Document ID 0455) and ToxaChemica, Inc. (2004, Document ID 0469) (following correction for minor data entry and rounding errors). An exception was the mean of the simulation coefficients based on the South Africa gold cohort (0.26), which was lower than the previously calculated exposure coefficient (0.582). ToxaChemica, Inc. (2004, Document ID 0469) concluded that this error source probably did not appreciably change the estimated exposure-response coefficient for the pooled data set.

To examine the effect of error in estimating gravimetric respirable crystalline silica exposures from historical dust concentration data (i.e., particle count data), ToxaChemica, Inc. (2004, Document ID 0469) used a procedure similar to that used to assess uncertainties in individual exposure value assignments. ToxaChemica, Inc. assumed that, for each job in the dataset, a specific conversion factor existed that related workers' exposures measured as particle concentrations to gravimetric respirable silica exposures, and that this conversion factor came from a normal distribution with a standard deviation σ = 1/2 its mean μ. The use of a normal distribution was a reasonable choice in that it allowed the sampled conversion factors to fall above or below the original values with equal probability, as the authors had no information to suggest that error in either direction was more likely. The normal distribution also assigned higher probability to conversion values closer to the original values. The choice of the normal distribution therefore reflected the study authors' judgment that their original conversion factors were more likely to be approximately correct than not, while allowing for the possibility of significant error in the original values.

A new conversion factor was then sampled for each job from the appropriate distribution, and the complete set of sampled conversion factors was then used to re-run the risk analysis used by Steenland et al. (2001a, Document ID 0452). The results were similar to the coefficients originally derived from each cohort; the only coefficient substantially affected by the procedure was that for the South African cohort, with an average value of 0.350 across ten runs compared to the original value of 0.582 (see Table II-5, Document ID 1711, p. 307). This suggests that the results of exposure-response analyses conducted using the South African cohort are sensitive to error in exposure estimates; therefore, there is greater uncertainty due to potential exposure estimation error in an exposure-response model based on this cohort than is the case for the other nine cohorts in Steenland et al's analysis.

To explore the potential effects of both kinds of random uncertainty described above, ToxaChemica, Inc. (2004, Document ID 0469) used the distributions representing the error in job-specific exposure assignment and the error in converting exposure metrics to generate 50 new exposure simulations for each cohort. A study-specific coefficient and a pooled coefficient were fit for each new simulation, with the assumption that the two sources of uncertainty were independent. The results indicated that the only cohort for which the mean of the exposure coefficients derived from the 50 simulations differed substantially from the previously calculated exposure Start Printed Page 16315coefficient was the South African gold cohort (simulation mean of 0.181 vs. original coefficient of 0.582). For the pooled analysis, the mean coefficient estimate from the simulations was 0.057, just slightly lower than the previous estimate of 0.060. Based on these results, OSHA concludes that random error in the underlying exposure estimates in the Steenland et al. (2001a, Document ID 0452) pooled cohort study of lung cancer is not likely to have substantially influenced the original risk estimates derived from the pooled data set, although the model coefficient for one of the ten cohorts (the South African gold miner cohort) appeared to be sensitive to measurement errors (see Table II-5, Document ID 1711, p. 307).

Drs. Steenland and Bartell also examined the effects of systematic bias in conversion factors, considering the possibility that these may have been consistently under-estimated or over-estimated for any given cohort. They addressed possible biases in either direction, conducting simulations where the true silica content was assumed to be either half or double the estimated silica content of measured exposures. For the conditional logistic regression model using log cumulative exposure with a 15-year lag, doubling or halving the exposure for a specific study resulted in virtually no change in the exposure-response coefficient for that study or for the pooled analysis overall. This is due to the use of log-transformed exposure metrics, which ensured that any multiplicative bias in exposure would have virtually no effect on conditional logistic regression coefficients (Document ID 0469, p. 17). That is, for this model, a systematic error in exposure estimation for any study had little effect on the lung cancer response rate for either the specific study or the pooled analysis overall.

ii. Silicosis Mortality

Following the procedures described above for the lung cancer analysis, Toxachemica, Inc. (2004, Document ID 0469) combined both sources of random measurement error in a Monte Carlo analysis of the silicosis mortality data from Mannetje et al. (2002b, Document ID 1089). Categorical analyses were performed with a nested case control model, in contrast to the Poisson model used previously by Mannetje et al. (2002b, Document ID 1089). The nested case control model was expected to control more effectively for age. This model yielded categorical rate ratio results using the original data (prior to simulation of measurement error) which were approximately 20-25 percent lower than those reported by Mannetje et al. (2002b, Document ID 1089). The silicosis mortality dataset thus appeared to be more sensitive to possible error in exposure measurement than the lung cancer dataset, for which the mean of the simulation coefficients was virtually identical to the original. OSHA notes that its risk estimates derived from the pooled analysis (Mannetje et al., 2002b, Document ID 1089), incorporated ToxaChemica, Inc.'s simulated measurement error (2004, Document ID 0469). More information is provided in the Preliminary QRA (Document ID 1711, pp. 310-314).

d. Renal Disease Mortality

i. Exposure-Response Studies

Steenland et al. (2002a, Document ID 0448) examined renal disease mortality in a pooled analysis of three cohorts, as discussed previously. These cohorts were chosen because data were available for both underlying cause mortality and multiple cause mortality. The combined cohort for the pooled analysis (Steenland et al., 2002a, Document ID 0448) consisted of 13,382 workers with exposure information available for 12,783 (95 percent). SMRs (compared to the U.S. population) for renal disease (acute and chronic glomerulonephritis, nephrotic syndrome, acute and chronic renal failure, renal sclerosis, and nephritis/nephropathy) were statistically significantly elevated using multiple cause data (SMR 1.29, 95% CI 1.10-1.47, 193 deaths) and underlying cause data (SMR 1.41, 95% CI 1.05-1.85, 51 observed deaths).

ii. Risk Estimates

As detailed in the Preliminary QRA, OSHA estimated that exposure to the previous (100 μg/m3) and revised (50 μg/m3) general industry PELs, over a 45-year working life, would result in a lifetime excess renal disease mortality risk of 39 and 32 deaths per 1,000 workers, respectively. For exposure to the previous construction/shipyard PELs, OSHA estimated the lifetime excess risk to range from 52 to 63 deaths per 1,000 workers at exposures of 250 and 500 μg/m3, respectively. These risks reflect the 1998 background all-cause mortality and renal mortality rates for U.S. males. Background rates were not adjusted for the renal disease risk estimates because the CDC significantly changed the classification of renal diseases after 1998; they are now inconsistent with those used by Steenland et al. (2002a, Document ID 0448) to ascertain the cause of death of workers in their study.

e. Silicosis Morbidity

i. Exposure-Response Studies

OSHA summarized, in its Preliminary QRA, the principal cross-sectional and cohort studies that quantitatively characterized relationships between exposure to crystalline silica and the development of radiographic evidence of silicosis (Document ID 1711). Each of these studies relied on estimates of cumulative exposure to evaluate the relationship between exposure and silicosis prevalence. The health endpoint of interest in these studies was the appearance of opacities on chest radiographs indicative of pulmonary fibrosis. Most of the studies reviewed by OSHA considered a finding consistent with an ILO classification of 1/1 to be a positive diagnosis of silicosis, although some also considered an x-ray classification of 1/0 or 0/1 to be positive. OSHA noted its belief, in the Preliminary QRA, that the most reliable estimates of silicosis morbidity, as detected by chest radiographs, come from the studies that evaluated radiographs over time, included radiographic evaluation of workers after they left employment, and derived cumulative or lifetime estimates of silicosis disease risk. OSHA also pointed out that the low sensitivity of chest radiography in detecting silicosis suggests that risk estimates derived from radiographic evidence likely underestimate the true risk.

Hnizdo and Sluis-Cremer (1993, Document ID 1052) described the results of a retrospective cohort study of 2,235 white gold miners in South Africa. A total of 313 miners had developed silicosis (x-ray with ILO 1/1 or greater) and had been exposed for an average of 27 years at the time of diagnosis. The average latency for the cohort was 35 years (range of 18-50 years) from the start of exposure to diagnosis. The average respirable dust exposure for the cohort overall was 290 μg/m3 (range 110-470), corresponding to an estimated average respirable silica concentration of 90 μg/m3 (range 33-140). The average cumulative dust exposure for the overall cohort was 6.6 mg/m3-yrs (range 1.2-18.7). Silicosis risk increased exponentially with cumulative exposure to respirable dust in models using log-logistic regression. Using the exposure-response relationship developed by Hnizdo and Sluis-Cremer (1993, Document ID 1052), and assuming a quartz content of 30 percent in respirable dust, Rice and Stayner (1995, Document ID 0418) estimated the risk of silicosis to be 13 percent for a 45-year exposure to 50 μg/m3 respirable crystalline silica.Start Printed Page 16316

Steenland and Brown (1995b, Document ID 0451) studied 3,330 South Dakota gold miners who had worked at least a year underground between 1940 and 1965. Chest x-rays were obtained in cross-sectional surveys in 1960 and 1976 and used along with death certificates to ascertain cases of silicosis; 128 cases were found via death certificate, 29 were found by x-ray (defined as ILO 1/1 or greater), and 13 were found by both. OSHA notes that the inclusion of death certificate diagnoses complicates interpretation of the risk estimate from this study since, as noted by Finkelstein (2000, Document ID 1015), it is not known how well such diagnoses correlate with ILO radiographic interpretations; as such, the risk estimates derived from this study may not be directly comparable to others that rely exclusively on radiographic findings to evaluate silicosis morbidity risk. The mean exposure concentration was 50 μg/m3 for the overall cohort, with those hired before 1930 exposed to an average of 150 μg/m3. The average duration of exposure for workers with silicosis was 20 years (s.d. = 8.7) compared to 8.2 years (s.d. = 7.9) for the rest of the cohort. This study found that cumulative exposure was the best disease predictor, followed by duration of exposure and average exposure. Lifetime risks were estimated from Poisson regression models using standard life table techniques; the results indicated an estimated risk of 47 percent associated with 45 years of exposure to 90 μg/m3 respirable crystalline silica, which reduced to 35 percent after adjustment for age and calendar time.

OSHA used the same life table approach as described for estimating lung cancer and NMRD mortality risks to estimate lifetime silicosis risk based on the silicosis rates, adjusted for age and calendar time, calculated by Steenland and Brown (1995b, Table 2, Document ID 0451). Silicosis risk was estimated through age 85, assuming exposure from age 20 through 65, and assuming that the silicosis rate remains constant after age 65. All-cause mortality rates to all males for calendar year 2006 were used to account for background competing risk. From this analysis, OSHA estimated the risk from exposure to the previous general industry PEL of 100 μg/m3 to be 43 percent; this is somewhat higher than estimated by Steenland and Brown (1995b) because of the use by OSHA of more recent mortality data and calculation of risk through age 85 rather than 75. For exposure to the revised PEL of 50 μg/m3, OSHA estimated the lifetime risk to be 7 percent. Since the time of the original analysis, NCHS has released updated all-cause mortality background rates from 2011; OSHA's final risk estimates, which incorporate these updated rates, are available in Section VI, Final Quantitative Risk Assessment and Significance of Risk.

Miller et al. (1995, Document ID 1097; 1998, 0374) and Buchanan et al. (2003, Document ID 0306) reported on a follow-up study conducted in 1990 and 1991 of 547 survivors of a 1,416 member cohort of Scottish coal workers from a single mine. These men all worked in the mine during a period between early 1971 and mid-1976, during which they had experienced “unusually high concentrations of freshly cut quartz in mixed coalmine dust” (Document ID 0374, p.52). Thus, this cohort allowed for the study of exposure-rate effects on the development of silicosis. The men all had radiographs dating from before, during, or just after this high concentration period, and the 547 participating survivors received follow-up chest x-rays between November 1990 and April 1991.

Buchanan et al. (2003, Document ID 0306) presented logistic regression models in stages. In the first stage they compared the effect of pre- vs. post-1964 cumulative quartz exposures on odds ratios; this yielded a statistically significant odds ratio estimate for post-1964 exposures. In the second stage they added total dust levels both pre- and post-1964, age, smoking status, and the number of hours worked pre-1954; only post-1964 cumulative exposures remained significant. Finally, in the third stage, they started with only the statistically significant post-1964 cumulative exposures, and separated these exposures into two quartz bands, one for exposure to concentrations less than 2,000 μg/m3 respirable quartz and the other for concentrations greater than or equal to 2,000 μg/m3. Both concentration bands were highly statistically significant in the presence of the other, with the coefficient for exposure concentrations greater than or equal to 2000 μg/m3 being three times that of the coefficient for concentrations less than 2000 μg/m3. From this, the authors concluded that their analysis showed that “the risks of silicosis over a working lifetime can rise dramatically with exposure to such high concentrations over a timescale of merely a few months” (Buchanan et al. 2003, Document ID 0306, p. 163). The authors then used the model to estimate the risk of acquiring a chest x-ray classified as ILO category 2/1+, 15 years after exposure, as a function of both low (<2000 μg/m3) and high (>2000 μg/m3) quartz concentrations. OSHA chose to use this model to estimate the risk of radiological silicosis consistent with an ILO category 2/1+ chest x-ray for several exposure scenarios; in each, it assumed 45 years of exposure, 2000 hours/year of exposure, and no exposure above a concentration of 2000 μg/m3. The results showed that occupational exposures to the revised PEL of 50 μg/m3 led to an estimated risk of 55 cases per 1,000 workers. Exposure at the previous general industry PEL of 100 μg/m3 increased the estimate to 301 cases per 1,000 workers. At higher exposure levels the risk estimates rose quickly to near certainty.

Chen et al. (2001, Document ID 0332) reported the results of a retrospective study of a Chinese cohort of 3,010 underground miners who had worked in tin mines at least one year between 1960 and 1965. They were followed through 1994, by which time 2,426 (80.6 percent) workers had either retired or died, and only 400 (13.3 percent) remained employed at the mines. Annual radiographs were taken beginning in 1963 and cohort members continued to have chest x-rays taken every 2 or 3 years after leaving work. Silicosis was diagnosed when at least 2 of 3 radiologists classified a radiograph as being a suspected case or at Stage I, II, or III under the 1986 Chinese pneumoconiosis roentgen diagnostic criteria, which the authors reported agreed closely with ILO categories 0/1, Category 1, Category 2, and Category 3, respectively. Silicosis was observed in 33.7 percent of the group; 67.4 percent of the cases developed after exposure ended.

Chen et al. (2001, Document ID 0332) found that a Weibull model provided the best fit to relate cumulative silicosis risk to eight categories of cumulative total dust exposure. The risk of silicosis was strongly related to cumulative silica exposure. The investigators predicted a 55-percent risk of silicosis associated with 45 years of exposure to 100 μg/m3. The paper did not report the risk associated with a 45-year exposure to 50 μg/m3, but OSHA estimated the risk to be about 17 percent (based on the parameters of the Weibull model).

In a later study, Chen et al. (2005, Document ID 0985) investigated silicosis morbidity risks among three cohorts to determine if the risk varied among workers exposed to silica dust having different characteristics. The cohorts consisted of 4,547 pottery workers, 4,028 tin miners, and 14,427 tungsten miners, all employed after January 1, 1950 and selected from a total of 20 workplaces. The approximate Start Printed Page 16317mean cumulative exposures to respirable silica for pottery, tin, and tungsten workers were 6.4 mg/m3-yrs, 2.4 mg/m3-yrs, and 3.2 mg/m3-yrs, respectively. Measurement of particle surface occlusion (presence of a mineral coating that may affect the biological availability of the quartz component) indicated that, on average, 45 percent of the surface area of respirable particles collected from pottery factory samples was occluded, compared to 18 percent of the particle surface area for tin mine samples and 13 percent of particle surface area for tungsten mines. When cumulative silica exposure was adjusted to reflect exposure to surface-active quartz particles (i.e., not occluded), the estimated cumulative risk among pottery workers more closely approximated those of the tin and tungsten miners, suggesting to the authors that alumino silicate occlusion of the crystalline particles in pottery factories at least partially explained the lower risk seen among pottery workers, despite their having been more heavily exposed. Based on Chen et al. (2005, Document ID 0985), OSHA estimated the cumulative silicosis risk associated with 45 years of exposure to 100 μg/m3 respirable crystalline silica to be 6 percent for pottery workers, 12 percent for tungsten miners, and 40 percent for tin miners. For 45 years of exposure to 50 μg/m3, cumulative silicosis morbidity risks were estimated to be 2 percent for pottery workers, 2 percent for tungsten miners, and 10 percent for tin miners.

ii. Risk Estimates

OSHA's risk estimates for silicosis morbidity ranged between 60 and 773 per 1,000 workers for a 45-year exposure to the previous general industry PEL of 100 μg/m3, and between 20 and 170 per 1,000 workers for a 45-year exposure to the revised PEL of 50 μg/m3, depending upon the study used. OSHA recognizes that actual risk, to the extent that workers are exposed for less than 45 years or intermittently, is likely to be lower, but also recognizes that silicosis can progress for years after exposure ends. Also, given the consistent finding of a monotonic exposure-response relationship for silicosis morbidity with cumulative exposure in the studies reviewed, OSHA continues to find that cumulative exposure is a reasonable exposure metric upon which to base risk estimates in the exposure range of interest.

D. Comments and Responses Concerning Silicosis and Non-Malignant Respiratory Disease Mortality and Morbidity

In this section, OSHA focuses on comments pertaining to the literature used by the Agency to assess risk for silicosis and non-malignant respiratory disease (NMRD) mortality and morbidity. As discussed in the Review of Health Effects Literature and Preliminary QRA (Document ID 1711) and in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA, of this preamble, OSHA used two studies (ToxaChemica, 2004, Document ID 0469; Park et al., 2002, 0405) to determine lifetime risk for silicosis and NMRD mortality and five studies (Buchanan et al., 2003, Document ID 0306; Chen et al., 2001, 0332; Chen et al., 2005, 0985; Hnizdo and Sluis-Cremer, 1993, 1052; and Steenland and Brown, 1995b, 0451) to determine cumulative risk for silicosis morbidity. OSHA discussed the reasons for selecting these scientific studies for quantitative risk assessment in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 340-342). Briefly, OSHA concluded that the aforementioned studies used scientifically accepted techniques to measure silica exposures and health effects in order to determine exposure-response relationships. The Agency believed, and continues to believe, that these studies, as a group, provide the best available evidence of the exposure-response relationships between silica exposure and silicosis morbidity, silicosis mortality, and NMRD mortality and that they constitute a solid and reliable foundation for OSHA's final risk assessment.

OSHA received both supportive and critical comments and testimony regarding these studies. Comments largely focused on how the authors of these studies analyzed their data, and concerns expressed by commenters generally focused on exposure levels and measurement, potential biases, confounding, statistical significance of study results, and model forms. This section does not include extensive discussion on exposure measurement error, potential biases, thresholds, confounding factors, and the use of the cumulative exposure metric, which are discussed in depth in other sections of this preamble, including V.J Comments and Responses Concerning Biases in Key Studies and V.K Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis. OSHA addresses comments on general model form and various other issues here and concludes that these comments do not meaningfully affect OSHA's reliance on the studies discussed herein or the results of the Agency's final risk assessment.

1. Silicosis and NMRD Mortality

There are two published studies that report quantitative risk assessments of silicosis and NMRD mortality (see Document ID 1711, pp. 292-298). The first is an exposure-response analysis of diatomaceous earth (DE) workers (Park et al., 2002, Document ID 0405). Park et al. quantified the relationship between cristobalite exposure and mortality caused by NMRD, which includes silicosis, pneumoconiosis, emphysema, and chronic bronchitis (Park et al. refers to these conditions as “lung disease other than cancer (LDOC),” while OSHA uses the term “NMRD”). Because NMRD captures much of the silicosis misclassification that results in underestimation of the disease and includes risks from other lung diseases associated with crystalline silica exposures, OSHA believes the risk estimates derived from the Park et al. study reasonably reflect the risk of death from silica-related respiratory diseases, including silicosis (Document ID 1711, pp. 297-298). The second study (Mannetje et al. 2002b, Document ID 1089) is a pooled analysis of six epidemiological studies that were part of an IARC effort. OSHA's contractor ToxaChemica later conducted a reanalysis and uncertainty analysis using these data (ToxaChemica, 2004, Document ID 0469). OSHA believes that the estimates from the pooled study represent credible estimates of mortality risk from silicosis across a range of industrial workplaces, but are likely to understate the actual risk because silicosis is under-reported as a cause of death.

a. Park et al. (2002)

The American Chemistry Council (ACC) submitted several comments pertaining to the Park et al. (2002, Document ID 0405) study, including comments on the cohort's exposure concentrations. In its post-hearing brief, the ACC noted that the mean crystalline silica exposure in Park's DE cohort was estimated to be more than three times the former general industry PEL of 100 μg/m3 and the mean estimated exposure of the workers with silicosis could have been close to 10 times that level. According to the ACC, extrapolating risks from the high exposure levels in this cohort to the much lower levels relevant to OSHA's risk assessment (the previous general industry PEL of 100 Start Printed Page 16318μg/m3 and the revised PEL of 50 μg/m3) is “fraught with uncertainty” (Document ID 4209, pp. 84-85).

OSHA acknowledges that there is some uncertainty in using models heavily influenced by exposures above the previous PEL due to potential deviance at areas of the relationship with fewer data points. However, OSHA believes that the ACC's characterization of exposures in the Park et al. (2002) study as vastly higher than the final and former PELs is incorrect. The ACC focused on mean exposure concentrations, reported by Park et al. as 290 μg/m3, to make this argument (Document ID 0405, p. 37). However, in the Park et al. study, the mean cumulative exposure of the cohort was 2.16 mg/m3-yrs, lower than what the final rule would permit over 45 years of exposure (2.25 mg/m3-yrs) (Document ID 0405, p. 37). Thus, whereas some participants in the Park et al. study had higher average-8-hour exposures than were typical under the previous PEL, they were quite comparable to the exposures workers might accumulate over their working lives under the final PEL of 50 μg/m3. In addition, as discussed in Section V.M, Comments and Responses Concerning Working Life, Life Tables, and Dose Metric, OSHA believes that the evidence in the rulemaking record, including comments and testimony from NIOSH (Document ID 3579, Tr. 127), Kyle Steenland, Ph.D. (Document ID 3580, Tr. 1227), and OSHA peer reviewer Kenneth Crump, Ph.D. (Document ID 1716, p. 166), points to cumulative exposure as a reasonable and appropriate dose metric for deriving exposure-response relationships. In sum, OSHA does not agree that the Park study should be discounted based on the ACC's concerns about the estimated exposure concentrations in the diatomaceous earth cohort.

The ACC also criticized the Park study for its treatment of possible confounding by smoking and exposure to asbestos. The ACC commented in its pre-hearing brief that data on smoking was available for only half of the cohort (Document ID 2307, Attachment A, p. 108). The Panel also wrote that, “while Park et al. dismissed asbestos as a potential confounder and omitted asbestos exposure in their final models, the situation is not as clear-cut as they would have one believe” (Document ID 2307, Attachment A, p. 109). The Panel highlighted that Checkoway et al. (1997), the study upon which Park relied to dismiss asbestos as a potential confounder, noted that “misclassification of asbestos exposure may have hindered our ability to control for asbestos as a potential confounder” (Document ID 0326, p. 685; 2307, Attachment A, p. 109).

OSHA has reviewed the ACC's concerns, and maintains that Park et al. adequately addressed the issues of possible confounding by smoking and exposure to asbestos in this data set. Smoking habits of a third of the individuals who died from NMRD were known in the Park et al. (2002) study. Based on that partial knowledge of smoking habits, Park et al. presented analyses indicating that confounding by smoking was unlikely to significantly impact the observed relationship between cumulative exposure to crystalline silica and NMRD mortality (Document ID 0405, p. 41). Specifically, Park et al. (2002) performed internally standardized analyses, which tend to be less susceptible to confounding by smoking since they compare the mortality experience of groups of workers within the cohort rather than comparing the mortality experience of the cohort with an external population (such as by using national mortality rates); the authors found that the internally standardized models yielded only slightly lower exposure-response coefficients than externally adjusted models (Document ID 0405; 1711, p. 302). These results suggested that estimates of NMRD mortality risks based on this cohort are not likely to be exaggerated due to cohort members' smoking habits. Park et al. also stated that the authors' findings regarding possible confounding by smoking were consistent with those of Checkoway et al., who also concluded there it was “very unlikely” that smoking could explain the association between mortality from NMRD and silica exposure in this cohort (Document ID 0405, p. 41; 0326, p. 687). NIOSH noted that “[r]esidual confounding from poorly characterized smoking could have an effect,” but that effect could be either positive or negative (Document ID 4233, pp. 32-33). While OSHA agrees that comprehensive smoking data would be ideal, the Agency believes that the approach taken by Park et al. to address this issue was reasonable.

Asbestos exposure was estimated for all workers in Park et al., which enabled the researchers to directly test confounding. They “found no confounding by asbestos” and, accordingly, omitted asbestos exposure in their final modeling (Document ID 0405, p. 41). As discussed in the Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 301-302), exposure to asbestos was particularly prevalent among workers employed prior to 1930; after 1930, asbestos was presumably no longer used in the process (Gibbs, 1998, Document ID 1024, p. 307; Checkoway et al., 1998, 0984, p. 309). Checkoway et al. (1998), who evaluated the issue of asbestos confounding for the same cohort used by Park et al., found that the risk ratio for the highest silica exposure group after excluding the workers employed before 1930 from the cohort (Relative Risk (RR) = 1.73) was almost identical to the risk ratio of the high-exposure group before excluding those same workers (RR = 1.74) (Document ID 0984, p. 309). In addition, Checkoway's reanalysis of the original cohort study (Checkoway et al., 1993) examined those members of the cohort for whom there was quantitative information on asbestos exposure, based on a mixture of historical exposure monitoring data, production records, and recorded quantities of asbestos included in mixed products of the plant (Checkoway et al., 1996, Document ID 0325). The authors found an increasing trend in lung cancer mortality with exposure to crystalline silica after controlling for asbestos exposure and found only minor changes in relative risk estimates after adjusting for asbestos exposure (1996, Document ID 0325). Finally, Checkoway et al. (1998) reported that the prevalence of pleural abnormalities (indicators of asbestos exposure) among workers hired before 1930 (4.2 percent) was similar to that of workers hired after 1930 who presumably had no asbestos exposure (4.9 percent), suggesting that asbestos exposure was not a confounder for lung abnormalities in this group of workers (Document ID 0984, p. 309). Therefore, Checkoway et al. (1998) concluded that asbestos was not likely to significantly confound the exposure-response relationship observed between lung cancer mortality and exposure to crystalline silica in diatomaceous earth workers.

Rice et al. also utilized Checkoway's (1997, Document ID 0326) data to test for confounding by asbestos in their Poisson and Cox proportional hazards models. Finding no evidence of confounding, Rice et al. did not include asbestos exposure as a variable in the final models presented in their 2001 paper (Document ID 1118, p. 41). Based on these numerous assessments of the effects of exposure to asbestos in the diatomaceous earth workers cohort used by Park et al. (2002), OSHA concludes that concerns about asbestos confounding in this cohort have been adequately addressed and that the additional analyses performed by Park et al. on this issue confirmed the findings of prior researchers that Start Printed Page 16319confounding by asbestos exposure was not likely to have a large effect on exposure-response relationships.

The ACC also expressed concern about model selection. Louis Anthony Cox, Jr., Ph.D., of Cox Associates, on behalf of the ACC, was concerned that the linear relative rate model was not appropriate because it is not designed to test for exposure-response thresholds and, similarly, the ACC has argued that threshold models are appropriate for crystalline silica-related diseases (Document ID 2307, Attachment 4, pp. 91). The ACC claimed that the Park et al. (2002) study is “fully consistent” with a threshold above the 100 μg/m3 concentration for NMRD, including silicosis, mortality (Document ID 2307, Attachment A, p. 107).

In its post-hearing comments, NIOSH explained that categorical analysis for NMRD indicated no threshold existed with cumulative exposure corresponding to 25 μg/m3 over 40 years of exposure, which is below the cumulative exposure equivalent to the new PEL over 45 years (Document ID 4233, p. 27). Park et al. did not estimate a threshold below that level because the data lacked the power needed to discern a threshold (Document ID 4233, p. 27). OSHA agrees with NIOSH's assessment. In addition, as discussed extensively in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases, OSHA has carefully reviewed the issue of thresholds and has concluded, based on the best available evidence, that workers with cumulative and average exposure levels permitted under the previous PEL of 100 μg/m3 are at risk of silica-related disease (that is, there is unlikely to be an exposure-response threshold at or near 100 μg/m3). For these reasons, OSHA disagrees with Dr. Cox's criticism of Park et al.'s reliance on the linear relative rate model.

The ACC then questioned the use of unlagged cumulative exposures as the metric in Park et al. (2002). Dr. Cox noted that “[u]nlagged models are not very biologically plausible for dust-related NMRD deaths (if any) caused by exposure concentrations in the range of interest. Unresolved chronic inflammation and degradation of lung defenses takes years to decades to manifest” (Document ID 2307, Attachment 4, p. 92). OSHA considers this criticism overstated. Park et al. considered a range of lag periods, from two years to 15. They found that “[u]nlagged models seemed to provide the best fit to the data in Poisson analyses although lagged models performed almost as well” (Document ID 0405, p. 37). Based on those findings, as well as acknowledgments that NMRD effects other than silicosis (e.g., chronic bronchitis) may be observable without a relatively long lag time (unlike cancer) and that the majority of deaths observed in the cohort were indeed NMRD other than silicosis, the researchers decided to use an unlagged model. Because Park found the differences between the lagged and unlagged models for this cohort and the NMRD endpoint to be insignificant, OSHA finds that Park's final choice to use an unlagged model does not detract from OSHA's decision to utilize lagged models in its risk assessment.

The ACC was also concerned about the truncation of cumulative exposures in the Park et al. (2002) paper. Peter Morfeld, Dr. rer. medic, stated that Park et al.:

suffers from a methodological drawback. . . . The authors truncated the cumulative RCS dust exposures before doing the final analyses based on their observation of where the cases were found. The maximum in the study was 62.5 mg/m3-years but exposures were only used up to 32 mg/m3-years because no LDOC deaths occurred at exposures higher than that level. Such a selection distorts the estimated exposure-response relationship because it is based on the outcome of the study and on the exposure variable. Because high exposures with no effects were deliberately ignored, the exposure-response effect estimates are biased upward (Document ID 2307, Attachment 2, p. 27).

OSHA acknowledges this concern about the truncation of data in the study, and asked Mr. Park about it at the public hearing. Mr. Park testified that there were good reasons to truncate the part of the exposed workforce at the high end of cumulative exposure. He noted several plausible reasons for the drop-off in the number of cases at high exposures (attenuation), including random variance in susceptibility to disease among different people and the healthy worker survivor effect [6] (Document ID 3579, Tr. 242-243). He also stated that this attenuation is a common occurrence in studies of workers (Document ID 3579, Tr. 242). Mr. Park then emphasized that how one describes the higher end of the exposure-response relationship is inconsequential for the risk assessment process because the relationship at the lower end of the spectrum, where the PEL was determined, is more important for rulemaking (Document ID 3579, Tr. 242-243). He also stated, in a post-hearing comment, that “[f]or the purpose of low exposure extrapolation, adding a quadratic term [to better describe the entirety of the exposure-response relationship] would result in loss of precision with no advantage [gained] over truncation of high cumulative exposure observation time” (Document ID 4233, p. 26). To summarize, Mr. Park stated that there are good scientific reasons to expect attenuation of exposure-response at the high end of the cumulative exposure range and that use of higher-exposure data affected by healthy worker survivor effect or other issues could reduce precision of the exposure-response model at the lower exposures that are more relevant to the final silica standard. OSHA finds that Mr. Park's approach in his study, along with his explanations in the rulemaking record, are reasonable and that he has fully responded to the concerns of the ACC.

Dr. Morfeld also noted that alternative techniques that do not require truncation are available to account for a healthy worker survivor effect (Document ID 2307, Attachment 2, pp. 27-28). OSHA believes such techniques, such as g-estimation, to be relatively new or not yet in standard use in occupational epidemiology. As discussed above, OSHA finds Mr. Park's approach in his study to be reasonable.

Finally, Dr. Cox stated in his comments that:

key studies relied on by OSHA, such as Park et al. (2002), do not correct for biases in reported ER [exposure-response] relations due to residual confounding by age (within age categories), i.e., the fact that older workers may tend to have both higher lung cancer risks and higher values of occupational exposure metrics, even if one does not cause the other. This can induce a non-causal association between the occupational exposure metrics and the risk of cancer (Document ID 2307, Attachment 4, p. 29).

Confounding occurs in an epidemiological study when the contribution of a causal factor cannot be separated from the effect of another variable (e.g., age) not accounted for in the analysis. Residual confounding occurs when attempts to control for confounding are not precise enough (e.g., controlling for age by using groups with age spans that are too wide), or subjects are misclassified with respect to confounders (Document ID 3607, p. 1). However, the Park et al. (2002) study of non-malignant respiratory disease mortality, which Dr. Cox cited as not Start Printed Page 16320considering residual confounding by age, actually addressed this issue by using 13 five-year age groups (<25, 25-29, 30-34, etc.) in the models (Document ID 0405, p. 37). Further discussion on residual confounding bias is found in Section V.J, Comments and Responses Concerning Biases in Key Studies.

The inclusion of Park et al. (2002) (Document ID 0405) in OSHA's risk assessment has additional support in the record. OSHA's expert peer-review panel supported including the Park et al. study in the risk assessment, with Gary Ginsberg, Ph.D., stating that it “represents a reasonable estimate of silica-induced total respiratory mortality” (Document ID 3574, p. 29). In addition, as OSHA noted in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 355-356), the Park et al. study is complemented by the Mannetje et al. multi-cohort silicosis mortality pooled study, which included several cohorts that had exposure concentrations in the range of interest for this rulemaking and also showed clear evidence of significant risk of silicosis and other NMRD at the previous general industry and construction PELs (2002b, Document ID 1089).

b. Mannetje et al. (2002b) and ToxaChemica (2004)

The ACC also submitted several comments on the Mannetje et al. (2002b) study of silicosis mortality; the data from Mannetje et al. were used in the ToxaChemica (2004) re-analysis. As noted above, the Mannetje et al. (2002b) study was a pooled analysis of silicosis mortality data from six epidemiological cohorts. This study showed a statistically significant association between silicosis mortality and workers' cumulative exposure, as well as with average exposure and exposure duration. The ACC's pre-hearing brief stated that the study “provided no justification for the relative rate model forms [Mannetje et al.] used to evaluate exposure-response” (Document ID 2307, Attachment A, p. 113). The concern expressed was that the study may not have considered all potential exposure-response relationships and was unable to discern differences between monotonic and non-monotonic characteristics (Document ID 2307, Attachment A, p. 113-114).

Mannetje et al. (2002b, Document ID 1089) did not discuss whether models other than relative rate models were tested. However, Mannetje's data was reexamined by ToxaChemica, Inc. on request from OSHA and the reexamined data was used by OSHA to help estimate lifetime risk for silicosis mortality (2004, Document ID 0469; 1711, pp. 310-314). The ToxaChemica reanalysis of the data included a categorical analysis and a five-knot restricted spline analysis, in addition to a logistic model, using the log of cumulative exposure (Document ID 0469, p. 50). ToxaChemica also corrected some errors found in the original data set and used a nested case-control approach, which they stated would control more precisely for age than the Poisson regression approach used by Mannetje et al. (Document ID 0469, p. 18). As shown in Figure 5 of ToxaChemica's report, the restricted spline model (which has considerable flexibility to represent non-monotonic features of exposure-response data) appeared to be monotonic, while the categorical analysis appeared largely monotonic but for one exposure group (Document ID 0469, p. 40, 50). When not adjusted for measurement error, the second highest exposure group deviated from the monotonic relationship existing between the other groups. However, the deviation was resolved when two sources of measurement error were accounted for (Document ID 0469, p. 40). The categorical analysis, restricted spline model, and logistic model yielded roughly similar exposure-response curves (Document ID 0469, p. 50). OSHA concludes that the ToxaChemica reanalysis addresses the concerns raised by the ACC by finding similar exposure-response relationships regardless of the model as well as providing greater validation of a monotonic curve.

The ACC next questioned the odds ratios generated in the Mannetje et al. (2002b) study (Document ID 2307, p. 114; 4209, p. 88). The Panel noted that “the exposure-response relationship is not even fully monotonic” and that the silica odds ratios in the pooled analysis have overlapping confidence intervals, suggesting no statistically significant difference (Document ID 2307, p. 114). The Panel concluded that “the data indicate that there is no clear effect of exposure on odds ratios over the entire range considered by the authors; hence, the study provides no basis for concluding that reducing exposures will reduce the odds ratio for silicosis mortality” (Document ID 4209, p. 88). Essentially, the ACC argued that the data do not appear to fit a monotonic relationship and that the confidence intervals for each exposure level overlap too much to discern any differences in risk ratios between those exposures.

OSHA believes that the ACC overstated its contention about confidence interval overlap between groups in the Mannetje et al. (2002b) paper. Although the original data set reported in the study lacks a monotonic relationship on the upper end of the exposure spectrum (>9.58 mg/m3-yrs) (possibly due to a healthy worker survivor effect, as explained above), OSHA notes that the 95 percent confidence intervals reported do not contradict the presence of a monotonic relationship (Document ID 1089). First, the confidence intervals of the lower exposed groups did not overlap with those of the higher exposed groups in that study (Document ID 1089). Second, even if they did, overlap in confidence intervals does not mean that there is not a significant difference between those groups. While it is true that, if 95 percent confidence intervals do not overlap, the represented populations are statistically significantly different, the converse—that, if confidence intervals do overlap, there is no statistically significant difference—is not always true (Nathaniel Schenker and Jane F. Gentleman. “On Judging the Significance of Differences by Examining the Overlap Between Confidence Intervals.” The American Statistician. 55(3): 2001. 182-186. (http://www.tandfonline.com/​doi/​abs/​10.1198/​000313001317097960).

Finally, as discussed above and in detail in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, the ToxaChemica et al. (2004) re-analysis of the corrected Mannetje et al. (2002b) data adjusting for two sources of measurement error resulted in a monotonic relationship for the risk ratios (Document ID 0469).

2. Silicosis Morbidity

OSHA relied on five studies for determining risk for silicosis morbidity: Buchanan et al., 2003 (Document ID 0306), Chen et al., 2001 (Document ID 0332), Chen et al., 2005 (Document ID 0985), Hnizdo and Sluis-Cremer, 1993 (Document ID 1052), and Steenland and Brown, 1995b (Document ID 0451). OSHA finds that the most reliable estimates of silicosis morbidity, as detected by chest radiographs, come from these five studies because they evaluated radiographs over time, included post-employment radiographic evaluations, and derived cumulative or lifetime estimates of silicosis disease risk. OSHA received several comments about these studies.

a. Buchanan et al. (2003)

Buchanan et al. (2003) reported on a cohort of Scottish coal workers (Document ID 0306). The authors found a statistically significant relationship Start Printed Page 16321between silicosis and cumulative exposure acquired after 1964 (Document ID 0306). They also found that the risks of silicosis over a working lifetime can rise dramatically with exposure to high concentrations over a timescale of merely a few months (Document ID 0306). In the Preliminary QRA, OSHA considered this study to be of the highest overall quality of the studies relied upon to assess silicosis morbidity risks, in large measure because the underlying exposure data was based on modern exposure measurement methods and avoided the need to estimate historical exposures. The risk estimates derived from this study were lower than those derived from any of the other studies criticized by the ACC. One reason for this is because Buchanan et al. only included cases with chest x-ray findings having an ILO score of 2/1 or higher, whereas the other studies included cases with less damage, having a lower degree of perfusion on x-ray (ILO 1/0 or 1/1) (Document ID 0306). Thus, OSHA considered the risk estimates derived from the Buchanan et al. study to be more likely to understate risks.

Dr. Cox commented that age needed to be included for modeling in Dr. Miller's 1998 paper, the data from which were used in the Buchanan et al. (2003) paper (Document ID 2307, Attachment 4, p. 97). However, the Miller et al. (1998) study explicitly states that age was one of several variables that were tried in the model but did not improve the model's fit, as was time spent working in the poorly characterized conditions before 1954 (Document ID 0374, p. 57). OSHA concludes that the original paper did assess these variables and how they related to the exposure-response relationship. Buchanan et al. (2003) also noted their own finding that differences in age and exposure both failed to improve fit, in agreement with Miller et al.'s conclusion (Document ID 0306, p. 161). OSHA therefore finds no credible reason that age should have been included as a variable in Miller et al. (1998).

Dr. Cox also questioned the modeling methods in the Buchanan paper, which presented logistic regression in progressive stages to search for significance (Document ID 2307, Attachment 4, pp. 97-98; 0306, pp. 161-163). Dr. Cox claimed that this is an example of uncorrected multiple testing bias where the post hoc selection of data, variables, and models can make independent variables appear to be statistically significant in the prediction model. He suggested that corrections for bias are needed to determine if the reported significance is causal or statistical (Document ID 2307, Attachment 4, pp. 97-98). OSHA peer reviewer Brian Miller, Ph.D., stated that Dr. Cox's claim that the model was affected by multiple testing bias is unfounded (Document ID 3574, pp. 31-32). He noted that the model was based on a detailed knowledge of the history of exposures at that colliery, and represented the researchers' attempt to build “a reality-driven and `best-fitting' model,” (Document ID 3574, p. 31, quoting 2307, Attachment 4, p. 4). Furthermore, none of OSHA's peer reviewers raised any concerns about the approach taken by Buchanan et al. to develop their exposure-response model and none suggested that corrections needed to be made for multiple testing bias; all of them supported the study's inclusion in OSHA's risk assessment (Document ID 3574). Finally, the cumulative risk for silicosis morbidity derived from this study is similar to values from other papers reported in the QRA (see OSHA's Final Quantitative Risk Assessment in Section VI). Therefore, for the reasons discussed above, OSHA is not convinced by Dr. Cox's arguments and finds no credible reason to remove Buchanan et al. (2003) from consideration.

b. Chen et al. (2001, 2005), Steenland and Brown (1995), and Hnizdo and Sluis-Cremer (1993)

The ACC also commented on several other studies used by OSHA to estimate silicosis morbidity risks; these were the studies by Chen et al. (2001, Document ID 0332; 2005, 0985), Steenland and Brown (1995b, Document ID 0451), and Hnizdo and Sluis-Cremer (1993, Document ID 1052). The ACC's comments focus on uncertainties in estimating the historical exposures of cohort members (Document ID 2307, Attachment A, pp. 117-122, 124-130, 132-136). Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, discusses the record in detail with respect to the general issue of uncertainties in estimating historical exposures to respirable crystalline silica in epidemiological studies. The issues specific to the studies relied upon by OSHA in its risk estimates for silicosis morbidity will be discussed below.

In the Chen et al. studies, which focused on mining (i.e., tin, tungsten) and pottery cohorts, high volume area samplers collected dust and the respirable crystalline silica concentration was determined from those samples (2001, Document ID 0332; 2005, 0985). However, according to the ACC, the rest of the collected dust was not assessed for chemicals that potentially could also cause radiographic opacities (Document ID 2307, Attachment A, pp. 132-135). Neither study expressed reason to be concerned about the non-silica portion of the dust samples. OSHA recognizes that uncertainty about potential unknown exposures exists in retrospective studies, which describes most epidemiological research. However, OSHA emphasizes that the risk values derived from the Chen et al. studies do not differ remarkably from other silicosis morbidity studies used in the risk assessment (Document ID 0306, 1052, 0451). Therefore, OSHA concludes that it is unlikely that an unknown compound significantly impacted the exposure-response relationships reported in both Chen studies.

The study on gold miners (Steenland and Brown, 1995b, Document ID 0451), which found that cumulative exposure was the best disease predictor, followed by duration of exposure and average exposure, was also criticized by the ACC, which alleged that the exposure assessment suffered from “enormous uncertainty” (Document ID 2307, Attachment A, pp. 146-147). The ACC noted that exposure measurements were not available for the years prior to 1937 or after 1975 and that this limitation of the exposure information may have resulted in an underestimation of exposures (Document ID 2307, Attachment A, pp. 124-126). OSHA agrees that these are potential sources of uncertainty in the exposure estimates, but recognizes exposure uncertainty to be a common occurrence in occupational epidemiology studies. OSHA believes that the authors used the best measurement data available to them in their study.

The ACC also took issue with Steenland and Brown's conversion factor for converting particle count to respirable silica mass (10 mppcf = 100 μg/m3), which was somewhat higher than that used in the Vermont granite worker studies (10 mppcf = 75 μg/m3) (Document ID 2307, Attachment A, p. 126). OSHA notes that the study's reasoning for adopting that specific particle count conversion factor was to address the higher percentage of silica found in the gold mine samples applicable to their cohort in comparison to the Vermont granite study (Document ID 0451, p. 1373). OSHA finds this decision, which was based on the specific known exposure conditions of this cohort, to be reasonable.

With respect to the Hnizdo and Sluis-Cremer (1993, Document ID 1052) Start Printed Page 16322study, which found that silicosis risk increased exponentially with cumulative exposure to respirable dust (Document ID 1052, p. 447), the ACC questioned three assumptions the study made about exposures. First, exposures were assumed to be static from the 1930s to the 1960s, based on measurements from the late 1950s to mid-1960s, an assumption that, according to the ACC, might underestimate exposure for workers employed before the late 1950s (Document ID 2307, Attachment A, pp. 117-119). Second, although respirable dust, by definition, includes particles up to 10 μm, the study only considered particles sized between 0.5 and 5 μm in diameter (Document ID 1052, p. 449). The ACC contends this exclusion may have resulted in underestimated exposure and overestimated risk (Document ID 2307, Attachment A, p. 119). OSHA agrees that uncertainty in exposure estimates is an important issue in the silica risk assessment, and generally discusses the issue of exposure measurement uncertainty in depth in a quantitative uncertainty analysis described in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis. As discussed there, after accounting for the likely effects of exposure measurement uncertainty in the risk assessment, OSHA affirms the conclusion of the risk assessment that there is significant risk of silicosis to workers exposed at the previous PELs.

Thirdly, the ACC challenged the authors' estimate of the quartz content of the dust as 30 percent when it should have been 54 percent (Document ID 1052, p. 450; 2307, Attachment A, p. 120). According to the ACC, the 30 percent estimate was based on an incorrect assumption that the samples had been acid-washed (resulting in a reduction in silica content) before the quartz content was measured (Document ID 2307, Attachment A, pp. 120-122). This assumption would greatly underestimate the exposures of the cohort and the exposures needed to cause adverse effects, thus overestimating actual risk (Document ID 2307, Attachment A, pp. 121-122). The ACC recommended that the quartz content in the Hnizdo and Sluis-Cremer study be increased from 30 to 54 percent, based on the Gibbs and Du Toit study (2002, Document ID 1025, p. 602).

OSHA considered this issue in the Preliminary QRA (Document ID 1711, p. 332). OSHA noted that the California Environmental Protection Agency's Office of Environmental Health Hazard Assessment reviewed the source data for Hnizdo and Sluis-Cremer, located in the Page-Shipp and Harris (1972, Document ID 0583) study, and compared them to the quartz exposures calculated by Hnizdo and Sluis-Cremer (OEHHA, 2005, Document ID 1322, p. 29). OEHHA concluded after analyzing the data that the samples likely were not acid-washed and that the Hnizdo and Sluis-Cremer paper erred in describing that aspect of the samples. Additionally, OEHHA reported data that suggests that the 30 percent quartz concentration may actually overestimate the exposure. It noted that recent investigations found the quartz content of respirable dust in South African gold mines to be less than 30 percent (Document ID 1322). In summary, OSHA concludes that no meaningful evidence was submitted to the rulemaking record that changes OSHA's original decision to include the Hnizdo and Sluis-Cremer study in its risk assessment.

Despite the uncertainties inherent in estimating the exposures of occupational cohorts in silicosis morbidity studies, the resulting estimates of risk for the previous general industry PEL of 100 μg/m3 are in reasonable agreement and indicate that lifetime risks of silicosis morbidity at this level, and, by extension, risks at the higher previous PELs for maritime and construction (see section VI, Final Quantitative Risk Assessment and Significance of Risk) are in the range of hundreds of cases per 1,000 workers. Even in the unlikely event that exposure estimates underlying all of these studies were systematically understated by several fold, the magnitude of resulting risks would likely still be such that OSHA would determine them to be significant.

3. Conclusion

After carefully considering all of the comments on the studies relied on by OSHA to estimate silicosis and NMRD mortality and silicosis morbidity risks, OSHA concludes that the scientific evidence used in its quantitative risk assessment substantially supports the Agency's finding of significant risk for silicosis and non-malignant respiratory disease. In its risk estimates in the Preliminary QRA, OSHA acknowledged the uncertainties raised by the ACC and other commenters, but the Agency nevertheless concluded that the assessment was sufficient for evaluating the significance of the risk. After evaluating the evidence in the record on this topic, OSHA continues to conclude that its risk assessment (see Final Quantitative Risk Assessment in Section VI.C of this preamble) provides a reasonable and well-supported estimate of the risk faced by workers who are exposed to respirable crystalline silica.

E. Comments and Responses Concerning Surveillance Data on Silicosis Morbidity and Mortality

As discussed above in this preamble, OSHA has relied on epidemiological studies to assess the risk of silicosis, a debilitating and potentially fatal occupationally-related lung disease caused by exposure to respirable crystalline silica. In the proposed rule (78 FR 56273, 56298; also Document ID 1711, pp. 31-49), OSHA also discussed data from silicosis surveillance programs that provide some information about the number of silicosis-associated deaths or the extent of silicosis morbidity in the U.S. (78 FR at 56298). However, as OSHA explained, the surveillance data are not sufficient for estimating the risks of health effects associated with exposure to silica, nor are they sufficient for estimating the benefits of any potential regulatory action. This is because silicosis-related surveillance data are only available from a few states and do not provide exposure data that can be matched to surveillance data. Consequently, there is no way of knowing how much silica a person was exposed to before developing fatal silicosis (78 FR at 56298).

In addition, the available data likely understate the resulting death and disease rates in U.S. workers exposed to crystalline silica (78 FR 56298). This understatement is due in large part to: (1) The passive nature of these surveillance systems, which rely on healthcare providers' awareness of a reporting requirement and submission of the appropriate information on standardized forms to health departments; (2) the long latency period of silicosis; (3) incomplete occupational exposure histories, and (4) other factors that result in a lack of recognition of silicosis by healthcare providers, including the low sensitivity, or ability of chest x-rays to identify cases of silicosis (78 FR 56298). Specific to death certificate data, information on usual industry and occupation are available from only 26 states for the period 1985 to 1999, and those codes are not verifiable (Document ID 1711). Added to these limitations is the “lagging” nature of surveillance data; it often takes years for cases to be reported, confirmed, and recorded. Furthermore, in many cases, the available surveillance systems lack information about actual exposures or even information about the usual occupation or industry of the deceased individual, which could provide some information about occupational Start Printed Page 16323exposure (see 78 FR at 56298). Therefore, the Agency did not use these surveillance data to estimate the risk of silicosis for the purpose of meeting its legal requirements to prove a significant risk of material impairment of health (see 29 U.S.C. 655(b)(5); Benzene, 448 U.S. 607, 642 (1980)).

Comments and testimony focusing on the silicosis surveillance data alleged that OSHA should have used the surveillance data in its risk estimates. Stakeholders argued that the declining numbers of reported silicosis deaths prove the lack of necessity for a new silica standard. Commenters also claimed that the surveillance data prove that OSHA overestimated both the risks at the former permissible exposure limits (PELs) and the benefits of the new rule.

After reviewing the rulemaking record, OSHA maintains its view that these silicosis surveillance data, although useful for providing context and an illustration of a significant general trend in the reduction of deaths associated with silicosis over the past 4-5 decades, are not sufficient for estimating the magnitude of the risk or the expected benefits. In the case of silicosis, surveillance data are useful for describing general trends nationally and a few states have the ability to use the data at the local or state level to identify “sentinel events” that would justify initiating an inspection of a workplace, for example. The overall data, however, are inadequate and inappropriate for estimating risks or benefits associated with various exposure levels, as is required of OSHA's regulatory process, in part because they significantly understate the extent of silicosis in workers in the United States and because they lack information about exposure levels, exposure sources (e.g., type of job), controls, and health effects that is necessary to examine the effects of lowering the PEL. Thus, for these reasons and the ones discussed below, OSHA has continued to rely on epidemiological data to meet its burden of demonstrating that workers exposed to respirable crystalline silica at the previous PELs face a significant risk of developing silicosis and that risk will be reduced when the new limit is fully implemented. Another related concern identified by stakeholders is the apparent inconsistency between surveillance data and risk and benefits estimates derived from modeling epidemiological data (Document ID 4194, pp. 7-10; 4209, pp. 3-4). However, this difference is not an inconsistency, but the result of comparing two distinctly different items. Surveillance data, primarily death certificate data, are known to be under-reported and lack associated exposure data necessary to model relationships between various exposure levels and observance of health effects. For these reasons, OSHA relied on epidemiologic studies with detailed exposure-response relationships to evaluate the significance of risk at the preceding and new PELs. Thus, the silicosis mortality data derived from death certificates and estimates of silica-related mortality risks derived from well-conducted epidemiologic studies cannot be directly compared in any meaningful way. With respect to silicosis morbidity, OSHA notes that the estimates by Rosenman et al. (2003, Document ID 0420) of the number of cases of silicosis estimated to occur in the U.S. (between 2,700 and 5,475 estimated to be in OSHA's jurisdiction (i.e., excluding miners)) each year is in reasonable agreement with the estimates derived from epidemiologic studies, assuming either a 13-year or 45-year working life (see Chapter VII, Table VII-2 of the FEA).

1. Surveillance Data on Silicosis Mortality

The principal source of data on annual silicosis mortality in the U.S. is the National Institute for Occupational Safety and Health (NIOSH) Work-Related Lung Disease (WoRLD) Surveillance System (e.g., NIOSH, 2008c, Document ID 1308), which compiles cause-of-death data from death certificates reported to state vital statistics offices and collected by the National Center for Health Statistics (NCHS). Paper copies were published in 2003 and 2008 (Document ID 1307; 1308) and data are updated periodically in the electronic version on the CDC Web site (http://www.cdc.gov/​eworld). NIOSH also developed and manages the National Occupational Respiratory Mortality System (NORMS), a data-storage and interactive data retrieval system that reflects death certificate data compiled by NCHS (http://webappa.cdc.gov/​ords/​norms.html).

From 1968 to 2002, silicosis was recorded as an underlying or contributing cause of death on 16,305 death certificates; of these, a total of 15,944 (98 percent) deaths occurred in males (CDC, 2005, Document ID 0319). Over time, silicosis-related mortality has declined in the U.S., but has not been eliminated. Based on the death certificate data, the number of recognized and coded deaths for which silicosis was an underlying or contributing cause decreased from 1,157 in 1968 to 161 in 2005, corresponding to an 86-percent decline (Document ID 1711, p. 33; 1308, p. 55) (http://wwwn.cdc.gov/​eworld). The crude mortality rate, expressed as the number of silicosis deaths per 1,000,000 general population (age 15 and higher) fell from about 8.9 per million to about 0.5 per million over that same time frame, a decline of 94 percent (Document ID 1711, p. 33; 1308, p. 55) (http://wwwn.cdc.gov/​eworld).

OSHA's Review of Health Effects Literature and Preliminary QRA included death certificate statistics for silicosis up to and including 2005 (Document ID 1711, p. 33). OSHA has since reviewed the more recent NORMS and NCHS data, up to and including 2013, which appear to show a general downward trend in mortality, as presented in Table V-1.

Start Printed Page 16324

However, more detailed examination of the most recent data collected through NCHS (Table V-2) indicates that the decline in the number of deaths with silicosis as an underlying or contributing cause has leveled off in more recent years, suggesting that the number of silicosis deaths being recorded and captured by death certificates may be stabilizing after 30 or more years of decline.

Robert Cohen, M.D., representing the American Thoracic Society, noted this apparent plateau effect, testifying that “[t]he data from the NIOSH work-related lung disease surveillance report and others show a plateau in silicosis Start Printed Page 16325mortality since the 1990s, and we are concerned that that has been the same without any further reduction for more than 20 years. So we think that we still have work to do” (Document ID 3577, p. 775).

Some commenters raised the question about whether decedents who died more recently were exposed to high levels of silica (pre-1970s) and therefore wouldn't necessarily reflect mortalities relevant to the current OSHA standard (Document ID 4194, p. 9; 4209, pp. 7-8). OSHA has no information on the age of these decedents, or the timing of their exposure to silica. If we assume that workers born in 1940-1950 would have started working around 1960, at the earliest, and into the 1970's, and life expectancy in general of 70 years, or 60-70 years to account for years of life lost due to silicosis, most of these workers' working life would have been spent after the 1971 PEL went into effect. It is likely that some of the more recent decedents were exposed to silica prior to 1971; however, it is less likely that all were exposed prior to 1971. At the end of the day, there is no actual exposure information on these decedents, and this generalization does not account for overexposures, which have persisted over time.

2. Surveillance Data on Silicosis Morbidity

There is no nation-wide system for collecting silicosis morbidity case data. The data available are from three sources: (1) The National Hospital Discharge Survey (Document ID 1711, p. 40-43); (2) the Agency for Healthcare Research and Quality's (AHRQ) Nationwide Inpatient Survey (Document ID 3425, p. 2; https://www.hcup-us.ahrq.gov/​nisoverview.jsp); and (3) states that administer silicosis and/or pneumoconiosis disease surveillance (see Document ID 1711, p. 40-43; http://www.cdc.gov/​niosh/​topics/​surveillance/​ords/​StateBasedSurveillance/​stateprograms.html).

Both of the first two sources of data on silicosis morbidity cases are surveys that provide estimates of hospital discharges. The first is the National Hospital Discharge Survey (NHDS), which was conducted annually from 1965-2010. The NHDS was a national probability survey designed to meet the need for information on characteristics of inpatients discharged from non-Federal short-stay hospitals in the United States (see http://www.cdc.gov/​nchs/​nhds.htm). Estimates of silicosis listed as a diagnosis on hospital discharge records are available from the NHDS for the years 1985 to 2010 (see http://www.cdc.gov/​nchs/​nhds.htm). National estimates were rounded to the nearest 1,000, and the NHDS has consistently reported approximately 1,000 discharges/hospitalizations annually since 1980 (e.g., Document ID 1307; 1308). The second survey, the National (Nationwide) Inpatient Sample (NIS), is conducted annually by the AHRQ. Dr. Kenneth Rosenman, Division Chief and Professor of Medicine at Michigan State University and who oversees one of the few occupational disease surveillance systems in the U.S., testified that data from the NIS indicated that the nationwide number of hospitalizations where silicosis was one of the discharge diagnoses has remained constant, with 2,028 hospitalizations reported in 1993 and 2,082 in 2011 (Document ID 3425, p. 2).

Morbidity data are also available from the states that administer silicosis and/or pneumoconiosis disease surveillance. These programs rely primarily on hospital discharge records and also may get some reports of cases from the medical community and workers' compensation programs. Currently, NIOSH funds the State-Based Occupational Safety and Health Surveillance cooperative agreements (Document ID 1711, p. 40-41; http://www.cdc.gov/​niosh/​topics/​surveillance/​ords/​StateBasedSurveillance.html). All states funded under a cooperative agreement conduct population-based surveillance for pneumoconiosis (hospitalizations and mortality), and a few states (currently Michigan and New Jersey) have expanded surveillance specifically for silicosis (Document ID 1711, p. 40-42; http://www.cdc.gov/​niosh/​topics/​surveillance/​ords/​StateBasedSurveillance/​stateprograms.html).

State-based hospital discharge data are a useful population-based surveillance data source for quantifying pneumoconiosis (including silicosis), even though only a small number of individuals with pneumoconiosis are hospitalized for that condition (Document ID 0996), and the data refer to hospitalizations with a diagnosis of silicosis, and not specific people. In addition to mortality data, NIOSH has updated its WoRLD Surveillance System with some state-based morbidity case data (http://wwwn.cdc.gov/​eworld/​Grouping/​Silicosis/​94). State-based surveillance systems can provide more detailed information on a few cases of silicosis.

NIOSH has published aggregated state case data in its WoRLD Reports (Document ID 1308; 1307) for two ten-year periods that overlap, 1989 to 1998 and 1993 to 2002. State morbidity case data are compiled and evaluated by variables such as ascertainment source, primary industry, and occupations. For the time period 1989 to 1998, Michigan reported 589 cases of silicosis, New Jersey 191 cases, and Ohio 400 cases (Document ID 1307, p. 69). In its last published report, for the later and partially overlapping time period 1993 to 2002, Michigan reported 465 cases, New Jersey 135, and Ohio 279 (Document ID 1308, p. 72). Data for the years 2003 to 2011, from the CDC/NIOSH electronic report, eWoRld, show a modest decline in the number of cases of silicosis in these three states; however, decreases are not nearly as substantial as are those seen in the mortality rates (see Table V-3). Annual averages for the two ten-year periods and the nine-year time period were calculated by OSHA solely for the purpose of comparing cases of silicosis reported over time.

Start Printed Page 16326

3. Critical Comments Received on Surveillance Data

Industry representatives, including ACC's Crystalline Silica Panel and Dr. Jonathan Borak, representing the Chamber of Commerce (Chamber), contended that the steep decline seen in the number and rate of silicosis deaths since 1968 proves that OSHA cannot meet its burden of demonstrating that a more protective standard is necessary (e.g., Document ID 4209, p. 10; 2376, p. 8; 4016, p. 9). Similarly, other commenters, such as the American Petroleum Institute, the Independent Petroleum Association of America, the National Mining Association, the American Foundry Society (AFS), the National Utility & Excavating Contractors Association, Acme Brick, the National Ready Mixed Concrete Association, and the Small Business Administration's Office of Advocacy stated that surveillance data demonstrate that the previous OSHA PEL was sufficiently effective in reducing the number of deaths from silicosis (Document ID 3589, Tr. 4041; 4122; 2301, pp. 3, 7-9; 2211, p. 2; 2379, pp. 23-25; 2171, p. 1; 3730, p. 5; 3586, Tr. 3358-3360; 3589, Tr. 4311; 2349, pp. 3-4). Industry commenters also argued that the number of recorded silicosis-related deaths in recent years, as reflected in the surveillance data, is far lower than the number of lives that OSHA projected would be saved by a more stringent rule, indicating that OSHA's risk assessment is flawed (e.g., Document ID 3578, Tr. 1074-1075; 4209, p. 3-4).

The Chamber, along with others, declared that OSHA ignored steep declines in silicosis mortality, which in its view indicates that there is no further need to reduce the PEL (Document ID 4194, pp. 7-8). OSHA has not ignored the fact that the available surveillance data indicate a decline in silicosis mortality. As discussed above and in the proposal, the Agency has acknowledged that the available surveillance data do show a decline in the silicosis mortality since 1968. Furthermore, OSHA has no information on whether underreporting has increased or decreased over time, and does not believe that differing rates of reporting and underreporting of silicosis on death certificates explains the observed decline in silicosis mortality. OSHA believes that the reductions in deaths attributable to silicosis are real, and not a statistical artifact. However, OSHA disagrees with commenters' argument that this trend shows the lack of a need for this new rule. First, as explained above, there is strong evidence that the death certificate data do not capture the entirety of silicosis mortality that actually exists, due to underreporting of silicosis as a cause of death. Second, the stakeholders' argument assumes that mortality will continue to decline, even in the absence of a stronger silica standard, and that OSHA and workers should wait for this decline to hit bottom (e.g., Document ID 4209, p. 7). However, testimony in the record suggests that the decline in the number of deaths has leveled off since 2000, probably because of the deaths of those historically exposed to higher levels of silica occurred before then (e.g., Document ID 3577, p. 775).

Third, the decline in silicosis deaths recorded over the past several decades cannot be solely explained by improved working conditions, but also reflects the decline in employment in industries that historically were associated with high workplace exposures to crystalline silica. One of OSHA's peer reviewers for the Review of Health Effects Literature and Preliminary QRA, Bruce Allen, commented that the observed decline in mortality “. . . in no way adjusts for the declining employment in jobs with silica exposure,” making “its interpretation problematic. To emphasize the contribution of historic declines in exposure as the underlying cause is spurious; no information is given to allow one to account for declining employment” (Document ID 3574, p. 7). The CDC/NIOSH also identified declining employment in heavy industries where silica exposure was prevalent as a “major factor” in the reduction over time in silicosis mortality (Document ID 0319, p. 2). As discussed below, however, some silica-generating operations or industries are new or becoming more prevalent.

In his written testimony, Dr. Rosenman pointed out that there are “two aspects to the frequency of occurrence of disease (1) . . . the risk of disease based on the level of exposure and (2) the number of individuals at risk” (Document ID 3425, pp. 3-4). Dr. Rosenman estimated the decline in the number of workers in Michigan foundries (75 percent) and the number of abrasive blasting companies in Michigan (71 percent), and then compared these percentages to the percentage decline in the number of recorded silicosis deaths (80 percent) over a similar time period. The similarities in these values led him to attribute “almost all” of the decrease in silicosis deaths to a decrease in the population at risk (Document ID 3425, pp. 3-4).

Finally, OSHA's reliance on epidemiological data for its risk assessment purposes does not suggest that the Agency ignored the available surveillance data. As discussed above, the data are inadequate and inappropriate for estimating risks or benefits associated with various exposure levels, as is required of OSHA's regulatory process. Even in the limited cases where surveillance data are available, OSHA generally relies on epidemiological data, to the extent they include sufficiently detailed information on exposures, exposure sources (e.g., type of job), and health effects, to satisfy its statutory requirement to use the best available evidence to evaluate the significance of risk associated with exposure to hazardous substances.

Some stakeholders provided comments to the rulemaking record consistent with OSHA's assessment. For example, Dr. Borak stated that the surveillance data “provide little or no basis” (Document ID 2376, p. 8) for OSHA to evaluate the protectiveness of the previous PELs. Similarly, NIOSH asserted that relying on the surveillance data to show that there is no need for a lower PEL or that there is no significant risk at 100 μg/m3 would be “a misuse of surveillance data” (Document ID 3579, Tr. 167). NIOSH also added that, because the surveillance data do not include information about exposures, it is not the kind of data that could be used for a quantitative risk assessment. NIOSH concluded that surveillance data are, in fact, “really not germane to the risk assessment” (Document ID 3579, Tr. 248). OSHA agrees with both Dr. Borak and NIOSH that the surveillance data cannot and do not inform the Agency on the need for a lower PEL, nor is there a role for surveillance data in making its significant risk findings. Therefore, for its findings of significant risk at the current PEL, the Agency relied on evidence derived from detailed exposure-response relationships from well-conducted epidemiologic studies, and not surveillance data, which have no associated exposure information. In this case, epidemiologic data provided the best available evidence.

In its testimony, the AFL-CIO pointed out that a recent U.S. Government Accountability Office (GAO) report on the Mine Safety and Health Administration's (MSHA) proposed coal dust standard references the National Academy of Sciences (NAS) conclusion that risk assessments based on epidemiological data, not surveillance data, were an appropriate means to assess risk for coal-dust exposures (Document ID 4204, p. 21; 4072, Attachment 48, pp. 7-8). The NAS Start Printed Page 16327emphasized that the surveillance data available to MSHA did not include individual miners' levels of exposure to coal mine dust and, therefore, could not be used for the purpose of estimating disease risk for miners. “Based on principles of epidemiology and statistical modeling, measures of past exposures to coal mine dust are critical to assessing the relationship between miners' cumulative coal mine dust exposure and their risk of developing [pneumoconiosis]” (Document ID 4072, Attachment 48, p. 8). The same rationale applies here. Thus, OSHA's decision to rely on epidemiological data is well supported by the record.

Commenters from companies and industry groups also argued that they had no knowledge of workers acquiring silicosis in their companies or industry (e.g., Document ID 2384, p. 2; 2338, p. 3; 2365, p. 2; 2185, p. 3; 2426, p. 1). OSHA received similar comments as part of a letter campaign in which over 100 letters from brick industry representatives claimed there to be little or no silicosis observed in the industry despite historical exposures above the PEL (e.g., Document ID 2009). OSHA considered these comments and believes that many companies, including companies in the brick industry, may not have active medical surveillance programs for silicosis. Silicosis may not develop until after retirement as a result of its long latency period. In addition, silica exposures in some workplaces may be well below the final PEL as a result of the environment in which workers operate, including existing controls. Thus, OSHA believes that it is difficult to draw conclusions about the rate of silicosis morbidity in specific workplaces without having detailed information on medical surveillance, silica exposures, and follow-up. This is why OSHA relies heavily on epidemiological studies with detailed exposure data and extended follow-up, and uses these data to evaluate exposure-response relationships to assess health risks at the preceding and new PELs.

Commenters also argued that, due to the long latency of the disease, silicosis cases diagnosed today are the result of exposures that occurred before the former PELs were adopted, and thus reflect exposures considerably higher than the previous PELs (e.g., Document ID 2376, p. 3; 2307, p. 12; 4194, p. 9; 3582, Tr. 1935). OSHA notes that the evidence shows that the declining trend in silicosis mortality does not provide a complete picture with regard to silicosis trends in the United States. Although many silicosis deaths reported today are likely the result of higher exposures (both magnitude and duration), some of which may have occurred before OSHA adopted the previous PELs, silicosis cases continue to occur today—some in occupations and industries where exposures are new and/or increasing. For example, five states reported cases of silicosis in dental technicians for the years 1994 to 2000 (CDC, MMWR Weekly, 2004, 53(09), pp. 195-197), for the first time. For the patients described in this report, the only identified source of crystalline silica exposure was their work as dental technicians. Exposure to respirable crystalline silica in dental laboratories can occur during procedures that generate airborne dust (e.g., mixing powders, removing castings from molds, grinding and polishing castings and porcelain, and using silica sand for abrasive blasting). In 2015, the CDC reported the first case of silicosis (progressive massive fibrosis) associated with exposure to quartz surfacing materials (countertop fabrication and installation) in the U.S. The patient was exposed to dust for 10 years from working with conglomerate or quartz surfacing materials containing 70%-90% crystalline silica. Cases had previously been reported in Israel, Italy and Spain (MMWR, 2015, 64(05); 129-130). Recently, hazardous silica exposures have been newly documented during hydraulic fracturing of gas and oil wells (Bang et al., MMWR, 2015, 64(05); 117-120).

Dr. Rosenman's testimony provides support for this point. He testified that newer industries with high silica exposures may also be under-recognized because workers in those industries have not yet begun to be diagnosed with silicosis due to the latency period (Document ID 3577, p. 858). Dr. Rosenman submitted to the record a study by Valiante et al. (2004, Document ID 3926) that identified newly exposed construction workers in the growing industry of roadway repair, which began using current methods for repair in the 1980s. These methods use quick-setting concrete that generates dust containing silica above the OSHA PEL when workers perform jackhammering, and sawing and milling concrete operations. State surveillance systems identified 576 confirmed silicosis cases in New Jersey, Michigan, and Ohio that were reported to NIOSH for the years 1993 through 1997. Of these, 45 (8 percent) cases were in construction workers, three of which had been engaged in highway repair.

Sample results for this study indicated a significant risk of overexposure to crystalline silica for workers who performed the five highway repair tasks involving concrete. Sample results in excess of the OSHA PEL were found for operating a jackhammer (88 percent of samples), sawing concrete and milling concrete tasks (100 percent of samples); cleaning up concrete tasks (67 percent of samples); and drilling dowels (100 percent of samples). No measured exposures in excess of the PEL were found for milling asphalt and cleaning up asphalt; however, of the eight samples collected for milling asphalt, six (55 percent) results approached the OSHA PEL, and one was at 92 percent of the PEL. No dust-control measures were in place during the sampling of these highway repair operations.

The authors pointed out that surveillance systems such as those implemented by these states are limited in their ability to detect diseases with long latencies in highway repair working populations because of the relatively short period of time that modern repair methods had been in use when the study was conducted. Nevertheless, a few cases were identified, although the authors explain that the work histories of these cases were incomplete, and the authors recommended ongoing research to evaluate the silicosis disease potential among this growing worker population (Document ID 3926, pp. 876-880). In construction, use of equipment such as blades used on handheld saws to dry-cut masonry materials have increased both efficiency and silica exposures for workers over the past few decades (Document ID 4223, p. 11-13). Exposure data collected by OSHA as part of its technological feasibility analysis demonstrates that exposures frequently exceed previous exposure limits for these operations when no dust controls are used (see Chapter IV of the FEA). Another operation seeing new and increasing exposures to respirable crystalline silica is hydraulic fracturing in the oil and gas industry (Document ID 3588, p. 3773). Information in the record from medical professionals noted that lung diseases caused by silica exposures are “not relics of the past,” and that they continue to see cases of silicosis and other related diseases, even among younger workers who entered the workforce after the former PEL was enacted (see Document ID 3577, Tr. 773).

Furthermore, the general declining trend seen in the death certificate data is considerably more modest in silicosis morbidity data. In his written testimony, Dr. Rosenman stated that the nationwide number of hospitalizations where silicosis was one of the discharge diagnoses has remained constant, with 2,028 hospitalizations reported in 1993 Start Printed Page 16328and 2,082 in 2011 (Document ID 3425, p. 2). It is the opinion of medical professionals including the American Thoracic Society and the American College of Chest Physicians that these hospitalizations likely represent “the tip of the iceberg” (of silicosis cases) since milder cases are not likely to be admitted to the hospital (Document ID 2175, p. 3). Again, this evidence shows that the declining trend observed in silicosis mortality statistics does not provide a complete picture with regard to silicosis trends in the United States. While silicosis mortality has decreased substantially since records were first available in 1968, the number of silicosis related deaths appears to have leveled off (see Table V-2; Document ID 3577, Tr. 775). Workers are still dying from silicosis today, and new cases are being identified by surveillance systems, where they exist.

Based on the testimony and evidence described above, OSHA finds that the surveillance data describing trends in silicosis mortality and morbidity provide useful evidence of a continuing problem, but are not suitable for evaluating either the adequacy of the previous PELs or whether a more protective standard is needed. In fact, it would not be possible to derive estimates of risk at various exposure levels from the available surveillance data for silica. OSHA therefore appropriately continues to rely on epidemiological data and its quantitative risk assessment to support the need to reduce the previous PELs in its final rule.

Commenters also argued that OSHA has failed to prove that a new standard is necessary because silica-associated deaths are due to existing exposures in excess of the previous PELs; therefore, the Agency should focus on better enforcing the previous PELs, rather than enacting a new standard (e.g., Document ID 2376, p. 8; 2307, p. 12; 4016, pp. 9-10; 3582, Tr. 1936). OSHA does not find this argument persuasive. First, many of the commenters used OSHA's targeted enforcement data to make this point. These data were obtained during inspections where OSHA suspected that exposures would be above the previous PELs. Consequently, the data by their very nature are skewed in the direction of exceeding the previous PELs, and such enforcement serves a deterrence function, encouraging future compliance with the PEL.

Second, not all commenters agreed that overexposures were “widespread.” A few other commenters (e.g., AFS) thought that OSHA substantially overstated the number of workers occupationally exposed above 100 μg/m3 in its PEA (Document ID 2379, p. 25). However OSHA's risk analyses evaluated various exposure levels in determining risks to workers, and did not rely on surveillance data, which rarely have associated exposure data. Although OSHA relied on exposure data from inspections to assess technological feasibility, it did not rely on inspection data for its risk assessment because these exposure data are not tied to specific health outcomes. Instead, the exposure data used for risk assessment purposes is found in the scientific studies discussed throughout this preamble section.

The surveillance data are also not comparable to OSHA's estimate of deaths avoided by the final rule because, as is broadly acknowledged, silicosis is underreported as a cause of death on death certificates. Thus, the surveillance data capture only a portion of the actual silicosis mortality. This point was raised by several rulemaking participants, including Dr. Rosenman; Dr. James Cone, MD, MPH, Occupational Medicine Physician at the New York City Department of Health, the AFL-CIO; and the American Thoracic Society (ATS) (Document ID 3425, p. 2; 3577, Tr. 855, 867; 4204, p. 17; 2175, p. 3; 3577, Tr. 772).

The rulemaking record includes one study that evaluated underreporting of silicosis mortality. Goodwin et al. (2003, Document ID 1030) estimated, through radiological confirmation, the prevalence of unrecognized silicosis in a group of decedents presumed to be occupationally exposed to silica, but whose causes of death were identified as respiratory diseases other than silicosis. In order to assess whether silicosis had been overlooked and under-diagnosed by physicians, the authors looked at x-rays of decedents whose underlying cause of death was listed as tuberculosis, cor pulmonale, chronic bronchitis, emphysema, or chronic airway obstruction, and whose usual industry was listed as mining, construction, plastics, soaps, glass, cement, concrete, structural clay, pottery, miscellaneous mineral/stone, blast furnaces, foundries, primary metals, or shipbuilding and repair.

Any decedent found to have evidence of silicosis on chest x-ray with a profusion score of 1/0 was considered to be a missed diagnosis. Of the 177 individuals who met study criteria, radiographic evidence of silicosis was found in 15 (8.5 percent). The authors concluded that silicosis goes undetected even when the state administers a case-based surveillance system. Goodwin et al. (2003, Document ID 1030) also cites mortality studies of Davis et al. (1983, Document ID 0999) and Hughes (1982, Document ID 0362) who reported finding decedents with past chest x-ray records showing evidence of silicosis but no mention of silicosis on the death certificate.

The Goodwin et al. (2003) study illustrates the importance of information about the decedent's usual occupation and usual industry on death certificates. Yet for the years 1985 to 1999, only 26 states coded this information for inclusion on death certificates. If no occupational information is available, recognizing exposure to silica, which is necessary to diagnose silicosis, becomes even more difficult, further contributing to possible underreporting.

Dr. Rosenman, a physician, epidemiologist and B-reader, testified that in his research he found silicosis recorded on only 14 percent of the death certificates of individuals with confirmed silicosis (Document ID 3425, p. 2; 3577, Tr. 854; see also 3756, Attachment 11). This means that as much as 86 percent of deaths related to silicosis are missing from the NIOSH WoRLD database, substantially compromising the accuracy of the surveillance information. Dr. Rosenman also found that silicosis is listed as the cause of death in a small percentage of individuals who have an advanced stage of silicosis; 18 percent in those with progressive massive fibrosis (PMF) and 10 percent in those with category 3 profusion.

As noted above, factors that contribute to underreporting by health care providers include lack of information about exposure histories and difficulty recognizing occupational illnesses that have long latency periods, like silicosis (e.g., Document ID 4214, p. 13; 3584, Tr. 2557). Dr. Rosenman's testimony indicated that many physicians are unfamiliar with silicosis and this lack of recognition is one factor that contributes to the low recording rate for silicosis on death certificates (Document ID 3577, Tr. 855). In order to identify cases of silicosis, a health care provider must be informed of the patient's history of occupational exposure to dust containing respirable silica, a critical piece of information in identifying and reporting cases of silicosis. However, information on a decedent's usual occupation and/or industry is often not available at the time of death or is too general to be useful. If the physician completing the death certificate is unaware of the decedent's occupational exposure history to crystalline silica, and does not have that information available to her/him on a medical record, a diagnosis of silicosis on the death certificate is Start Printed Page 16329unlikely. According to a study submitted by the Laborers' Health and Safety Fund of North America, (Wexelman et al., 2010), a sample of physician residents surveyed in New York City did not believe that cause of death reporting is accurate; this was a general finding, and not specific to silicosis (Document ID 3756, Attachment 7).

The ATS and the American College of Chest Physicians commented that physicians often fail to recognize or misdiagnose silicosis as another lung disease on the death certificate, leading to under-reporting on death certificates (3577, Tr. 821, 826-827) and under-recognize and underreport cases of silicosis (Document ID 2175, p. 3). As Dr. Weissman from NIOSH responded:

. . . it's well known that death certificates don't capture all of the people that have a condition when they pass away, and so there would be many that probably would not be captured if the silicosis didn't directly contribute to the death and depending on who filled out the death certificate, and the conditions of the death and all those kinds of things. So it's an under-representation of people who die with the condition . . . . (Document ID 3579, pp. 166-167).

Although there is little empirical evidence describing the extent to which silicosis is underreported as a cause of death, OSHA finds, based on this evidence as well as on testimony in the record, that the available silicosis surveillance data are likely to significantly understate the number of deaths that occur in the U.S. where silicosis is an underlying or contributing cause. This is in large part due to physicians and medical residents who record causes of death not being familiar or having access to the patient's work or medical history (see Wexelman et al., 2010, Document ID 3756, Attachment 7; Al-Samarri et al., Prev. Chronic Dis. 10:120210,2013). According to Goodwin et al. (2003, Document ID 1030, p. 310), most primary care physicians do not take occupational histories, nor do they receive formal training in occupational disease. They further stated that, since it is likely that a person would not retain the same health care provider over many years, even if the presence of silicosis in a patient might have been known by a physician who cared for them, it would not necessarily be known by another physician or resident who recorded cause of death years or decades later and who did not have access to the patient's medical or work history. OSHA finds the testimony of Dr. Rosenman compelling, who found that silicosis was not recorded as an underlying or contributing cause of death even where there was chest x-ray evidence of progressive massive fibrosis related to exposure to crystalline silica.

Some commenters stated that the decline in silicosis mortality demonstrates that there is a threshold for silicosis above the prior PEL of 100 μg/m3 (Document ID 4224, p. 2-5; 3582, Tr. 1951-1963). OSHA finds this argument irrelevant as the threshold concept does not apply to historical surveillance data. As noted above and discussed in Section V.I, Comments and Responses Concerning Threshold for Silica-Related Diseases, OSHA believes that surveillance data should not be used for quantitative risk analysis (including determination of threshold effects) because it lacks an exposure characterization based on sampling. Thus, the surveillance data cannot demonstrate the existence of a population threshold.

There is also evidence in the record that silicosis morbidity statistics reviewed earlier in this section are underreported. This can be due, in part, to the relative insensitivity of chest roentgenograms for detecting lung fibrosis. Hnizdo et al. (1993) evaluated the sensitivity, specificity and predictive value of radiography by correlating radiological and pathological (autopsy) findings of silicosis. “Sensitivity” and “specificity” refer to the ability of a test to correctly identify those with the disease (true positive rate), and those without the disease (true negative). Because pathological findings are the most definitive for silicosis, findings on biopsy and autopsy provide the best comparison for determining sensitivity and specificity of chest imaging.

The study used three readers and defined a profusion score of 1/1 as positive for silicosis. Sensitivity was defined as the probability of a positive radiological reading (ILO category >1/1) given that silicotic nodules were found in the lungs at autopsy. Specificity was defined as the probability of a negative radiological reading (ILO category <1/1) given that no, or only an insignificant number of silicotic nodules were found at autopsy. The average sensitivity values were low for each of the three readers (0.39, 0.37, and 0.24), whereas the average specificity values were high (0.99, 0.97, and 0.98). For all readers, the proportion of true positive readings (i.e., the sensitivity) increased with the extent of silicosis found at autopsy (Document ID 1050).

In the only published study that quantified the extent of underreporting of silicosis mortality and morbidity, Rosenman et al. estimated the number of new cases of silicosis occurring annually in the U.S. at between 3,600 and 7,300 based on the ratio of living to deceased persons identified and confirmed as silicotics in the Michigan surveillance data and extrapolating that ratio using the number of deaths due to silicosis for the U.S. as a whole (2003, Document ID 0420). OSHA reviewed the study in its Review of the Health Effects Literature (Document ID 1711, p. 48). Patrick Hessel, Ph.D., criticized the methods used by Dr. Rosenman, and deemed the resulting estimates unreliable, stating that the actual number of new silicosis cases arising each year is likely to be lower than the authors estimated (Document ID 2332, p. 2; 3576, Tr. 323-331).

OSHA disagrees with the criticisms that Dr. Hessel, commenting on behalf of the Chamber, offered on the study by Rosenman et al. (2003, Document ID 0420). Specifically, Dr. Hessel argued: (1) That the silicosis-related deaths used by Rosenman et al. occurred during the period 1987 through 1996, and do not reflect the declining numbers after that time period; (2) that the Michigan surveillance system relied on a single B-reader who was biased toward finding silicosis in patients who were brought to his attention for suspected silicosis; and (3) that the Michigan population was not representative of the rest of the country, since about 80 percent of the workers diagnosed with silicosis worked in foundries, which are not prevalent in most other states. Finally, in his hearing testimony, Dr. Hessel criticized the capture-recapture analysis used by Rosenman et al. to estimate the extent of underreporting of cases, stating that a number of underlying assumptions used in the analysis were not met (Document ID 3576, Tr. 323-332).

Dr. Rosenman addressed many of these criticisms in the study and at the rulemaking hearing. Regarding the fact that the number of silicosis-related deaths does not reflect the decline in deaths after 1996, Dr. Rosenman testified that, although the number of recorded silicosis deaths have declined since then, the ratio of cases to deaths has increased because the number of cases has not declined. “The living to dead ratio that we reported in our published study in 2003 was 6.44. This ratio has actually increased in recent years to 15.2. A similar ratio . . . [was] found in the New Jersey surveillance data, which went from 5.97 to 11.5 times” (Document ID 3577, Tr. 854). If one were to apply the more recent ratio from Michigan (more than double the ratio used by Rosenman et al.) to the more recent number of deaths in the country (about half that recorded in the mid-1990s; see Table V-1) to extrapolate Start Printed Page 16330the number of silicosis cases for the U.S. overall, the result would be even greater than the estimate in Rosenman et al. (2003).

At the hearing, Dr. Rosenman testified that he was the sole B-reader of lung x-rays for the study, and that he received the x-ray films from other radiologists who suspected but did not confirm the presence of silicosis (Document ID 3577, Tr. 877-878). Dr. Rosenman, while acknowledging that there could be differences between readers in scoring x-ray films, argued that such differences in scoring—for example, whether a film is scored a 3/3, 3/2, or 2/3—did not affect this study since the study design only required that a case be identified and confirmed (diagnosis requires a chest radiograph interpretation showing rounded opacities of 1/0 or greater profusion) (Document ID 3577, Tr. 877-878; 0420, p. 142).

Dr. Rosenman also addressed the criticism that Michigan's worker population with silica exposure is significantly different from the rest of the country. In the study, Rosenman et al. reported that the ratio of cases to deaths was about the same for Ohio as for Michigan and, during the public hearing, Dr. Rosenman testified that the ratio of cases to deaths for New Jersey was also similar to Michigan's (11.5 vs. 15.2) (Document ID 0420, p. 146; 3577, Tr. 854). This similarity was despite the fact that New Jersey had a different industrial mix, with fewer foundries (Document ID 3577, Tr. 878). Furthermore, the estimates made by Rosenman et al. depended on the ratio of cases to deaths in Michigan, rather than just the number of cases in that state. The authors believed that the ratio would be unaffected by the level of industrialization in Michigan (Document ID 0420, p. 146).

Finally, regarding the capture-recapture analysis, OSHA notes that Dr. Hessel acknowledged that this technique has been used in epidemiology to estimate sizes of populations identified from multiple overlapping sources (Document ID 2332, p. 2), which is the purpose for which Rosenman et al. used the approach. In addition, the Rosenman et al. study noted that the assumptions used in capture-recapture analysis could not be fully met in most epidemiological study designs, but that the effect of violating these assumptions was either negligible or was evaluated using interaction terms in the regression models employed. The investigators also reported that the capture-recapture analysis used on Ohio state surveillance data found that the total number of cases estimated for the state was between 3.03 and 3.18 times the number of cases identified, a result that is comparable to that for Michigan (Document ID 0420, pp. 146-147). After considering Dr. Hessel's written testimony, Dr. Rosenman testified that “. . . overall I don't think his comments make a difference in my data” (Document ID 3577, Tr. 877).

OSHA finds all of Dr. Rosenman's responses to Dr. Hessel's criticisms to be reasonable. And based on Dr. Rosenman's comments and testimony, OSHA continues to believe that the Rosenman et al. (2003) analysis and resulting estimates of the number of new silicosis cases that arise each year are reasonable. Additionally, Dr. Rosenman, in updating his data for his testimony for this rulemaking, found that the ratio had increased from 6.44 in the published study to 15.2 times in more recent years (Document ID 3577, Tr. 854). The study supports OSHA's hypothesis that silicosis is a much more widespread problem than the surveillance data suggest and that OSHA's estimates of the non-fatal illnesses that will be avoided as a result of this new silica standard are not unreasonable. Regardless, even assuming commenters' criticisms have merit, they do not significantly affect OSHA's own estimates from the epidemiological evidence of the risks of silicosis.

Accordingly, after careful consideration of the available surveillance data, stakeholders' comments and testimony, and the remainder of the record as a whole, OSHA has determined that the available silicosis surveillance data are useful for providing context and an illustration of a significant general trend in the reduction of deaths associated with silicosis over the past four to five decades. As discussed above, and in large part because the data themselves are limited and incomplete, OSHA believes reliance upon them for the purpose of estimating the magnitude of the risk would be inappropriate. The Agency has chosen instead to follow its well-established practice of relying on epidemiological data to meet its burden of demonstrating that workers exposed to respirable crystalline silica at the previous PELs face a significant risk of developing silicosis and that such risk will be reduced when the new limit is fully implemented.

F. Comments and Responses Concerning Lung Cancer Mortality

OSHA received numerous comments regarding the carcinogenic potential of crystalline silica as well as the studies of lung cancer mortality that the Agency relied upon in the Preliminary Quantitative Risk Assessment (QRA). Many of these comments, particularly from the ACC, asserted that (1) OSHA should have relied upon additional epidemiological studies, and (2) the studies that the Agency did rely upon (Steenland et al., 2001a, as re-analyzed in ToxaChemica, 2004; Rice et al., 2001; Attfield and Costello, 2004; Hughes et al., 2001; and Miller and MacCalman, 2009) were flawed or biased. In this section, OSHA presents these comments and its responses to them.

1. Carcinogenicity of Crystalline Silica

As discussed in the Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 76-77), in 1997, the World Health Organization's International Agency for Research on Cancer (IARC) conducted a thorough expert committee review of the peer-reviewed scientific literature and classified crystalline silica dust, in the form of quartz or cristobalite, as Group 1, “carcinogenic to humans” (Document ID 2258, Attachment 8, p. 211). IARC's overall finding for silica was based on studies of nine occupational cohorts that it considered to be the least influenced by confounding factors (Document ID 1711, p. 76). In March of 2009, 27 scientists from eight countries participated in an additional IARC review of the scientific literature and subsequently, in 2012, IARC reaffirmed that respirable crystalline silica dust is a Group 1 human carcinogen that causes lung cancer (Document ID 1473, p. 396). Additionally, in 2000, the National Toxicology Program (NTP) of HHS concluded that respirable crystalline silica is a known human carcinogen (Document ID 1164, p. 1).

The ACC, in its pre-hearing comments, questioned the carcinogenic potential of crystalline silica, asserting that IARC's 1996 recommendation that crystalline silica be classified as a Group 1 carcinogen was controversial (Document ID 2307, Attachment A, p. 29). The ACC cited Dr. Patrick Hessel's 2005 review of epidemiological studies, published after the initial IARC determination, in which he concluded that “the silica-lung cancer hypothesis remained questionable” (Document ID 2307, Attachment A, p. 31). The ACC reasserted this position in its post-hearing brief, contending that “epidemiological studies have been negative as often as they have been positive” (Document ID 4209, pp. 33-34).

After the publication of Dr. Hessel's 2005 review article, IARC reaffirmed in 2012 its earlier Group 1 classification for crystalline silica dust (Document ID 1473). As pointed out by Steenland and Start Printed Page 16331Ward, IARC is one of “2 agencies that are usually considered to be authoritative regarding whether a substance causes cancer in humans,” the other being the NTP, which has also determined crystalline silica to be carcinogenic on two separate occasions (2013, article included in Document ID 2340, p. 5). David Goldsmith, Ph.D., who coauthored one of the first published articles linking silica exposure to lung cancer, echoed Steenland and Ward:

It is important to recognize that evidence for silica's carcinogenicity has been reviewed three times by the International Agency for Research on Cancer, once in 1987, 1997, and 2012. It has been evaluated by California's Proposition 65 in 1988, by the National Toxicology Program in 2000 and reaffirmed in 2011, and by the National Institute for Occupational Safety and Health in 2002 (Document ID 3577, Tr. 861-862).

Multiple organizations with great expertise in this area, including the American Cancer Society, submitted comments supporting the thorough and authoritative nature of IARC's findings regarding silica's carcinogenicity (e.g., Document ID 1171; 1878). OSHA likewise places great weight on the IARC and NTP classifications and, based on their findings, concludes that the carcinogenic nature of crystalline silica dust has been well established. Further support for this finding is discussed in Section V.L, Comments and Responses Concerning Causation.

2. Silicosis and Lung Cancer

In addition to debating the conclusions of IARC, Peter Morfeld, Dr. rer. medic, testifying on behalf of the ACC Crystalline Silica Panel, concluded that OSHA's risk estimates for lung cancer are “unreliable” because they “ignore threshold effects and the apparent mediating role of silicosis” (Document ID 2307, Attachment 2, p. 16). Dr. Morfeld argued that silicosis is a necessary prerequisite for silica-related lung cancer. Commenters' arguments about silicosis being a prerequisite for lung cancer and silicosis having a threshold are linked; if it were shown both that silicosis requires a certain threshold of exposure and that only persons with silicosis get lung cancer, then silica-related lung cancer would also have an exposure threshold. As discussed in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases, commenters claimed that there is a threshold for silicosis above the previous PEL for general industry, which would make any threshold for lung cancer above that level as well. OSHA discusses these comments in detail in that section, and has determined that even if lung cancer does not occur in the absence of silicosis, the record strongly supports the conclusion that workers exposed to respirable crystalline silica would still be at risk of developing lung cancer as a result of their exposure because silicosis can develop among workers whose average and cumulative exposures are below the levels permitted by the previous PELs.

OSHA received comments from other stakeholders, including Robert Glenn, representing the Brick Industry Association, and the AFS on the possible mediating role of silicosis in the development of lung cancer (Document ID 2307, pp. 29-35; 2343, Attachment 1, pp. 42-45; 2379, Attachment 2, pp. 24-25). The ACC cited several review articles in support of its claim that “silica exposures have not been shown to increase the risk of lung cancer in the absence of silicosis” (Document ID 2307, Attachment A, pp. 29, 32, 35). These articles included: A 2004 review of studies by Kurihara and Wada that found that while silicosis is a risk factor for lung cancer, exposure to silica itself may not be a risk factor (Document ID 1084); a 2006 review by Pelucchi et al. that determined that the issue of whether silica itself increases lung cancer risk in the absence of silicosis has not been resolved (Document ID 0408); and a 2011 review by Erren et al. that concluded it is unclear whether silica causes lung cancer in persons who do not already have silicosis (Document ID 3873). Similarly, the AFS cited a review by the Health and Safety Executive (2003) that concluded that increased risks of lung cancer are restricted to those groups with the highest cumulative exposures, with evidence tending to show that excess lung cancer mortality is restricted to those with silicosis (Document ID 2379, Attachment 2, pp. 24-25). Having reviewed the studies cited by commenters, OSHA has come to the conclusion that none of the cited studies demonstrates that silicosis is a necessary precursor to lung cancer, but acknowledges that uncertainty remains about what percentage of lung cancers in silica-exposed workers are independent of silicosis.

Similarly, the ACC stated that none of the studies of lung cancer mortality that OSHA relied upon in the Preliminary QRA demonstrates that silica exposure causes lung cancer in the absence of silicosis (Document ID 2307, Attachment A, p. 66). During the rulemaking hearing, NIOSH scientists addressed the issue of whether silicosis is a necessary precursor to the development of lung cancer. They stated that it is a difficult issue to resolve because the two diseases may have a similar pathway, such that they can develop independently but still appear correlated. Mr. Robert Park also added that:

[S]ilicosis isn't detectable until there's splotches on the lung that are visible in x-rays. So prior to that point, somebody could have [been] developing lung disease and you just can't see it. So, of course, people that have silicosis are going to have higher lung cancer, and it's going to look like a threshold because you didn't see the silicosis in other people that have lower lung cancer risk. To really separate those two, you'd have to do a really big study. You'd have to have some measures, independent measures of lung physiological pathology, and see what's going on with silicosis as a necessary condition for development of lung cancer (Document ID 3579, Tr. 245-247).

Similarly, David Weissman, MD, concurred that “there's quite a bit of reason as Bob [Park] said to think that the two processes [development of silicosis and development of lung cancer] don't require each other, and it would be extraordinarily difficult to sort things out in human data” (Document ID 3579, Tr. 247). Indeed, Checkoway and Franzblau (2000) reviewed the epidemiological literature addressing this topic, and found that the “limitations of existing epidemiologic literature that bears on the question at hand suggest that prospects for a conclusive answer are bleak” (Document ID 0323, p. 257). The authors concluded that silicosis and lung cancer should be treated in risk assessments as “separate entities whose cause/effect relations are not necessarily linked” (Document ID 0323, p. 257). Brian Miller, Ph.D., a peer reviewer of OSHA's Review of Health Effects Literature and Preliminary QRA, likewise wrote in his post-hearing comments, “I consider this issue unanswerable, given that we cannot investigate for early fibrotic lesions in the living, but must rely on radiographs” (Document ID 3574, p. 31).

During the public rulemaking hearing, several stakeholders pointed to a recent study of Chinese pottery workers and miners by Liu et al. (2013, article included in Document ID 2340) as evidence that exposure to crystalline silica is associated with lung cancer even in the absence of silicosis (Document ID 3580, Tr. 1232-1235; 3577, Tr. 803-804, 862-863). In this study, the authors excluded 15 percent of the cohort (including 119 lung cancer deaths) with radiographic evidence of silicosis and found that the risk of lung cancer mortality still increased with cumulative exposure to crystalline silica, suggesting that clinically-Start Printed Page 16332apparent silicosis is not a prerequisite for silica-related lung cancer (article included in Document ID 2340, pp. 3, 7).

The ACC argued that it is “premature to draw that conclusion,” stating that the Liu study's conclusions are not supported by the data and raising questions about uncertainty in the exposure estimates, modeling and statistics, confounding, and the silicosis status of cohort members (Document ID 2307, Attachment A, p. 48; 4027, pp. 35-36; 4209, pp. 40-51). With regard to exposure estimates, the ACC had a number of concerns, including that conversion factors determined by side-by-side sampling in 1988-1989 were used to convert Chinese total dust concentrations to respirable crystalline silica exposures (Document ID 4209, pp. 40-41). Dr. Cox expressed concern that these conversion factors from 1988-1989 might not have been applicable to other time periods, as particle size distributions could change over time (Document ID 4027, p. 32). OSHA acknowledges this concern, but given the “insufficient historical particle size data . . . to analyze whether there were changes in particle size distributions from the 1950s to the 1990s,” believes that the authors were justified in making their exposure assumptions (Document ID 4027, p. 32). Dr. Cox's concerns involving modeling and statistics (see Document ID 4027, pp. 33-36) in the study, including the absence of model diagnostics, the use of inappropriate or misspecified models, the lack of a discussion of residual confounding and model uncertainty, and the use of inappropriate data adjustments and transformations, are discussed in detail in Section V.J, Comments and Responses Concerning Biases in Key Studies.

On the issue of confounding, the ACC noted that Liu et al. (2013) used a subcohort of 34,018 participants from 6 tungsten mines, 1 iron mine, and 4 potteries derived from a total cohort of 74,040 participants from 29 mines and pottery factories studied previously by Chen et al. (2007, Document ID 1469; 2307, Attachment A, pp. 48-50). Liu et al. (2013) excluded participants in the original cohort if detailed information on work history or smoking was not available, or if they worked in copper mines or tin mines where the analysis could be confounded by other exposures, namely radon and carcinogenic polycyclic aromatic hydrocarbons (PAHs) in the former and arsenic in the latter (article included in Document ID 2340, p. 2). The ACC's main concern was that Liu et al. (2013) did not adjust for these confounders in their analyses, but rather claimed that there were no confounding exposures in their smaller cohort on the basis of the exclusion criteria (Document ID 2307, Attachment A, p. 49).

The ACC also noted that Chen et al. (2007) stated that the Chinese pottery workers were exposed to PAHs, and some of the iron-copper miners were exposed to PAHs and radon progeny (Document ID 2307, Attachment A, p. 49). Chen et al. (2007) initially found an association between respirable silica and lung cancer mortality in the pottery workers and iron-copper miners, but it disappeared after adjusting for PAH exposures (Document ID 1469). In the tungsten miners, Chen et al. (2007) found no significant association for lung cancer mortality, while Liu et al. (2013) did. Similarly, the ACC pointed out that a subsequent study by Chen et al. (2012, article included in Document ID 2340) also failed to find a statistically significant increase in the hazard ratio for lung cancer, meaning that there was no significant positive exposure-response relationship between cumulative silica exposure and lung cancer mortality (Document ID 4209, p. 45). Dr. Morfeld concluded, “Unless and until these issues are resolved, Liu et al. (2013) should not be used to draw conclusions regarding exposure-response relationships between RCS [respirable crystalline silica], silicosis and lung cancer risk” (Document ID 2307, Attachment 2, pp. 15-16).

During the public hearing, counsel to the ACC asked Dr. Steenland, a co-author on the Liu et al. (2013) study, if he would provide measurement data on the PAH exposures in the potteries, as well as present the data from the Liu et al. (2013) study separately for pottery factories and tungsten mines, as they were in Chen et al. (2007, Document ID 1469) (Document ID 3580, Tr. 1237-1240). Dr. Steenland subsequently provided the requested data for inclusion in the rulemaking record (Document ID 3954).

With respect to the PAH data for the potteries, Dr. Weihong Chen, the study's first author, reported that, in measurements in 1987-1988 in the four potteries that were excluded from the Liu et al. (2013) analysis, the mean total PAHs was 38.9 µg/m3 and the mean carcinogenic PAHs was 4.7 µg/m3. In the four potteries that were included in the Liu et al. (2013) analysis, the mean total and carcinogenic PAHs, as measured in 1987-1988, were substantially lower at 11.6 and 2.5 µg/m3, respectively. When the measurements were repeated in 2006, the mean total and carcinogenic PAHs in the four potteries included in the analysis were still lower, at 2.2 and 0.08 µg/m3, levels that were “not much higher than environmental PAH in many [Chinese] cities” (Document ID 3954, p. 2). Dr. Chen also reported that, when comparing levels within six job titles, there was no significant correlation between total or carcinogenic PAHs (based on the 2006 measurements) and respirable silica dust. When the results were presented separately for the mines and potteries, in analyses using continuous cumulative exposure, the relationship between silica exposure and lung cancer mortality remained significant for the pottery factories, but not the metal mines. In the categorical analyses using quartiles of cumulative exposure, the results were mixed: The association between silica exposure and lung cancer mortality was statistically significant in some exposure quartiles for both metal mines and pottery factories (Document ID 3954, p. 2).

Based upon these subsequent data, the ACC concluded that PAHs were likely present in the potteries but not in the mines (Document ID 4209, p. 45). OSHA believes this conclusion, although plausible, to be speculative. What is known is that the potteries that were excluded had a higher average level of PAHs, and that a significant association between cumulative silica exposure and lung cancer mortality remained in the included potteries even after the analysis was separated by potteries and mines. However, the association was less clear in the metal mines.

The ACC also raised concerns about the silicosis status of lung cancer cases in the Liu cohort, asserting that some workers may not have had post-employment radiography given that social health insurance only recently began to pay for it. As such, the ACC asserted that some workers who developed lung cancer post-employment may have also had undiagnosed silicosis (Document ID 4209, pp. 49-50). OSHA acknowledges the limitations of the study, as with any retrospective study, but also notes that no evidence was put forth to indicate that workers with silicosis were misclassified in the study as workers without silicosis. Further, Dr. Goldsmith testified that the method used by Liu et al. for excluding workers with silicosis (x-ray findings) was “very eminently reasonable,” given that the only foolproof means of proving the absence of silicosis—autopsy—was not available for this particular cohort (Document ID 3577, Tr. 874-875).

Thus, OSHA concludes that the Liu et al. (2013) study preliminarily suggests Start Printed Page 16333that silicosis is not required for the development of lung cancer; however, no one study will settle the question of the role of silicosis in the carcinogenicity of crystalline silica. As acknowledged by Dr. Cox, the Agency did not rely upon the Liu et al. (2013) study in its preliminary or final QRA (Document ID 2307, Attachment 4, p. 37).

Overall, after giving lengthy consideration to all evidence in the record regarding whether silicosis is a necessary precursor to the development of lung cancer, including the Liu study, the NIOSH testimony, and the mechanistic evidence for the carcinogenicity of crystalline silica discussed in Section V.H, Mechanisms of Silica-Induced Adverse Health Effects, OSHA concludes that the mediating role of silicosis in the development of lung cancer is not “apparent,” as suggested by Dr. Morfeld and the ACC (Document ID 2307, Attachment 2, p. 16). As such, OSHA continues to believe that substantial evidence supports the Agency's decision to consider lung cancer as a separate, independent health endpoint in its risk analysis. The Agency also notes that even if lung cancer does not occur in the absence of silicosis, the record strongly supports the conclusion that workers exposed to respirable crystalline silica would still be at risk of developing lung cancer as a result of their exposure because silicosis can develop from average and cumulative exposures below the levels allowed at the previous PEL (see Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases.)

3. Additional Studies

Stakeholders also suggested several additional studies that they believe OSHA should include in its QRA on lung cancer. The AFS commented that OSHA's Preliminary QRA overlooked a 2003 report by the Health and Safety Executive (HSE, Document ID 1057), asserting that over 40 percent of the references cited by HSE were omitted in OSHA's review (Document ID 4035, p. 2). OSHA disagrees with this assessment of overlooking the report, noting that the Agency reviewed and referenced the HSE report in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711, p. 77). As discussed in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA, OSHA used a weight-of-evidence approach to evaluate the scientific studies in the literature to determine their overall quality. In so doing, OSHA thoroughly reviewed approximately 60 published, peer-reviewed primary epidemiological studies covering more than 30 occupational cohorts in over a dozen industrial sectors, as well as the IARC pooled study and several meta-analyses (Document ID 1711, pp. 75-172).

The AFS also submitted a 2011 review of 30 foundry epidemiology studies by the Industrial Industries Advisory Council (IIAC) and noted that only 7 of those 30 studies were included in OSHA's Review of Health Effects Literature and Preliminary QRA (Document ID 2379, p. 24). AFS wrote:

The PQRA largely dismisses the foundry epidemiology studies, based on assertions of positive confounding. However, a study showing that there is no adverse effect despite a positive confounder is not only still relevant to the question, but should be more persuasive than a study without positive confounders because the data then show that even with an additive risk, there is no increase in effect at the reported exposure levels (Document ID 2379, p. 24).

In response to this comment, OSHA gathered the remaining 23 foundry studies cited in the submitted report and placed them in the rulemaking docket during the post-hearing comment period. OSHA notes, in the first instance, that most of these studies were not designed to study the effects of silica exposure on foundry workers, and did not even attempt to do so; rather, their purpose was to examine lung cancer mortality and/or morbidity in foundry work, which involves many toxic and otherwise harmful substances besides silica. Therefore, OSHA would likely be unable to suitably use these studies as a basis for a quantitative risk assessment regarding respirable crystalline silica by itself.

With respect to AFS's assertions of studies showing “no adverse effect,” OSHA notes that the summary section of the IIAC review report, submitted as evidence by AFS, stated that, “The cohort mortality studies and two morbidity studies suggest an increased risk of lung cancer in foundry workers when considered overall, but do not support a doubling of risk. . . . Findings in the case-control studies, the majority of which adjust for the effects of smoking . . . tend to support those of the cohort studies” (Document ID 3991, p. 5). As such, this review of 30 foundry epidemiology studies showed an increased excess risk of lung cancer from foundry work; the fact that the excess risk was not increased by a factor of two is irrelevant to the current proceedings. The factor of two appears to be used by the IIAC in determining whether monetary benefits should be paid to foundry workers in Great Britain and is completely unrelated to OSHA's statutory requirements for determining whether workers exposed to silica are at a significant risk of material impairment of health. Given that excess lung cancer was observed in many of these studies, OSHA rejects the AFS's assertion that, even with positive confounding, there was no increase in adverse effect (i.e., lung cancer).

OSHA also notes that the IIAC's finding of an elevated risk of lung cancer in foundries is not surprising. As Dr. Mirer stated during his testimony, IARC categorized foundry work as Group 1, carcinogenic to humans, in 1987 based on observed lung cancer (Document ID 2257, Attachment 3, p. 5). IARC reaffirmed its Group 1 classification for foundry work in 2012 (Document ID 4130). However, as noted by OSHA in its Review of Health Effects Literature, the foundry epidemiology studies were profoundly confounded by the presence of exposures to other carcinogens, including PAHs, aromatic amines, and metals (Document ID 1711, p. 264). Because of this confounding, as well as the fact that most of these studies did not specifically study the effects of silica exposure on foundry workers, OSHA has decided not to include them in its QRA.

The ACC likewise cited several individual studies that it believed found no relationship between silica exposure and lung cancer risk (Document ID 2307, Attachment A, pp. 33-35). These included studies by: (1) Yu et al. (2007), which found no consistent exposure-response relationship between silica exposure and lung cancer death in workers with silicosis in Hong Kong (Document ID 3872); (2) Chen et al. (2007), which found, as mentioned in relation to the Liu et al. (2013) study, no relationship between silica exposure and lung cancer after adjusting for confounders in a study of Chinese tungsten miners, tin miners, iron-copper miners, and pottery workers (Document ID 1469); (3) Birk et al. (2009), which found the standardized mortality ratio (SMR) for lung cancer was not elevated in a subgroup of men who worked in areas of German porcelain plants with the highest likely silica exposures (Document ID 1468); (4) Mundt et al. (2011), which found, in a subsequent analysis of the German porcelain industry, that cumulative silica exposure was not associated with lung cancer mortality, mortality from kidney cancer, or any other cause of death other than silicosis (Document ID 1478); and (5) Westberg et al. (2013), which found that cumulative silica exposure was not associated with lung cancer morbidity (Document ID 4054).Start Printed Page 16334

Briefly, Chen et al. (2007) examined a cohort of male workers in 29 Chinese mines and factories, and initially found a significant trend between cumulative silica exposure and lung cancer mortality in pottery workers and tin miners; this trend was no longer significant after adjustment for occupational confounders (carcinogenic PAHs in potteries, arsenic in tin mines) (Document ID 1469, pp. 320, 323-324). On the contrary, Liu et al. (2013) demonstrated a statistically significant association between cumulative silica exposure and lung cancer mortality after excluding mines and factories with confounding exposures (article included in Document ID 2340). As noted previously, there are questions of how confounding exposures to radon, PAHs, and arsenic were handled in both the Chen et al. (2007) and Liu et al. (2013) studies. One important difference between the two studies, however, was the follow-up time. While Chen et al. (2007) had follow-up to 1994 and identified 511 lung cancer deaths in a cohort of 47,108 workers (Document ID 1469, pp. 321-322), Liu et al. (2013) had follow-up to 2003 and identified 546 lung cancer deaths in a cohort of 34,018 workers (article included in Document ID 2340, pp. 2-4).

OSHA discussed the Birk et al. (2009, Document ID 1468) and Mundt et al. (2011, Document ID 1478) studies of the German porcelain industry in its Supplemental Literature Review, noting several limitations that are applicable to both studies and might preclude the conclusion that there was no association between silica exposure and lung cancer (Document ID 1711, Attachment 1, pp. 6-12). One such limitation was the mean age of subjects—35 years—at the start of follow-up, making this a relatively young cohort in which to observe lung cancer. The mean follow-up period of 19 years per subject was also a limitation, given the long latency for lung cancer and the young age of the cohort at the start of follow-up; only 9.2 percent of the cohort was deceased by the end of the follow-up period. OSHA noted that Mundt et al. (2011) acknowledged that additional follow-up of the cohort may be valuable (Document ID 1711, Attachment 1, pp. 10-11; 1478, p. 288). In addition, Mundt et al. (2011) had only 74 male lung cancer deaths, some of whom had possible or probable prior silica exposure that could have resulted in cumulative exposure misclassification (Document ID 1478, pp. 285, 288). The authors also reported statistically significantly elevated lung cancer hazard ratios for some categories of average silica exposure, but did not present any trend analysis data (Document ID 1478, p. 285). It also does not appear that Mundt et al. performed any lagged analyses for lung cancer to account for the latency period of lung cancer.

Following the ACC's citation of the Yu et al. (2007) and Westberg et al. (2013) studies in its pre-hearing comments, OSHA obtained and reviewed these studies, and added them to the rulemaking docket (Document ID 3872; 4054). Yu et al. (2007) followed a cohort of 2,789 workers in Hong Kong diagnosed with silicosis between 1981 and 1998. The average follow-up time was 9 years, with 30.6 percent of the cohort deceased when the study ended in 1999. The SMR for lung cancer was not statistically significantly elevated following indirect adjustment for cigarette smoking; similarly, the authors did not find a significant exposure-response relationship between cumulative silica exposure and lung cancer mortality (Document ID 3872). Westberg et al. (2013) studied a group of 3,045 male Swedish foundry workers to determine lung cancer incidence and morbidity. Although the lung cancer incidence was statistically significantly elevated, the authors did not find a significant exposure-response relationship with cumulative quartz exposure (Document ID 4054, p. 499).

Regarding these studies, OSHA notes that the Westberg et al. (2013) study, like other foundry studies, is confounded by other carcinogenic substances present in foundries, including, as the authors pointed out, phenol, formaldehyde, furfuryl alcohols, PAHs, carbon black, isocyanates, and asbestos (Document ID 4054, p. 499). The Yu et al. (2007) study had an average follow-up period of only 9 years (Document ID 3872, p. 1058, Table 1), which is a short follow-up period when considering the latency period for the development of cancer. In addition, the Yu et al. study (2007), as described in the earlier Tse et al. (2007) study, used a job exposure matrix developed from expert opinion to assign estimated past levels of silica exposure to individuals based on self-reported work history; changes in exposure intensity with calendar year were not considered because of limited data (Document ID 3841, p. 88; 3872, p. 1057). OSHA notes that this exposure estimation may have included considerable misclassification due to inaccuracies in self-reported work history, the use of expert opinion to estimate past exposure levels rather than actual measurements for the subjects under study, and the failure to incorporate any changes in exposure levels over calendar time into the exposure estimates. Although these exposure estimates were used in an analysis that found a significant exposure-response for NMRD mortality among workers with silicosis (Tse et al., 2007, Document ID 3841), an exposure-response for lung cancer mortality may not be as strong and may be harder to detect, requiring more accurate exposure information. OSHA also notes that NMRD mortality is likely to be a competing cause of death with lung cancer, such that some workers may have died from NMRD before developing lung cancer. The workers with silicosis in this study also had high exposures (mean cumulative exposure of 10.89 mg/m3-yrs) (Document ID 3872, p. 1058), possibly making it difficult to detect an exposure-response for lung cancer when exposures are relatively homogenous and high. Selection effects would have been extreme in these highly-exposed workers, whose all-cause mortality was double what would be expected (853 deaths observed, 406 expected) in the general population of males in Hong Kong and whose respiratory disease mortality was an astounding six times the expected level (445 deaths observed, 75 expected) (Document ID 3872, p. 1059).

OSHA acknowledges that not every study reaches the same results and conclusions. This is typically true in epidemiology, as there are different cohorts, measurements, study designs, and analytical methods, among other factors. As a result, scientists critically examine the studies, both individually and overall, in the body of literature to draw weight-of-evidence conclusions. IARC noted, with respect to its 1997 carcinogenicity determination:

[N]ot all studies reviewed demonstrated an excess of cancer of the lung and, given the wide range of populations and exposure circumstances studied, some non-uniformity of results had been expected. However, overall, the epidemiological findings at the time supported an association between cancer of the lung and inhaled crystalline silica (α-quartz and cristobalite) resulting from occupational exposure (Document ID 1473, p. 370).

Given IARC's re-affirmation of this finding in 2012, OSHA does not believe that the individual studies mentioned above fundamentally change the weight of evidence in the body of literature supporting the carcinogenicity of crystalline silica. The best available evidence in the rulemaking record continues to indicate that exposure to respirable crystalline silica causes lung cancer. OSHA acknowledges, however, that there is some uncertainty with respect to the exact magnitude of the Start Printed Page 16335lung cancer risk, as each of the key studies relied upon provides slightly different risk estimates, as indicated in Table VI-1.

Further, the ACC focused extensively on and advocated for a study by Vacek et al. (2011) that found no significant association between respirable silica exposure and lung cancer mortality in a cohort of Vermont granite workers (Document ID 1486, pp. 75-81). Included in the rulemaking docket are the peer-reviewed published version of the study (Document ID 1486) and the earlier Final Report to the ACC, whose Crystalline Silica Panel funded the study (Document ID 2307, Attachment 6), as well as comments from two of the authors of Vacek et al. (2011) responding to OSHA's treatment of the study in its Supplemental Literature Review (Document ID 1804). The ACC stated:

Perhaps of most interest and relevance for present purposes—because the cohort has been studied so extensively in the past and because the present PEL is based indirectly on experience in the Vermont granite industry—is the mortality study of Vermont granite workers published in 2011. While the Vermont granite workers cohort has been studied on a number of previous occasions, this is the most comprehensive mortality study of Vermont granite workers to date (Document ID 2307, Attachment A, p. 36).

The ACC criticized OSHA for rejecting the Vacek et al. (2011) study in its Supplemental Literature Review and instead relying upon the Attfield and Costello (2004, Document ID 0284) study of Vermont granite workers (Document ID 2307, Attachment A, pp. 36-47; 4209, pp. 34-36). The ACC asserted several differences between the studies. First, while Attfield and Costello had 5,414 workers (201 lung cancer deaths) in the cohort, Vacek et al. had 7,052 workers (356 lung cancer deaths) as they extended the follow-up period by 10 years to 2004. Vacek et al. also claimed to have more complete mortality data, finding that “162 workers, whom Attfield assumed were alive in 1994, had died before that time and some decades earlier” (Document ID 2307, Attachment A, p. 38). In addition, Vacek et al. used exposure measurements and raw data not used by Attfield and Costello; for example, Vacek et al. used pension records and interviews from other studies to account for gaps in employment and changes in jobs, while Attfield and Costello assumed that a person remained in the same job between chest x-rays at the Vermont Department of Industrial Health surveillance program. Different conversion factors to estimate gravimetric concentrations from particle count data were also used: Attfield and Costello used a factor of 10 mppcf = 75 µg/m3 while Vacek et al. used a factor of 10 mppcf = 100 µg/m3 (Document ID 2307, Attachment A, pp. 36-39; 1804, p. 3). OSHA notes that this discrepancy in gravimetric conversion factors should not affect the detection of an exposure-response relationship, as all exposures would differ by a constant factor.

The ACC also pointed out that Attfield and Costello's exposure estimate for sandblasters was 60 µg/m3 prior to 1940, 50 µg/m3 from 1940-1950, and 40 µg/m3 after 1950, maintaining these numbers were too low compared to Vacek et al.'s estimates of 240, 160, and 70 µg/m3, respectively (Document ID 2307, Attachment A, p. 39; 1486, p. 313). Attfield and Costello took these estimates for sand blasters from the Davis et al. (1983, Document ID 0999) study, discussed in detail below; the estimates were based on six published industrial hygiene measurement studies.

Lastly, the ACC posited that Attfield and Costello inappropriately excluded the highest exposure group, stating:

Vacek et al. used all their data in evaluating potential E-R [exposure-response] trends with increasing exposure. Attfield and Costello did not. Instead, on a post hoc basis, they excluded the highest exposure category from their analysis when they discovered that the E-R trend for lung cancer was not significant if that group was included (even though the trends for non-malignant respiratory diseases were significant when all the data were used). This is an example of both data selection bias and confirmation bias (Document ID 2307, Attachment A, p. 40).

Based upon these assertions, the ACC concluded, “In sum, when judged without a result-oriented confirmation bias, the larger, more recent, more comprehensive, and more detailed study by Vacek et al. (2011) must be deemed to supersede Attfield and Costello (2004) as the basis for evaluating potential silica-related lung cancer risks in the Vermont granite industry” (Document ID 2307, Attachment A, p. 41).

OSHA initially discussed some issues surrounding the Vacek et al. (2011) study in its Supplemental Literature Review (Document ID 1711, Attachment 1, pp. 2-5). Specifically, OSHA noted that (1) the cumulative exposure quintiles used in the Vacek et al. (2011) analysis were higher than the values used in the Attfield and Costello (2004) analysis; (2) the regression models used in the Vacek et al. (2011) study exhibited signs of uncontrolled confounding, as workers in the second lowest cumulative exposure stratum in the models (except for silicosis) exhibited a lower risk than those in the lowest stratum, while all outcomes (except NMRD) in the highest exposure stratum showed a decline in the odds ratio (a measure of the association between silica exposure and health outcome) compared to the next lower stratum; and (3) Vacek et al. (2011) found a statistically significant excess of lung cancer (SMR = 1.37, with almost 100 excess lung cancer deaths) in the cohort when compared to U.S. white males (Document ID 1486, p. 315). Regarding the excess lung cancer deaths, although they were unable to obtain information on smoking for many of the cohort members, Vacek et al. suggested that the elevated SMR for lung cancer was due, at least in part, to the differences between the smoking habits of the cohort and reference populations (Document ID 1486, p. 317). OSHA noted that although the SMR for other NMRD was elevated, there was no significant SMR elevation for other smoking-associated diseases, including cancers of the digestive organs, larynx, and bladder, as well as bronchitis, emphysema, and asthma (Document ID 1711, Attachment 1, p. 5). Elevated SMRs for these diseases would be expected if workers in the study population smoked more than those in the reference population; in fact, for all heart disease, the mortality in the study population (SMR = 0.89) was statistically significantly lower than the reference population (Document ID 1486, p. 315). These data do not support Vacek et al.'s assertion that smoking was responsible for the increased lung cancer SMR in the cohort. In addition, Davis et al. (1983) noted that granite shed workers employed during the 1970's smoked only slightly more than U.S. white males (Document ID 0999, p. 717). OSHA also pointed out that the SMR may have been understated, as Vacek et al. did not account for a healthy worker effect (HWE).

The ACC did not agree with OSHA's review of the Vacek et al. study, noting that OSHA “rejects Vacek et al. (2011) on grounds that are confusing and unfounded” (Document ID 2307, Attachment A, p. 41). The ACC argued that the quintiles of cumulative exposure used by Vacek et al. were not higher than typical values for lung cancer, and that OSHA, in its Supplemental Literature Review, compared the Vacek et al. quintiles of cumulative exposure for silicosis with the Attfield and Costello groups used for both silicosis and lung cancer (Document ID 2307, Attachment A, pp. 41-42). OSHA acknowledges this discrepancy and, given that Vacek et al. Start Printed Page 16336used quintiles of cumulative exposure that differed for each health endpoint, agrees that the quintiles for lung cancer used by Vacek et al. were not appreciably higher than the exposure groups used by Attfield and Costello, though the Agency recognizes that there may be alternative explanations for the patterns observed in the Vacek et al. data. Regarding uncontrolled confounding, the ACC stated that “The Vermont granite worker cohort, after all, supposedly is free of confounding exposures,” (Document ID 2307, Attachment A, p. 43 (citing Attfield and Costello, 2004, 0284)). Vacek et al. also pointed out that although the odds ratios for the second lowest exposure stratums were lower than those for the lowest categories for each of the diseases, they were not statistically significantly lower (Document ID 1804, pp. 1-2).

Although OSHA notes that this latter phenomenon, in which the odds ratio for the second lowest exposure stratum is lower than that for the lowest stratum, is commonly observed and often attributable to some form of selection confounding, the Agency recognizes that there may be alternative explanations for the patterns observed in the Vacek et al. data. One such explanation for the decreased odds ratios in the highest exposure group is potential attenuation resulting from a HWE.

The HWE, as defined by Stayner et al. (2003), has two components: (1) A healthy initial hire effect, in which bias is “introduced by the initial selection of workers healthy enough to work . . . and the use of general population rates for the comparison group, which includes people who are not healthy enough to work,” and (2) a healthy worker survivor effect, referring “to the tendency of workers with ill health to drop from the workforce and the effect this dropout may have on exposure-response relationships in which cumulative exposure is the measure of interest” (Document ID 1484, p. 318). Thus, the healthy initial hire effect occurs in the scenario in which the death rate in a worker group is compared to that in the general population; because the general population has many people who are sick, the death rate for workers may be lower, such that a direct comparison of the two death rates results in a bias. The healthy worker survivor effect occurs in the scenario in which less healthy workers transfer out of certain jobs into less labor-intensive jobs due to decreased physical fitness or illness, or leave the workforce early due to exposure-related illness prior to the start of follow-up in the study. As a result, the healthier workers accumulate the highest exposures such that the risk of disease at higher exposures may appear to be constant or decrease.

OSHA disagrees with the ACC's statement that “the possibility of a potential HWE in this cohort could not have affected the E-R analyses” in Vacek et al. (2011) (Document ID 2307, Attachment A, p. 46), and with the similar statement by study authors Pamela Vacek, Ph.D. and Peter Callas, Ph.D., both of the University of Vermont, who asserted that the HWE could not have impacted their exposure-response analyses “because they were not based on an external reference population” (Document ID 1804, p. 2). This explanation only considers one component of the HWE, the healthy initial hire effect. An internal control analysis, such as that performed by Vacek et al., will generally minimize the healthy initial hire effect but does not address the healthy worker survivor effect (see Document ID 1484, p. 318 (Stayner et al. (2003)). Thus, the statement by the ACC that there could be no HWE in the internal case control analysis of Vacek et al. (2011) is incorrect, as it considered only the healthy initial hire effect and not the healthy worker survivor bias.

In contrast, Attfield and Costello's stated rationale for excluding the highest exposure group is related to the healthy worker survivor effect:

We do know that this group is distinctive in entering the cohort with substantial exposures—83% had worked for 20 years or more in the high dust levels prevalent prior to controls. They were, therefore, a highly selected healthy worker group. A further reason may be that in the days when tuberculosis and silicosis were the main health concerns in these workers, lung cancer may have been obscured in this group as a cause of death in some cases” (Document ID 0284, p. 136).

Support for Attfield and Costello's reasoning is provided by a study by Applebaum et al. (2007), which re-analyzed the data from the Attfield and Costello (2004) paper and concluded that there was a healthy worker survivor effect present (study cited by Vacek et al., 2009, Document ID 2307, Attachment 6, p. 3). Applebaum et al. (2007) split the cohort of Vermont granite workers into two groups: (1) Those that began working before the start of the study follow-up, i.e., prevalent hires; and (2) those that began working after the start of the study follow-up, i.e., incident hires. The rationale for splitting the cohort into these two groups was to examine if a healthy worker survivor effect was more likely in the prevalent hire group, as this group would be affected by workers that were more susceptible to health effects and left the industry workforce prior to the start of the study follow-up (Applebaum et al., 2007, pp. 681-682). Using spline models to examine exposure-response relationships without forcing a particular form (e.g., linear, linear-quadratic) on the observed data, the authors found that the inclusion of prevalent hires in the analysis weakened the association between cumulative silica exposure and lung cancer because of bias from the healthy worker survivor effect. The bias can be reduced by including only incident hires, or keeping the date of hire close to the start of follow-up (Applebaum et al., 2007, pp. 685-686). An alternative explanation for this trend offered by Applebaum et al. may be that, assuming that there was more measurement error in the older data, the prevalent hires had more exposure misclassification (2007, p. 686); in such a case, however, the inclusion of prevalent hires would still bias the results towards the null. Given the findings of the Applebaum et al. (2007) study, OSHA believes that Attfield and Costello (2004) had good reasons for removing the highest exposure group, which was composed mostly of prevalent workers (83 percent of workers in the highest exposure group had worked at least 20 years prior to the start of the follow-up period) (Document ID 0284, p. 136).

Vacek et al. (2011), on the other hand, excluded 609 workers in the design of their study cohort due to insufficient information. However, the majority of the workers excluded from the cohort were incident hires who began work after 1950 (Document ID 2307, Attachment 6, p. 12; 1486, p. 314). The final Vacek et al. (2011) cohort included 2,851 prevalent hires (began employment before 1950) compared to 4,201 incident hires (began employment in or after 1950) (Document ID 2307, Attachment 6, p. 12; 1486, p. 314). By composing about 40 percent of their cohort with prevalent hires and excluding many incident hires, Vacek et al. (2011) may have introduced additional healthy worker survivor effect bias into their study. Interestingly, Vacek et al. described the Applebaum et al. (2007) results in their 2009 report, stating, “They [Applebaum et al.] found that decreasing the relative proportion of prevalent to incident hires [in the data used by Attfield and Costello] resulted in a stronger association between cumulative silica exposure and lung cancer mortality” (Document ID Start Printed Page 163372307, Attachment 6, p. 3). Despite their acknowledgement of the Applebaum et al. (2007) findings, Vacek et al. (2011) did not conduct any analysis of only the incident hires, or use statistical methods to better determine the presence and effect of a healthy worker survivor effect in their study.

The ACC also commented on Vacek et al.'s suggestion that the elevated SMR observed for lung cancer in the cohort (when compared to a reference population of U.S. white males) was due to differences in the smoking habits of the cohort and reference population, which OSHA criticized in its Supplemental Literature Review (Document ID 1486, p. 317; 1711, Attachment 1, p. 5). The ACC stated, “OSHA suggests that the lack of complete smoking data for the cohort is a problem and contends that smoking could not explain the elevated SMR for lung cancer. This criticism, as Dr. Vacek explains, is overstated, and, in any event, does not detract from the study's findings regarding the absence of an association between silica exposure and lung cancer” (Document ID 2307, Attachment A, pp. 46-47; 1804, p. 2).

Vacek et al. (2011) estimated the relative smoking prevalence in the cohort to be 1.35 times that in the reference population; using this estimated relative smoking prevalence, the authors estimated that “the expected number of lung cancer deaths in the cohort after adjusting the reference rates for smoking would be 353, yielding a [non-significant] SMR of 1.02” (Document ID 1486, p. 317). OSHA notes that this method used by Vacek et al. to adjust the SMR for smoking neglects the healthy worker survivor effect (i.e., smokers may leave the workforce sooner than nonsmokers because smoking is a risk factor for poor health). Absent control for the healthy worker survivor effect, smoking would (and perhaps did) become a negative confounder because long duration—high cumulative exposure—workers would tend toward lower smoking attributes. The method used by Vacek et al. is also inconsistent with the frequently cited Axelson (1978) method, which is used to adjust the SMR when the exposed population has a higher percentage of smokers than the reference population (Checkoway et al. 1997, Document ID 0326; Chan et al. 2000, 0983). As a result, Vacek et al. (2011) likely overestimated the confounding effect of smoking in this cohort.

In addition, as previously noted by OSHA, the SMRs for cancers largely attributable to smoking, such as those of the buccal cavity and pharynx (SMR = 1.01), larynx (SMR = 0.99), and esophagus (SMR = 1.15) were not significant in the Vacek et al. study (Document ID 1486, p. 315; 2307, Attachment 6, p. 14). The SMR of 0.94 for bronchitis, emphysema, and asthma also was not significant. If smoking were truly responsible for the highly statistically significant SMR (1.37) observed for lung cancer, the SMRs for these other diseases should be significant as well. OSHA likewise notes that other studies have found that smoking does not have a substantial impact on the association between crystalline silica exposure and lung cancer mortality (e.g., Checkoway et al., 1997, Document ID 0326; Steenland et al., 2001a, 0452, p. 781) and that crystalline silica is a risk factor for lung cancer independent of smoking (Kachuri et al., 2014, Document ID 3907, p. 138; Preller et al., 2010, 4055, p. 657).

OSHA is also concerned about some features of the study design and exposure assessment in Vacek et al. (2011). Regarding the study design, in their nested case-control analyses, Vacek et al. sorted cases into risk sets based on year of birth and year of death, and then matched three controls to each risk set; from the data presented in Table 5 of the study, the actual number of controls per lung cancer case can be calculated as 2.64 (Document ID 1486, p. 316). Vacek et al.'s decision to use such a small number of controls per case was unnecessarily restrictive, as there were additional cohort members who could have been used as controls for the lung cancer deaths. Typically, if the relevant information is available, four or more (or all eligible) controls are used per case to increase study power to detect an association. OSHA notes that Steenland et al. (2001a), in their nested case-control pooled analysis, used 100 controls per case (Document ID 0452, p. 777).

In addition, Vacek et al. stated that for the categorical analysis, cut points on cumulative exposure were based on quintiles of the combined distribution for cases and controls (Document ID 1486, p. 314). Therefore, there should be an approximately equal total number of subjects (cases plus controls) in each group (or quintile). OSHA's examination of Table 5 in the Vacek et al. (2011) study shows that there is an approximately equal distribution of subjects for all endpoints except lung cancer; for example, the silicosis groups each had 43-44 subjects, the NMRD groups each had 125-130 subjects, the kidney cancer groups each had 22-23 subjects, and the kidney disease groups each had 25 subjects. However, the lung cancer groups, ranging from the lowest to the highest exposure, had 325, 232, 297, 241, and 202 subjects (Document ID 1486, p. 316). OSHA could find no explanation for this discrepancy in the text of the Vacek et al. (2011) study, and questions how the lung cancer groups were composed.

With respect to the different job exposure matrices, OSHA has reason to believe that the exposure data reported in the Attfield and Costello study are more accurate than the data Vacek et al. used. OSHA is particularly concerned that Vacek et al.'s pre-1940 exposure estimate of 150 µg/m3 for one job (channel bar operator) was much lower than Attfield and Costello's estimate, from the Davis et al. (1983) matrix, of 1070 µg/m3 (Document ID 1486, p. 313; 0284, p. 131). As NIOSH observed in its post-hearing comments, changing the exposure estimate for channel bar operators could have “major consequences” on the exposure-response analysis, as the job occurred frequently (Document ID 4233, p. 22). NIOSH then pointed out that the Attfield and Costello (2004) exposure estimate for channel bar operators was based on multiple exposure measurements conducted by Davis et al. (1983), whereas Vacek et al. based their exposure estimate “on only three dust measurements” in which “only wet drilling was used. Thus, their study used not only very limited sampling data but also values that were biased towards low levels, since the samples were taken when water was being used to control dust,” a practice that was not typically used for this occupation at the time (Document ID 4233, p. 22). In fact, photographs from Hosey et al. (1957) showed channel bar drilling in 1936 and 1937 with and without dust control; the caption for the photo without dust control states that the “operator in background is barely visible through dust cloud” (Document ID 4233, p. 24, citing 3998, Attachment 14b). As NIOSH explained,

If there is a true [linear] relationship between exposure to silica dust and lung cancer mortality, classifying highly exposed workers incorrectly as low-exposed shifts the elevated risks to the low exposure range. The impact is to spuriously elevate risks at low exposures and lower them at high exposures, resulting in the exposure-response trend being flattened or even obscured. Ultimately, the true relationship may not be evident, or if it is, may be attenuated (Document ID 4233, p. 22, n. 1).

Vacek et al. reported in their study that they conducted a sensitivity analysis that did not change the exposure-response relationship between silica exposure and lung cancer risk, Start Printed Page 16338even when Attfield and Costello's pre-1940 exposure estimates were used for channel bar operators (Document ID 2340, pp. 317-318; 2307, Attachment 6, p. 31). Part of the problem may be the way that channel bar operators were defined by Vacek et al. As noted by NIOSH, “Leyner driller and channel bar operator or driller are synonyms” (Document ID 4233, p. 22, n. 3). Attfield and Costello defined channel bar operators in that way, with a pre-1940 exposure estimate of 1070 µg/m3 (Document ID 0284, p. 131). Vacek et al., on the contrary, assigned channel bar operators to a category called “channel bar (wet)” and assigned a pre-1940 exposure estimate of 150 µg/m3 (Document ID 2307, Attachment 6, Appendix B, pp. 7, 15). They included Leyner drillers under a general category called “driller” with a pre-1940 exposure estimate of 1070 µg/m3 (Document ID 2307, Attachment 6, Appendix B, pp. 7, 15). Included in the Vacek et al. (2009) category of “drillers” were plug drillers (Document ID 2307, Attachment 6, Appendix B, p. 15); OSHA notes that Attfield and Costello used a lower pre-1940 exposure estimate of 650 µg/m3 for plug drillers, as defined by Davis et al. (1983). OSHA believes that Vacek et al. underestimated the exposures of some channel bar operators, and overestimated the exposures of plug drillers, which may have contributed to the lack of association, and that the categorization used by Attfield and Costello, with the synonymous channel bar operators and Leyner drillers in one category, and plug drillers in a separate category, was more appropriate. Thus, even in Vacek et al' s sensitivity analysis, in which they used Attfield and Costello's exposure estimate of 1070 µg/m3 for channel bar operators and drillers, the plug drillers would still have had a higher exposure estimate (1070 µg/m3 versus Attfield and Costello's 650 µg/m3), making the analysis different from that of Attfield and Costello.

For the reasons discussed herein, OSHA has decided not to reject the Attfield and Costello (2004) study in favor of the Vacek et al. (2011) study as a basis for risk assessment. OSHA maintains that it has performed an objective analysis of the Attfield and Costello (2004) and Vacek et al. (2011) studies. OSHA agrees with some of the ACC's criticisms regarding the Agency's initial evaluation of the exposure groupings and confounding in the Vacek et al. (2011) study. OSHA is concerned, however, as discussed above, about several aspects of Vacek et al. (2011), including a potential bias from the healthy worker survivor effect, which was shown to exist in this cohort (see Applebaum et al., 2007, cited in Document ID 2307, Attachment 6, p. 3), as well as about job categorization that may have resulted in exposure misclassification for certain job categories (e.g., the synonymous channel bar operators and Leyner drillers). Despite its concerns with the Vacek et al. study, OSHA acknowledges that comprehensive studies, such as Attfield and Costello (2004) and Vacek et al. (2011), in the Vermont granite industry have shown conflicting results with respect to lung cancer mortality (Document ID 0284; 1486). As discussed earlier, conflicting results are often observed in epidemiological studies due to differences in study designs, analytical methods, exposure assessments, populations, and other factors. In addition, the exposure-response relationship between silica and lung cancer may be easily obscured by bias, as crystalline silica is a comparably weaker carcinogen (i.e., the increase in risk per unit exposure is smaller) than other well-studied, more potent carcinogens such as hexavalent chromium (Steenland et al., 2001, Document ID 0452, p. 781). Although OSHA believes that the Attfield and Costello (2004) study is the most appropriate Vermont granite study to use in its QRA, the Agency notes that, even in the absence of the Attfield and Costello (2004) study, the risk estimates for lung cancer mortality based on other studies still provide substantial evidence that respirable crystalline silica poses a significant risk of serious health conditions to exposed workers.

4. Comments on Specific Studies Relied Upon by OSHA in Its QRA

a. Attfield and Costello (2004)

As stated above, OSHA disagrees with the ACC's contention that Vacek et al. provides a more reliable scientific basis for estimating risk than Attfield and Costello. While it is true that the final risk estimate (54 deaths per 1,000 workers) derived from the Attfield and Costello study for an exposure level of 100 µg/m3 is the highest when compared to the other studies, it is not true that the final risk estimate (22 deaths per 1,000 workers) derived from the Attfield and Costello study is the highest for the final rule's PEL of 50 µg/m3. In fact, it is within the range of risk estimates derived from the ToxaChemica (2004) pooled analysis of 16 to 23 deaths per 1,000 workers at the final PEL. Thus OSHA has decided to retain its reliance on the Attfield and Costello (2004) study and, again, notes that, even without the Attfield and Costello (2004) study, all of the other studies in the Final QRA demonstrate a clearly significant risk of lung cancer mortality (11 to 54 deaths per 1,000 workers) at an exposure level of 100 µg/m3, with a reduced, albeit still significant, risk (5 to 23 deaths per 1,000 workers) at an exposure level of 50 µg/m3 (see Table VI-1 in Section VI, Final Quantitative Risk Assessment and Significance of Risk). Excluding Attfield and Costello (2004), in other words, would not change OSHA's final conclusion regarding the risk of death from lung cancer.

b. Miller and MacCalman (2009)

According to the ACC, OSHA's risk estimates based on the Miller and MacCalman (2009, Document ID 1306) study are “more credible than the others—because [the study] involved a very large cohort and was of higher quality in terms of design, conduct, and detail of exposure measurements,” and also adjusted for smoking histories (Document ID 2307, Attachment A, p. 73). Although the risk estimates generated from the Miller and MacCalman data were the lowest of the lung cancer mortality estimates, the ACC next asserted that they were biased upwards for several reasons. First, the ACC stated that exposure information was lacking for cohort members after the mines closed in the mid-1980's, and quoted OSHA as stating, “Not accounting for this exposure, if there were any, would bias the risk estimates upwards” (Document ID 2307, Attachment A, p. 74 (quoting 1711, p. 289)). OSHA, however, does not believe there to have been additional substantial quartz exposures. As the study authors wrote, “Because of the steep decline of the British coal industry, the opportunities for further extensive coal mine exposure were vanishingly small” (Document ID 1306, p. 11). Thus OSHA believes it to be unlikely that the risk estimates are biased upwards to any meaningful degree based on lack of exposure information at the end of the study period.

The ACC also stated that the unrestricted smoking of cohort members after the closure of the mines would have resulted in risk estimates that were biased upwards (Document ID 2307, Attachment A, p. 74). OSHA has no reason to believe, nor did the ACC submit any evidence in support of its contention, that unrestricted smoking occurred, however, and notes that the authors stated that the period after the mines closed was one of “greater anti-Start Printed Page 16339smoking health promotion campaigns” (Document ID 1306, p. 11).

Finally, the ACC noted that Miller and MacCalman did not adjust significance levels for the multiple comparisons bias with respect to lag selection that Dr. Cox alleged affected their study (Document ID 2307, Attachment A, p. 74). Dr. Cox claimed that trying multiple comparisons of alternative approaches, such as different lag periods, and then selecting a final choice based on the results of these multiple comparisons, leads to a multiple comparisons bias that could result in false-positive associations (Document ID 2307, Attachment 4, p. 28; see Section V.J, Comments and Responses Concerning Biases in Key Studies). He argued that the authors should have reduced the significance level (typically p = 0.05) at which a result is considered to be significant. “Lag” refers to the exclusion of the more recent years of exposure (e.g., 10-year lag, 15-year lag) to account for the fact that diseases like cancer often have a long latency period (i.e., that the cancer may not be detected until years after the initiating exposure, and exposures experienced shortly before detection probably did not contribute to the development of disease). “Lag selection,” therefore, refers to the choice of an appropriate lag period. As addressed later in the Section V.J, Comments and Responses Concerning Biases in Key Studies, OSHA does not necessarily believe such an adjustment of significance levels to be appropriate, based upon the testimony of Mr. Park of NIOSH, nor is it typically performed in the occupational epidemiology literature (Document ID 3579, Tr. 151-152). Similarly, the ACC stated that the confidence intervals are overly narrow because they ignore model uncertainty, and that multiple imputation of uncertain exposure values should have been performed (Document ID 2307, Attachment A, p. 75). OSHA rejects this assertion on the grounds that the authors used detailed exposure estimates that the ACC recognized raised the credibility of the study; the ACC wrote, regarding the study, “it involved a very large cohort and was of higher quality in terms of design, conduct, and detail of exposure measurements” (Document ID 2307, Attachment A, p. 73). Lastly, the ACC argued that an exposure threshold should have been examined (Document ID 2307, Attachment A, p. 75). OSHA discusses at length this issue of thresholds, and the difficulty in ruling them in or out at low exposures, in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases.

In summary, OSHA notes that the ACC has not provided any non-speculative evidence to support its claims that the risk estimates derived from the Miller and MacCalman (2009) study are biased upwards. As stated in the Review of Health Effects Literature and Preliminary QRA, and acknowledged by the ACC (Document ID 2307, p. 73), OSHA believes these risk estimates to be very credible, as the study was based on well-defined union membership rolls with good reporting, had over 17,000 participants with nearly 30 years of follow-up, and had detailed exposure measurements of both dust and quartz, as well as smoking histories (Document ID 1711, pp. 288-289).

c. Steenland (2001a) and ToxaChemica (2004)

OSHA also received several comments on the ToxaChemica (2004, Document ID 0469) analysis, which was based on the Steenland et al. (2001a, Document ID 0452) pooled analysis. First, the ACC claimed that there is significant heterogeneity in the exposure-response coefficients, derived from the individual studies. Because the risk estimates based on these coefficients differ by almost two orders of magnitude, the ACC suggested that these models are misspecified for the data (Document ID 2307, Attachment A, pp. 75-76). Essentially, the ACC claimed that the exposure-response coefficients differ too much among the individual studies, and asserted that it is therefore inappropriate to use the pooled models. Dr. Cox wrote: “Steenland et al. did not address the heterogeneity, but artificially suppressed it by unjustifiably applying a log transformation. This is not a valid statistical approach for exposure estimates with substantial estimation errors” (Document ID 2307, Attachment 4, p. 75). During the public hearing, however, Dr. Steenland explained to OSHA's satisfaction how the data in his study was transformed, using accepted statistical methods. Specifically, referring to his use of a log transformation to address the heterogeneity, Dr. Steenland testified:

[I]t reduces the effect of the very highest exposures being able to drive an exposure-response curve because those exposures are often [skewed] way out—skewed to the right, because occupational exposure data is often log normal. With some very high exposures, they are sort of extreme, and that can drive your exposure-response curve. And you take the log, it pulls them in, and so therefore gives less influence to those high data points. And I think those high data points are often measured with more error (Document ID 3580, Tr. 1265-1266).

OSHA finds this testimony to be persuasive and, therefore, believes that Dr. Steenland's use of a log transformation to address the heterogeneity was appropriate. The log transformation also permits a better model fit when attenuation of the response is observed at high cumulative exposures.

Dr. Morfeld commented that Steenland et al. did not take into account smoking, which could explain the observed excess lung cancer of 20 percent (SMR = 1.2). Dr. Morfeld stated, “Thus, lung cancer excess risks were demonstrated only under rather high occupational exposures to RCS dust, and, even then, an upward bias due to smoking and a necessary intermediate role for silicosis could not be ruled out” (Document ID 2307, Attachment 2, p. 10). Dr. Steenland addressed the concern about a potential smoking bias during his testimony:

We concluded that this positive exposure response was not likely due to different smoking habits between high exposed and low exposed workers. And the reason we did that was twofold. First, workers tend to smoke similar amounts regardless of their exposure level in general. We often worry about comparing workers to the general population because workers tend to smoke more than the general population. But, in internal analyses, we don't have this problem very often. When we have smoking data, we see that it is not related to exposure, so a priori we don't think it is likely to be a strong confounder in internal analyses. Secondly, a number of the studies we used in our pool[ed] cohort had smoking data, either for the whole cohort or partially. And when they took that into account, their results did not change. In fact, they also found that smoking was not related to exposure in their studies, which means that it won't affect the exposure-disease relationship because if it is going to do that, it has to differ between the high exposed and the low exposed, and it generally did not (Document ID 3580, Tr. 1227-1228).

In addition, Brown and Rushton (2009), in their review article submitted to the rulemaking record by Dr. Morfeld, appeared to agree with Dr. Steenland, stating, “This [Steenland et al.] internal analysis removed the possibility of confounding by smoking” (Document ID 3573, Attachment 5, p. 150). Thus, OSHA rejects Dr. Morfeld's assessment that the risk estimates may be biased upwards due to smoking.

The ACC also commented that exposure misclassification due to uncertain exposure estimates in Steenland's pooled cohort could have created the appearance of a monotonic relationship, in which the response Start Printed Page 16340increases with the exposure, even if the true response was not monotonic (Document ID 2307, Attachment A, p. 76). The ACC, along with Dr. Borak (representing the U.S. Chamber of Commerce) and others, likewise cited OSHA's statement from the Review of Health Effects Literature and Preliminary QRA, in which the Agency acknowledged that uncertainty in the exposure estimates that underlie each of the 10 studies in the pooled analysis was likely to represent one of the most important sources of uncertainty in the risk estimates (Document ID 1711, p. 292; 2376, p. 16). Dr. Borak also quoted Mannetje et al. (2002), who developed quantitative exposure data for the pooled analysis, as stating, “While some measurement error certainly occurred in our estimates, a categorical analysis based on broad exposure groups should not be much affected by the resulting level of misclassification” (Document ID 2376, p. 17, quoting 1090, p. 84). From this statement, Dr. Borak concluded that the researchers themselves believed the data were only adequate for “categorical analyses which might lead to qualitative conclusions” (Document ID 2376, p. 17).

OSHA disagrees with Dr. Borak's interpretation of the Mannetje et al. statement, as categorical analyses are typically quantitative in nature, with the data being used to draw quantitative conclusions. However, OSHA recognized the possibility for uncertainty in the exposure estimates, and it is for this reason that OSHA commissioned a quantitative analysis of uncertainty in Steenland's pooled study (ToxaChemica, 2004, Document ID 0469). This analysis suggested that exposure misclassification had little effect on the pooled exposure coefficient (and the variance around that estimate) for the lung cancer risk model (Document ID 1711, pp. 313-314). Given this analysis, OSHA also disagrees with the ACC's statement that “it is virtually certain that substantial exposure estimation error infused the pooled analysis, resulting in exposure misclassification that would create a false appearance of a monotonically increasing exposure-response even where none exists” (Document ID 2307, Attachment A, p. 78). OSHA notes that this statement is not supported with any evidence from the Steenland et al. (2001) study. In addition, as discussed at length in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, exposure estimation error can also bias results towards the null (weaken or obscure the exposure-response relationship) (Document ID 3580, Tr. 1266-67; 3576, Tr. 358-359; 3574, p. 21). Other criticisms from the ACC concerning alleged modeling errors and biases in the Steenland study and the alleged threshold for the health effects of silica exposure are discussed generally in Section V.J, Comments and Responses Concerning Biases in Key Studies, and Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases. Dr. Cox's and Dr. Morfeld's criticisms of the uncertainty analysis performed by Toxachemica are addressed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis. For the reasons stated in those sections, OSHA is unpersuaded by these criticisms.

The ACC concluded:

For all these reasons, the pooled analysis by Steenland et al. (2001) does not yield credible or reliable estimates of silica-related lung cancer risk. But, even if risk estimates based on Steenland et al. (2001) were not so problematic, that study would not demonstrate that reducing the PEL from 0.1 mg/m3 [100 µg/m3] to 0.05 mg/m3 [50 µg/m3] will result in a substantial reduction in the risk of lung cancer (Document ID 2307, Attachment A, p. 81).

The ACC then discussed the ToxaChemica report (2004), which the ACC claimed shows that “under the spline model (which the authors prefer over the log cumulative model because of biological plausibility)” reducing the PEL from 100 µg/m3 to 50 µg/m3 would negligibly reduce the excess risk of lung cancer mortality from 0.017 (17/1,000) to 0.016 (16/1,000), “risk values that are indistinguishable given the overlapping confidence limits of the two estimates” (Document ID 2307, Attachment A, p. 81). In addition, the ACC noted that the excess risk at 150 µg/m3 and 250 µg/m3 in the spline model is the same as the excess risk at 50 µg/m3, while that at 200 µg/m3 is lower. “Estimates of lung cancer risk in the neighborhood of the current general industry PEL are hugely uncertain—with the data suggesting that a greater reduction in lung cancer risk could be achieved by doubling the PEL to 200 µg/m3 than by cutting it in half to a level of 50 µg/m3” (Document ID 2307, Attachment A, pp. 81-82).

OSHA notes that these risk estimates cited by the ACC were the original estimates for the spline model provided to OSHA by ToxaChemica in its 2004 report (Document ID 0469). These are not the risk estimates used by OSHA. Instead, to estimate the risks published in this final rule, the Agency used the exposure-response coefficients from the study in an updated life table analysis using background all-cause mortality and lung cancer mortality rates from 2006 and 2011, respectively. The risk estimates using the 2011 background data are the most updated numbers with which to make the comparisons ACC has suggested. With the 2011 background data, the estimated excess risk is 20 deaths per 1,000 workers at 100 µg/m3, and 16 deaths per 1,000 workers at 50 µg/m3, a reduction of 4 deaths. OSHA's estimated excess risk at 250 µg/m3 is 24 deaths per 1,000 workers, an increase in 8 deaths when compared to 50 µg/m3. Thus it is not the case, as ACC suggested, that increasing the PEL would cause a reduction in lung cancer mortality risk.

In addition, the linear spline model employed by Steenland et al. (2001) was only one of three models used by OSHA to estimate quantitative risks from the pooled analysis. OSHA also used the log-linear model with log cumulative exposure as well as the linear model with log cumulative exposure (see Section VI, Final Quantitative Risk Assessment and Significance of Risk). OSHA notes that all three models indicated a reduction in risk when comparing an exposure level of 100 µg/m3 to 50 µg/m3.

In summary, OSHA disagrees with the ACC's assertion that the Steenland et al. pooled analysis does not yield credible risk estimates for lung cancer mortality. Dr. Morfeld's assertion that the risk estimates were biased upwards due to smoking is quite unlikely to be true, given that the study was an internal (worker to worker) analysis. The ACC's claim that exposure estimation error resulted in false exposure-response relationships was not supported by any actual data; as discussed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, exposure estimation error can also bias results towards the null (weaken or obscure the exposure-response relationship) (Document ID 3580, Tr. 1266-67; 3576, Tr. 358-359; 3574, p. 21). For these reasons, OSHA rejects the ACC's claims that the Steenland study of lung cancer mortality does not yield credible risk estimates. Rather, based upon its review, OSHA believes this pooled analysis to be of high quality. As Dr. Steenland testified during the informal public hearings, this pooled analysis, with its more than 60,000 workers and 1,000 lung cancer deaths, involved “a rich dataset with high statistical power to see anything, if there was anything to see” (Document ID 3580, Tr. 1227). In fact, OSHA believes the Steenland et al. (2001a) study to be among the best available studies in the peer-reviewed literature on the topic of Start Printed Page 16341silica exposure and its relationship to lung cancer mortality.

d. Rice et al. (2001)

The ACC also commented on the Rice et al. (2001, Document ID 1118) study of diatomaceous earth workers, which found a significant risk of lung cancer mortality that increased with cumulative silica exposure in a cohort of diatomaceous earth workers. The ACC claimed that it had a high likelihood of exposure misclassification. Dr. Cox contended that the practice of “[a]ssigning each worker a single estimated cumulative exposure based on estimated mean values produces biased results and artificially narrow confidence intervals (and hence excess false-positive associations)” (Document ID 2307, Attachment 4, p. 76). OSHA notes that Rice et al. (2001) described the exposure estimation procedure in their paper. There were more than 6,000 measurements of dust exposure taken from 1948-1988; particle count data were converted to gravimetric data using linear regression modeling. Cumulative exposures to respirable crystalline silica were then estimated for each worker using detailed employment records (Document ID 1118, p. 39). OSHA concludes it is highly unlikely that the exposure estimates are biased to such an extent, as Dr. Cox suggests, that they would produce false-positive associations.

The ACC also noted that the mean crystalline silica exposure in the diatomaceous earth worker cohort was 290 μg/m3, approximately three times the former PEL for general industry (Document ID 2307, Attachment A, p. 83). OSHA, however, believes that the cumulative respirable crystalline silica dust concentration is the metric of concern here, as that is what was used in the regression models. The mean cumulative respirable crystalline silica dust concentration in the study was 2.16 mg/m3-yrs, which is a very realistic cumulative exposure for many workers (Document ID 1118, p. 39).

The ACC also stated that the results of the Rice study were confounded by smoking and possibly asbestos exposure (Document ID 2307, Attachment A, p. 83). OSHA previously addressed the possible confounding in this cohort in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 139-143). Rice et al. (2001) used the same cohort originally reported on by Checkoway et al. (1993, Document ID 0324; 1996, 0325; 1997, 0326). The Rice study discussed the smoking confounding analysis performed by Checkoway et al. (1997), in which the Axelson method (1978) was used to make a worst case estimate (assuming 20 times greater lung cancer risk in smokers compared to non-smokers) and indirectly adjust the relative risk (RR) estimates for lung cancer for differences in smoking rates (Document ID 1118, pp. 40-41). With exposures in the Checkoway study lagged 15 years to account for the latency period, the worst case effect was to reduce the RR for lung cancer in the highest exposure group from 2.15 to 1.67. Checkoway et al. concluded that the association between respirable silica exposure and lung cancer was unlikely to be confounded by cigarette exposure (Document ID 0326, pp. 684, 687). Regarding confounding by asbestos exposure, Rice et al. (2001) stated:

Checkoway et al. found no evidence that exposure to asbestos accounted for the observed association between mortality from lung cancer and cumulative exposure to silica. Our analyses of their data also found no evidence of confounding by asbestos in the Poisson regression or Cox's proportional hazards models regardless of lag period; therefore, exposure to asbestos was not included in the models presented in this paper (Document ID 1118, p. 41).

Based upon these analyses, OSHA rejects the ACC's unsupported assertion that the results of Rice et al. (2001) were confounded by smoking and asbestos exposure.

Lastly, Dr. Cox asserted that there were several biases in Rice et al. (2001), including multiple-testing bias from testing multiple lag periods, exposure groupings, and model forms; model specification bias; and a lack of model diagnostics (Document ID 2307, Attachment 4, pp. 63-64, 77). OSHA addressed these issues generally in Section V.J, Comments and Responses Concerning Biases in Key Studies, and rejects these assertions for the same reasons. OSHA also discussed regression diagnostics at length in the same section. In summary, despite the criticisms directed at the Rice et al. study by the ACC, OSHA continues to believe that the quantitative exposure-response analysis by Rice et al. (2001) is of high quality and appropriate for inclusion in the QRA (Document ID 1711, p. 143).

e. Hughes et al. (2001)

The ACC, through the comments of Dr. Cox, presented a similar critique of the study of North American industrial sand workers by Hughes et al. (2001, Document ID 1060). This study found a statistically significant association (increased odds ratios) between lung cancer mortality and cumulative silica exposure as well as average silica concentration (Document ID 1060). In this study, according to Dr. Cox, “The selected model form guarantees a monotonic exposure-response relation, independent of the data. Model uncertainty and errors in exposure estimates have both been ignored, so the slope estimate from Hughes et al. (2001), as well as the resulting excess risk estimates, are likely to be biased and erroneous” (Document ID 2307, Attachment 4, p. 85). The ACC also noted that this cohort had incomplete smoking information, with the proportion of “ever smokers” significantly higher in cases than in controls. In addition, the ACC asserted that asbestos exposure may have also occurred, as three death certificates listed mesothelioma as the cause of death (Document ID 2307, Attachment A, pp. 85-86).

OSHA discussed the Hughes et al. (2001, Document ID 1060) study in its Review of Health Effects Literature and Preliminary QRA, highlighting as strengths the individual job, exposure, and smoking histories that were available (Document ID 1711, p. 285). Exposure levels over time were estimated via a job exposure matrix constructed by Rando et al. (2001, Document ID 0415) utilizing substantial exposure data, including 14,249 respirable dust and silica samples taken from 1974 to 1998 in nine plants (Document ID 1711, pp. 88, 124-128; 1060, 202). Smoking data were collected from medical records supplemented by information from next of kin or living subjects for 91 percent of cases and controls (Document ID 1060, p. 202). OSHA believes these smoking histories allowed the authors to adequately control for confounding by smoking in their analyses. Regarding the three death certificates listing mesothelioma, McDonald et al. (2001) explained that two were for workers not included in the case/control study because they were hired at or after age 40 with less than 10 years of work time; the third was for a worker hired at age 19 who then accumulated 32 years of experience in maintenance jobs (Document ID 1091, p. 195). As such, OSHA does not believe it likely that asbestos exposure was a large source of confounding in typical industrial sand operations in this study. OSHA also notes that the positive findings of this study were consistent with those of other studies of workers in this cohort, including Steenland and Sanderson (2001, Document ID 0455) and McDonald et al. (2005, Document ID 1092).

The ACC also noted that there was no consistent correlation in Hughes et al. (2001) between employment duration Start Printed Page 16342and lung cancer risk (Document ID 2307, Attachment A, p. 86), with Dr. Cox suggesting that model specification error was to blame (Document ID 2307, Attachment 4, p. 86). OSHA believes that cumulative exposure is a more appropriate metric for determining risk than is duration of exposure because the cumulative exposure metric considers both the duration and intensity of exposure. For example, some workers may have been employed for a very long duration with low exposures, whereas others may have been employed for a short duration but with high exposures; both groups could have similar cumulative exposures.

In summary, OSHA considers the Hughes et al. (2001) study to be of high enough quality to provide risk estimates for excess lung cancer from silica exposure, as the study is unlikely to be substantially confounded. For these reasons, the Agency finds the assertion that the risk estimates based on this study are erroneous to be unconvincing.

Overall, regarding all of the studies upon which OSHA relied in its Preliminary QRA, the ACC concluded, “In sum, none of the studies on which OSHA relies is inconsistent with a concentration threshold above 100 μg/m3 for any risk of silica-related lung cancer; none demonstrates an increased lung cancer risk in the absence of silicosis; and none provides a sound basis for estimating lung cancer risks at RCS [respirable crystalline silica] exposure levels of 100 μg/m3 and below” (Document ID 2307, Attachment A, p. 87).

OSHA is not persuaded that the evidence presented by the ACC supports these conclusions. On the contrary, as OSHA discussed in the Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases, demonstrating the absence of a threshold is not a feasible scientific pursuit, and some models produce threshold estimates well below the PELs. Similarly, the ACC has not put forward any study that has proven that silicosis must be a precursor for lung cancer and, as discussed in Section V.H, Mechanisms of Silica-Induced Adverse Health Effects, some studies have shown genotoxic mechanisms by which exposure to crystalline silica may lead to lung cancer. The strong epidemiological evidence for carcinogenicity, supported by evidence from experimental animal and mechanistic studies, allowed IARC to conclude on multiple occasions that respirable crystalline silica is a Group I carcinogen. OSHA places great weight on this conclusion given IARC's authority and standing in the international scientific community. In addition, all of the lung cancer studies relied upon by OSHA used models that allow for the estimation of lung cancer risks at crystalline silica exposure levels of 100 μg/m3 and below. OSHA believes these studies (Steenland et al., 2001a, Document ID 0452, as re-analyzed in ToxaChemica, 2004, 0469; Rice et al., 2001, 1118; Attfield and Costello, 2004, 0284; Hughes et al., 2001, 1060; and Miller and MacCalman, 2009, 1306) are of high quality and contain well-supported findings. Thus, OSHA continues to rely upon these studies for deriving quantitative risk estimates in its QRA and continues to believe that workers exposed to respirable crystalline silica at levels at or near the previous and new PELs are faced with a significant risk of dying from lung cancer. As such, the Agency believes it would be irresponsible as a scientific matter, and inconsistent with its statutory obligations to issue standards based on the best available evidence after conducting an extensive rulemaking, to retain the regulatory status quo.

G. Comments and Responses Concerning Renal Disease Mortality

OSHA estimated quantitative risks for renal disease mortality (Document ID 1711, pp. 314-316) using data from a pooled analysis of renal disease, conducted by Steenland et al. (2002a, Document ID 0448). As illustrated in Table VI-1, the lifetime renal disease mortality risk estimate for 45 years of exposure to the previous general industry PEL (100 μg/m3 respirable crystalline silica) is 39 deaths per 1,000 workers. However, for the final PEL (50 μg/m3), it is 32 deaths per 1,000 workers. Although OSHA acknowledges that there are considerably less data for renal disease mortality, and thus the risk findings based on them are less robust than those for silicosis, lung cancer, and non-malignant respiratory disease (NMRD) mortality, the Agency believes the renal disease risk findings are based on credible data. Indeed, the Steenland et al. pooled analysis had a large number of workers from three cohorts with sufficient exposure data, and exposure matrices for the three cohorts had been used in previous studies that showed positive exposure-response trends for silicosis morbidity or mortality, thus tending to validate the underlying exposure and work history data (see Document ID 1711, pp. 215-216). Nevertheless, OSHA received comments that were critical of its risk estimates for renal disease mortality. Based upon its review of the best available evidence, OSHA finds that these comments do not alter its overall conclusions on renal disease mortality. In addition, OSHA notes that even if the risk of renal disease mortality is discounted, there would remain clearly significant risks of lung cancer mortality, silicosis and NMRD mortality, and silicosis morbidity, with more robust risk estimates based upon a larger amount of data from numerous studies (see Table VI-1).

OSHA received several comments from the ACC regarding the Agency's quantitative risk estimates for renal disease mortality. Specifically, the ACC argued that: (1) The pooled study (Steenland et al., 2002a, Document ID 0448) that OSHA relied upon did not provide sufficient data to estimate quantitative risks; (2) the individual studies included in the pooled study had several limitations; and (3) most epidemiological studies have not demonstrated a statistically significant association between silica exposure and renal disease mortality (Document ID 2307, Attachment A, pp. 139-157; 4209, pp. 92-96). As explained below, and as stated above, although the Agency acknowledges there is greater uncertainty in the risk estimates related to renal disease than other silica-related diseases, the best available evidence is of sufficient quality to quantify the risk of renal disease in the final risk assessment.

1. Pooled Study

Some commenters expressed concern about the Steenland et al. (2002a, Document ID 0448) pooled study of renal disease mortality, which OSHA and its contractor, ToxaChemica, used to calculate quantitative risk estimates. Specifically, the ACC questioned why the analysis only used three studies (Homestake, North Dakota gold miners, Steenland and Brown, 1995a, Document ID 0450; U.S. industrial sand workers, Steenland et al., 2001b, Document ID 0456; Vermont granite workers, Costello and Graham, 1988, Document ID 0991) out of the ten originally used in the pooled study of lung cancer mortality (Steenland et al., 2001a, Document ID 0452). Peter Morfeld, Dr. rer. medic., representing the ACC, wrote in his written testimony that although Steenland et al. (2002a, Document ID 0448) indicated that the three studies were selected because they were the only ones to have information on multiple cause mortality, all 10 studies had information on renal disease as an underlying cause of death (Document ID 2308, Attachment 4, pp. 24-25). Since ToxaChemica focused on underlying cause results in their discussion, Dr. Morfeld argued that not having used all Start Printed Page 1634310 studies in the pooled analysis “raises a suspicion of study selection bias” (Document ID 2308, Attachment 4, pp. 24-25).

OSHA finds this assertion of study selection bias by the ACC and Dr. Morfeld to be unpersuasive because Steenland et al.'s explanation (2002a) for including only three studies in the pooled analysis was sound. The authors reported in their pooled study that both underlying cause and multiple cause mortality were available for only three cohorts of silica-exposed workers, and “multiple cause (any mention on the death certificate) was of particular interest because renal disease is often listed on death certificates without being the underlying cause” (Document ID 0448, p. 5). The authors likewise cited a study (Steenland et al., 1992), indicating that the ratio of chronic renal disease mortality shown anywhere on a U.S. death certificate versus being shown as an underlying cause is 4.75 (Document ID 0453, Table 2, pp. 860-861). Indeed, in their pooled analysis of renal disease mortality, Steenland et al. noted that there were 51 renal disease deaths when using underlying cause, but 204 when using multiple cause mortality (Document ID 0448, p. 5). As renal disease is a serious disabling disease, the use of multiple cause mortality gives a much better sense of the burden of excess disease than does the use of underlying cause of death as an endpoint. As such, Steenland et al. calculated odds ratios by quartile of cumulative silica exposure for renal disease in a nested case-control analysis that considered any mention of renal disease on the death certificate as well as underlying cause. For multiple-cause mortality, the exposure-response trend was statistically significant for both cumulative exposure (p = 0.004) and log cumulative exposure (p = 0.0002); whereas for underlying cause mortality, the trend was statistically significant only for log cumulative exposure (p = 0.03) (Document ID 1711, p. 315). Thus, OSHA believes that Steenland et al. (2002a, Document ID 0448) were justified in including only the three cohorts with all-cause mortality in their pooled analysis.

Concern was also expressed about the model selection in the pooled analysis. Dr. Morfeld noted that a statistically significant association between exposure to crystalline silica and renal disease mortality was only found in the underlying cause analysis in which the model was logged (p = 0.03) (Document ID 2308, Attachment 4, p. 25). Dr. Morfeld commented, “The authors stated that the log-model fit better, but evidence was not given (e.g., information criteria), and it is unclear whether the results are robust to other transformations” (Document ID 2308, Attachment 4, p. 25).

OSHA disagrees with this criticism because a log transformation of the cumulative exposure metric is reasonable, given that exposure variables are often lognormally distributed in epidemiological studies, as discussed in Section V.J, Comments and Responses Concerning Biases in Key Studies. Also, while it is true that Steenland et al. (2002a) only found a statistically significant association in the continuous underlying cause analysis when the cumulative exposure metric was logged (p = 0.03), OSHA notes that the authors also found a statistically significant association in the highest quartile of unlogged cumulative silica exposure (1.67 + mg/m3-yr) in the categorical underlying cause analysis (95% confidence interval: 1.31-11.76) (Document ID 0448, Table 2, p. 7). Thus, for the highest cumulative exposures, there was a significant association with renal disease mortality even without a log transformation of the exposure metric. Dr. Morfeld also failed to mention that Steenland et al. (2002a) found statistically significant associations in the continuous analyses (for both untransformed and log-transformed cumulative exposure) using any mention of renal disease on the death certificate, which adds weight to the study's findings that exposure to respirable crystalline silica is associated with renal disease mortality (Document ID 0448, Table 2, p. 7). In light of this, OSHA concludes that Dr. Morfeld's criticism of the pooled analysis is without merit.

The ACC also noted that the authors of this study, Drs. Kyle Steenland and Scott Bartell, acknowledged the limitations of the data in their 2004 ToxaChemica report to OSHA. Specifically, in reference to the 51 renal deaths (underlying cause) and 23 renal cases in the pooled study, Drs. Steenland and Bartell wrote, “This amount of data is insufficient to provide robust estimates of risk” (Document ID 2307, Attachment A, p. 139, citing 0469, p. 27). Given this acknowledgement, the ACC concluded that OSHA's inclusion of the renal disease mortality risk estimates in the significant risk determination and calculation of expected benefits was speculative (Document ID 2307, Attachment A, pp. 139-140). During the hearing, Dr. Steenland further explained, “I think there is pretty good evidence that silica causes renal disease. I just think that there is not as big a database as there is for lung cancer and silicosis. And so there is more uncertainty” (Document ID 3580, Tr. 1245). OSHA agrees with Dr. Steenland and acknowledges, as it did in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711, p. 357), that its quantitative risk estimates for renal disease mortality have more uncertainty and are less robust than those for the other health effects examined (i.e., lung cancer mortality, silicosis and NMRD mortality, and silicosis morbidity). However, OSHA disagrees with the ACC's suggestion that the Agency's renal disease risk estimates are “rank speculation” (Document ID 4209, pp. 95-96), as these estimates are based on the best available evidence in the form of a published, peer-reviewed pooled analysis (Steenland et al. 2002a, Document ID 0448) that uses sound epidemiological and statistical methods. Thus, OSHA believes that it is appropriate to present the risk estimates along with the associated uncertainty estimate (e.g., 95% confidence intervals) (see Document ID 1711, p. 316).

2. Individual Studies in the Pooled Study

The ACC also identified limitations in each of the three epidemiological studies included in the Steenland et al. (2002a, Document ID 0448) pooled study. First, with respect to the Steenland and Brown (1995a, Document ID 0450) study of North Dakota gold miners, the ACC noted there was a significantly elevated standardized mortality ratio (SMR) for chronic renal disease only in the men hired prior to 1930. It noted that there were no silica exposure measurement data available for this early time period, such that Steenland and Brown (1995a, Document ID 0450) instead estimated a median exposure (150 μg/m3) that was seven times higher for men hired prior to 1930, versus men hired after 1950 (20 μg/m3) (Document ID 2307, Attachment A, p. 147). The ACC maintained that these exposure estimates were likely to be understated and not credible, while also suggesting “the existence of an average exposure threshold ≥150 μg/m3 for any risk of silica-related renal disease mortality” (Document ID 2307, Attachment A, p. 147).

OSHA finds the ACC's suggestion of a threshold to be unpersuasive, as the ACC provided no analysis to indicate a threshold in this study. OSHA addresses the Steenland and Brown (1995a, Document ID 0450) exposure assessment in Section V.D, Comments and Responses Concerning Silicosis and Non-Malignant Respiratory Disease Mortality and Morbidity. The ACC also Start Printed Page 16344ignored the alternative explanation, that elevated chronic renal disease mortality may have only been seen in the workers hired prior to 1930 because they had a higher cumulative exposure than workers hired later, not because there was necessarily a threshold.

The ACC had a similar criticism of the Steenland et al. (2001b, Document ID 0456) study of North American industrial sand workers. The ACC posited that the exposure estimates were highly uncertain and likely to be understated (Document ID 2307, Attachment A, p. 149). The ACC noted that these exposure estimates, developed by Sanderson et al. (2000, Document ID 0429), were considerably lower than those developed by Rando et al. (2001, Document ID 0415) for another study of North American industrial sand workers (Document ID 2307, Attachment A, p. 149). After discussing several differences between these two exposure assessments, the ACC pointed to OSHA's discussion in the lung cancer section of the preamble to the Proposed Rule (78 FR at 56302) in which the Agency acknowledged that McDonald et al. (2001, Document ID 1091), Hughes et al. (2001, Document ID 1060) and Rando et al. (2001, Document ID 0415) had access to smoking histories, plant records, and exposure measurements that allowed for the development of a job exposure matrix, while Steenland and Sanderson (2001, Document ID 0455) had limited access to plant facilities, less detailed historic exposure data, and used MSHA enforcement records for estimates of recent exposure (Document ID 2307, Attachment A, pp. 149-151). The ACC then noted that the McDonald et al. study (2005, Document ID 1092), using the Rando et al. (2001, Document ID 0415) exposure assessment, found no association between end-stage renal disease or renal cancer and cumulative silica exposure (Document ID 2307, Attachment A, pp. 149, 152).

The ACC also noted that, based on underlying cause of death, the SMR for acute renal death in the Steenland et al. (2001b, Document ID 0456) study was not significant (95% confidence interval: 0.70-9.86), and the SMR for chronic renal disease was barely significant (95% confidence interval: 1.06-4.08) (Document ID 2307, Attachment A, p. 151). In light of this, the ACC maintained that Steenland et al. based their exposure-response analyses on multiple-cause mortality data, using all deaths with any mention of renal disease on the death certificate even if it was not listed as the underlying cause. The ACC asserted that “only the underlying cause data involve actual deaths from renal disease” (Document ID 2307, Attachment A, p. 152).

OSHA does not find this criticism persuasive. For regulatory purposes, multiple-cause mortality data is, if anything, more relevant because renal disease constitutes the type of material impairment of health that the Agency is authorized to protect against through regulation regardless of whether it is determined to be the underlying cause of a worker's death. Moreover, the discrepancy in the renal disease mortality findings is a moot point, as only the model in the pooled study with renal disease as an underlying cause was used to estimate risks in the Preliminary QRA (Document ID 1711, p. 316). In any event, OSHA notes an important difference between the Steenland et al. study (2001b, Document ID 0456) and the McDonald study (2005, Document ID 1092): They did not look at the same cohort of North American industrial sand workers. Steenland et al. (2001b) examined a cohort of 4,626 workers from 18 plants; the average year of first employment was 1967, with follow-up through 1996 (Document ID 0456, pp. 406-408). McDonald et al. (2005) examined a cohort of 2,452 workers employed between 1940 and 1979 at eight plants, with follow-up through 2000 (Document ID 1092, p. 368). Although there was overlap of about six plants in the studies (Document ID 1711, p. 127), these were clearly two fairly different cohorts of industrial sand workers. These differences in the cohorts might explain the discrepancy in the studies' results. In addition, OSHA notes that McDonald et al. (2005, Document ID 1092) observed statistically significant excess mortality from nephritis/nephrosis in their study that was not explained by the findings of their silica exposure-response analyses (Document ID 1092, p. 369).

The ACC further argued that the Steenland et al. (2002a, Document ID 0448) pooled study is inferior to the Vacek et al. (2011, Document ID 2340) study of Vermont granite workers, which found no association between cumulative silica exposure and mortality from either kidney cancer or non-malignant kidney disease and which it contended has better mortality and exposure data (Document ID 2307, Attachment A, p. 154) (citing Vacek et al. (2011, Document ID 2340). In particular, it argued that the Vacek et al. study is more reliable for this purpose than the unpublished Attfield and Costello data (2004, Document ID 0285) on Vermont granite workers, which Steenland et al. relied on in finding an association between silica exposure and renal disease.

OSHA notes that Steenland et al. acknowledged in their pooled study that that unpublished data had not undergone peer review (Document ID 0448, p. 5). Despite this limitation, OSHA is also unpersuaded that the Vacek et al. study, although it observed no increased kidney disease mortality (Document ID 2340, Table 3, p. 315), negates Steenland et al.'s overall conclusions. OSHA discussed several substantial differences between these two studies in Section V.F, Comments and Responses Concerning Lung Cancer Mortality.

3. Additional Studies

The ACC also submitted to the record several additional studies that did not show a statistically significant association between exposure to crystalline silica and renal disease mortality. These included the aforementioned studies by McDonald et al. (2005, Document ID 1092) and Vacek et al. (2011, Document ID 2340), as well as studies by Davis et al. (1983, Document ID 0999), Koskela et al. (1987, Document ID 0363), Cherry et al. (2012, article included in Document ID 2340), Birk et al. (2009, Document ID 1468), Mundt et al. (2011, Document ID 1478), Steenland et al. (2002b, Document ID 0454), Rosenman et al. (2000, Document ID 1120), and Calvert et al. (2003, Document ID 0309) (Document ID 2307, Attachment A, pp. 140-145). In light of its assertions on the limitations of the three studies in the pooled analysis, and because the three studies “run counter to a larger number of studies in which a causal association between silica exposure and renal disease was not found,” the ACC concluded that “the three studies relied on by OSHA do not provide a reliable or supportable basis for projecting any risk of renal disease mortality from silica exposure” (Document ID 4209, p. 94). Similarly, the AFS argued that renal disease was only “found in a couple of selected studies and not observed in most others,” including no foundry studies (Document ID 2379, Attachment 1, pp. 1-3).

In light of the analysis contained in the Review of Health Effects Literature and Preliminary QRA, and OSHA's confirmation of its preliminary findings through examination of the record, OSHA finds these claims to be lacking in merit (Document ID 1711, pp. 211-229). In the Review of Health Effects Literature and Preliminary QRA, OSHA presented a comprehensive analysis of several studies that showed an association between crystalline silica Start Printed Page 16345and renal disease, as well as discussing other studies that did not (Document ID 1711, pp. 211-229). Based upon its overall analysis of the literature, including the negative studies, OSHA concluded that there was substantial evidence suggesting an association between exposure to crystalline silica and increased risks of renal disease. This conclusion was supported by a number of case reports and epidemiological studies that found statistically significant associations between occupational exposure to silica dust and chronic renal disease (Calvert et al., 1997, Document ID 0976), subclinical renal changes (Ng et al., 1992c, Document ID 0386), end-stage renal disease morbidity (Steenland et al., 1990, Document ID 1125), end-stage renal disease incidence (Steenland et al. 2001b, Document ID 0456), chronic renal disease mortality (Steenland et al., 2002a, 0448), and granulomatosis with polyangitis (Nuyts et al., 1995, Document ID 0397). In other findings, silica-exposed individuals, both with and without silicosis, had an increased prevalence of abnormal renal function (Hotz et al., 1995, Document ID 0361), and renal effects were reported to persist after cessation of silica exposure (Ng et al., 1992c, Document ID 0386). While the mechanism of causation is presently unknown, possible mechanisms suggested for silica-induced renal disease included a direct toxic effect on the kidney, deposition in the kidney of immune complexes (IgA) following silica-related pulmonary inflammation, or an autoimmune mechanism (Calvert et al., 1997, Document ID 0976; Gregorini et al., 1993, 1032).

From this review of the studies on renal disease, OSHA concluded that there were considerably less data, and thus the findings based on them were less robust, than the data available for silicosis and NMRD mortality, lung cancer mortality, or silicosis morbidity. Nevertheless, OSHA concluded that the Steenland et al. (2002a, Document ID 0448) pooled study had a large number of workers and validated exposure information, such that it was sufficient to provide useful estimates of risk of renal disease mortality. With regard to the additional negative studies presented by the ACC, OSHA notes that it discussed the Birk et al. (2009, Document ID 1468) and Mundt et al. (2011, Document ID 1478) studies in the Supplemental Literature Review of the Review of Health Effects Literature and Preliminary QRA, noting the short follow-up period as a limitation, which makes it unlikely to observe the presence of renal disease (Document ID 1711, Supplement, pp. 6-12). OSHA likewise discussed the Vacek et al. (2011, Document ID 2340) study earlier in this section, and notes that Cherry et al. reported a statistically significant excess of non-malignant renal disease mortality in the cohort for the period 1985-2008, with an unexplained cause (2012, p. 151, article included in Document ID 2340). Although these latter two studies did not find a significant association between silica exposure and renal disease mortality, OSHA does not believe that they substantially change its conclusions on renal disease mortality from the Preliminary QRA, given the number of positive studies presented and the limitations of those two studies.

Thus, OSHA recognizes that the renal risk estimates are less robust and have more uncertainty than those for the other health endpoints for which there is a stronger case for causality (i.e., lung cancer mortality, silicosis and NMRD mortality, and silicosis morbidity). But, for the reasons stated above, OSHA believes that the evidence supporting causality regarding renal risk outweighs the evidence casting doubt on that conclusion. Scientific certainty is not the legal standard under which OSHA acts. OSHA is setting the standard based upon the clearly significant risks of lung cancer mortality, silicosis and NMRD mortality, silicosis morbidity, and renal disease mortality at the previous PELs; even if the risk of renal disease mortality is discounted, the conclusion would not change that regulation is needed to reduce the significant risk of material impairment of health (see Society of the Plastics Industry, Inc. v. OSHA, 509 F.2d 1301, 1308 (2d Cir. 1975)).

H. Mechanisms of Silica-Induced Adverse Health Effects

In this section, OSHA describes the mechanisms by which silica exposure may cause silica-related health effects, and responds to comments criticizing the Agency's analysis on this topic. In the proposal as well as this final rule, OSHA relied principally on epidemiological studies to establish the adverse health effects of silica exposure. The Agency also, however, reviewed animal studies (in vivo and in vitro) as well as in vitro human studies that provide information about the mechanisms by which respirable crystalline silica causes such effects, particularly silicosis and lung cancer. OSHA's review of this material can be found in the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (QRA), which provided background and support for the proposed rule (Document ID 1711, pp. 229-261).

As described in the Review of Health Effects Literature, OSHA performed an extensive evaluation of the scientific literature pertaining to inhalation of respirable crystalline silica (Document ID 1711, pp. 7-265). Due to the lack of evidence of health hazards from dermal or oral exposure, the Agency focused solely on the studies addressing the inhalation hazards of respirable crystalline silica. OSHA determined, based on the best available scientific information, that several cellular events, such as cytotoxicity (i.e., cellular damage), oxidative stress, genotoxicity (i.e., damage to cellular DNA), cellular proliferation, and inflammation can contribute to a range of neoplastic (i.e., tumor-forming) and non-neoplastic health effects in the lung. While the exact mechanisms have yet to be fully elucidated, they are likely initiated by damage to lung cells from interaction directly with the silica particle itself or through silica particle activation of alveolar macrophages following phagocytosis (i.e., engulfing particulate matter in the lung for the purpose of removing or destroying foreign particles). The crystalline structure and unusually reactive surface properties of the silica particle appear to cause the early cellular effects. Silicosis and lung cancer share common features that arise from these early cellular interactions but OSHA, in its Review of Health Effects Literature and Preliminary QRA, “preliminarily conclude[d] that available animal and in vitro studies have not conclusively demonstrated that silicosis is a prerequisite for lung cancer in silica-exposed individuals” (Document ID 1711, p. 259). Although the health effects associated with inhalation of respirable crystalline silica are seen primarily in the lung, other observed health effects include kidney and immune dysfunctions.

Below, OSHA reviews the record evidence and responds to comments it received on the mechanisms underlying respirable crystalline silica-induced lung cancer and silicosis. The Agency also addresses comments regarding the use of animal studies to characterize adverse health effects in humans caused by exposure to respirable crystalline silica.

1. Mechanisms for Silica-Related Health Effects

In 2012, IARC reevaluated the available scientific information regarding respirable crystalline silica and lung cancer and reaffirmed that crystalline silica is carcinogenic to Start Printed Page 16346humans, i.e., a Group 1 carcinogen (Document ID 1473, p. 396). OSHA's review of all the evidence now in the rulemaking record, including the results of IARC's reevaluation, indicates that silica may lead to increased risk of lung cancer in humans by a multistage process that involves a combination of genotoxic (i.e., causing damage to cellular DNA) and non-genotoxic (i.e., not involving damage to DNA) mechanisms. Respirable crystalline silica may cause genotoxicity as a result of reactive oxygen species (ROS) produced by activated alveolar macrophages and other lung cells exposed to crystalline silica particles during phagocytosis. ROS have been shown to damage DNA in human lung cells in vitro (see Document ID 1711, pp. 236-239). This genotoxic mechanism is believed to contribute to neoplastic transformation and silica-induced carcinogenesis. ROS is not only produced during the early cellular interaction with crystalline silica but also produced by PMNs (polymorphonuclear leukocytes) and lymphocytes recruited during the inflammatory response to crystalline silica. In addition to genotoxicity contributed by ROS, it is also plausible that reactive molecules on the surface of crystalline silica itself may bind directly to DNA and result in genotoxicity (Document ID 1711, p. 236). It should be noted that the mechanistic evidence summarized above suggests that crystalline silica may cause early genotoxic events that are independent of the advanced chronic inflammatory response and silicosis (Document ID 1473, pp. 391-392).

Non-genotoxic mechanisms are also believed to contribute to the lung cancer caused by respirable crystalline silica. Phagocytic activation as well as silica-induced cytotoxicity trigger release of the aforementioned ROS, cytokines (e.g., TNFα), and growth factors (see Document ID 1711, pp. 233-235). These agents are able to cause cellular proliferation, loss of cell cycle regulation, activation of oncogenes (genes that have the potential to cause cancer), and inhibition of tumor suppressor genes, all of which are non-genotoxic mechanisms known to promote the carcinogenic process. It is plausible that these mechanisms may be involved in silica-induced tumorigenesis. The biopersistence and cytotoxic nature of crystalline silica leads to a cycle of cell death (i.e., cytotoxicity), activation of alveolar macrophages, recruitment of inflammatory cells (e.g., PMNs, leukocytes), and continual release of the non-genotoxic mediators (i.e., ROS, cytokines) able to promote carcinogenesis. The non-genotoxic mechanisms caused by early cellular responses (e.g., phagocytic activation, cytotoxicity) are regarded, along with genotoxicity, as important potential pathways that lead to the development of tumors (Document ID 1711, pp. 232-239; 1473, pp. 394-396).

The same non-genotoxic processes that may cause lung cancer from respirable crystalline silica exposure are also believed to lead to chronic inflammation, lung scarring, fibrotic lesions, and eventually silicosis. This would occur when inflammatory cells move from the alveolar space through the interstitium of the lung as part of the clearance process. In the interstitium, respirable crystalline silica-laden cells—macrophages and neutrophils—release ROS and TNF-α, as well as other cytokines, stimulating the proliferation of fibroblasts (i.e., the major lung cell type in silicosis). Proliferating fibroblasts deposit collagen and connective tissue, inducing the typical scarring that is observed with silicosis. Alternatively, alveolar epithelial cells containing respirable crystalline silica die and may be replaced by fibroblasts due to necrosis of the epithelium. This allows for uninhibited growth of fibroblasts and formation of connective tissue where scarring proliferates (i.e., silicosis). As scarring increases, there is a reduction in lung elasticity concomitant with a reduction of the lung surface area capable of gas exchange, thus reducing pulmonary function and making breathing more difficult (Document ID 0314; 0315). It should be noted that silicosis involves many of the same mechanisms that occur during the early cellular interaction with crystalline silica. Therefore, it is plausible that development of silicosis may also potentially contribute to silica-induced lung cancer. However, the relative contributions of silicosis-dependent and silicosis-independent pathways are not known.

Although it is clear that exposure to respirable crystalline silica increases the risk of lung cancer in exposed workers (see Section VI, Final Quantitative Risk Assessment and Significance of Risk), some commenters claimed that such exposure cannot cause lung cancer independently of silicosis (i.e., only those workers who already have silicosis can get lung cancer) (Document ID 2307, Attachment A, p. 53). This claim is inconsistent with the credible scientific evidence presented above that genotoxic and non-genotoxic mechanisms triggered by early cellular responses to crystalline silica prior to development of silicosis may contribute to crystalline silica-induced carcinogenesis. OSHA finds, based on its review of all the evidence in the rulemaking record, that workers without silicosis, as well as those with silicosis, are at risk of lung cancer if regularly exposed to respirable crystalline silica at levels permitted under the previous and new PELs. The Agency also emphasizes that, regardless of the mechanism by which respirable crystalline silica exposure increases lung cancer risk, the fact remains that workers exposed to respirable crystalline silica continue to be diagnosed with lung cancer at a higher rate than the general population. Therefore, as discussed in section VI, Final Quantitative Risk Assessment and Significance of Risk, OSHA has met its burden of proving that workers exposed to previously allowed levels of respirable crystalline silica are at significant risk, by one or more of these mechanisms, of serious and life-threatening health effects, including both silicosis and lung cancer.

2. Relevance of Animal Models to Humans

Animal data has been used for decades to evaluate hazards and make inferences regarding causal relationships between human health effects and exposure to toxic substances. The National Academies of Science has endorsed the use of well-conducted animal studies to support hazard evaluation in the risk assessment process (Document ID 4052, p. 81) and OSHA's policy has been to rely on such studies when regulating carcinogens. In the case of respirable crystalline silica, OSHA has used evidence from animal studies, along with human epidemiology and other relevant information, to establish that occupational exposure is associated with silicosis, lung cancer, and other non-malignant respiratory diseases, as well as renal and autoimmune effects (Document ID 1711, pp. 261-266). Exposure to various forms of respirable crystalline silica by inhalation and intratracheal instillation has consistently caused lung cancer in rats (IARC, 1997, Document ID 1062, pp. 150-163). These results led IARC and NTP to conclude that there is sufficient evidence in experimental animals to demonstrate the carcinogenicity of crystalline silica in the form of quartz dust. IARC also concluded that there is sufficient evidence in human studies for the carcinogenicity of crystalline silica in the form of quartz or cristobalite.Start Printed Page 16347

In its pre-hearing comments and post-hearing brief, the ACC noted that increased lung cancer risks from exposure to respirable crystalline silica have not been found in animal species other than rats, and questioned the relevance of the rat model for evaluating potential lung carcinogenicity in humans (Document ID 2307, Attachment A, p. 30; 4209, p. 32). Specifically, the ACC highlighted studies by Holland (1995) and Saffiotti et al. (1996) indicating that bioassays in respirable crystalline silica-exposed mice, guinea pigs, and Syrian hamsters have not found increased lung cancer (Document ID 2307, Attachment A, p. 30, f. 51).

The ACC proposed that the increased lung cancer risk in respirable crystalline silica-exposed rats is due to a particle overload phenomenon, in which lung clearance of nonfibrous durable particles initiates a non-specific response that results in intrapulmonary lung tumors (Document ID 2307, Attachment A, p. 30, n. 51). Dr. Cox, on behalf of the ACC, citing Mauderly (1997, included in Document ID 3600), Oberdorster (1996, Document ID 3969), and Nikula et al. (1997, included in Document ID 3600), likewise commented that rats are “uniquely sensitive to particulate pollution, for species-specific reasons that do not generalize to other rodents or mammals, including humans” (Document ID 2307, Attachment 4, p. 83). OSHA reviewed the three studies referenced by Dr. Cox and notes that two actually appear to support the use of the rat model and the third does not reject it. Mauderly (1997) noted that the rat model was the only one to correctly predict carcinogenicity after inhalation exposure to several types of asbestos, and highlighted the shortcomings of other models, such as those using hamsters, which are highly insensitive to particle-induced lung cancers (article included in Document ID 3600, pp. 1339-1343). While Mauderly (1997) advised caution when using the rat because it is the most sensitive rodent species for lung cancer, he concluded that “there is evidence supporting continued use of rats in exploration of carcinogenic hazards of inhaled particles,” and that the other test species are problematic because they provide too many false negatives to be predictive (article included in Document ID 3600, p. 1343). Similarly, Oberdorster (1996), in discussing particle parameters used in the evaluation of exposure-dose-relationships of inhaled particles, stated that “the rat model should not be dismissed prematurely” (Document ID 3969, p. 73). Oberdorster (1996) postulated that humans and rats have very similar responses to particle-induced effects when analyzing the exposure-response relationship using particle surface area, rather than particle mass, as the exposure metric. Oberdorster concluded that there simply was not enough known regarding exact mechanisms to reject the model outright (Document ID 3969, pp. 85-87). The remaining paper cited by Dr. Cox, Nikula et al. (1997), evaluated the anatomical differences between primate and rodent responses to inhaled particulate matter and the role of clearance patterns and physiological responses to inhaled toxicants. The study noted that the differences between primate clearance patterns and rat clearance patterns may play a role in the pathogenesis from inhaled poorly soluble particles but did not dismiss the rat model as irrelevant to humans (Nikula, 1997, included in Document ID 3600, pp. 83, 93, 97).

Thus, OSHA finds that the Mauderly (1997) and Oberdorster (1996) articles generally support the rat as an appropriate model for qualitatively assessing the hazards associated with particle inhalation. OSHA likewise notes that the rat model is a common and well-accepted toxicological model used to assess human health effects from toxicant inhalation (ILSI, 2000, Document ID 3906, pp. 2-9). OSHA evaluated the available studies in the record, both positive and non-positive, and believes that it is appropriate to regard positive findings in experimental studies using rats as supportive evidence for the carcinogenicity of crystalline silica. This determination is consistent with that of IARC (Document ID 1473, p. 388) and NTP (Document ID 1164, p. 1), which also regarded the significant increases in incidence of malignant lung tumors in rats from multiple studies by both inhalation and intratracheal instillation of crystalline silica to be sufficient evidence of carcinogenicity in experimental animals and, therefore, to contribute to the evidence for carcinogenicity in humans.

3. Hypothesis That Lung Cancer Is Dependent on Silicosis

The ACC asserted in its comments that “if it exists at all, silica-related carcinogenicity most likely arises through a silicosis pathway or some other inflammation-mediated mechanism, rather than by means of a direct genotoxic effect” (Document ID 2307, Attachment A, p. 52; 4209, p. 51; 2343, Attachment 1, pp. 40-44). It explained that the “silicosis pathway” means that lung cancer stems from chronic inflammatory lung damage, which in turn, “implies that there is a threshold for any causal association between silica exposure and risk of lung cancer” (Document ID 2307, Attachment A, pp. 52-53). The ACC went on to state that a mechanism that involves ROS, growth factors, and inflammatory cytokines from alveolar macrophages is “most consistent” with development of advanced chronic inflammation (e.g., epithelial hyperplasia, lung tissue damage, fibrosis, and silicosis). According to this hypothesis, silica-related lung cancer is restricted to people who have silicosis (Document ID 2307, Attachment 2, p. 7). Regarding this hypothesis, the ACC concluded, “[t]his view of the likely mechanism for silica-related lung cancer is widely accepted in the scientific community, including by OSHA's primary source of silica-related health risk estimates, Dr. Kyle Steenland. OSHA appears to share this view as well” (Document ID 2307, Attachment A, p. 54).

The ACC statement regarding acceptance by OSHA and the scientific community is inaccurate. It implies scientific consensus, as well as OSHA's concurrence, that the chronic inflammation from silicosis is the only mechanism by which crystalline silica exposure results in lung cancer. The ACC has over-simplified and neglected the findings of the mechanistic studies that show activation of phagocytic and epithelial cells to be an early cellular response to crystalline silica prior to chronic inflammation (see Document ID 1711, pp. 234-238). As discussed previously, alveolar macrophage activation leads to initial production of ROS and release of cytokine growth factors that could contribute to silica-induced carcinogenicity through both genotoxic and non-genotoxic mechanisms. The early cellular response does not require chronic inflammation and silicosis to be present, as postulated by the ACC. It is possible that the early mechanistic influences that increase cancer risk may be amplified by a later severe chronic inflammation or silicosis, if such a condition develops. However, as Brian Miller, Ph.D., stated “this issue of silicosis being a precursor for lung cancer is unanswerable, given that we cannot investigate for early fibrotic lesions in the living, but must rely on radiographs.” (Document ID 3574, Tr. 31).

In pre-hearing comments the ACC commented, as proof of silicosis being linked to lung cancer, that fibrosis was linked to adenocarcinomas (Document ID 2307, Attachment A, p. 61). This statement is misleading. As explained Start Printed Page 16348earlier, silicosis results from stimulation of fibroblast cells that cause lung fibrosis. Adenocarcinomas, a hallmark tumor type in respirable crystalline silica-induced lung cancer, are tumors that arise not from fibroblasts, but exclusively from lung epithelial cells (IARC, 2012, Document ID 1473, pp. 381-389, 392). These tumors may be linked to the genotoxic and non-genotoxic mechanisms that occur prior to fibrosis, not secondary to the fibrotic process itself.

OSHA also received some comments that questioned the existence of a direct genotoxic mechanism. Jonathan Borak, M.D., on behalf of the U.S. Chamber of Commerce, commented, “there is no direct evidence that silica causes cancer by means of a directly DNA-reactive mechanism” (Document ID 2376, p. 21). Dr. Peter Morfeld, on behalf of the ACC, as well as Peter Valberg, Ph.D., and Christopher M. Long, Sc.D., of Gradient Corporation, on behalf of the U.S. Chamber of Commerce, cited a scientific article by Borm et al. (2011, included in Document ID 3573) which reported finding evidence against a genotoxic mechanism and in favor of a mechanism secondary to chronic inflammation (Document ID 3458, pp. 5-7; 4016, pp. 5-6; 4209, p. 51). Borm et al. (2011, included in Document ID 3573) analyzed 245 published studies from 1996 to 2008 identified using the search terms “quartz” and `toxicity” in conjunction with “surface,” “inflammation,” “fibrosis,” and “genotoxicity.” The authors then estimated the lowest dose (in units of micrograms per cell surface area) to consistently induce DNA damage or induce markers of inflammation (e.g., IL-8 upregulation) in in vitro studies. They adjusted the in vitro doses for the lung surface area encountered in vivo and found the crystalline silica dose that produced primary genotoxicity was 60-120 times higher than the dose that produced inflammatory cytokines (Borm et al., 2011, included in Document ID 3573, p. 762). Drs. Valberg and Long concluded that Borm et al. demonstrated that genotoxicity was a secondary response to chronic inflammation, except at very high exposures at which genotoxicity independent of inflammation might occur. They also maintained that lung cancer as a secondary response to chronic inflammation is considered to have a threshold (Document ID 4016, p. 6).

OSHA reviewed the Borm et al. study (2011, Document ID 3889), and notes several limitations. The authors examined the findings from various genotoxic assays (comet assay, 8-OH-dG, micronucleus test) (Borm et al., 2011, 3889, p. 758). They reported that 40 μg/cm2 was the lowest dose in vitro to produce significant direct DNA damage from crystalline silica. This genotoxic dose appears to be principally obtained from a study of a specific quartz sample (i.e., DQ12) in a single human alveolar epithelial cell line (i.e., A549 cells), even though Appendix Table 3 cited in vitro studies using other cells (e.g., fibroblasts) and other types of quartz (e.g., MinUsil) that produced direct genotoxic effects at lower doses (Borm et al., 2011, Document ID 3889, pp. 760, 769-770). This is especially pertinent since Borm et al. state that in vitro systems utilizing single-cell cultures are generally much less sensitive than in vivo systems, especially if attempting to determine oxidative stress-induced effects, since many cell culture systems use reagents that can scavenge ROS (Borm et al. 2011, Document ID 3889, p. 760). There was no indication that the authors accounted for this deficiency. They go on to conclude that their work shows a large-scale variation in hazard across different forms of quartz with regard to effects such as DNA breakage (e.g., genotoxicity) and inflammation (Borm et al. 2011, Document ID 3889, p. 762).

The extreme variation in response along with reliance on an insensitive genotoxicity test system could overestimate the appropriate genotoxic dose in human lung cells in vivo. In addition, Borm et al. used the dose sufficient to initiate production of an inflammatory cytokine (i.e., IL-8) in the A549 cell-line as the threshold for inflammation. It is not clear that an early cellular response, such as IL-8 production necessarily reflects a sustained inflammatory response. In summary, OSHA finds inconsistencies in this analysis, leaving some questions regarding the study's conclusion that silica induces genotoxicity only as a secondary response to an inflammation-driven mechanism. While the in vitro dose comparisons in this study fail to demonstrate that genotoxicity is secondary to the inflammatory response, the study findings do indicate that cellular responses to crystalline silica that drive inflammation may also lead to tumorigenesis through both genotoxic and non-genotoxic mechanisms.

Dr. Morfeld, in his hearing testimony on behalf of the ACC, referred to the paper by Borm et al. (2011) as reaching the conclusion that the mechanism of silica-related lung cancer is secondary inflammation-driven genotoxicity. As summarized by the ACC in post-hearing comments, he observed that “there are no crystalline silica particles found in the nucleus of the cells. There is nothing going on with particles in the epithelial cells inside the lung” (Document ID 4209, p. 52). In hearing testimony, however, Dr. Morfeld acknowledged that the Borm paper had limitations on extrapolating from in vitro to in vivo and cited a study by Donaldson et al. (2009), which discussed some of the limitations and the need for caution in extrapolating from in vitro to in vivo (Document ID 3582, Tr. 2076-2077; 3894, pp. 1-2). In considering this testimony, OSHA notes that the Donaldson et al. (2009) study, which includes the same authors as the Borm et al. (2011) study, acknowledged that direct interaction between respirable crystalline silica and epithelial cellular membranes induces intracellular oxidative stress which is capable of being genotoxic (Document ID 3894, p. 3). This is consistent with the OSHA position as well as the most recent IARC reevaluation of the cancer hazard from crystalline silica dust. As IARC stated in its most recent evaluation of the carcinogenicity of respirable crystalline silica under a section on direct genotoxicity and cell transformation (Document ID 1473, section 4.2.2, pp. 391-393):

Reactive oxygen species are generated not only at the particle surface of crystalline silica, but also by phagocytic and epithelial cells exposed to quartz particles. . . . Oxidants generated by silica particles and by the respiratory burst of silica-activated phagocytic cells may cause cellular and lung injury, including DNA damage (Document ID 1473, p. 391).

Given the IARC determination as well as the animal and in vitro studies reviewed herein, OSHA finds that there is no conclusive evidence that silica-related lung cancer only occurs as a secondary response to chronic inflammation, or that silicosis is a necessary prerequisite for lung cancer. Instead, OSHA finds support in the scientific literature for a conclusion that tumors may form through genotoxic as well as non-genotoxic mechanisms that result from respirable crystalline silica interaction with alveolar macrophages and other lung cells prior to onset of silicosis.

4. Hypothesis That Crystalline Silica-Induced Lung Disease Exhibits a Threshold

It is well established that silicosis arises from an advanced chronic inflammation of the lung. As noted above, a common hypothesis is that pathological conditions that depend on chronic inflammation may have a threshold. The exposure level at which silica-induced health effects might begin Start Printed Page 16349to appear, however, is poorly characterized in the literature (see Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases). The threshold exposure level required for a sustained inflammatory response is dependent upon multiple pro- and anti-inflammatory factors that can be quite variable from individual to individual and from species to species (Document ID 3896).

Discounting or overlooking the evidence that respirable crystalline silica may be genotoxic in the absence of chronic inflammation, Drs. Valberg and Long commented that crystalline silica follows a threshold paradigm for poorly soluble particles (PSPs). PSPs are defined generally as nonfibrous particles of low acute toxicity, which are not directly genotoxic (ILSI, 2000, Document ID 3906, p. 1). Specifically, Drs. Valberg and Long stated:

Mechanisms whereby lung cells respond to retention of a wide variety of PSPs, including crystalline silica, follow a generally accepted threshold paradigm, where the initiation of a chronic inflammatory response is a necessary step in the disease process, and the inflammatory response does not become persistent until particle retention loads become sufficient to overwhelm lung defense mechanisms. This overall progression from increased but controlled pulmonary inflammation across a threshold exposure that leads to lung damage has been described by a number of investigators (Mauderly and McCunney, 1995; ILSI, 2000; Boobis et al., 2009; Porter et al. 2004) (Document ID 2330, p. 19).

Similarly, Dr. Cox, in his post-hearing comments, discussed his 2011 article describing a quantifiable exposure-response threshold for lung diseases induced by inhalation of respirable crystalline silica (Document ID 4027, p. 29). Dr. Cox hypothesized the existence of an exposure threshold such that exposures to PSPs, which he described as including titanium dioxide, carbon black, and crystalline silica, must be intense enough and last long enough to disrupt normal homeostasis (i.e., normal cellular functions) and overwhelm normal repair processes. Under the scenario he described, a persistent state of chronic, unresolved inflammation results in a disruption of macrophage and neutrophil ability to clear silica and other foreign particles from the lung (Document ID 1470, pp. 1548-1551, 1555-1556).

OSHA disagrees with these characterizations about exposure thresholds because, among other reasons, respirable crystalline silica is not generally considered to be in the class of substances defined as PSPs.[7] Specifically, regarding the comments of Drs. Valberg and Long, OSHA notes that the two cited documents (Mauderly and McCunney, 1995, and ILSI, 2000) summarizing workshops on PSPs did not include crystalline silica in the definition of PSP and the lung “overload” concept, instead highlighting silica's cytotoxic and genotoxic mechanisms. Mauderly and McCunney (1995) stated, “[i]t is generally accepted that the term `overload' should be used in reference to particles having low cytotoxicity, which overload clearance [mechanisms] by virtue of the mass, volume, or surface area of the deposited material (Morrow, 1992)” (p. 3, article cited in Document ID 2330, p. 19). Mauderly specifically cited quartz as a cytotoxic particle that may fall outside this definition (p. 24, article cited in Document ID 2330, p. 19). The International Life Science Institute's (ILSI) Workshop Report (2000) intended only to address particles of “low acute toxicity,” such as carbon black, coal dust, soot, and titanium dioxide (Document ID 3906, p. 1). OSHA believes that the cytotoxic nature of crystalline silica would exclude it from the class of rather nonreactive, non-toxic particles mentioned above. Therefore, the Agency concludes that most scientific experts would not include crystalline silica in the class of substances known as PSPs, nor intend for findings regarding PSPs to be extrapolated to crystalline silica.

During the public hearing, OSHA questioned Dr. Morfeld about the relevance of the rat overload response and whether he considered crystalline silica to be like other PSPs such as carbon black. Dr. Morfeld replied that he was well aware of the literature and indicated that crystalline silica was not considered one of the PSPs (specifically not like carbon black) that these reports reviewed (Document ID 3582, Tr. 2072-2074). OSHA also notes a report of the European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC), which was cited by the ACC (Document ID 4209, p. 32) and stated that “particles exhibiting significant surface related (cyto)toxicity like crystalline silica (quartz) and/or other specific toxic properties do not fall under this definition [of PSPs]” (Document ID 3897, p. 5).

Respirable crystalline silica differs from PSPs because it does not require particle overload to induce the same response typical of PSPs. “Overload” refers to the consequence of exposure that results in a retained lung burden of particles that is greater than the steady-state burden predicted from deposition rates and clearance kinetics (Document ID 4174, p. 20). This is a result of a volumetric over-exposure of dust in the lung, which overwhelms macrophage function. Respirable crystalline silica does not operate on this mechanism since macrophage function is inhibited by the cytotoxic nature of respirable crystalline silica rather than a volumetric overload (Oberdorster, 1996, Document ID 3969). Therefore, respirable crystalline silica does not require particle overload to induce the same response. Studies have found that the respirable crystalline silica exposure levels required to induce tumor formation in some animal studies are similar to those observed in human studies, whereas studies involving PSPs tend to show responses at much higher levels of exposure (Muhle et al., 1991, Document ID 1284; Muhle et al., 1995, 0378; Saffiotti and Ahmed, 1995, 1121).

A study by Porter et al. (2004) demonstrated that pulmonary fibrosis induction does not require silica particle overload (Document ID 0410, p. 377). The ACC cited this study in its post-hearing brief, stating, “Porter . . . noted that the response of the rat lung to inhaled crystalline silica particles is biphasic, with a below-threshold phase characterized by increased but controlled pulmonary inflammation” (Document ID 4209, p. 52). OSHA notes that this biphasic response is due in part to the cytotoxic nature of crystalline silica, which disrupts macrophage clearance of silica particles leading to a chronic inflammatory response at less than overload conditions. While there are some mechanistic similarities, OSHA believes that the argument that crystalline silica operates on the basis of lung overload is erroneous and based on false assumptions that ignore toxicological properties unique to crystalline silica, such as cytotoxicity and the generation of intracellular ROS (Porter et al., 2002, Document ID 1114; Porter et al., 2004, 0410). As previously discussed, the generation of ROS could Start Printed Page 16350potentially damage cellular DNA by a genotoxic mechanism that may not exhibit a threshold.

OSHA thoroughly reviewed Dr. Cox's 2011 article (Document ID 1470), in which he proposed a threshold for crystalline silica, in its Supplemental Literature Review (Document ID 1711, Attachment 1, pp. 37-39). OSHA concluded that the evidence used to support Cox's assertion that the OSHA PEL was below a threshold for lung disease in humans was not supported by the evidence presented (Document ID 1470, p. 1543; 1711, Attachment 1). Specifically, Cox (2011) modelled a threshold level for respirable crystalline silica using animal studies of PSPs. This approach, according to the ILSI report (2000) and ECETOC report (2013), is clearly not appropriate since the cytotoxic nature of crystalline silica is not consistent with the low-toxicity PSPs (Document ID 3906, p. 1; 3897, p. 5). Dr. Cox (2011) categorized crystalline silica incorrectly as a PSP and ignored the evidence for cytotoxicity and genotoxicity associated with crystalline silica. He further failed to consider or include studies indicating a tumor response at exposure levels below that leading to an excessive chronic inflammatory response, such as Porter et al. (2002) and Muhle et al. (1995) (Document ID 1114; 0378). Thus, OSHA considers the threshold model designed by Dr. Cox (2011, Document ID 1470) and referenced by Drs. Valberg and Long (Document ID 2330) to be contradicted by the best available evidence regarding the toxicological properties of respirable crystalline silica. Although OSHA acknowledges the possible existence of a threshold for an inflammatory response, the Agency believes that the threshold is likely much lower than that advocated by industry representatives such as the ACC and the Chamber of Commerce (see Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases).

OSHA concludes that a better estimate of a threshold effect for inflammation and carcinogenesis was done by Kuempel et al. (2001, Document ID 1082). These researchers studied the minimum human exposures necessary to achieve adverse functional and pathological evidence of inflammation. They employed a physiologically-based lung dosimetry model, included more relevant studies, and considered a genotoxic effect for lung cancer (Kuempel et al., 2001, Document ID 1082; see 1711, pp. 231-232). Briefly, Kuempel et al. evaluated both linear and nonlinear (threshold) models and determined that the average minimum critical quartz lung burden (Mcrit) in rats associated with reduced pulmonary clearance and increased neutrophil inflammation was 0.39 mg quartz/g lung tissue. Mcrit is based on the lowest observed adverse effect level in a study in rats (Kuempel, 2001, Document ID 1082, pp. 17-23). A human lung dosimetry model, developed from respirable coal mine dust and quartz exposure and lung burden data in UK coal miners (Tran and Buchanan, 2001, Document ID 1126), was then used to estimate the human-equivalent working lifetime exposure concentrations associated with lung doses. An 8-hour time-weighted average (TWA) concentration of 0.036 mg/m3 (36 μg/m3) over a 45-year working lifetime was estimated to result in a human-equivalent lung burden to the average Mcrit in rats (Document ID 1082, pp. 24-26). OSHA peer reviewer Gary Ginsburg, Ph.D., summarized, “the Kuempel et al. (2001, 2001b) rat analysis of lung threshold loading and extrapolation to human dosimetry leads to the conclusion that in the median case this threshold is approximately 3 times below the current [now former] OSHA PEL” (Document ID 3574, pp. 23). This estimated threshold would be significantly below the final PEL of 50 μg/m3.

In pre-hearing comments, ACC stated that some health organizations suggested a silicosis-dependent threshold exists for lung cancer (ACC, Document ID 2307, Attachment A, pp. 60-62). Specifically, ACC cited Environment and Health Canada as stating:

Although the mechanism of induction for the lung tumours has not been fully elucidated, there is sufficient supportive mode of action evidence from the data presented to demonstrate that a threshold approach to risk assessment is appropriate based on an understanding of the key events in the pathogenesis of crystalline silica induced lung tumours (pp. 49-51 as cited by ACC, Document ID 2307, p. 62).

In addition to the statement submitted by ACC, Environment and Health Canada also stated that:

While there is sufficient evidence to support key events in a threshold mode of action approach for lung tumours, the molecular mechanism is still not fully elucidated. Also, despite the fact that the effects seen in rats parallel the effects observed in human studies, additional mechanistic studies could further clarify why lung tumours are not seen in all experimental animals . . . Thus, the question of whether silica exposure, in the absence of silicotic response, results in lung tumours remains unanswered.” (pp. 51-52 as cited by ACC, Document ID 2307, pp. 59-61).

It should be noted that the Environment and Health Canada report was to determine general population risk of exposure to respirable crystalline silica as a fraction of PM10. Environment and Health Canada found that levels 0.1-2.1 μg/m3 respirable crystalline silica were sufficiently protective for the general population because they represented a margin of exposure (MOE) 23-500 times lower than the 50 μg/m3 quartz concentration associated with silicosis in humans (pp. 50-51 as cited by ACC, Document ID 2307, pp. 59-61).

A report by Mossman and Glenn (2013) reviewed the findings from several international OEL setting panels (Document ID 4070). The report cites findings from the European Commission's Scientific Committee on Occupational Exposure Limits for respirable crystalline silica. The findings “acknowledged a No Observed Adverse Exposure Level (NOAEL) for respirable crystalline silica in the range below 0.020 mg/m3, but stated that a clear threshold for silicosis could not be identified” (Mossman and Glen, 2013; Document ID 4070, p. 655). The report went on to state that SCOEL (2002) recommended that an OEL should lie below 50 μg/m3 (Document ID 4070, p. 655). Therefore, even if silica-induced lung cancer were limited only to a mechanism that involved an inflammation-dependent threshold, OSHA concludes that exposure threshold would likely be lower than the final PEL.

5. Renal Disease and Autoimmunity

While mechanistic data is limited, other observed health effects from inhalation of respirable crystalline silica include kidney and autoimmune effects. Translocation of particles through the lymphatic system and filtration through the kidneys may induce effects in the immune and renal systems similar to the types of changes observed in the lung (Miller, 2000, Document ID 4174, pp. 40-45). A review of the available literature indicates that respirable crystalline silica most likely induces an oxidative stress response in the renal and immune cells similar to that described above (Donaldson et al., 2009, Document ID 3894).

6. Conclusion

OSHA has reviewed and responded to the comments received on the mechanistic studies of respirable crystalline silica-induced lung cancer and silicosis, as well as comments that the mechanistic data imply the existence of an exposure threshold. OSHA concludes that: (1) Lung cancer likely results from both genotoxic and non-genotoxic mechanisms that arise during early cellular responses as well Start Printed Page 16351as during chronic inflammation from exposure to crystalline silica; (2) there is not convincing data to demonstrate that silicosis is a prerequisite for lung cancer; (3) experimental studies in rats are relevant to humans and provide supporting evidence for carcinogenicity; (4) crystalline silica does not behave like PSPs such as titanium dioxide; and (5) any threshold for an inflammatory response to respirable crystalline silica is likely several times below the final PEL of 50 μg/m3. Thus, the best available evidence on this issue supports OSHA's findings that respirable crystalline silica increases the risk of lung cancer in humans, even in the absence of silicosis, and that lung cancer risk can be increased by exposure to crystalline silica at or below the new OSHA PEL of 50 μg/m3.

I. Comments and Responses Concerning Thresholds for Silica-Related Diseases

In this section, OSHA discusses comments focused on the issue of exposure-response thresholds for silica exposure. In the comments received by OSHA on this topic, an exposure-response “threshold” for silica exposure typically refers to a level of exposure such that no individual whose exposure is below that level would be expected to develop an adverse health effect. Commenters referred to thresholds both in terms of concentration and cumulative exposure (i.e., a level of cumulative exposure below which an individual would not be expected to develop adverse health effects). In addition to individual thresholds, some commenters referred to a “population average threshold,” that is, the mean or median value of individual thresholds across a population of workers. There is significant scientific controversy over whether any such thresholds exist for silicosis and lung cancer, as well as the cumulative exposure level or concentration at which a threshold effect may occur and whether certain statistical modeling approaches can be used to identify threshold effects.

OSHA has reviewed the evidence in the record pertaining to thresholds, and has determined that the best available evidence supports the Agency's use of non-threshold exposure-response models in its risk assessments for silicosis and lung cancer. The voluminous scientific record accrued by OSHA in this rulemaking supports lowering the existing PEL to 50 μg/m3. Rather than indicating a threshold of risk that starts above the previous general industry PEL, the weight of this evidence, including OSHA's own risk assessment models, supports a conclusion that there continues to be significant, albeit reduced, risk at the 50 μg/m3 exposure limit. OSHA's evaluation of the best available evidence on thresholds indicates that there is considerable uncertainty about whether there is any threshold below which silica exposure causes no adverse health effects; but, in any event, the weight of evidence supports the view that, if there is a threshold of exposure for the health effects caused by respirable crystalline silica, it is likely lower than the new PEL of 50 μg/m3. Commenters have not provided convincing evidence of a population threshold (e.g., an exposure level safe for all workers) above the revised PEL. In addition, OSHA's final risk assessment demonstrates that achieving this limit—which OSHA separately concludes is overall the lowest feasible level for silica-generating operations—will result in significant reductions in mortality and morbidity from occupational exposure to respirable crystalline silica.

1. Thresholds—General

In the Preliminary Quantitative Risk Assessment (QRA) (Document ID 1711, pp. 275, 282-285), OSHA reviewed evidence on thresholds from a lung dosimetry model developed by Kuempel et al. (2001, Document ID 1082) and from epidemiological analyses conducted by Steenland and Deddens (2002, Document ID 1124). As discussed in the Preliminary QRA, Kuempel et al. (2001) used kinetic lung models for both rats and humans to relate lung burden of crystalline silica and estimate a minimum critical lung burden (Mcrit) of quartz above which particle clearance begins to decline and lung inflammation begins to increase (early steps in the process of developing silica-related disease). The Mcrit would be achieved by a human equivalent airborne exposure to 36 μg/m3 for 45 years, based on the authors' rat-to-human lung model conversion. Exposures below this level would not lead to an excess lung cancer risk in the average individual, if it were assumed that cancer is strictly a secondary response to persistent inflammation. OSHA notes, however, that if some of the silica-related lung cancer risk occurs as a result of direct genotoxicity from early cellular interaction with respirable silica particles, then this threshold value may not be applicable. Since silicosis is caused by persistent lung inflammation, this exposure level could be viewed as a possible average threshold level for that disease as well (Document ID 1711, p. 284). As 36 μg/m3 is well below the previous general industry PEL of 100 μg/m3 and below the final PEL of 50 μg/m3, the Kuempel et al. study showed no evidence of an exposure-response threshold high enough to impact OSHA's choice of PEL.

Steenland and Deddens (2002, Document ID 1124) examined a pooled lung cancer study originally conducted by Steenland et al. (2001a). They found that a threshold model based on the log of cumulative dose (15-year lag) fit better than a no-threshold model, with the best threshold at 4.8 log mg/m3-days (representing an average exposure of 10 μg/m3 over a 45-year working lifetime). OSHA preliminarily concluded that, in the Kuempel et al. (2001) study and among the studies evaluated by Steenland et al. (2001a) in the pooled analysis, there was no empirical evidence of a threshold for lung cancer in the exposure range represented by the previous and final PELs (i.e., at 50 μg/m3 or higher) (Document ID 1711, pp. 275, 284). Thus, based on these two studies, workers exposed at or below the new PEL of 50 μg/m3 over a working lifetime still face a risk of developing silicosis and lung cancer because their exposure would be above the supposed exposure threshold.

In its prehearing comments, the ACC argued that OSHA's examination of the epidemiological evidence, along with animal studies and mechanistic considerations, “has not shown that reducing exposures below currently permitted exposure levels would create any additional health benefits for workers. OSHA's analysis and the studies on which it relies have not demonstrated the absence of an exposure threshold above 100 μg/m3 for the various adverse health effects considered in the QRA” (Document ID 2307, Attachment A, p. 26; also 2348, Attachment 1, p. 33). According to the ACC, an exposure threshold above OSHA's previous general industry PEL of 100 μg/m3 means that workers exposed below that level will not get sick, negating the need to lower the PEL (Document ID 2307, Attachment A, p. 91).

Members of OSHA's peer review panel for the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (Document ID 1711) rejected the ACC's comments as unsupportable. Peer reviewer Mr. Bruce Allen stated: “it is essentially impossible to distinguish between dose-response patterns that represent a threshold and those that do not” in epidemiological data (Document ID 3574, p. 8). Peer reviewer Dr. Kenneth Crump similarly commented:

OSHA is on very solid ground in the [Preliminary QRA's] statement that “available information cannot firmly establish a threshold exposure for silica-Start Printed Page 16352related effects” . . . the hypothesis that a particular dose response does not have a threshold is not falsifiable. Similarly, the hypothesis that a particular dose response does have a threshold is not falsifiable (Document ID 3574, p. 17).

Dr. Cox, representing the ACC, agreed with Dr. Crump that “it's impossible to prove a negative, empirically . . . you could never rule out that possibility” of a threshold at a low level of exposure (Document ID 3576, Tr. 402). However, he contended that it is possible to rule out a threshold in the higher-level range of observed exposures based on observed illness: “I think that there are plenty of chemicals for which the hypothesis of a threshold exist[ing] at or above current standards could be ruled out because you see people getting sick at current levels” (Document ID 3576, Tr. 403). Other commenters stated their belief that workers recently diagnosed with silicosis must have had exposures above the previous general industry PEL and, based on this supposition, concluded that OSHA has not definitively proven risk to workers exposed below the previous general industry PEL (Document ID 4224, pp. 2-5; Tr. 3582, pp. 1951-1963).

OSHA agrees with Dr. Cox that observation of workers “getting sick at current levels” can rule out a threshold effect at those levels. As is discussed below, there is evidence that workers exposed to silica at cumulative or average exposure levels permitted under the previous PELs have become ill and died as a result of their exposure. OSHA thus strongly disagrees with any implication from commenters that the Agency should postpone reducing a PEL until it has extensive documentation of sick and dying workers to demonstrate that the current PEL is not sufficiently protective (see Section II, Pertinent Legal Authority, and Section VI, Final Quantitative Risk Assessment and Significance of Risk).

The ACC's and Chamber's comments on this issue essentially argue that the model OSHA used to assess risk was inadequate to assess whether a threshold of risk exists and, if one does exist, at what level (Document ID 2307, Attachment A, pp. 52-65; 2376, pp. 20-22; 2330, pp. 17-21). According to OSHA peer reviewer Dr. Crump, however, the analytical approach taken by OSHA in the Preliminary QRA was appropriate. Considering the inherent limitations of epidemiological data:

an attempt to distinguish between threshold and non-threshold dose responses is not even a scientific exercise . . . The best that can be done is to attempt to place bounds on the amount of risk at particular exposures consistent with the available data, which is what OSHA had done in their risk assessment (Document ID 3574, p. 17).

A further source of uncertainty in investigating thresholds was highlighted by Dr. Mirer, on behalf of the AFL-CIO (Document ID 3578, Tr. 988-989) and by peer reviewer Dr. Andrew Salmon, who stated:

[m]any of the so-called thresholds seen in epidemiological studies represent thresholds of observability rather than thresholds of disease incidence . . . studies (and anecdotal observations) with less statistical power and shorter post-exposure followup (or none) will necessarily fail to see the less frequent and later-appearing responses at lower doses. This creates an apparent threshold which is higher in these studies than the apparent threshold implied by studies with greater statistical power and longer follow-up (Document ID 3574, p. 37).

Peer reviewer Dr. Gary Ginsberg suggested that, recognizing these inherent limitations, OSHA should characterize the body of evidence and argument surrounding thresholds by discussing the following factors related to whether a threshold for silica-related health effects exists at exposure levels above the previous general industry PEL:

the choices relative to the threshold concept for the silica dose response . . . [including] specific dose response datasets that are consistent with a linear or a threshold-type model, if a threshold seems likely, where was it seen relative to the current and proposed PEL, and a general discussion of mechanism of action, measurement error and population variability as concepts that can help us understand silica dose response for cancer and non-cancer endpoints (Document ID 3574, p. 24).

Following Dr. Ginsberg's suggestion, OSHA has, in its final health and risk analysis, considered the epidemiological evidence relevant to possible threshold effects for silicosis and lung cancer. As discussed below, first in “Thresholds—Silicosis and NMRD” and then in “Thresholds—Lung Cancer,” OSHA has carefully considered comments about statistical methods, exposure measurement uncertainty, and variability as they pertain to threshold effects. The discussion addresses the epidemiological evidence with respect to both cumulative and concentration thresholds. For reference, a working lifetime (45 years) of exposure to silica at the previous general industry PEL (100 μg/m3) and the final PEL (50 μg/m3) yield cumulative exposures of 4.5 mg/m3-yrs and 2.25 mg/m3-yrs, respectively. Other sections with detailed discussions pertinent to threshold issues include Section V.H, Mechanisms of Silica-Induced Adverse Health Effects, and Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis.

2. Thresholds—Silicosis and NMRD

OSHA has determined that the studies most relevant to the threshold issue in this rulemaking are those of workers who have cumulative exposures or average exposure concentrations below the levels associated with the previous general industry PEL (100 μg/m3, or cumulative exposure of 4.5 mg/m3-yrs). Contrary to comments that OSHA only relied on studies involving exposures far above the levels of interest to OSHA in this rulemaking, and then extrapolated exposure-response relationships down to relevant levels (e.g., Document ID 2307, Attachment A, pp. 94-95; 4226, p. 2), a number of silicosis studies included workers who were exposed at levels close to or below the previous OSHA PEL for general industry. For example, four of the six cohorts of workers in the pooled silicosis mortality risk analysis conducted by Mannetje et al. (2002) had median cumulative exposures below 2.25 mg/m3-yrs., and three had median silica concentrations below 100 μg/m3 (Mannetje et al., 2002, Document ID 1089, p. 724). Other silicosis studies with significant numbers of relatively low-exposed workers include analyses of German pottery workers (Birk et al., 2009, Document ID 4002, Attachment 2; Mundt et al., 2011, 1478; Morfeld et al., 2013, 3843), Vermont granite workers (Attfield and Costello, 2004, Document ID 0285; Vacek et al., 2011, 1486), and industrial sand workers (McDonald et al., 2001, Document ID 1091; Hughes et al., 2001, 1060; McDonald et al., 2005, 1092). In this section, OSHA will discuss each of them in relationship to whether they suggest the existence of a threshold above 100 μg/m3, the previous PEL for general industry.

a. Mannetje et al. Pooled Study and Related Analyses

Mannetje et al. (2002b, Document ID 1089) estimated excess lifetime risk of silicosis based on six of the ten cohorts that were part of the IARC multi-center exposure-response study (Steenland et al., 2001a, Document ID 0452). The six cohorts were U.S. diatomaceous earth (DE) workers, Finnish granite workers, U.S. granite workers, U.S. industrial sand workers, U.S. gold miners, and Australian gold miners. Together, the cohorts included 18,634 subjects and 170 silicosis deaths. All cohorts except the Finnish granite workers and Australian gold miners had significant numbers of workers with median Start Printed Page 16353cumulative and/or average exposures below the levels associated with OSHA's previous general industry PEL. Checking for nonlinearities in their exposure-response model, Mannetje et al. found that a five-knot cubic spline model (which allows for deviations, such as thresholds, from a linear relationship) did not fit the data better than the linear model used in their main analysis. The result of this attempt to check for nonlinearities suggests that there is no threshold effect in the relationship between cumulative silica exposure and silicosis risk in the study. Significantly, NIOSH stated that the results of Mannetje et al.'s analysis “suggest the absence of threshold at the lowest [cumulative] exposure analyzed . . . in fact, the trend for silicosis mortality risk extends down almost linearly to the lowest cumulative exposure stratum”, in which “the average cumulative exposure is the equivalent of 45 years of exposure at 11.1 μg/m3 silica” (Document ID 4233, pp. 34-35). This level is significantly below the new OSHA PEL of 50 μg/m3.

As discussed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, OSHA commissioned Drs. Kyle Steenland and Scott Bartell to examine the potential effects of exposure measurement error on the mortality risk estimates derived from the pooled studies of lung cancer (Steenland et al., 2001, Document ID 0452) and silicosis (Mannetje et al., 2002b, Document ID 1089). Their analysis of the pooled data, using a variety of standard statistical techniques (e.g., regression analysis), also found the data either consistent with the absence of a threshold or inconsistent with the existence of a threshold [8] (Document ID 0469). Thus, neither Mannetje et al. nor Steenland and Bartell's analyses of the pooled cohorts suggested the existence of a cumulative exposure threshold effect; in fact, they suggested the absence of a threshold. Given the predominance in these studies of cohorts where at least half of the workers had cumulative exposures below 4.5 mg/m3-yrs, OSHA believes these results constitute strong evidence against an exposure threshold above the level of cumulative exposure resulting from long-term exposure at the previous PEL of 100 μg/m3.

b. Vermont Granite Workers

As discussed in the Supplemental Literature Review of Epidemiological Studies, Vacek et al. (2011, Document ID 1486) examined exposures from 1950 to 1999 for a group of 7,052 workers in the Vermont granite industry (Document ID 1711, Attachment 1, pp. 2-5). The exposure samples show relatively low exposures for the worker population. For the period 1950 to 2004, Verma et al. (2012), who developed the job exposure matrix used by Vacek et al., estimated that average exposure concentrations in 21 of 22 jobs were below 100 μg/m3, and 11 of the 22 job classes were at 50 μg/m3 or below. The remaining job category, laborer, had an estimated average exposure concentration of exactly 100 μg/m3 (Verma et al., 2011, Document ID 1487, p. 75).

Six of the 5,338 cohort members hired in or after 1940, when Vermont's dust control program was in effect, were identified as having died of silicosis by the end of the follow-up period (Vacek et al., Document ID 1486, p. 314). The frequency of observed silicosis mortality in the population is significant by OSHA standards (1.1 per 1,000 workers), and may be underestimated due to under-reporting of silicosis as a cause of death (see Section V.E, Comments and Responses Concerning Surveillance Data on Silicosis Morbidity and Mortality). This observed silicosis mortality shows that deaths from silicosis occurred among workers hired after silica concentrations were reduced below OSHA's previous general industry PEL. It therefore demonstrates that a threshold for silicosis above 100 μg/m3 is unlikely.

In terms of morbidity, Graham et al.'s study of radiographic evidence of silicosis among retired Vermont granite workers found silicosis in 5.7 percent of workers hired after 1940 (equivalent to 57/1,000 workers) (Graham et al., 2004, Document ID 1031, p. 465). OSHA concludes that these studies of low-exposed workers in the Vermont granite industry show significant risk of silicosis—both mortality and morbidity—at concentrations below the previous PELs. These studies also indicate that a threshold at an exposure concentration significantly above the previous PEL for general industry, as posited by industry representatives, is unlikely.

c. U.S. Industrial Sand Workers

In an exposure-response study of 4,027 workers in 18 U.S. industrial sand plants, Steenland and Sanderson (2001) reported that approximately three-quarters of the workers with complete work histories had cumulative exposures below 1.28 mg/m3-yrs, well below the cumulative exposure of 2.25 mg/m3-yrs associated with a working lifetime of exposure at the final PEL of 50 μg/m3 (Document ID 0455, p. 700). The study identified fourteen deaths from silicosis and unspecified pneumoconiosis (~3.5 per 1,000 workers) (Document ID 0455, p. 700), of which seven occurred among workers with cumulative exposures below 1.28 mg/m3-yrs. As with other reports of silicosis mortality, this figure may underestimate the true rate of silicosis mortality in this worker population.

Hughes et al. (2001) reported 32 cases of silicosis mortality in a cohort of 2,670 workers at nine North American industrial sand plants (~12 per 1,000) (Document ID 1060, p. 203). The authors developed a job-exposure matrix based on exposure samples collected by the companies and by MSHA between 1973 and 1994, along with the 1946 exposure survey used by Steenland and Sanderson (2001, Document ID 0455; 2307, Attachment 7, p. 6). Job histories were available for 29 workers who died of silicosis. Of these, fourteen had estimated cumulative exposure less than or equal to 5 mg/m3-yrs, and seven had cumulative exposures less than or equal to 1.5 mg/m3-yrs (Document ID 1060, p. 204). Both studies clearly showed silicosis risk among workers whose cumulative exposures were comparable to those that workers could experience under the final PEL (Document ID 0455, p. 700; 1060, p. 204), indicating that a threshold above this level of cumulative exposure is unlikely.

d. German Porcelain Workers

A series of papers by Birk et al. (2009, Document ID 4002, Attachment 2; 2010, Document ID 1467), Mundt et al. (2011, Document ID 1478), and Morfeld et al. (2013, Document ID 3843) examined silicosis mortality and morbidity in a population of over 17,000 workers in the German porcelain industry. Cohort members' annual average concentrations of respirable quartz dust were reconstructed from detailed work histories and dust measurements collected in the industry from 1951 onward (Birk et al., 2009, Document ID 4002, Attachment 2, pp. 374-375). Morfeld et al. observed 40 silicosis morbidity cases (ILO profusion category 1/1 or greater), and noted that additional Start Printed Page 16354follow-up of the cohort might be necessary due to the long latency period of silicosis (2013, Document ID 3843, p. 1032).

Follow-up time is a critical factor for detection of silicosis, which has a typical latency of 20-30 years (see Morfeld et al., 2013, Document ID 3843, p. 1028). As stated in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA, the disease latency for silicosis can extend to around 30 years. Follow-up was extremely limited in the German porcelain workers silicosis morbidity analysis, with a mean of 7.5 years of follow up for the study population (Document ID 3843). Despite the limited follow-up time, the cohort showed evidence of silicosis morbidity among low-exposed workers: 17.5 percent of cases occurred among workers whose highest average silica exposure in any year (“highest annual”) was estimated by the authors to be less than 250 μg/m3, and 12.5 percent of cases occurred among workers whose highest annual silica exposure was estimated at less than 100 μg/m3 (Document ID 3843).

The lead author of the study, Dr. Peter Morfeld, testified at the public hearings on behalf of the ACC Crystalline Silica Panel. In his post-hearing comments, Dr. Morfeld stated that “[m]echanistic considerations imply that we should not expect to see a threshold for cumulative exposure” in silicosis, but that the question of whether a threshold concentration level may exist remains (Document ID 4003, p. 3). The study by Morfeld et al. “focused on the statistical estimation of a concentration threshold . . . [and] simultaneously took into account the cumulative exposure to respirable crystalline silica dust as a driving force of the disease” (Document ID 4003, p. 3). Morfeld et al. applied a technique developed by Ulm et al. (1989, 1991) to estimate a concentration threshold. In this method a series of candidate exposure concentration values are subtracted from the estimated annual mean concentration data. Using the recalculated exposure estimates for the study population, regression analyses for each candidate are run to identify the best fitting model, using the Akaike Information Criterion (AIC) to evaluate model fit (Document ID 3843, p. 1029). According to Morfeld, the best fitting model in their study estimated a threshold concentration of 250 μg/m3 (AIC = 488.3) with a 95 percent confidence interval of 160 to 300 μg/m3. A second model with very similar fit (AIC = 488.8) estimated a threshold concentration of 200 μg/m3 with a 95 percent confidence interval of 57 μg/m3 to 270 μg/m3. A third model with a poorer fit (AIC=490.6) estimated a threshold concentration of 80 μg/m3 with a 95 percent confidence interval of 0.2 μg/m3 to 210 μg/m3 (Document ID 3843, Table 3, p. 1031).

In the Final Peer Review Report, Dr. Crump stated that Morfeld et al.'s modeling approach, like “all such attempts statistically to estimate a threshold,” is “not reliable because the threshold estimates so obtained are highly unstable” (Document ID 3574, p. 17). Dr. Morfeld's co-author, Dr. Mundt, stated in the public hearings:

I'll be the first one to tell you there is a lot of imprecision and, therefore, say confidence intervals or uncertainty should be respected, and that the—I'm hesitant to just focus on a single point number like the .25 [250 μg/m3], and prefer that you encompass the broader range that was reported in the Morfeld, on which I was an author and consistently brought this point to the table (Document ID 3577, Tr. 645).

NIOSH submitted post-hearing comments on the analysis in Morfeld et al. (2013). NIOSH pointed out that the exposure measurements in the analysis were based on German dust samplers, which for pottery have been shown to collect approximately twice as much dust as U.S. samplers. Therefore, “when Dr. Morfeld cited 0.15 mg/m3 (150 μg/m3) as the lower 95% confidence limit for the threshold, that would convert to 0.075 mg/m3 (75 μg/m3) in terms of equivalent measurements made with a U.S. sampler” (Document ID 4233, p. 21). Similarly, the U.S. equivalent of each of the other threshold estimates and confidence limits presented in Morfeld et al.'s analysis would be about half the reported exposure levels. NIOSH also commented that Morfeld et al.'s analysis appears to be consistent with both threshold and non-threshold models (Document ID 4233, p. 55). Furthermore, NIOSH observed that Morfeld et al. did not account for uncertainty in the values of one of their model parameters (ε); therefore their reported threshold confidence limits of 0.16-0.30 are too narrow (Document ID 4233, p. 56). More generally, NIOSH noted that Morfeld et al. did not quantitatively evaluate how uncertainty in exposure estimates may have impacted the results of the analysis; Morfeld agreed that he had not performed a “formal uncertainty analysis” (Document ID 4233, p. 58; 3582, Tr. 2078-2079). NIOSH concluded, “it is our firm recommendation to discount results based on the model specified in [Morfeld et al. Eq. 3] . . . including all results related to a threshold” (Document ID 4233, p. 58). OSHA has evaluated NIOSH's comments on the analysis and agrees that the issues raised by NIOSH raise serious questions about Morfeld et al.'s conclusions regarding a silica threshold.

OSHA's greater concern with Dr. Morfeld's estimate of 250 μg/m3 as a threshold concentration for silicosis is the fact that a substantial proportion of workers with silicosis in Dr. Morfeld's study had no estimated exposure above the threshold suggested by the authors; this threshold was characterized by commenters, including the Chamber of Commerce (Chamber), as a concentration “below which the lung responses did not progress to silicosis” (Document ID 4224, Attachment 1, p. 3). This point was emphasized by Dr. Brian Miller in the Final Peer Review Report (Document ID 3574, p. 57) and by NIOSH (Document ID 4233, p. 57). In the study, 17.5 percent of workers with silicosis were classified as having no exposure above Morfeld et al.'s estimated threshold of 250 μg/m3, (Document ID 3843, p. 1031) and 12.5 percent of these workers were classified as having no exposure above 100 μg/m3. OSHA believes the presence of these low-exposed workers with silicosis clearly contradicts the authors' estimate of 250 μg/m3 as a level of exposure below which no worker will develop silicosis (see Document ID 4233, p. 57).

In a post-hearing comment, Dr. Morfeld offered a different interpretation of his results, describing his threshold estimate as a “population average” which would not be expected to characterize risk for all individuals in a population. Rather, according to Dr. Morfeld “we expect to see differences in response thresholds among subjects” (Document ID 4003, p. 5). OSHA agrees with this interpretation, which was similarly expressed in several comments from OSHA's peer reviewers on the subject of thresholds (e.g., Document ID 3574, pp. 13, 21-22). Consistent with its peer reviewers' opinions, OSHA draws the conclusion from the data and discussion concerning population averages that these “differences in response thresholds among subjects” support setting the PEL at 50 μg/m3 in order to protect the majority of workers in the population of employees exposed to respirable crystalline silica. OSHA's review of the Morfeld et al. data on German porcelain workers thus reinforces its view that reducing exposures to this level will benefit the many workers who would develop silicosis at exposure levels below that of the “average” worker.

Dr. Morfeld's discussion of his estimate as a “population average” among workers with different individual responses to silica exposure Start Printed Page 16355echoes several comments from OSHA's peer reviewers on the subject of thresholds. In the Final Peer Review Report, Dr. Ginsberg observed that a linear exposure-response model may reflect a distribution of individual “thresholds,” such that “the population can be characterized as having a distribution of vulnerability. This distribution may be due to differences in levels of host defenses that come with differences in age, co-exposure to other chemicals, the presence of interacting background disease processes, non-chemical stressors, and a variety of other host factors” (Document ID 3574, p. 21). Given the number of factors that may influence vulnerability to certain diseases in a population of workers, Dr. Ginsberg continued:

it is logical for OSHA to strongly consider inter-subject variability . . . as the reason for linearly-appearing regression slopes in silica-related non-cancer and cancer studies. This explanation does not imply an artifact [that is, a false appearance of linear exposure-response] but that the linear (or log linear) regression coefficient extending down to low dose reflects the inherent variability in susceptibility such that the effect of concern . . . may occur in some individuals at doses well below what might be a threshold in others (Document ID 3574, pp. 21-22).

Peer reviewer Mr. Bruce Allen agreed that “[i]t makes no sense to discuss a single threshold value . . . Given, then, that thresholds must be envisioned as a distribution in the population, then there is substantial population-level risk even at the mean threshold value, and unacceptably high risk levels at exposures far below the mean threshold.” He further stated:

It is NOT, therefore, inappropriate to model the population-level observations using a non-threshold model . . . In fact, I would claim that it is inappropriate to include ANY threshold models (i.e., those that assume a single threshold value) when modeling epidemiological data. A non-threshold model for characterizing the population dose-response behavior is theoretically and practically the optimal approach (Document ID 3574, p. 13).

OSHA concludes that this German porcelain workers cohort shows evidence of silicosis among workers exposed at levels below the previous PELs, and that continued follow-up of this cohort would be likely to show greater silicosis risk among low-exposed workers due to the short follow-up time. Furthermore, the Chamber's characterization of Dr. Morfeld's result as “a threshold concentration of 250 μg/m3 below which the lung responses did not progress to silicosis” (Document ID 4224, p. 3) is plainly inaccurate, as the estimated exposures of a substantial proportion of the workers with silicosis in the data set did not exceed this level.

e. Park et al. (2002)

The ACC submitted comments on the Park et al. (2002, Document ID 0405) study which examined silicosis and lung disease other than cancer (i.e., NMRD) in a cohort of diatomaceous earth workers. The ACC's comments on this study are discussed in detail in Section V.D, Comments and Responses Concerning Silicosis and Non-Malignant Respiratory Disease Mortality and Morbidity, including comments relating to exposure-response thresholds in this study. Briefly, the ACC claimed that the Park et al. (2002) study is “fully consistent” with Morfeld's estimate of a threshold above the 100 μg/m3 concentration for NMRD, including silicosis, mortality (Document ID 2307, Attachment A, p. 107). However, NIOSH explained in its post-hearing brief that categorical analysis for NMRD indicated no threshold existed at or above a cumulative exposure corresponding to 25 μg/m3 over 40 years of exposure, which is below the cumulative exposure equivalent to the new PEL over 45 years (Document ID 4233, p. 27). Park et al. did not attempt to estimate a threshold below that level because the data lacked the power needed to discern a threshold (Document ID 4233, p. 27). OSHA agrees with NIOSH's assessment, which indicates that, if there is a cumulative exposure threshold for NMRD, including silicosis, it is significantly below the final PEL of 50 μg/m3.

f. Conclusion—Silicosis and NMRD

OSHA concludes that the body of epidemiological literature clearly demonstrates risk of silicosis and NMRD morbidity and mortality among workers who have been exposed to cumulative exposures or average exposure concentrations at or below the levels associated with the previous general industry PEL (100 μg/m3, or cumulative exposure of 4.5 mg/m3-yrs). Thus, OSHA does not agree with commenters who have stated that the previous general industry PEL is fully protective and that reducing it will yield no health benefits to silica-exposed workers (e.g., Document ID 4224, p. 2-5; Tr. 3582, pp. 1951-1963). Instead, the Agency finds that the evidence is at least as consistent with a finding that no threshold is discernible as it is with a finding that a threshold exists at some minimal level of exposure. The best available evidence also demonstrates silicosis morbidity and mortality below the previous PEL of 100 μg/m3, indicating that any threshold for silicosis (understood as an exposure level below which no one would develop disease), if one exists, is below that level. Even if the conclusion reached by Dr. Morfeld that a population average threshold exists above the level of the previous PEL is accurate, there will still be a substantial portion of the population who will develop silicosis from exposures below the identified “threshold.” These findings support OSHA's action in lowering the PEL to 50 μg/m3.

3. Thresholds—Lung Cancer

OSHA's Preliminary QRA and supplemental literature review included several studies that provide information on possible threshold effects for lung cancer. OSHA has determined that the epidemiological studies most relevant to the threshold issue are those with workers who have cumulative exposures or average exposure concentrations below the levels associated with the previous general industry PEL (100 μg/m3, or cumulative exposure of 4.5 mg/m3-yrs). As with the silicosis studies previously discussed, contrary to comments that OSHA only relied on studies involving exposures far above the levels of interest to OSHA in this rulemaking (e.g., Document ID 2307, Attachment A, pp. 94-95; 4226, p. 2), a number of lung cancer studies included workers who were exposed at levels close to or below the previous general industry PEL. Five of the 10 cohorts of workers in the pooled lung cancer risk analysis conducted by Steenland et al. (2001a) had median cumulative exposures below 4.5 mg/m3-yrs (the cumulative level associated with a working lifetime of exposure at the previous general industry PEL); four were also below 2.25 mg/m3-yrs (the cumulative level associated with a working lifetime of exposure at the revised PEL) and three had median silica concentrations below 100 μg/m3 (Document ID 0452, p. 775). Other lung cancer studies with significant numbers of relatively low-exposed workers include analyses of the Vermont granite workers (Attfield and Costello, 2004, Document ID 0285; Vacek et al., 2011, 1486) and industrial sand workers (McDonald et al., 2001, Document ID 1091; Hughes et al., 2001, 1060; McDonald et al., 2005, 1092) described in the previous discussion on silicosis. In addition to the epidemiological studies discussed here, in Section V.H, Mechanisms of Silica-Induced Adverse Health Effects, OSHA discussed studies that have shown direct genotoxic mechanisms by which exposure to crystalline silica at any level, with no threshold effect, may lead to lung cancer.Start Printed Page 16356

a. Steenland et al. Pooled Lung Cancer Study and Related Analyse

Steenland et al. (2001a) estimated excess lifetime risk of lung cancer based on a 10-cohort pooled study, which included several cohorts with significant numbers of workers with median cumulative and average exposures below those allowed by the previous general industry PEL (Document ID 0452). Results indicated that 45 years of exposure at 0.1 mg/m3 (100 μg/m3) would result in a lifetime risk of 28 excess lung cancer deaths per 1,000 workers (95% confidence interval (CI) 13-46 per 1,000). An alternative (non-linear) model yielded a lower risk estimate of 17 per 1,000 (95% CI 2-36 per 1,000).

A follow-up letter by Steenland and Deddens (2002, Document ID 1124) addressed the possibility of an exposure threshold effect in the pooled lung cancer analysis conducted by Steenland et al. in 2001. According to Dr. Steenland, “We further investigated whether there was a level below which there was no increase in risk, the so-called threshold. So we fit models that had a threshold versus those that didn't, and we explored various thresholds that might apply” (Document ID 3580, Tr. 1229). Threshold models using average exposure and cumulative exposure failed to show a statistically significant improvement in fit over models without a threshold. However, the authors found that when they used the log of cumulative exposure (a transformation commonly used to reduce the influence of high exposure points on a model), a threshold model with a 15-year lag fit better than a no-threshold model. The authors reported the best threshold estimate at 4.8 log mg/m3-days (Document ID 1124, p. 781), or an average exposure of approximately 10 μg/m3 over a 45-year working lifetime, one-fifth of the final PEL. Dr. Steenland explained what his analysis indicated regarding a cumulative exposure threshold for lung cancer: “we found, in fact, that there was a threshold model that fit better than a no-threshold model, not enormously better but better statistically, but that threshold was extremely low . . . far below the . . . silica standard proposed by OSHA” (Document ID 3580, Tr. 1229).

In response to comments from ACC Panel members Dr. Valberg and Dr. Long that the analysis presented by Steenland et al. showed a clear threshold at a level of cumulative exposure high enough to bear on OSHA's choice of PEL (Document ID 2330, p. 20), Dr. Steenland explained that their conclusion was based on a misreading of an illustration in his study:

[I]f you look at the figure, you see that the curve of the spline [a flexible, nonlinear exposure-response model] starts to go up around four on the log scale of microgram per meter cubed days. And if you transform that from the log to the regular scale, that is quite consistent with the threshold we got when we did a formal analysis using the log transform model [discussed above] (Document ID 3580, Tr. 1255).

The ACC representatives' comments do appear to be based on a misunderstanding of the figure in question, due to an error in Dr. Steenland's 2001 publication in which the axis of the figure under discussion was incorrectly labeled. This error was later corrected in an erratum (Document ID 3580, Tr. 1257; Steenland et al., 2002, Erratum. Cancer Causes Control, 13:777).

In addition, at OSHA's request, Drs. Steenland and Bartell (ToxaChemica, 2004, Document ID 0469) conducted a quantitative uncertainty analysis to examine the effects of possible exposure measurement error on the pooled lung cancer study results (see Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis). These analyses showed no evidence of a threshold effect for lung cancer at the final or previous PELs. Based on Dr. Steenland's work, therefore, OSHA believes that no-threshold models are appropriate for evaluating the exposure-response relationship between silica exposure and lung cancer. Even if commenters are correct that threshold models are preferable, the threshold is likely at a level of cumulative exposure significantly below what a worker would accumulate in 45 years of exposure at the final PEL, and is therefore immaterial to this rulemaking (see Document ID 1124, p. 781).

b. Vermont Granite Workers

In the Preliminary QRA and supplemental literature review, OSHA reviewed several studies on lung cancer among silica-exposed workers in the Vermont granite industry, whose exposures were reduced to relatively low levels due to a program for dust control initiated in 1938-1940 by the Vermont Division of Industrial Hygiene (Document ID 1711, pp. 97-102; 1711, Attachment 1, pp. 2-5; 1487, p. 73). As discussed above, Verma et al. (2012) reported that all jobs in the industry had average exposure concentrations at or below 100 μg/m3—most of them well below this level—in the time period 1950-2004 after implementation of exposure controls (Document ID 1487, Table IV, p. 75).

Attfield and Costello (2004) examined a cohort of 5,414 Vermont granite workers, including 201 workers who died of lung cancer (Document ID 0285, pp. 130, 134). In this study, cancer risk was elevated at cumulative exposure levels below 4.5 mg/m3-yrs, the amount of exposure that would result from a 45-year working lifetime of exposure at the previous PEL. The authors reported elevated lung cancer in all exposure groups, observing statistically significant elevation among workers with cumulative exposures between 0.5 and 1 mg/m3-yrs (p < 0.05), cumulative exposures between 2 and 3 mg/m3-yrs (p < 0.01), and cumulative exposures between 3 and 6 mg/m3-yrs (p < 0.05) (Document ID 0285, p. 135). These findings indicate that a threshold in exposure-response for lung cancer is unlikely at cumulative exposure levels associated with 45 years of exposure at the previous PEL and below.

Vacek et al. (2011) examined a group of 7,052 men, overlapping with the Attfield and Costello cohort, who worked in the Vermont granite industry at any time between January 1, 1947 and December 31, 1998 (Document ID 1486). Like Attfield and Costello, Vacek et al. reported significantly elevated lung cancer (p < 0.01) (Document ID 1486, p. 315). Most of the lung cancer cases in Vacek et al. (305/356) had cumulative exposures less than or equal to 4.1 mg/m3-yrs (Document ID 1486, p. 316), below the cumulative exposure level of 4.5 mg/m3-yrs associated with 45 years of exposure at the previous PEL and below. However, unlike Attfield and Costello, Vacek et al. did not find a statistically significant relationship of increasing lung cancer risk with increasing silica exposure, leading Vacek et al. to conclude that increased lung cancer mortality in the cohort may not have been due to silica exposure (Document ID 1486, p. 312).

The strengths and weaknesses of both studies and the differences between them that could account for their conflicting conclusions were discussed in great detail in Section V.F, Comments and Responses Concerning Lung Cancer Mortality. For the purpose of evaluating the effects of low concentrations of silica exposure, as well as whether a threshold exposure exists, OSHA believes the Attfield and Costello study may merit greater weight than Vacek et al. As discussed in Section V.F, Comments and Responses Concerning Lung Cancer Mortality, OSHA believes Attfield and Costello's choice to exclude the highest exposure group from their analysis likely improved their study's Start Printed Page 16357estimate of the exposure-response relationship at lower exposures; by making this choice, they limited the influence of highly uncertain exposure estimates at higher levels and helped to reduce the impact of the healthy worker survivor effect. The Agency acknowledges the strengths of the Vacek et al. analysis as well, including longer follow-up of workers.

In conclusion, OSHA does not find compelling evidence in these studies of Vermont granite workers of a cumulative exposure threshold for lung cancer in the exposure range below the previous general industry PEL. This conclusion is based on the statistically significant elevations in lung cancer reported in both cohorts described above, which were composed primarily of workers whose cumulative exposures were below the level associated with a working lifetime of exposure. However, OSHA acknowledges that a strong conclusion regarding a threshold is difficult to draw from these studies, due to the disagreement between Attfield and Costello and Vacek et al. regarding the likelihood that excess lung cancer among Vermont granite workers was due to their silica exposures.

c. Industrial Sand Workers

OSHA's Preliminary QRA (Document ID 1711, pp. 285-287) evaluated a 2001 case-control analysis of industrial sand workers including 2,640 men employed before 1980 for at least three years in one of nine North American sand-producing plants. One of the sites was a large associated office complex where workers' exposures were lower than those typically experienced by production workers (Hughes et al., 2001, Document ID 1060). A later update by McDonald et al. (2005, Document ID 1091) eliminated one plant, following 2,452 men from the 8 remaining U.S. plants. Both cohorts overlapped with an earlier industrial sand cohort, including 4,626 workers at 18 plants, which was included in Steenland et al.'s pooled analysis (2001a, Document ID 0452). OSHA noted that these studies (Hughes et al., 2001, Document ID 1060; McDonald et al., 2005, 1092; Steenland and Sanderson, 2001, 0455) showed similar exposure-response patterns of increased lung cancer mortality with increased exposure.

In the Final Peer Review Report, Dr. Ginsberg commented on the relevance of the industrial sand cohort studies, which included low-exposed workers with exceptionally well-characterized exposures, for threshold issues:

With respect to the body of silica epidemiology literature, perhaps the case with the least amount of measurement error is of US industrial sand workers wherein many measurements were made with filter samples and SRD determination of crystalline silica and in which there was very careful estimation of historical exposure for both silica and smoking (MacDonald et al. 2005; Steenland and Sanderson 2001; Hughes et al. 2001) (Document ID 3574, pp. 22-23).

OSHA agrees with Dr. Ginsberg's assessment of these studies and has found them to be particularly high quality. Thus, the Agency was especially interested in the studies' findings, which showed that cancer risk was elevated at cumulative exposure levels below 4.5 mg/m3-yrs, the amount of exposure that would result from a 45-year working lifetime of exposure at the previous PEL. OSHA believes these results provide strong evidence against a threshold in cumulative exposure at any level high enough to impact OSHA's choice of PEL. Dr. Ginsberg agrees with OSHA's conclusion (Document ID 3574, p. 23).

d. Other Studies

Comments submitted by the ACC briefly mentioned several epidemiological studies that, they claim, “suggest the existence of a threshold for any increased risk of silica-related lung cancer,” including studies by Sogl et al. (2012), Mundt et al. (2011), Pukkala et al. (2005), Calvert et al. (2003), Checkoway et al. (1997), and Steenland et al. (2001a). OSHA previously reviewed several of these studies in the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment, and the Supplemental Literature Review, though not with specific attention to their implications for exposure-response thresholds (Document ID 1711, pp. 139-155; 1711, Attachment 1, pp. 6-12). The studies cited by ACC are discussed below, with the exception of Steenland et al. (2001a), which was previously reviewed in this section.

e. German Porcelain Workers

OSHA reviewed Mundt et al. (2011, Document ID 1478) in its Supplemental Literature Review (Document ID 1711, Attachment 1, pp. 6-12). As discussed there, Mundt et al. examined the risks of silicosis morbidity and lung cancer mortality in a cohort of 17,644 German porcelain manufacturing workers who had participated in medical surveillance programs for silicosis between 1985 and 1987. This cohort was also examined in a previous paper by Birk et al. (2009, Document ID 4002, Attachment 2).

Quantitative exposure estimates for this cohort showed an average annual exposure of 110 μg/m3 for workers hired prior to 1960 and an average of 30 μg/m3 for workers hired after 1960. More than 40 percent of the cohort had cumulative exposures less than 0.5 mg/m3-yrs at the end of follow-up, and nearly 70 percent of the cohort had average annual exposures less than 50 μg/m3 (Mundt et al., 2011, Document ID 1478, pp. 283-284).

The lung cancer mortality hazard ratios (HRs) associated with average annual exposure were statistically significant in two of the four average annual exposure groups: 2.1 (95% CI 1.1-4.0) for average annual exposure group >50-100 μg/m3 and 2.4 (95% CI 1.1-5.2) for average annual exposure group >150-200 μg/m3, controlling for age, smoking, and duration of employment. In contrast, the HRs for lung cancer mortality associated with cumulative exposure were not statistically elevated after controlling for age and smoking.

The authors suggested the possibility of a threshold for lung cancer mortality. However, no formal threshold analysis for lung cancer was conducted in this study or in the follow-up threshold analysis conducted on this population by Morfeld et al. for silicosis (2013, Document ID 4175). Having reviewed this study carefully, OSHA believes it is inconclusive on the issue of thresholds due to the elevated risk of lung cancer seen among low-exposed workers (for example, those with average exposures of 50-100 μg/m3), which is inconsistent with the ACC's claim that a threshold exists at or above the previous PEL of 100 μg/m3, and due to several limitations which may preclude detection of a relationship between cumulative exposure and lung cancer in this cohort. As discussed in the Preliminary QRA, these include: (1) A strong healthy worker effect observed for lung cancer; (2) Mundt et al. did not follow the typical convention of considering lagged exposures to account for disease latency; and (3) the relatively young age of this cohort (median age 56 years old at time of silicosis determination) (Document ID 1478, p. 288) and limited follow-up period (average of 19 years per subject) (Birk et al. 2009, Document ID 4002, Attachment 2, p. 377). Only 9.2 percent of the cohort was deceased by the end of the follow up period. Mundt et al. (2011) acknowledged this limitation, stating that the lack of increased risk of lung cancer was a preliminary finding (Document ID 1478, p. 288).

f. German Uranium Miners

In pre-hearing comments, Dr. Morfeld described a study of 58,677 German uranium miners by Sogl et al. (2012, Start Printed Page 16358Document ID 3842; 2307, Attachment 2, p. 11). Dr. Morfeld noted that the study was based on a detailed exposure assessment of respirable crystalline silica (RCS) dust. According to Dr. Morfeld, Sogl et al. “showed that no lung cancer excess risk was observed at RCS dust exposure levels below 10 mg/m3-years” (Document ID 2307, Attachment 2, p. 11). OSHA's review of this publication confirmed that the authors reported a spline function with a single knot at 10 mg/m3-yrs, which Morfeld interprets to suggest a threshold for lung cancer of approximately 250 μg/m3 average exposure concentration for workers exposed over the course of 40 years. However, the authors also noted that an increase in risk below this level could not be ruled out due to strong confounding with radon, resulting in possible over-adjustment (Sogl et al., Document ID 3842, p. 9). That is, because workers with high exposures to silica would also have had high exposures to the lung carcinogen radon, the models used by Sogl et al. may have been unable to detect a relationship between silica and lung cancer in the presence of radon. As described previously, excess lung cancer has been observed among workers with lower cumulative exposures than the Sogl et al. “threshold” in other studies which do not suffer from confounding from potent lung carcinogens other than silica (for example, industrial sand workers), and which are, therefore, likely to provide more reliable evidence on the issue of thresholds. OSHA concludes that the Sogl et al. study does not provide convincing evidence of a cumulative exposure threshold for lung cancer.

g. U.S. Diatomaceous Earth Workers

Checkoway et al. (1997) investigated the risk of lung cancer among diatomaceous earth (DE) workers exposed to respirable cristobalite (a type of silica found in DE) (Document ID 0326; 1711, pp. 139-143). Exposure samples were collected primarily at one of the two plants in the study by plant industrial hygienists over a 40-year timeframe from 1948 to 1988 and used to estimate exposure for each individual in the cohort (Seixas et al., 1997, Document ID 0431, p. 593). Based on 77 deaths from cancer of the trachea, lung, and bronchus, the standardized mortality ratios (SMR) were 129 (95% CI 101-161) and 144 (95% CI 114-180) based on rates for U.S. and local county males, respectively (Document ID 0326, pp. 683-684). The authors found a positive, but not monotonic, exposure-response trend for lung cancer. The risk ratios for lung cancer with increasing quintiles of respirable crystalline silica exposure were 1.00, 0.96, 0.77, 1.26 and 2.15 with a 15-year exposure lag. Lung cancer mortality was thus elevated for workers with cumulative exposures greater than 2.1 mg/m3-yrs, but was only statistically significantly elevated for the highest exposure category (RR = 2.15; 95% CI 1.08-4.28) (Document ID 0326, p. 686). OSHA notes that this highest exposure category includes cumulative exposures only slightly higher than 4.5 mg/m3-yrs, the level of cumulative exposure resulting from a 45-year working lifetime at the previous PEL of 100 μg/m3. OSHA does not believe that the appearance of a statistically significantly elevated lung cancer risk in the highest category should be interpreted as evidence of an exposure-response threshold, especially in light of the somewhat elevated risk seen at lower exposure levels. OSHA believes it is more likely to reflect limited power to detect excess risk at lower exposure levels, a common issue in epidemiological studies which was emphasized by peer reviewer Dr. Andrew Salmon in relation to purported thresholds (Document ID 3574, p. 37).

h. Finnish Nationwide Job Exposure Matrix

OSHA reviewed Pukkala et al. (2005, Document ID 0412) in the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (Document ID 1711, pp. 153-154). As discussed there, Pukkala et al. (2005) evaluated the occupational silica exposure among all Finns born between 1906 and 1945 who participated in a national population census on December 31, 1970. Follow-up of the cohort was through 1995. Between 1970 and 1995, there were 30,137 cases of incident lung cancer among men and 3,527 among women. Exposure data from 1972 to 2000 was collected by the Finnish Institute of Occupational Health (FIOH). Cumulative exposure categories for respirable quartz were defined as: <1.0 mg/m3-yrs (low), 1.0-9.9 mg/m3-yrs (medium) and >10 mg/m3-yrs (high). For men, over 18 percent of the 30,137 lung cancer cases worked in occupations with potential exposure to silica dust. The cohort showed statistically significantly increased lung cancer among men in the lowest occupationally exposed group (those with less than 1.0 mg/m3-yrs cumulative silica exposure), as well as for men with exposures in the two higher groups (1.0-9.9 mg/m3-yrs and >10 mg/m3-yrs). For women, the cohort showed statistically significantly increased lung cancer among women with at least 1.0 mg/m3-yrs cumulative silica exposure. Given these results, it is unclear why ACC stated that Pukkula's results suggest that “excess risk of lung cancer is mainly attributable to . . . cumulative exposure exceeding 10 mg/m3-years” (Document ID 4209, p. 54). Indeed, Pukkula's analysis appears to show excess risk of lung cancer among men with any level of occupational exposure and among women whose cumulative exposures were quite low (at least equivalent to about 25 μg/m3 over 45 years). It does not support the ACC's contention that lung cancer is seen primarily in workers with exposures greater than 200 μg/m3 (Document ID 4209, p. 54), but rather suggests that any threshold for lung cancer risk would likely be well below 100 μg/m3.

i. U.S. National (27 states) Case-Control Study

As discussed in the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (Document ID 1711, pp. 152-153), Calvert et al. (2003, Document ID 3890) conducted a case-control study using 4.8 million death certificates from the National Occupational Mortality Surveillance data set. Death certificates were collected from 27 states covering the period from 1982 to 1995. Cases were persons who had died from any of several diseases of interest: Silicosis, tuberculosis, lung cancer, chronic obstructive pulmonary disease (COPD), gastrointestinal cancers, autoimmune-related diseases, or renal disease. Worker exposure to crystalline silica was categorized as no/low, medium, high, or super-high based on their industry and occupation. The authors acknowledged the potential for confounding by higher smoking rates for cases compared to controls, and partially controlled for this by eliminating white-collar workers from the control group in the analysis. Following this adjustment, the authors reported weak, but statistically significantly elevated, lung cancer mortality odds ratios (OR) of 1.07 (95% CI 1.06-1.09) and 1.08 (95% CI 1.01-1.15) for the high- and super-high exposure groups, respectively (Calvert et al., 2003, Document ID 3890, p. 126). Upon careful review of this study, OSHA maintains its position that it should not be used for quantitative risk analysis (including determination of threshold effects) because it lacks an exposure characterization based on sampling. Any determination regarding the existence or location of a threshold based on Calvert et al. (2003) must, therefore, be considered highly speculative.Start Printed Page 16359

j. Conclusion—Lung Cancer

In conclusion, OSHA has determined that the best available evidence on the issue of a threshold for silica-related lung cancer does not support the ACC's contention that an exposure-response threshold, below which respirable crystalline silica exposure is not expected to cause cancer, exists at or above the previous general industry PEL of 100 μg/m3. While there are some studies that claim to point to thresholds above the previous general industry PEL, multiple studies contradict this evidence, most convincingly through evidence that cohort members with low cumulative silica exposures suffered from lung cancer as a result of their exposure. These studies indicate that there is either no threshold for silica-related lung cancer, or that this threshold is at such a low level that workers cumulatively exposed at or below the level allowed by the new PEL of 50 μg/m3 will still be at risk of developing lung cancer. Thus, OSHA does not agree with commenters who have stated that the previous general industry PEL is fully protective and that reducing it will yield no health benefits to silica-exposed workers (e.g., Document ID 4224, p. 2-5; Tr. 3582, pp. 1951-1963).

4. Exposure Uncertainty and Thresholds

In his pre-hearing comments, Dr. Cox stated that the observation of a positive and monotonic exposure-response relationship in epidemiological studies “does not constitute valid evidence against the hypothesis of a threshold,” and that OSHA's findings of risk at exposures below the previous PEL for general industry “could be due simply to exposure misclassification” in studies of silica-related health effects in exposed workers (Document ID 2307, Attachment 4, pp. 41-42). His statements closely followed his analyses from a 2011 paper, in which Cox presented a series of simulation analyses designed to show that common concerns in epidemiological analyses, such as uncontrolled confounding, errors in exposure estimates, and model specification errors, can obscure evidence of an exposure-response threshold, if such a threshold exists (Document ID 3600, Attachment 7). Dr. Cox concluded that the currently available epidemiological studies “do not provide trustworthy information about the presence or absence of thresholds in exposure-response relations” with respect to an exposure concentration threshold for lung cancer (Document ID 3600, Attachment 7, p. 1548).

OSHA has reviewed Dr. Cox's comments and testimony, and concludes that uncertainty about risk due to exposure estimation and confounding cannot be resolved through the application of the statistical procedures recommended by Dr. Cox. (Similar comments from Dr. Cox about alleged biases in the studies relied upon are addressed in the next section, where OSHA reaches similar conclusions). A reviewer on the independent peer review panel, Dr. Ginsberg, commented that:

epidemiology studies will always have issues of exposure misclassification or other types of error that may create uncertainty when it comes to model specification. However, these types of error will also bias correlations to the null such that if they were sufficiently influential to obscure a threshold they may also substantially weaken regression results and underestimate the true risk (Document ID 3574, p. 23).

OSHA agrees with Dr. Ginsberg. As discussed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis, a “gold standard” exposure sample is not available for the epidemiological studies in the silica literature, so it is not possible to determine the direction or magnitude of the effects of exposure misclassification on OSHA's risk estimates. The silica literature is not unique in this sense. As stated by Mr. Robert Park of NIOSH, “modeling exposure uncertainty as described by Dr. Cox . . . is infeasible in the vast majority of retrospective observational studies. Nevertheless, mainstream scientific thought holds that valid conclusions regarding disease causality can still be drawn from such studies” (Document ID 4233, p. 32).

For the reasons discussed throughout this analysis of the scientific literature, OSHA concludes that, even acknowledging a variety of uncertainties in the studies relied upon, these uncertainties are, for the most part, typical or inherent in these types of studies. OSHA therefore finds that the weight of evidence in these studies, representing the best available evidence on the health effects of silica exposure, strongly supports the findings of significant risk from silicosis, NMRD, lung cancer, and renal disease discussed in this section and in the quantitative risk assessment that follows in the next section (see Benzene, 448 U.S. at 656 (“OSHA is not required to support its finding that a significant risk exists with anything approaching scientific certainty. Although the Agency's findings must be supported by substantial evidence, 29 U.S.C. 655(f), 6(b)(5) specifically allows the Secretary to regulate on the basis of the `best available evidence.' ”)).

5. Conclusion

In summary, OSHA acknowledges that common issues with epidemiological studies limit the Agency's ability to determine whether and where a threshold effect exists for silicosis and lung cancer. However, as shown in the foregoing discussion, there is evidence in the epidemiological literature that workers exposed to silica at concentrations and cumulative levels allowable under the previous general industry PEL not only develop silicosis, but face a risk of silicosis high enough to be significant ( >1 per 1,000 exposed workers). Although the evidence is less clear for lung cancer, studies nevertheless show excess cases of lung cancer among workers with cumulative exposures in the range of interest to OSHA. Furthermore, the statistical model-based approaches proposed in public comments do not demonstrate the existence or location of a “threshold” level of silica exposure below which silica exposure is harmless to workers. The above considerations lead the Agency to conclude that any possible exposure threshold is likely to be at a low level, such that some workers will continue to suffer the health effects of silica exposure even at the new PEL of 50 μg/m3.

There is a great deal of argument and analysis directed at the question of thresholds in silica exposure-response relationships, but nothing like a scientific consensus about the appropriate approach to the question has emerged. If OSHA were to accept the ACC's claim that exposure to 100 μg/m3 silica is safe for all workers (due to a threshold at or above an exposure concentration of 100 μg/m3) and set a PEL at 100 μg/m3 for all industry sectors, and if that claim is in fact erroneous, the consequences of that error to silica-exposed workers would be grave. A large population of workers would remain at significant risk of serious occupational disease despite feasible options for exposure reduction.

J. Comments and Responses Concerning Biases in Key Studies

OSHA received numerous comments and testimony, particularly from representatives of the ACC, regarding biases in the data that the Agency relied upon to conduct its Preliminary Quantitative Risk Assessment (Preliminary QRA). In this section, OSHA focuses on these comments regarding biases, particularly with respect to how such biases may have affected the data and findings from the Start Printed Page 16360key peer-reviewed, published studies that OSHA relied upon in its Preliminary QRA.

The data utilized by OSHA to conduct its Preliminary QRA came from published studies in the peer-reviewed scientific literature. When developing health standards, OSHA is not required or expected to conduct original research or wait for better data or new studies (see 29 U.S.C. 655(b)(5); e.g., United Steelworkers v. Marshall, 647 F.2d 1189, 1266 (D.C. Cir. 1980), cert. denied, 453 U.S. 913 (1981)). Generally, OSHA bases its determinations of significant risk of material impairment of health on the cumulative evidence found in a number of studies, no one of which may be conclusive by itself (see Public Citizen Health Research Group v. Tyson, 796 F.2d 1479, 1495 (D.C. Cir. 1986) (reviewing courts do not “seek a single dispositive study that fully supports the Administrator's determination . . . Rather, [OSHA's] decision may be fully supportable if it is based . . . on the inconclusive but suggestive results of numerous studies.”). OSHA's critical reading and interpretation of scientific studies is thus appropriately guided by the instructions of the Supreme Court's Benzene decision that “so long as they are supported by a body of reputable scientific thought, OSHA is free to use conservative assumptions in interpreting the data with respect to carcinogens, risking error on the side of overprotection rather than underprotection” (Industrial Union Dep't v. American Petroleum Inst., 448 U.S. 607, 656 (1980)).

Since OSHA is not a research agency, it draws from the best available existing data in the scientific literature to conduct its quantitative risk assessments. In most cases, with the exception of certain risk and uncertainty analyses prepared for OSHA by its contractor ToxaChemica, OSHA had no involvement in the data generation or analyses reported in those studies. Thus, in calculating its risk estimates, OSHA used published regression coefficients or equations from key peer-reviewed, published studies, but had no control over the actual published data; nor did the Agency have access to the raw data from such studies.

As discussed throughout Section V of this preamble, the weight of scientific opinion indicates that respirable crystalline silica is a human carcinogen that causes serious, life-threatening disease at the previously-permitted exposure levels. Under its statutory mandate, the Agency can and does take into account the potential for statistical and other biases to skew study results in either direction. However, the potential biases of concern to the commenters are well known among epidemiologists. OSHA therefore believes that the scientists who conduct the studies and subject them to peer review before publication have taken the potential for biases into account in evaluating the quality of the data and analysis. As discussed further below, OSHA heard testimony from David Goldsmith, Ph.D., describing how scientists use “absolutely the best evidence they can lay their hands on” and place higher value on studies that are the least confounded by other factors that, if unaccounted for, could contribute to the effect (e.g., lung cancer mortality). (Document ID 3577, Tr. 894-895). Dr. Goldsmith also testified that many of the assertions of biases put forth in the rulemaking docket are speculative in nature, with no actual evidence presented (Document ID 3577, Tr. 901). Thus, while taking seriously the critiques of the “body of reputable scientific thought” OSHA has used to support this final silica standard, the Agency finds no reason, as discussed below, to consider discredited in any material way its key conclusions regarding causation or significant risk of harm.

In his pre-hearing comments, Dr. Cox, on behalf of the ACC, claimed that the Preliminary QRA did not address a number of sources of potential bias:

The Preliminary QRA and the published articles that it relies on do not correct for well-known biases in modeling statistical associations between exposures and response. (These include study, data, and model selection biases; model form specification and model over-fitting biases; biases due to residual confounding, e.g., because age is positively correlated with both cumulative exposure and risk of lung diseases within each age category (typically 5 or more years long); and biases due to the effects of errors in exposure estimates on shifting apparent thresholds to lower concentrations). As a result, OSHA has not demonstrated that there is any non-random association between crystalline silica exposure and adverse health responses (e.g., lung cancer, non-malignant respiratory disease, renal disease) at exposure levels at or below 100 [µg/m3]. The reported findings of such an association, e.g., based on significantly elevated relative risks or statistically significant positive regression coefficients for exposed compared to unexposed workers, are based on unverified modeling assumptions and on ignoring uncertainty about those assumptions (Document ID 2307, Attachment 4, pp. 1-2).

These biases, according to Dr. Cox, nearly always result in false positives, i.e., finding that an exposure-response relationship exists when there really is no such relationship (Document ID 3576, Tr. 380). Although his comments appear to be directed to all published, peer-reviewed studies relied upon by OSHA in estimating risks, Dr. Cox admitted at the hearing that his statements about false positives were based on his review of the Preliminary QRA with relation to lung cancer only, and that he “[didn't] really know” whether the same allegations of bias he directed at the lung cancer studies are relevant to the studies of silica's other health risks (Document ID 3576, Tr. 426). In his comments, Dr. Cox discussed each source of bias in detail; OSHA will address them in turn. The concerns expressed by commenters, including Dr. Cox, about exposure uncertainty—another potential source of bias—are addressed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis.

1. Model Specification Bias

Dr. Cox stated that model specification error occurs when the model form, such as the linear absolute risk model, does not correctly describe the data (Document ID 2307, Attachment 4, p. 21). Using a simple linear regression example from Wikipedia, Dr. Cox asserted that common indicators of goodness-of-fit, including sum of square residuals and correlation coefficients, can be weak in identifying “nonlinearities, outliers, influential single observations, and other violations of modeling assumptions” (Document ID 2307, Attachment 4, pp. 52-53). He advocated for the use of diagnostic tests to check that a model is a valid and robust choice, stating, “[u]nfortunately, OSHA's Preliminary QRA and the underlying papers and reports on which it relies are not meticulous in reporting the results of such model diagnostics, as good statistical and epidemiological practice requires” (Document ID 2307, Attachment 4, p. 21). In his post-hearing brief, Dr. Cox further described these diagnostic tests to include plots of residuals, quantification of the effects of removing outliers and influential observations, and comparisons of alternative model forms using model cross-validation (Document ID 4027, p. 2). He also suggested using Bayesian Model Averaging (BMA) or other model ensemble methods to quantify the effects of model uncertainty (Document ID 4027, p. 3).

OSHA believes that guidelines for which diagnostic procedures should be performed, and whether and how they are reported in published papers, are best determined by the scientific community through the pre-publication peer review process. Many studies in Start Printed Page 16361the silica literature did not report the results of diagnostic tests. For example, the Vacek et al. (2009) study of lung cancer and silicosis mortality, which was submitted to the rulemaking record by the ACC to support its position, made no mention of the results of model diagnostic tests; rather, the authors simply stated that models were fitted by maximum likelihood, with the deviance used to examine model fitting (Document ID 2307, Attachment 6, pp. 11-12). As illustrated by this example, authors of epidemiological studies do not normally report the results of diagnostic tests; nor do such authors publish their raw data. Therefore, there is no data readily available to OSHA with which it could perform the diagnostic analysis that Dr. Cox states is necessary. If the suggestion is that no well-conducted epidemiological study that failed to report a battery of diagnostic tests or disclose what they showed should be relied upon for regulatory purposes, there would be virtually no body of scientific study left for OSHA to consider, raising the legal standard for issuing toxic substance standards far above what the Benzene decision requires. Despite this, OSHA maintains that, given the large number of peer-reviewed studies in the published scientific literature on crystalline silica, subjecting each model in each study to diagnostic testing along the lines advocated by Dr. Cox would not fundamentally change the collective conclusions when examining the literature base as a whole. Despite Dr. Cox's criticisms, the scientific literature that OSHA reviewed to draw its conclusions regarding material impairment of health and used in its quantitative risk assessment, constitutes the best available evidence upon which to base this toxic substance standard, in accordance with 29 U.S.C. 655(b) and the Benzene decision and subsequent case law.

Dr. Cox's other suggested approach to addressing model uncertainty, BMA, can be used to construct a risk estimate based on multiple exposure-response models. Unlike BMA, standard statistical practice in the epidemiological literature is to evaluate multiple possible models, identify the model that best represents the observations in the data set, and use this model to estimate risk. In some cases, analysts may report the results of two or more models, along with their respective fit statistics and other information to aid model selection for risk assessment and show the sensitivity of the results to modeling choices (e.g., Rice et al., 2001, Document ID 1118). These standard approaches were used in each of the studies relied on by OSHA in its Preliminary QRA.

In contrast, BMA is a probabilistic approach designed to account for uncertainty inherent in the model selection process. The analyst begins with a set of possible models (Mi) and assigns each a prior probability (Pr[Mi]) that reflects the analyst's initial belief that model Mi represents the true exposure-response relationship. Next, a data set is used to update the probabilities assigned to the models, generating the posterior probability for each model. Finally, the models are used in combination to derive a risk estimate that is a composite of the risk estimates from each model, weighted by each model's posterior probability (see Viallefont et al., 2001, Document ID 3600, Attachment 34, pp. 3216-3217). Thus, BMA combines multiple models, and uses quantitative weights accounting for the analyst's belief about the plausibility of each model, to generate a single weighted-average risk estimate. These aspects of BMA are regarded by some analysts as improvements to the standard approaches to exposure-response modeling.

However, Kyle Steenland, Ph.D., Professor, Department of Environmental Health, Rollins School of Public Health, Emory University, the principal author of a pooled study that OSHA heavily relied upon, noted that BMA is not a standard method for risk assessment. “[Bayesian] model averaging, to my knowledge, has not been used in risk assessment ever. And so, sure, you could try that. You could try a million things. But I think OSHA has correctly used standard methods to do their risk assessment and [BMA] is not one of those standard methods” (Document ID 3580, Tr. 1259).

Indeed, BMA is a relatively new method in risk analysis. Because of its novelty, best practices for important steps in BMA, such as defining the class of models to include in the analysis, and choosing prior probabilities, have not been developed. Until best practices for BMA are established, it would be difficult for OSHA to conduct and properly evaluate the quality of BMA analyses. Evaluation of the quality of available analyses is a key step in the Agency's identification of the best available evidence on which to base its significant risk determination and benefits analysis.

OSHA also emphasizes that, as noted by Dr. Steenland, scientifically accepted and standard practices were used to estimate risk from occupational exposure to crystalline silica (Document ID 3580, Tr. 1259). Thus OSHA has decided that it is not necessary to use BMA in its QRA, and that the standard statistical methods used in the studies it relies upon to estimate risk are appropriate as a basis for risk estimation. OSHA notes that it is possible to incorporate risk estimates based on more than one model in its risk assessment by presenting ranges of risk, a strategy often used by OSHA when the best available evidence includes more than one model, analytical approach, or data set. In its Preliminary QRA, OSHA presented ranges of risks for silica-related lung cancer and silicosis based on different data sets and models, thus further lessening the utility of using more complex techniques such as BMA. OSHA continued this practice in its final risk assessment, presented in Section VI, Final Quantitative Risk Assessment and Significance of Risk.

2. Study Selection Bias

Another bias described by Dr. Cox is study selection bias, which he stated occurs when only studies that support a positive exposure-response relationship are included in the risk assessment, and when criteria for the inclusion and exclusion of studies are not clearly specified in advance (Document ID 2307, Attachment 4, pp. 22-23). Dr. Cox noted the criteria used by OSHA to select studies, as described in the Supplemental Literature Review of Epidemiological Studies on Lung Cancer Associated with Exposure to Respirable Crystalline Silica (Supplemental Literature Review) (Document ID 1711, Attachment 1, p. 29). Dr. Cox, however, claimed that OSHA did not apply these criteria consistently, in that there may still be exposure misclassification or confounding present in the studies OSHA relied upon to estimate the risk of the health effects evaluated by the Agency (Document ID 2307, Attachment 4, pp. 24-25). Similarly, the American Foundry Society (AFS), in its post-hearing brief, asserted that, “No formal process is described for search criteria or study selection” and that OSHA's approach of identifying studies based upon the IARC (1997) and NIOSH (2002) evaluations of the literature “is a haphazard approach that is not reproducible and is subject to bias. Moreover it appears to rely primarily on information that is more than 10 years old” (Document ID 4229, p. 4).

OSHA disagrees with the arguments presented by Dr. Cox and the AFS, as did some commenters. The American Public Health Association (APHA), in its post-hearing brief, expressed strong Start Printed Page 16362support for OSHA's study selection methods. Dr. Georges Benjamin, Executive Director, wrote, “APHA recognizes that OSHA has thoroughly reviewed and evaluated the peer-reviewed literature on the health effects associated with exposure to respirable crystalline silica. OSHA's quantitative risk assessment is sound. The agency has relied on the best available evidence and acted appropriately in giving greater weight to those studies with the most robust designs and statistical analyses” (Document ID 2178, Attachment 1, p. 1). Similarly, Dr. Steenland testified that “OSHA has done a very capable job in conducting the summary of the literature” (Document ID 3580, Tr. 1235).

In response to the criticisms by Dr. Cox and the AFS, OSHA notes that the silica literature was exhaustively reviewed by IARC in 1997 and NIOSH in 2002 (Document ID 1062; 1110). As a result, there was no need for OSHA to initiate a new review of the historical literature. Instead, OSHA used the IARC and NIOSH reviews as a starting point for its own review. As recognized by the APHA, OSHA evaluated and summarized many of the studies referenced in the IARC and NIOSH reviews, and then performed literature searches to identify new studies published since the time of the IARC and NIOSH reviews. OSHA clearly described this process in its Review of Health Effects Literature: “OSHA has included in its review all published studies that the Agency deems relevant to assessing the hazards associated with exposure to respirable crystalline silica. These studies were identified from numerous scientific reviews that have been published previously such as the IARC (1997) and NIOSH (2002) evaluations of the scientific literature as well as from literature searches and contact with experts and stakeholders” (Document ID 1711, p. 8). For its Preliminary QRA, OSHA relied heavily on the IARC pooled exposure-response analyses and risk assessment for lung cancer in 10 cohorts of silica-exposed workers (Steenland et al., 2001a, Document ID 0452) and multi-center study of silicosis mortality (Mannetje et al., 2002b, Document ID 1089). As stated in the Review of Health Effects Literature, these two studies “relied on all available cohort data from previously published epidemiological studies for which there were adequate quantitative data on worker exposures to crystalline silica to derive pooled estimates of disease risk” (Document ID 1711, p. 267).

In addition to relying on these two pooled IARC multi-center studies, OSHA also identified single cohort studies with sufficient quantitative information on exposures and disease incidence and mortality rates. As pointed out by Dr. Cox, OSHA described the criteria used for selection of the single cohort studies of lung cancer mortality:

OSHA gave studies greater weight and consideration if they (1) included a robust number of workers; (2) had adequate length of follow-up; (3) had sufficient power to detect modest increases in lung cancer incidence and mortality; (4) used quantitative exposure data of sufficient quality to avoid exposure misclassification; (5) evaluated exposure-response relationships between exposure to silica and lung cancer; and (6) considered confounding factors including smoking and exposure to other carcinogens (Document ID 1711, Attachment 1, p. 29).

Using these criteria, OSHA identified four single-cohort studies of lung cancer mortality that were suitable for quantitative risk assessment; two of these cohorts (Attfield and Costello, 2004, Document ID 0285; Rice et al., 2001, 1118) were included among the 10 used in the IARC multi-center study and two appeared later (Hughes et al., 2001, Document ID 1060; Miller and MacCalman, 2009, 1306) (Document ID 1711, p. 267). For NMRD mortality, in addition to the IARC multi-center study (Mannetje et al., 2002b, Document ID 1089), OSHA relied on Park et al. (2002) (Document ID 0405), who presented an exposure-response analysis of NMRD mortality (including silicosis and other chronic obstructive pulmonary diseases) among diatomaceous earth workers (Document ID 1711, p. 267). For silicosis morbidity, several single-cohort studies with exposure-response analyses were selected (Chen et al., 2005, Document ID 0985; Hnizdo and Sluis-Cremer, 1993, 1052; Steenland and Brown, 1995b, 0451; Miller et al., 1998, 0374; Buchanan et al., 2003, 0306) (Document ID 1711, p. 267).

With respect to Dr. Cox's claim that OSHA did not apply its criteria consistently, on the basis that there may still be exposure misclassification or confounding present, OSHA notes that it selected studies that best addressed the criteria; OSHA did not state that it only selected studies that addressed all of the criteria. Given the fact that some of the epidemiological studies concern exposures of worker populations dating back to the 1930's, there is always some potential for exposure misclassification or the absence of information on smoking. When this was the case, OSHA discussed these limitations in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711). For example, OSHA discussed the lack of smoking information for cases and controls in the Steenland et al. (2001a, Document ID 0452) pooled lung cancer analysis (Document ID 1711, pp. 150-151).

With respect to the AFS's claim that OSHA relied on studies that were more than 10 years old, OSHA again notes that it reviewed, in its Review of Health Effects Literature and its Supplemental Literature Review, the studies in the silica literature and selected the ones that best met the criteria described above (Document ID 1711; 1711, Attachment 1). It would be improper to only select the most recent studies, particularly if the older studies are of higher quality based on the criteria. Furthermore, the studies OSHA relied upon in its Preliminary QRA were published between 1993 and 2009; the claim that OSHA primarily relied on older studies is thus misleading, when the studies were of relatively recent vintage and determined to be of high quality based on the criteria described above. The AFS also suggested that OSHA examine several additional foundry studies of lung cancer (Document ID 2379, Attachment 2, p. 24); OSHA retrieved all of these suggested studies, added them to the rulemaking docket following the informal public hearings, and discusses them in Section V.F, Comments and Responses Concerning Lung Cancer Mortality.

3. Data Selection Bias

A related bias presented by Dr. Cox is data selection bias, which he stated occurs when only a subset of the data is used in the analysis “to guarantee a finding of a positive” exposure-response relationship (Document ID 2307, Attachment 4, p. 26). He provided an example, the Attfield and Costello (2004, Document ID 0285) study of lung cancer mortality, which excluded data as a result of attenuation observed in the highest exposure group (Document ID 2307, Attachment 4, pp. 26-27). Attenuation of response means the exposure-response relationship leveled off or decreased in the highest exposure group. Referring to another study of the same cohort, Vacek et al. (2009, Document ID 2307, Attachment 6; 2011, 1486), Dr. Cox stated, “OSHA endorses the Attfield and Costello findings, based on dropping cases that do not support the hypothesis of an ER [exposure-response] relation for lung cancer, while rejecting the Vacek et al. study that included more complete data (that was not subjected to post hoc subset selection) but that did not find a significant ER [exposure-response] Start Printed Page 16363relation” (Document ID 2307, Attachment 4, pp. 26-27).

OSHA believes there are very valid reasons for the observance of attenuation of response in the highest exposure group that would justify the exclusion of data in Attfield and Costello (2004, Document ID 0285) and other studies. This issue was discussed by Gary Ginsberg, Ph.D., an OSHA peer reviewer from the Connecticut Department of Public Health, in his post-hearing comments. Dr. Ginsberg noted that several epidemiological studies have found an attenuation of response at higher doses, with possible explanations including: (1) Measurement error, which arises from the fact that the highest doses are associated with the oldest datasets, which are most prone to measurement error; (2) “intercurrent causes of mortality” from high dose exposures that result in death to the subject prior to the completion of the long latency period for cancer; and (3) the healthy worker survivor effect, which occurs when workers with ill health leave the workforce early (Document ID 3574, p. 24). As discussed in Section V.F, Comments and Responses Concerning Lung Cancer Mortality, OSHA disagrees strongly with Dr. Cox's assertion that data were excluded to ensure a positive exposure-response relationship (Document ID 2307, Attachment 4, p. 26). In addition, as detailed in Section VI, Final Quantitative Risk Assessment and Significance of Risk, OSHA calculated quantitative risk estimates for lung cancer mortality from several other studies that did not rely on a subset of the data (Rice et al., 2001, Document ID 1118; Hughes et al., 2001, 1060; Miller and MacCalman, 2009, 1306; ToxaChemica, 2004, 0469; 1711, p. 351). These studies also demonstrated positive exposure-response relationships.

4. Model Selection Bias

Another selection bias presented by Dr. Cox is model selection bias, which he said occurs when many different combinations of models, including alternative exposure metrics, different lags, alternative model forms, and different subsets of data, are tried with respect to their “ability to produce `significant'-looking regression coefficients” (Document ID 2307, Attachment 4, p. 27). This is another aspect of model specification error, as discussed above under model averaging. Dr. Cox wrote:

This type of multiple testing of hypotheses and multiple comparisons of alternative approaches, followed by selection of a final choice based [on] the outcomes of these multiple attempts, completely invalidates the claimed significance levels and confidence intervals reported for the final ER [exposure-response] associations. Trying in multiple ways to find a positive association, and then selecting a combination that succeeds in doing so and reporting it as `significant,' while leaving the nominal (reported) statistical significance level of the final selection unchanged (typically at p=0.05), is a well-known recipe for producing false-positive associations (Document ID 2307, Attachment 4, p. 28).

Dr. Cox further stated that unless methods of significance level reduction (i.e., reducing the nominal statistical significance level of the final selection) are used, the study is biased towards false-positive results (Document ID 2307, Attachment 4, p. 28).

During the informal public hearings, counsel for the ACC asked Mr. Park of NIOSH's Risk Evaluation Branch about this issue, i.e., trying a number of modeling choices, including exposure metrics, log-transformations, lag periods, and model subsets (Document ID 3579, Tr. 149-150). Mr. Park's reply supports the use of multiple modeling choices in the risk assessment as a form of sensitivity analysis:

Investigations like this look at a number of options. They come into the study not totally naïve. They, in fact, have some very strong preference even before looking at the data based on prior knowledge. So cumulative exposure, for example, is a generally very high confidence choice in a metric. Trying different lags is interesting. It helps validate the study because you know what it ought to look like sort of. And in many cases, the choice does not make a lot of difference. So it's kind of a robust test, and similarly, the choice of the final model is not just coming in naïve. A linear exposure response has a lot of biological support in many different contexts, but it could be not the best choice (Document ID 3579, Tr. 150-151).

ACC counsel further asked, “And does one at the end of this process, though, make any adjustment in what you consider to be the statistically significant relationship in light of the fact that you've looked at so many different models and arrangements?” (Document ID 3579, Tr. 151-152). Mr. Park replied, “No, I don't think that's a legitimate application of a multiple comparison question” (Document ID 3579, Tr. 152). OSHA agrees with Mr. Park that significance level reduction is not appropriate in the context of testing model forms for risk estimation, and notes that, in the Agency's experience, significance level reduction is not typically performed in the occupational epidemiology literature. In addition, OSHA notes that, in many of the key studies relied upon by the Agency to estimate quantitative risks, the authors presented the results of multiple models that showed statistically significant exposure-response relationships. For example, Rice et al. (2001) presented the results of six model forms, with all except one being significant (Table 1, Document ID 1118, p. 41). Attfield and Costello (2004) presented the results of their model with and without a 15-year lag and log transformation, with many results being significant (Table VII, Document ID 0285, p. 135). Thus, OSHA concludes that model selection bias is not a problem in its quantitative risk assessment.

Furthermore, OSHA disagrees with Dr. Cox's assertion that modeling choices are used to “produce `significant'-looking regression coefficients” (Document ID 2307, Attachment 4, p. 27). OSHA believes that the investigators of the studies it relied upon in its Preliminary, and now final, QRA made knowledgeable modeling choices based upon the exposure distribution and health outcome being examined. For example, in long-term cohort studies, such as those of lung cancer mortality relied upon by OSHA, most authors relied upon cumulative exposure (mg/m3-yrs or mg/m3-days), i.e., the concentration of crystalline silica exposure (mg/m3) multiplied by the duration of exposure (years or days), as an exposure metric. Consistent with standard statistical techniques used in epidemiology, the cumulative exposure metric may then be log-transformed to account for an asymmetric distribution with a long right tail, or attenuation, and the metric may be lagged by several years to account for the long latency period between the exposure and the development of lung cancer. When investigators use subsets of the data, they typically explain the rationale and the effect of using the subset in the analysis. These choices all have important justifications and are not used purely to produce the authors' desired results, as Dr. Cox suggested (Document ID 2307, Attachment 4, p. 27).

5. Model Uncertainty Bias

Related to model selection bias is Dr. Cox's assertion of model uncertainty bias, which he said occurs when many different models are examined and then one is selected on which to base risk calculations; this approach “treats the finally selected model as if it were known to be correct, for purposes of calculating confidence intervals and significance levels. But, in reality, there remains great uncertainty about what the true causal relation between exposure and response looks like (if there is one)” (Document ID 2307, Start Printed Page 16364Attachment 4, pp. 28-29). He further stated that ignoring this bias leads to artificially narrow confidence intervals, which bias conclusions towards false-positive findings. He then cited a paper (Piegorsch, 2013, included in Document ID 3600) describing statistical methods for overcoming this bias by “including multiple possible models in the calculation of results” (Document ID 2307, Attachment 4, p. 29). OSHA concludes this bias is really an extension of model specification error and model selection bias, previously discussed, and maintains that best practices for model averaging have not yet been established, making it difficult for the Agency to conduct and properly evaluate the quality of BMA analyses.

6. Model Over-Fitting Bias

Next, Dr. Cox discussed model over-fitting bias, which he said occurs when the same data set is used both to fit a model and to assess the fit; this “leads to biased results: Estimated confidence intervals are too narrow (and hence lower confidence limits on estimated ER [exposure-response] slopes are too high); estimated significance levels are too small (i.e., significance is exaggerated); and estimated measures of goodness-of-fit overstate how well the model fits the data” (Document ID 2307, Attachment 4, p. 39). He suggested using appropriate statistical methods, such as “k-fold cross-validation,” to overcome the bias (Document ID 2307, Attachment 4, p. 39).

OSHA does not agree that using the same data set to fit and assess a model necessarily results in an over-fitting bias. The Agency understands over-fitting to occur when a model is excessively complex relative to the amount of data available such that there are a large number of predictors relative to the total number of observations available. For survival models, it is the number of events, i.e., deaths, that is relevant, rather than the size of the entire sample (Babyak, 2004, included in Document ID 3600, p. 415). If the number of predictors (e.g., exposure, age, gender) is small relative to the number of events, then there should be no bias from over-fitting. In an article cited and submitted to the rulemaking docket by Dr. Cox, Babyak (2004) discussed a simulation study that found, for survival models, an unacceptable bias when there were fewer than 10 to 15 events per independent predictor (included in Document ID 3600, p. 415). In the studies that OSHA relied on in its Preliminary QRA, there were generally a large number of events relative to the small number of predictors. For example, in the Miller and MacCalman (2009) study of British coal miners, in the lung cancer model using both quartz and coal dust exposures, there was a large number of events (973 lung cancer deaths) relative to the few predictors in the model (quartz exposure, coal dust exposure, cohort entry date, smoking habits at entry, cohort effects, and differences in regional background cause-specific rates) (Document ID 1306, pp. 6, 9). Thus, OSHA does not agree the studies it relied upon were substantially influenced by over-fitting bias. OSHA also notes that k-fold cross-validation, as recommended by Dr. Cox, is not typically reported in published occupational epidemiology studies, and that the studies the Agency relied upon in the Preliminary QRA were published in peer-reviewed journals and used statistical techniques typically used in the field of occupational epidemiology and epidemiology generally.

7. Residual Confounding Bias

Dr. Cox also asserted a bias due to residual confounding by age. Bias due to confounding occurs in an epidemiological study, in very general terms, when the effect of an exposure is mixed together with the effect of another variable (e.g., age) not accounted for in the analysis. Residual confounding occurs when additional confounding factors are not considered, control of confounding is not precise enough (e.g., controlling for age by using groups with age spans that are too wide), or subjects are misclassified with respect to confounders (Document ID 3607, p. 1). Dr. Cox stated in his comments that:

key studies relied on by OSHA, such as Park et al. (2002), do not correct for biases in reported ER [exposure-response] relations due to residual confounding by age (within age categories), i.e., the fact that older workers may tend to have both higher lung cancer risks and higher values of occupational exposure metrics, even if one does not cause the other. This can induce a non-causal association between the occupational exposure metrics and the risk of cancer (Document ID 2307, Attachment 4, p. 29).

The Park et al. (2002) study of non-malignant respiratory disease mortality, which Dr. Cox cited as not considering residual confounding by age, used 13 five-year age groups (<25, 25-29, 30-34, etc.) in the models (Document ID 0405, p. 37). Regarding this issue in the Park et al. (2002) study, in its post-hearing comments, NIOSH stated:

This is a non-issue. The five-year categorization was used only for deriving the expected numbers of cases as an offset in the Poisson analysis using national rates which typically are classified in five-year intervals (on age and chronological time). The cumulative exposures were calculated with a 10-day resolution over follow-up and then averaged across observation time within 50 cumulative exposure levels cross-classified with the five-year age-chronological time cells of the classification table. There would be virtually no confounding between age and exposure [using this approach] (Document ID 4233, p. 33).

OSHA agrees with this assessment, noting that it appears that age groups were adequately constructed to prevent residual confounding. OSHA thus rejects this assertion of residual confounding by age in the Park et al. (2002) study.

8. Summary of Biases

In summary, OSHA received comments and heard testimony on potential biases in the studies upon which it relied for its QRA. The ACC's Dr. Cox, in particular, posited a long list of biases, including model form specification bias, study selection bias, data selection bias, model selection bias, model over-fitting bias, model uncertainty bias, residual confounding bias, and bias as a result of exposure measurement error. OSHA, in this section, has specifically addressed each of these types of bias (except for bias due to exposure estimation error, which is addressed in Section V.K, Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis).

In addition, OSHA heard testimony that countered the claims of biases and their potential to cause false positive results. When asked about the biases alleged by Dr. Cox and Dr. Long, Dr. Goldsmith testified, “All of these other things, it seems to me, are smoke screens for an inability to want to try and see what the body of evidence really shows” (Document ID 3577, Tr. 895-896). Later in his testimony, when asked about exposure misclassification, Dr. Goldsmith similarly noted, “[a]nd for a lot of the arguments that are being put forward by industry, they are speculating that there is the potential for these biases, but they haven't gotten, [from] my perspective, the actual evidence that this is the case” (Document ID 3577, Tr. 901). Similarly, OSHA has reviewed the record evidence extensively and is not aware of any specific, non-speculative evidence of biases in the studies that it relied upon.

There also is a question of the extent to which Dr. Cox actually reviewed all of the studies that he asserted to be biased. Upon questioning from Anne Ryder, Attorney in the Office of the Solicitor, Department of Labor, Dr. Cox admitted that he had not examined the Start Printed Page 16365issue of silica and silicosis, and that his statements about false positives were based on his review of the Preliminary QRA with relation to lung cancer only:

MS. RYDER: . . . You talked a little bit earlier about the false positives that are . . . present with a lot of the studies on lung cancer. And, but I believe, in your comment you didn't say that there are any of those same false positives with studies dealing with silicosis and silica exposure. Is that correct?

DR. COX: I don't think I opined on that. So—and I really haven't looked carefully at the question. I do take it as given that silica at sufficiently high and prolonged exposures causes silicosis. I've not really examined that literature.

MS. RYDER: So you don't think that those studies have the same issues that some of the lung cancer studies have?

DR. COX: I don't really know (Document ID 3576, Tr. 426).

Dr. Cox further testified, regarding the likelihood that the conclusions of the Preliminary QRA for silicosis are correct, “I expect that the evidence is much stronger for silica and silicosis. But I haven't reviewed it, so I can't testify to it” (Document ID 3576, Tr. 427).

OSHA believes this testimony to be inconsistent with some of the broad conclusions in Dr. Cox's pre-hearing written submission to the rulemaking record, in which he claimed that all adverse outcomes in the Preliminary QRA may have been affected by false positives. Dr. Cox concluded in this submission that:

These multiple uncontrolled sources of false-positive bias can generate findings of statistically “significant” positive ER [exposure-response] associations even in random data, or in data for which there is no true causal relation between exposure and risk of adverse health responses. Because OSHA's Preliminary QRA and the studies on which it relies did not apply appropriate technical methods (which are readily available, as discussed in the references) to diagnose, avoid, or correct for these sources of false-positive conclusions, the reported findings of “significantly” positive ER [exposure-response] associations between crystalline silica exposures at and below the current PEL and adverse outcomes (lung cancer, non-malignant lung disease, renal disease) are not different from what might be expected in the absence of any true ER [exposure-response] relations. They therefore provide no evidence for (or against) the hypothesis that a true ER [exposure-response] relation exists. Thus, OSHA has not established that a non-random association exists between crystalline silica exposures at or below the current PEL and the adverse health effects on which it bases its determination of significant risk and calculates supposed health effect benefits (Document ID 2307, Attachment 4, pp. 29-30).

OSHA notes that “non-malignant lung disease” includes silicosis, studies of which Dr. Cox subsequently testified that he did not examine.

In conclusion, the studies relied upon by OSHA for its risk assessment were peer-reviewed and used methods for epidemiology and risk assessment that are commonly used. Dr. Cox provided no study-specific evidence (e.g., data re-analysis) to support his comments that the studies OSHA relied upon were adversely affected by numerous different types of bias. As described above, OSHA recognizes that there are uncertainties associated with the results of the studies relied on for its risk assessment, as is typically the case for epidemiological studies such as these. Nevertheless, as previously stated, OSHA maintains that it has used a body of peer-reviewed scientific literature that, as a whole, constitutes the best available evidence of the relationship between respirable crystalline silica exposure and silicosis, lung cancer, and the other health effects studied by the Agency in promulgating this final rule.

K. Comments and Responses Concerning Exposure Estimation Error and ToxaChemica's Uncertainty Analysis

Exposure estimation error, a typical feature of epidemiological studies, occurs when the authors of an exposure-response study construct estimates of the study subjects' exposures using uncertain or incomplete exposure data. Prior to the publication of its Preliminary Quantitative Risk Assessment (Preliminary QRA), the Agency commissioned an uncertainty analysis conducted by Drs. Kyle Steenland and Scott Bartell, through its contractor, ToxaChemica, Inc., to address exposure estimation error in OSHA's risk assessment, and incorporated the results into the Preliminary QRA. After reviewing comments submitted to the record on the topic of exposure estimation error, OSHA maintains that it has relied upon the best available evidence by: (1) Using high-quality exposure-response studies and modeling approaches; (2) performing an uncertainty analysis of the effect of exposure estimation error on the risk assessment results; and (3) further submitting that analysis to peer review. OSHA concludes from its uncertainty analysis that exposure estimation error did not substantially affect the results in the majority of studies examined (Document ID 1711, pp. 299-314).

Furthermore, having carefully considered the public comments criticizing ToxaChemica's uncertainty analysis, OSHA has concluded that it was not necessary to conduct additional analyses to modify the approach adopted by Drs. Steenland and Bartell in the uncertainty analysis. Nor was it necessary to incorporate additional sources of uncertainty in the analysis. Also, given the evidence in the rulemaking record that these estimation errors bias results towards underestimating rather than overestimating the risks from exposure in many circumstances, it is very unlikely that regression coefficients and risk estimates from all of the different studies relied on in the Preliminary QRA were biased upward. Accordingly, OSHA remains convinced that the conclusions of the Agency's risk assessment are correct and largely unaffected by potential error in exposure measurement.

OSHA received significant comments on the topic of exposure estimation error in the studies it relied on in its Review of Health Effects Literature and Preliminary QRA (Document ID 1711). A number of commenters discussed the importance of accounting for exposure estimation error. Dr. Cox, representing the ACC, described exposure estimation error as perhaps the “most quantitatively important” issue in the studies OSHA relied upon (Document ID 2307, Attachment 4, p. 40). Similarly, Christopher M. Long, Sc.D., Principal Scientist at Gradient, representing the U.S. Chamber of Commerce (Chamber), testified that exposure measurement error is a “common source of uncertainty in most occupational and environmental epidemiologic studies” (Document ID 3576, Tr. 298). According to Dr. Long, this type of error can lead to inaccurate risk estimates by creating error in the exposure-response curve derived from a data set and obscuring the presence of a threshold (Document ID 3576, Tr. 300; see Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases, for further discussion on thresholds). Dr. Long further stated that exposure measurement error can lead to over- or under-estimation of risk: “the impact of exposure measurement error . . . can bias either high or low. It can bias towards the null. It can be a source of positive bias.” (Document ID 3576, Tr. 358-359). A bias to the null in an exposure-response model used in a quantitative risk assessment is an underestimation of the relationship between exposure level and the rate of the disease or health effect of interest, and results in underestimation of risk.

OSHA agrees with the assessments of the ACC and the Chamber with respect to the importance of exposure Start Printed Page 16366measurement error. Indeed, OSHA peer reviewer, Dr. Gary Ginsberg, in his peer review comments (Document ID 3574, p. 21), and OSHA's risk assessment contractor, Dr. Steenland, in his hearing testimony (Document ID 3580, Tr. 1266-1267), noted the potential for exposure measurement error to bias exposure-response coefficients towards the null. Dr. Steenland explained: “misclassification I would say in general tends to bias things to the null. It's harder to see positive exposure-response trends in the face of misclassification. It depends partly on the type of error. . . . But, on the whole, I would say that exposure measurement tends to bias things down rather than up” (Document ID 3580, Tr. 1266-1267). Fewell et al., the authors of a paper on residual confounding submitted by the ACC, wrote, “It is well recognized that under certain conditions, nondifferential measurement error in the exposure variable produces bias towards the null” (2007, Document ID 3606, p. 646).

Several commenters representing the ACC challenged the methods used in ToxaChemica's uncertainty analysis on the grounds that the analysis failed to adequately address exposure estimation error. In spite of their criticisms, critics were unable to supply better studies than those OSHA used. Indeed, when asked during the hearing, Dr. Long was unable to identify any studies that the Agency could use that acceptably account for the impact of exposure measurement error on exposure-response associations for crystalline silica (Document ID 3576, Tr. 356-357), and none was supplied following the hearings.

Taking into account the record evidence discussed above, OSHA concludes that it is possible for exposure measurement error to lead to either over- or under-estimation of risk and that this issue of exposure measurement error is not specific to the silica literature. It further concludes that industry representatives could not identify, and failed to submit, any published epidemiological studies of occupational disease that corrected for such bias to their satisfaction (Document ID 3576, Tr. 356-357).

Nevertheless, because OSHA agreed that an analysis of exposure estimation error as a source of uncertainty is important, it commissioned the uncertainty analysis discussed above to explore the potential effects of exposure measurement error on the conclusions of OSHA's risk assessment (Document ID 0469). The analysis examined the potential effects of exposure measurement error on the mortality risk estimates derived from the pooled studies of lung cancer (Steenland et al. 2001a, Document ID 0452) and silicosis (Mannetje 2002b, Document ID 1089). This included the effects of estimation error on the detection and location of a possible threshold effect in exposure-response models.

The uncertainty analysis OSHA commissioned from Drs. Steenland and Bartell (2004, Document ID 0469) addressed possible error in silica exposure estimates from: (1) Random error in individual workers' exposure estimates and (2) error in the conversion of dust measurements (typically particle count concentrations) to gravimetric respirable silica concentrations, which could have affected estimates of average exposure for job categories in the job-exposure matrices used to estimate workers' silica exposure. To address possible error in individual workers' exposure estimates, the analysts performed a Monte Carlo analysis, a type of simulation analysis which varies the values of an uncertain input to an analysis (in this case, exposure estimates) to explore the effects of different values on the outcome of the analysis. The Monte Carlo analysis sampled new values for workers' job-specific exposure levels from distributions they believed characterized the exposures of individual workers in each job. In each run of the Monte Carlo analysis, the sampled exposure values were used to calculate new estimates of each worker's cumulative exposures, and the resulting set was used to fit a new exposure-response model.

Similarly, the analysts performed a Monte Carlo analysis to address the issue of uncertainty in conversion from dust to respirable silica exposure, sampling new conversion factors from a normal distribution with means equal to the original conversion factor, calculating new estimates of workers' cumulative exposures, and re-fitting the exposure-response model for each Monte Carlo run. To examine the sensitivity of the model to the joint effects of both error types, the analysts ran 50 Monte Carlo simulations using the sampling procedure for both individual exposures and job-specific conversion factors. They also examined the effects of systematic bias in conversion factors, considering that these may have been consistently under-estimated or over-estimated for any given cohort. They addressed possible biases in either direction, conducting 20 simulations where the true silica content was assumed to be either half or double the estimated silica content of measured exposures.

The results of their analysis indicated that the conclusions of the pooled lung cancer study conducted previously by Steenland et al. (Document ID 0452) and included in OSHA's Preliminary QRA were unlikely to be affected by the types of exposure estimation error examined by Drs. Steenland and Bartell, whose analysis of the underlying data was itself reviewed by OSHA's peer review panel. As explained below, after reviewing comments critical of the uncertainty analysis, OSHA reaffirms its conclusion that workers exposed to silica at the previous PELs are at significant risk of disease from their exposure.

Drs. Long and Valberg, representing the Chamber, commented that Drs. Steenland and Bartell's uncertainty analysis did not address all potential sources of error and variability in exposure measurement, such as possible instrument error; possible sampling error; random variability in exposure levels; variability in exposure levels resulting from changes in worker job functions during work shifts, production process changes, or control system changes; variability in sampler type used; variability in laboratory methods for determining sampling results and laboratory errors; variability in duration of exposure sampling; variability in sampling locations; variability in reasons for sample data collection (e.g., compliance sampling, periodic sampling, random survey sampling); variability in type of samples collected (e.g., bulk samples, respirable dust samples); variation among workers and over time in the size distribution, surface area, recency of fracture, and other characteristics of the particles inhaled; and extrapolation of exposure sampling data to time periods for which sampling data are not available (Document ID 2330, pp. 4-5). OSHA notes that these sources of potential error and variability are common in occupational exposure estimation, and are sources of uncertainty in most epidemiological studies, a point with which Drs. Valberg and Long agree (Document ID 2330, p. 14).

OSHA has determined that its reliance on the best available evidence provided it with a solid, scientifically sound foundation from which to conclude that exposure to crystalline silica poses a significant risk of harm, notwithstanding the various uncertainties inherent in epidemiology generally or potentially affecting any given study and that no studies exist entirely free from the types of data limitations or error and variability Drs. Valberg and Long identified. During the public hearing Dr. Long acknowledged Start Printed Page 16367that OSHA had not overlooked studies that he believed adequately addressed the sources of error cited in his comments. He was also unable to provide examples of such analyses in the silica literature, or in any other area of occupational epidemiology (Document ID 3576, Tr. 355-358; see also Document ID 3577, Tr. 641, 648 (testimony of Dr. Kenneth Mundt)). Additionally, Drs. Valberg and Long's critique of Drs. Steenland and Bartell's uncertainty analysis ignores constraints on the available data and reasonable limits on the analysts' ability to investigate the full variety of possible errors and their potential effects on OSHA's risk assessment.

OSHA additionally notes that Dr. Kenneth Crump, an OSHA peer reviewer, in his examination of ToxaChemica's (Document ID 0469) study of exposure uncertainty in the Steenland et al. pooled study, opined that it was sound. He further observed that the “analysis of error conducted by [ToxaChemica] is a very strong effort. The assumptions are clearly described and the data upon [which] they are based appear to be appropriate and appropriately applied.” Dr. Crump was careful to note, however, that “there are questions, as there will always be with such an analysis . . . A major source of error that apparently was not accounted for is in assuming that the average measure of exposure assigned to a job is the true average” (Document ID 3574, pp. 161-162). Dr. Cox referenced Dr. Crump's comment in his own pre-hearing comments, in the context of a discussion on the importance of exposure uncertainty in OSHA's risk analysis (Document ID 2307, p. 40). OSHA addressed this particular criticism in the Review of Health Effects Literature and Preliminary QRA. There, it stated that it is possible that some job exposure estimates were above or below the true average for a job; however, there was no “gold standard” measurement available to appropriately test or adjust for this potential source of error (Document ID 1711, p. xv). The Agency further stated that the uncertainty, or sensitivity, analysis included potential error in job averages, and found that most cohorts in the lung cancer and silicosis mortality pooled studies were not highly sensitive to random or systematic error in job-average exposure estimates (Document ID 1711, pp. 303-314). In his final evaluation of OSHA's response to his comments of 2009, Dr. Crump stated, “I believe that my comments have been fairly taken into account in the current draft and I have no further comments to make” (Document ID 3574, p. 17).

Similarly, Dr. Morfeld, representing the ACC, criticized Drs. Steenland and Bartell for performing only 50 simulations of workplace exposures as part of the uncertainty analysis (Document ID 2307, Attachment 2, p. 10). Peer reviewer Mr. Bruce Allen also remarked that this type of uncertainty analysis typically requires more than 50 simulations (Document ID 3574, p. 114). However, as stated by OSHA in the response to peer review section of the Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 379-400), the results did not appear to change much with an increased number of simulations. Thus, OSHA has concluded that the sensitivity findings would not have changed substantially by running more simulations. Indeed, in the final peer review report conveying his evaluation of OSHA's response to his comments of 2009, Mr. Allen stated that OSHA adequately addressed his comments in the updated risk assessment (Document ID 3574, p. 5).

The overall salient conclusion that OSHA draws from this peer-reviewed analysis is that even in those cohorts where exposure error had some impact on exposure-response models for lung cancer or silicosis, the resulting risk estimates at the previous and new PELs remain clearly significant. Therefore, OSHA continues to rely on, and have confidence in, the risk analysis it had performed. In particular, OSHA concludes that Drs. Steenland and Bartell's modeling choices were based on the best available data from a variety of industrial sources and, through their uncertainty analysis, reached conclusions that survive the ACC and Chamber criticisms of the study methodology. OSHA further concludes that it is not necessary to conduct additional analysis to modify the approach adopted by Drs. Steenland and Bartell or to incorporate additional sources of exposure estimation uncertainty in the analysis.

OSHA also disagrees with other specific criticisms that Drs. Long and Valberg made concerning the uncertainty analysis. Dr. Long testified that “there are no formal analyses conducted to determine the error structures of the three sources of exposure measurement error included in the sensitivity analyses; for example, without any formal analysis, the OSHA assessment simply assumed a purely Berkson type error structure from the assignment of job-specific average exposure levels for individual exposures” (Document ID 3576, 304-305).[9] Dr. Cox expressed a similar concern that

OSHA has not developed an appropriate error model specifically for the exposure estimates in the crystalline silica studies and has not validated (e.g., using a validation subset) that any of the ad hoc error models that they discuss describes the real exposure estimate errors of concern. They have also provided no justification for ToxaChemica's assumption of a log-normal distribution without outliers or mixtures of different distributions . . . and have provided no rationale for the assumption that a=0.8*p (Document ID 2307, Attachment 4, p. 45).

OSHA disagrees with Dr. Long's and Dr. Cox's characterizations, which implies that Drs. Steenland and Bartell did not adequately investigate the patterns of error in the data available to them. As noted in their 2004 report and by Dr. Steenland during the public hearings, ToxaChemica did not have the internal validation data (true exposures for a subset of the data set) that would be required to conduct formal analyses or validation of the error structure within each cohort of the pooled analysis (Document ID 0469, p. 16; 3580, pp. 1229-1231). Such data are not often available to analysts. However, Drs. Steenland and Bartell researched and reviewed worker exposure and dust composition data from several worksites to inform the error structures used in their analyses. For example, their analysis of individual workers' exposure data from the pooled analyses' industrial sand cohort formed the basis of the equation used for the exposure error simulation, which Dr. Cox represented as an assumption lacking any rationale. Drs. Steenland and Bartell also reviewed a number of studies characterizing the distribution of conversion factors across and within jobs at different worksites. OSHA concludes that Drs. Steenland and Bartell made a strong effort to collect data to inform their modeling choices, and that their choices were based on the Start Printed Page 16368best available information on error structure.

Dr. Long stated that “another limitation of the [ToxaChemica uncertainty] assessment was its assumption of log-linear . . . types of models, including log linear models with log-transformed exposure variables, and it focused on cumulative measures of silica exposure that obscure both within-person and between-person variability in exposure rates” (Document ID 3576 pp. 305-306). Dr. Long's assertion regarding the choice of exposure models is incorrect, as the sensitivity analysis was not limited to log-linear models. It included models with flexibility to capture nonlinearities in exposure-response, including spline analyses and categorical analyses, and log-transformation of the exposure variable was used only in the lung cancer analysis where it was shown in the original pooled analysis to better fit the data and address issues of heterogeneity between cohorts (Document ID 0469). Drs. Steenland and Bartell found only slight differences between the adjusted exposure-response estimates for each type of model.

Drs. Long and Valberg also contended that the cumulative exposure metric used in the Steenland and Bartell pooled study did not sufficiently allow for examination of the effects of exposure measurement uncertainty on the results of OSHA's risk assessment, because other exposure metrics could be more relevant. OSHA disagrees. As discussed in Section V.M, Comments and Responses Concerning Working Life, Life Tables, and Dose Metric, cumulative exposure is widely acknowledged by health experts as a driver of chronic diseases such as silicosis and lung cancer, has been found to fit the exposure-response data well in many studies of silicosis and lung cancer in the silica literature, and best fit the exposure-response data in the underlying pooled data sets to which Drs. Steenland and Bartell applied their subsequent uncertainty analyses. Thus, OSHA believes it was appropriate for this investigation of exposure estimation error to focus on the cumulative exposure metric, for reasons including data fit and general scientific understanding of this disease.

Furthermore, Dr. Long's concern that the choice of cumulative silica exposure might “obscure within-person variability in exposure rates” is not well supported in the context of lung cancer and silicosis mortality. Because death from these diseases typically occurs many years after the exposure that caused it, and complete records of past exposures do not typically exist, it is very difficult, using any metric, to trace within-person exposure variability (that is, changes in a person's exposure over time); these factors, not the choice of cumulative exposure metric, make it difficult to address variability in individuals' exposures over time and their effects on risk. OSHA notes that some analysts have explored the use of other exposure metrics in threshold analyses, submitting studies to the record which the Agency has reviewed and discussed in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases.

Dr. Long also testified that “[t]here's very little discussion in the OSHA report regarding the potential impacts of exposure measurement error on identification of thresholds . . . [ToxaChemica's 2004 report] noted that exposure-response threshold estimates are imprecise and appear to be highly sensitive to measurement errors” (Document ID 3576 p. 306). Dr. Cox further noted that exposure misclassification can “create the appearance of a smooth, monotonically increasing estimated ER [exposure-response] relation” and shift thresholds to the left (Document ID 2307, Attachment 4, pp. 41-42); that is, create the appearance that a threshold effect occurs at a lower exposure level than would be seen in a data set without exposure misclassification.

In their uncertainty analysis, Drs. Steenland and Bartell estimated an exposure-response threshold for the pooled cohorts in each of the 50 runs conducted for their lung cancer analysis. They defined the “threshold” as the highest cumulative exposure for which the estimated odds ratio was less than or equal to 1.0, reporting a mean value of 3.04 mg/m3-days and median of 33.5 mg/m3-days across the 50 runs (Document ID 0469, p. 15). The authors observed that “[t]hese estimates are somewhat lower than the original estimate (Steenland and Deddens 2002) of a threshold at 121 mg/m3-days (4.8 on the log scale), which translates to about 0.01 mg/m3 [10 µg/m3] over a working 30-year lifetime (considering a 15-year lag), or 0.007 [7µg/m3] over a 45-year lifetime without considering a 15-year lag” (Document ID 0469, p. 15). These exposure levels are about one-fifth the PEL of 50 μg/m3 included in the final standard.

As noted by Dr. Long, the threshold estimates were highly variable across the 50 iterations (SD of 1.64 on the log scale), in keeping with other comments received by OSHA that estimates of exposure-response thresholds based on epidemiological data tend to be highly sensitive to sources of measurement error and other issues common to epidemiological investigations (see Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases). However, the Agency notes that the results of the uncertainty analysis, suggesting a possible cumulative exposure threshold at approximately one-fifth the final 50 μg/m3 PEL, provide no cause to doubt OSHA's determination that significant risk exists at both the previous and the revised PEL.

An additional concern raised by Dr. Cox was based on his misunderstanding that the equation used to characterize the relationship between true and observed exposure in Drs. Steenland and Bartell's simulation, “Exposuretrue = Exposureobserved + E”, concerned cumulative exposure. Dr. Cox stated that the equation is “inappropriate for cumulative exposures [because] both the mean and the variance of actual cumulative exposure received typically increase in direct proportion to duration” (Document ID 2307, Attachment 4, p. 45). That is, the longer period of time over which a cumulative exposure is acquired, the higher variance is likely to be, because cumulative exposure is the sum of the randomly varying exposures received on different days. However, the exposures referred to in the equation are the mean job-specific concentrations recorded in the job-exposure matrix (Exposureobserved) and individuals' actual exposure concentrations from each job worked (Exposuretrue), not their cumulative exposures (Document ID 0469, p. 11). Therefore, Dr. Cox's criticism is unfounded.

Dr. Cox additionally criticized the simulation analysis on the basis that “[t]he usual starting point for inhalation exposures [is] with the random number of particles inhaled per breath modeled as a time-varying (non-homogenous) Poisson process . . . It is unclear why ToxaChemica decided to assume (and why OSHA accepted the assumption) of an underdispersed distribution . . . rather than assuming a Poisson distribution” (Document ID 2307, Attachment 4, pp. 45-46). OSHA believes this criticism also reflects a misunderstanding of Drs. Steenland and Bartell's analysis. While it could be pertinent to an analysis of workers' silica dose (the amount of silica that enters the body), the analysis addresses the concentration of silica in the air near a worker's breathing zone, not internal dose. The worker's airborne concentration is the regulated exposure endpoint and the exposure of interest for OSHA's risk assessment. Thus, the uncertainty analysis does not need to Start Printed Page 16369account for the number of particles inhaled per breath.

More broadly, Dr. Cox asserted that the Monte Carlo analysis “is an inappropriate tool for analyzing the effects of exposure measurement error on estimated exposure-response data,” citing a paper by Gryparis et al. (2009) (Document ID 2307, Attachment 4, p. 44). This paper indicates that by randomly simulating exposure measurement error, the Monte Carlo approach can introduce classical error (Document ID 3870, p. 262). Peer reviewer Dr. Noah Seixas similarly commented that “[t]he typical Monte Carlo simulation, which is what appears to have been done, would introduce classical error,” that is, error which is independent of the unobserved variable (in this case, the true exposure value). He explained that, as a result, “the estimated risks [from the simulation analyses] are most likely to be underestimates, or conservatively estimating risk. This is an important aspect of measurement error with significant implications for risk assessment and should not be overlooked.” (Document ID 3574, pp. 116-117). Addressing Dr. Cox's broader point, Dr. Seixas in his peer review stated that the “simulation of exposure measurement error in assessing the degree of bias that may have been present is a reasonable approach to assessing this source of uncertainty” (Document ID 3574, pp. 116). Dr. Crump similarly characterized the uncertainty analysis used in the Steenland and Bartell study as “a strong effort” that “appropriately applied” this method (Document ID 3574, pp. 161-162). In this regard, OSHA generally notes that the advantages and limitations of various methods to address exposure measurement error in exposure-response models is an area of ongoing investigation in risk assessment. As shown by the comments of OSHA's peer reviewers above, there is no scientific consensus to support Dr. Cox's opinion that the Monte Carlo analysis is an inappropriate approach to analyze the effects of exposure measurement error.

In conclusion, through use of high quality studies and modeling, performance of an uncertainty analysis, and submission of the results of that analysis to peer review, OSHA maintains that it has relied upon the best available evidence. In addition, OSHA has carefully considered the public comments criticizing ToxaChemica's uncertainty analysis and has concluded that exposure estimation error did not substantially affect the results in the majority of studies examined (Document ID 1711, pp. 299-314). As a result, it was not necessary to conduct additional analyses modifying the approach adopted by Drs. Steenland and Bartell. Accordingly, OSHA reaffirms its determination that the conclusions of the Agency's risk assessment are correct and largely unaffected by potential error in exposure measurement.

L. Comments and Responses Concerning Causation

As discussed in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA, OSHA finds, based upon the best available evidence in the published, peer-reviewed scientific literature, that exposure to respirable crystalline silica increases the risk of silicosis, lung cancer, other non-malignant respiratory disease (NMRD), and renal and autoimmune effects. Exposure to respirable crystalline silica causes silicosis and is the only known cause of silicosis. For other health endpoints like lung cancer that have both occupational and non-occupational sources of exposure, OSHA used a comprehensive weight-of-evidence approach to evaluate the published, peer-reviewed scientific studies in the literature to determine their overall quality and whether there is substantial evidence that exposure to respirable crystalline silica increases the risk of a particular health effect. For example, with respect to lung cancer, OSHA reviewed 60 epidemiological studies covering more than 30 occupational groups in over a dozen industrial sectors and concluded that exposure to respirable crystalline silica increases the risk of lung cancer (Document ID 1711, pp. 77-170). This conclusion is consistent with that of the World Health Organization's International Agency for Research on Cancer (IARC), HHS' National Toxicology Program (NTP), the National Institute for Occupational Safety and Health (NIOSH), and many other organizations and individuals, as evidenced in the rulemaking record and discussed throughout this section.

In spite of this, and in addition to asserting that OSHA's Preliminary QRA was affected by many biases, Dr. Cox, on behalf of the ACC, argued that OSHA failed to conduct statistical analyses of causation, which led to inaccurate conclusions about causation. He specifically challenged OSHA's reliance upon the IARC determination of carcinogenicity, as discussed in Section V.F, Comments and Responses Concerning Lung Cancer Mortality, and its use of the criteria for evaluating causality developed by the noted epidemiologist Bradford Hill (Document ID 2307, Attachment 4, pp. 13-14; 4027, p. 28). The Hill criteria are nine aspects of an association that should be considered when examining causation: (1) The strength of the association; (2) the consistency of the association; (3) the specificity of the association; (4) the temporal relationship of the association; (5) the biological gradient (i.e., dose-response curve); (6) the biological plausibility of the association; (7) coherency; (8) experimentation; and (9) analogy (Document ID 3948, pp. 295-299).

Instead, Dr. Cox suggested that OSHA use the methods listed in Table 1 of his 2013 paper, “Improving causal inferences in risk analysis,” which he described as “the most useful study designs and methods for valid causal analysis and modeling of causal exposure-response (CER) relations” (Document ID 2307, Attachment 4, p. 11). Because OSHA did not use these methods, Dr. Cox maintained that the Agency's Preliminary QRA “asserts causal conclusions based on non-causal studies, data, and analyses” (Document ID 2307, Attachment 4, p. 3). He also contended that OSHA “ha[d] conflated association and causation, ignoring the fact that modeling choices can create findings of statistical associations that do not predict correctly the changes in health effects (if any) that would be caused by changes in exposures” (Document ID 2307, Attachment 4, p. 3). He claimed that “[t]his lapse all by itself invalidates the Preliminary QRA's predictions and conclusions” (Document ID 2307, Attachment 4, p. 3). As discussed below, since OSHA's methodology and conclusions regarding causation are based on the best available evidence, they are sound. Consequently, Dr. Cox's contrary position is unpersuasive.

1. IARC Determination

Dr. Cox asserted that OSHA erred in its reliance on the IARC determination of carcinogenicity for crystalline silica inhaled in the forms of quartz or cristobalite. He believed OSHA only relied on the IARC findings because they aligned with the Agency's opinion, noting that the “IARC analysis involved some of the same researchers, same methodological flaws, and same gaps in explicit, well-documented derivations of benefits and conclusions as OSHA's own preliminary QRA” (Document ID 2307, Attachment 4, pp. 13-14). OSHA, however, relied on IARC's determination to include lung cancer in its quantitative risk assessment because it constitutes the best available evidence. For this reason, Dr. Cox's position is without merit and OSHA's Start Printed Page 16370findings are supported by substantial evidence in the record and reasonable.

As discussed in Section V.F, Comments and Responses Concerning Lung Cancer Mortality, the IARC classifications and accompanying monographs are well recognized in the scientific community, and have been described by scientists as “the most comprehensive and respected collection of systematically evaluated agents in the field of cancer epidemiology” (Demetriou et al., 2012, Document ID 4131, p. 1273). IARC's conclusions resulted from a thorough expert committee review of the peer-reviewed scientific literature, in which crystalline silica dust, in the form of quartz or cristobalite, was classified as Group 1, “carcinogenic to humans,” in 1997 (Document ID 2258, Attachment 8, p. 210). Since the publication of these conclusions, the scientific community has reaffirmed their soundness. In March of 2009, 27 scientists from eight countries participated in an additional IARC review of the scientific literature and reaffirmed that crystalline silica dust is a Group 1 carcinogen, i.e., “carcinogenic to humans” (Document ID 1473, p. 396). Additionally, the HHS' U.S. National Toxicology Program also concluded that respirable crystalline silica is a known human carcinogen (Document ID 1164, p. 1).

Further supporting OSHA's reliance on IARC's determination of carcinogenicity for its quantitative risk assessment is testimony offered by scientists during the informal public hearings. This testimony highlighted IARC's carcinogenicity determinations as very thorough examinations of the scientific literature that demonstrate that exposure to respirable crystalline silica causes lung cancer. For example, when asked about Dr. Cox's causation claims during the informal public hearings, David Goldsmith, Ph.D., noted that causation was very carefully examined by IARC. He believed that IARC, in its 1997 evaluation of evidence for cancer and silica, “. . . chose . . . the best six studies that were the least confounded for inability to control for smoking or other kinds of hazardous exposures like radiation and asbestos and arsenic . . .” (Document ID 3577, Tr. 894-896). He also believed it “. . . crucial . . . that we pay attention to those kinds of studies, that we pay attention to the kinds of studies that were looked at by the IARC cohort that Steenland did from 2001. That's where they had the best evidence” (Document ID 3577, Tr. 894-896).

Regarding IARC's evaluation of possible biases and confounders in epidemiological studies, as well as its overall determination, Frank Mirer, Ph.D., of CUNY School of Public Health, representing the AFL-CIO, testified:

IARC has active practicing scientists review—I've been on two IARC monographs, but not these monographs, monograph working groups. It's been dealt with. It's been dealt with over a week of intense discussion between the scientists who are on these committees, as to whether there's chance bias in confounding which might have led to these results, and by 1987 for foundries and 1997 for silica, and it's been decided and reaffirmed.

So people who don't believe it are deniers, pure and simple. This is the scientific consensus. I was on the NTP Board of Scientific Counselors when we reviewed the same data. Known to be a human carcinogen. Once you know it's a human carcinogen from studies in humans, you can calculate risk rates (Document ID 3578, Tr. 937).

That OSHA relied on the best available evidence to draw its conclusions was also affirmed by Dr. Cox's inability to provide additional studies that would have cast doubt on the Agency's causal analysis. Indeed, during the informal public hearings, Kenneth Crump, Ph.D., an OSHA peer reviewer from the Louisiana Tech University Foundation, asked Dr. Cox if he could identify “any causal studies of silica that they [OSHA] should have used but did not use?” Dr. Cox responded: “I think OSHA could look at a paper from around 2007 of Brown's, on some of the issues and causal analysis, but I think the crystalline silica area has been behind other particulate matter areas . . . in not using causal analysis methods. So no, I can't point to a good study that they should have included but didn't” (Document ID 3576, Tr. 401-402). In light of the above, OSHA maintains that in relying on IARC's determination of carcinogenicity, its conclusions on causation are rooted in the best available evidence.

2. Bradford Hill Criteria and Causality

Dr. Cox also challenged OSHA's use of Hill's criteria for causation. He claimed that the Bradford Hill considerations were neither necessary nor sufficient for establishing causation, which was his reason for failing to include them in the statistical methods listed in Table 1 of his written comments for objectively establishing evidence about causation (Document ID 4027, p. 28). As explained below, based on its review of the record, OSHA finds this position meritless, as it is unsupported by the best available evidence.

As a preliminary matter, Hill's criteria for causation (Document ID 3948) are generally accepted as a gold standard for causation in the scientific community. Indeed, OSHA heard testimony during the informal public hearings and received post-hearing comments indicating that Dr. Cox's assertion that statistical methods should be used to establish causality is not consistent with common scientific practice. For example, Andrew Salmon, Ph.D., an OSHA peer reviewer, wrote:

The identification of causality as opposed to statistical association is, as described by Bradford Hill in his well-known criteria, based mainly on non-statistical considerations such as consistence, temporality and mechanistic plausibility: the role of statistics is mostly limited to establishing that there is in fact a quantitatively credible association to which causality may (or may not) be ascribed. OSHA correctly cites the substantial body of evidence supporting the association and causality for silicosis and lung cancer following silica exposure, and also quotes previous expert reviews (such as IARC). The causal nature of these associations has already been established beyond any reasonable doubt, and OSHA's analysis sufficiently reflects this (Document ID 3574, p. 38).

Similarly, Kyle Steenland, Ph.D., Professor, Department of Environmental Health, Rollins School of Public Health, Emory University, in response to a question about Dr. Cox's testimony on causation from Darius Sivin, Ph.D., of the UAW Health and Safety Department, stated that the Bradford Hill criteria are met for lung cancer and silicosis:

[M]ost of the Bradford Hill criteria apply here. You know you can never prove causality. But when the evidence builds up to such an extent and you have 100 studies and they tend to be fairly consistent, that's when we draw a causal conclusion. And that was the case for cigarette smoke in lung cancer. That was the case for asbestos in lung cancer. And when the evidence builds up to a certain point, you say, yeah, it's a reasonable assumption that this thing causes, X causes Y (Document ID 3580, pp. 1243-1244).

As a follow-up, OSHA asked if Dr. Steenland felt that the Bradford Hill criteria were met for silica health endpoints. Dr. Steenland replied, “For silicosis or for lung cancer. I had said they're met for both” (Document ID 3580, p. 1262).

Gary Ginsberg, Ph.D., an OSHA peer reviewer, agreed with Dr. Steenland, remarking to Dr. Cox during questioning, “I'm a little dumbfounded about the concern over causality, given all the animal evidence” (Document ID 3576, Tr. 406). Mr. Park from NIOSH's Risk Evaluation Branch, in his question to Dr. Cox, echoed the sentiments of Dr. Ginsberg, stating:

Start Printed Page 16371

It's ludicrous to hear someone question causality. There's 100 years of research in occupational medicine, in exposure assessment. People here even in industry would agree that silica they say causes silicosis, which causes lung cancer. There's some debate about whether the middle step is required. There's no question that there's excess lung cancer in silica-exposed populations. We look at literature, and we identify what we call good studies. Good studies are ones that look at confounding, asbestos, whatever. We make judgments. If there's data that allows one to control for confounding, that's part of the analysis. If there is confounding that we can't control for, we evaluate it. We ask how bad could it be? There's a lot of empirical judgment from people who know these populations, know these exposures, know these industries, who can make very good judgments about that. We aren't stupid. So I don't know where you're coming from (Document ID 3576, Tr. 410-411).

Indeed, Kenneth Mundt, Ph.D., testifying on behalf of the International Diatomite Producers Association (part of the ACC Crystalline Silica Panel, which included Dr. Cox), and whose research study was the basis for the Morfeld et al. (2013, Document ID 3843) paper that reportedly identified a high exposure threshold for silicosis, also appeared to disagree with Dr. Cox's view of causation. Dr. Mundt testified that while he thought he could appreciate Dr. Cox's testimony, at some point there is sufficiently accumulated evidence of a causal association; he concluded, “I think here, over time, we've had the advantage with the reduction of exposure to see reduction in disease, which I think just makes it a home run that the diseases are caused by, therefore can be prevented by appropriate intervention” (Document ID 3577, Tr. 639-640).

OSHA notes that Dr. Cox, upon further questioning by Mr. Park, appeared to concede that exposure to respirable crystalline silica causes silicosis; Dr. Cox stated, “I do not question that at sufficiently high exposures, there are real effects” (Document ID 3576, Tr. 412). Later, when questioned by Anne Ryder, an attorney in the Solicitor of Labor's office, he made a similar statement: “I do take it as given that silica at sufficiently high and prolonged exposures causes silicosis” (Document ID 3576, Tr. 426). Based upon this testimony of Dr. Cox acknowledging that silica exposure causes silicosis, OSHA interprets his concern with respect to silicosis to be not one of causation, but rather a concern with whether there is a silicosis threshold (i.e., that exposure to crystalline silica must generally be above some level in order for silicosis to occur). Indeed, OSHA peer reviewer Brian Miller, Ph.D., noted in his post-hearing comments that Dr. Cox, when challenged, accepted that silica was causal for silicosis, “but questioned whether there was evidence for increased risks at low concentrations; i.e. whether there was a threshold” (Document ID 3574, p. 31). Thresholds for silicosis are addressed in great detail in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases.

Based on the testimony and written comments of numerous scientists representing both public health and industry—all of whom agree that causation is established by applying the Bradford Hill criteria and examining the totality of the evidence—OSHA strongly disagrees with Dr. Cox's claims that the Bradford Hill criteria are inadequate to evaluate causation in epidemiology and that additional statistical techniques are needed to establish causation. OSHA defends its reliance on the IARC determination of 1997 and re-determination of 2012 that crystalline silica is a causal agent for lung cancer. OSHA's own Review of Health Effects Literature further demonstrates the totality of the evidence supporting the causality determination (Document ID 1711). Indeed, other than Dr. Cox representing the ACC, no other individual or entity questioned causation with respect to silicosis. Even Dr. Cox's questioning of causation for silicosis appears to be more of a question about thresholds, which is discussed in Section V.I, Comments and Responses Concerning Thresholds for Silica-Related Diseases.

3. Dr. Cox's Proposed Statistical Methods

OSHA reviewed the statistical methods provided by Dr. Cox in Table 1 of his 2013 paper, “Improving causal inferences in risk analysis,” (Document ID 2307, Attachment 4, p. 11), and explains below why the Agency did not adopt them. For example, Intervention Time Series Analysis (ITSA), as proposed by Dr. Cox in his Table 1, is a method for assessing the impact of an intervention or shock on the trend of outcomes of interest (Gilmour et al., 2006, cited in Document ID 2307, Attachment 4, p. 11). Implementing ITSA requires time series data before and after the intervention for both the dependent variable (e.g., disease outcome) and independent variables (e.g., silica exposure and other predictors), as well as the point of occurrence of the intervention. Although time-series data are frequently available in epidemiological studies, for silica we do not have a specific “intervention point” comparable to the implementation of a new OSHA standard that can be identified and analyzed. Rather, changes in exposure controls tend to be iterative and piecemeal, gradually bringing workers' exposures down over the course of a facility's history and affecting job-specific exposures differently at different points in time. Furthermore, individual workers' exposures change continually with new job assignments and employment. In addition, in a situation where the intervention really reduces the adverse outcome to a low level, such as 1/1000 lifetime excess risk, ITSA would require an enormous observational database in order to be able to estimate the actual post-intervention level of risk. OSHA believes the standard risk analysis approach of estimating an exposure-response relationship based on workers' exposures over time and using this model to predict the effects of a new standard on risk appropriately reflects the typical pattern of multiple and gradual changes in the workers' exposures over time found in most industrial facilities.

Another method listed in Dr. Cox's Table 1, marginal structural models (MSM), was introduced in the late 1990s (Robins, 1998, cited in Document ID 2307, Attachment 4, p. 11) to address issues that can arise in standard modeling approaches when time-varying exposure and/or time-dependent confounders are present.[10] These methods are actively being explored in the epidemiological literature, but have not yet become a standard method in occupational epidemiology. As such, OSHA faces some of the same issues with MSM as were previously noted with BMA: Published, peer-reviewed studies using this approach are not available for the silica literature, and best practices are not yet well established. Thus, the incorporation of MSM in the silica risk assessment is not possible using the currently available literature and would be premature for OSHA's risk assessment generally.

In addition, in his post-hearing brief, Dr. Cox contended that “[a] well-done QRA should explicitly address the causal fraction (and explain the value used), rather than tacitly assuming that it is 1” (Document ID 4027, p. 4). However, this claim is without grounds. OSHA understands Dr. Cox's reference to the “causal fraction” to mean that, Start Printed Page 16372when estimating risk from an exposure-response model, only a fraction of the total estimated risk should be attributed to disease caused by the occupational exposure of interest. The Agency notes that the “causal fraction” of risk is typically addressed through the use of life table analyses, which incorporate background rates for the disease in question. Such analyses, which OSHA used in its Preliminary QRA, calculate the excess risk, over and above background risk, that is solely attributable to the exposure in question. Thus, there is no need to estimate a causal fraction due to exposure. These approaches are further discussed in Section V.M, Comments and Responses Concerning Working Life, Life Tables, and Dose Metric. Furthermore, nowhere in the silica epidemiological literature has the use of an alternative “causal fraction” approach to ascribing the causal relationship between silica exposure and silicosis and lung cancer been deemed necessary to reliably estimate risk.

4. The Assertion That the Silica Scientific Literature May Be False

Dr. Cox also asserted that the same biases and issues with causation in OSHA's Quantitative Risk Assessment (QRA) were likewise present in the silica literature. He wrote, “In general, the statistical methods and causal inferences described in this literature are no more credible or sound than those in OSHA's Preliminary QRA, and for the same reasons” (Document ID 2307, Attachment 4, p. 30).

The rulemaking record contains evidence that contradicts Dr. Cox's claims with respect to the scientific foundation of the QRA. Such evidence includes scientific testimony and the findings of many expert bodies, including IARC, the HHS National Toxicology Program, and NIOSH, concluding that exposure to respirable crystalline silica causes lung cancer. At the public hearing, Dr. Steenland, Professor at Emory University, testified that the body of evidence pertaining to silica was of equal quality to that of other occupational health hazards (Document ID 3580, pp. 1245-1246). Dr. Goldsmith similarly testified:

Silica dust . . . is like asbestos and cigarette smoking in that exposure clearly increases the risk of many diseases. There have been literally thousands of research studies on exposure to crystalline silica in the past 30 years. Almost every study tells the occupational research community that workers need better protection to prevent severe chronic respiratory diseases, including lung cancer and other diseases in the future. What OSHA is proposing to do in revising the workplace standard for silica seems to be a rational response to the accumulation of published evidence (Document ID 3577, Tr. 865-866).

OSHA agrees with these experts, whose positive view of the science supporting the need for better protection from silica exposures stands in contrast to Dr. Cox's claim regarding what he believes to be the problematic nature of the silica literature. Dr. Cox asserted in his written statement:

Scientists with subject matter expertise in areas such as crystalline silica health effects epidemiology are not necessarily or usually also experts in causal analysis and valid causal interpretation of data, and their causal conclusions are often mistaken, with a pronounced bias toward declaring and publishing findings of `significant' effects where none actually exists (false positives). This has led some commentators to worry that `science is failing us,' due largely to widely publicized but false beliefs about causation (Lehrer, 2012); and that, in recent times, `Most published research findings are wrong' (Ioannadis, 2005), with the most sensational and publicized claims being most likely to be wrong. (Document ID 2307, Attachment 4, pp. 15-16).

Moreover, during the public hearing, Dr. Cox stated that, with respect to lung cancer in the context of crystalline silica, the literature base may be false:

MR. PERRY [OSHA Director of the Directorate of Standards and Guidance]: So as I understand it, you basically think there's a good possibility that the entire literature base, with respect to lung cancer now, I'm talking about, is wrong?

DR. COX: You mean with respect to lung cancer in the context of crystalline silica?

MR. PERRY: Yes, sir.

DR. COX: I think that consistent with the findings of Lauer [Lehrer] and Ioannidis and others, I think that it's very possible and plausible that there is a consistent pattern of false positives in the literature base, yes. And that implies, yes, they are wrong. False positives are false (Document ID 3576, Tr. 423).

The Ioannidis paper (Document ID 3851) used mathematical constructs to purportedly demonstrate that most claimed research findings are false, and then provided suggestions for improvement (Document ID 3851, p. 0696). Two of his suggestions appear particularly relevant to the silica literature: “Better powered evidence, e.g., large studies or low-bias meta-analyses, may help, as it comes closer to the unknown `gold' standard. However, large studies may still have biases and these should be acknowledged and avoided”; and “second, most research questions are addressed by many teams, and it is misleading to emphasize the statistically significant findings of any single team. What matters is the totality of the evidence” (Document ID 3851, pp. 0700-0701). OSHA finds no merit in the claim that most claimed research findings are false. Instead, it finds that the silica literature for lung cancer is overall trustworthy, particularly because the “totality of the evidence” characterized by large studies demonstrates a causal relationship between crystalline silica exposure and lung cancer, as IARC determined in 1997 and 2012 (Document ID 2258, Attachment 8, p. 210; 1473, p. 396).

OSHA likewise notes that there was disagreement on Ioannidis' methods and conclusions. Jonathan D. Wren of the University of Oklahoma, in a correspondence to the journal that published the paper, noted that Ioannidis, “after all, relies heavily on other studies to support his premise, so if most (i.e., greater than 50%) of his cited studies are themselves false (including the eight of 37 that pertain to his own work), then his argument is automatically on shaky ground” (Document ID 4087, p. 1193). In addition, Steven Goodman of Johns Hopkins School of Medicine and Sander Greenland of the University of California, Los Angeles, performed a substantive mathematical review (Document ID 4081) of the Ioannidis models and concluded in their correspondence to the same journal that “the claims that the model employed in this paper constitutes `proof' that most published medical research claims are false, and that research in `hot' areas is most likely to be false, are unfounded” (Document ID 4095, p. 0773).

Christiana A. Demetriou, Imperial College London, et al. (2012), analyzed this issue of potential false positive associations in the field of cancer epidemiology (Document ID 4131). They examined the scientific literature for 509 agents classified by IARC as Group 3, “not classifiable as to its carcinogenicity to humans” (Document ID 4131). Of the 509 agents, 37 had potential false positive associations in the studies reviewed by IARC; this represented an overall frequency of potential false positive associations between 0.03 and 0.10 (Document ID 4131). Regarding this overall false positive frequency of about 10 percent, the authors concluded, “In terms of public health care decisions, given that the production of evidence is historical, public health care professionals are not expected to react immediately to a single positive association. Instead, they are likely to wait for further support or enough evidence to reach a consensus, and if a hypothesis is repeatedly tested, then any initial false-positive results will be quickly undermined” (Document ID 4131, p. 1277). The Start Printed Page 16373authors also cautioned that “Reasons for criticisms that are most common in studies with false-positive findings can also underestimate an association and in terms of public health care, false-negative results may be a more important problem than false-positives” (Document ID 4131, pp. 1278-1279). Thus, this study suggested that the false positive frequency in published literature is actually rather low, and stressed the importance of considering the totality of the literature, rather than a single study.

Given these responses to Ioannidis, OSHA fundamentally rejects the claim that most published research findings are false. The Agency concludes that, most likely, where, as here, there are multiple, statistically significant positive findings of an association between silica and lung cancer made by different researchers in independent studies looking at distinct cohorts, the chances that there is a consistent pattern of false positives are small; OSHA's mandate is met when the weight of the evidence in the body of science constituting the best available evidence supports such a conclusion.

M. Comments and Responses Concerning Working Life, Life Tables, and Dose Metric

As discussed in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA, OSHA presented risk estimates associated with exposure over a working lifetime to 25, 50, 100, 250, and 500 μg/m3 respirable crystalline silica (corresponding to cumulative exposures over 45 years to 1.125, 2.25, 4.5, 11.25, and 22.5 mg/m3-yrs). For mortality from silica-related disease (i.e., lung cancer, silicosis and non-malignant respiratory disease (NMRD), and renal disease), OSHA estimated lifetime risks using a life table analysis that accounted for background and competing causes of death. The mortality risk estimates were presented as excess risk per 1,000 workers for exposures over an 8-hour working day, 250 days per year, and a 45-year working lifetime. This is a legal standard that OSHA typically uses in health standards to satisfy the statutory mandate to “set the standard which most adequately assures, to the extent feasible, that no employee will suffer material impairment of health or functional capacity even if such employee has regular exposure to the hazard dealt with by such standard for the period of his working life.” 29 U.S.C. 655(b)(5). For silicosis morbidity, OSHA based its risk estimates on cumulative risk models used by various investigators to develop quantitative exposure-response relationships. These models characterized the risk of developing silicosis (as detected by chest radiography) up to the time that cohort members (including both active and retired workers) were last examined. Thus, risk estimates derived from these studies represent less-than-lifetime risks of developing radiographic silicosis. OSHA did not attempt to estimate lifetime risk (i.e., up to age 85) for silicosis morbidity because the relationships between age, time, and disease onset post-exposure have not been well characterized.

OSHA received critical comments from representatives of the ACC and the Chamber. These commenters expressed concern that (1) the working lifetime exposure of 45 years was not realistic for workers, (2) the use of life tables was improper and alternative methods should be used, and (3) the cumulative exposure metric does not consider the exposure intensity and possible resulting dose-rate effects. OSHA examines these comments in detail in this section, and shows why they do not alter its conclusion that the best available evidence in the rulemaking record fully supports the Agency's use of a 45-year working life in a life table analysis with cumulative exposure as the exposure metric of concern.

1. Working Life

The Chamber commented that 45-year career silica exposures do not exist in today's working world, particularly in “short term work-site industries” such as construction and energy production (Document ID 4194, p. 11; 2288, p. 11). The Chamber stated that careers in these jobs are closer to 6 years, pointing out that OSHA's contractor, ERG, estimated a 64 percent annual turnover rate in the construction industry. Referring to Section 6(b)(5) of the Occupational Safety and Health (OSH) Act of 1970, the Chamber concluded, “OSHA improperly inflates risk estimates with its false 45-year policy, contradicting the Act, which requires standards based on actual, `working life' exposures—not dated hypotheticals” (Document ID 4194, pp. 11-12; 2288, pp. 11-12).

As stated previously, OSHA believes that the 45-year exposure estimate satisfies its statutory obligation to evaluate risks from exposure over a working life, and notes that the Agency has historically based its significance-of-risk determinations on a 45-year working life from age 20 to age 65 in each of its substance-specific rulemakings conducted since 1980. The Agency's use of a 45-year working life in risk assessment has also been upheld by the DC Circuit (Bldg & Constr. Trades Dep't v. Brock, 838 F.2d 1258, 1264-65 (D.C. Cir. 1988)) (also see Section II, Pertinent Legal Authority). Even if most workers are not exposed for such a long period, some will be, and OSHA is legally obligated to set a standard that protects those workers to the extent such standard is feasible. For reasons explained throughout this preamble, OSHA has set the PEL for this standard at 50 µg/m3 TWA. In setting the PEL, the Agency reasoned that while this level does not eliminate all risk from 45 years of exposures for each employee, it is the lowest level feasible for most operations.

In addition, OSHA heard testimony and received several comments with accompanying data that support a 45-year working life in affected industries. For example, six worker representatives of the International Union of Bricklayers and Allied Craftworkers (BAC), which represents a portion of the unionized masonry construction industry (Document ID 4053, p. 2), raised their hands in the affirmative when asked if they had colleagues who worked for longer than 40 years in their trade (Document ID 3585, Tr. 3053). Following the hearings, BAC reviewed its International Pension Fund and counted 116 members who had worked in the industry for 40 years or longer. It noted that this figure was likely an understatement, as many workers had previous experience in the industry prior to being represented by BAC, and many BAC affiliates did not begin participation in the Fund until approximately a decade after its establishment in 1972 (Document ID 4053, p. 2).

OSHA heard similar testimony from representatives of other labor groups and unions. Appearing with the Laborers' Health and Safety Fund of North America (LHSFNA), Eddie Mallon, a long-time member of the New York City tunnel workers' local union, testified that he had worked in the tunnel business for 50 years, mainly on underground construction projects (Document ID 3589, Tr. 4209). Appearing with the United Steelworkers, Allen Harville, of the Newport News Shipbuilding Facility and Drydock, testified that there are workers at his shipyard with more than 50 years of experience. He also believed that 15 to 20 percent of workers had 20 to 40 years of experience (Document ID 3584, Tr. 2571).

In addition, several union representatives appearing with the Building and Construction Trades Department (BCTD) of the American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) also Start Printed Page 16374commented on the working life exposure estimate. Deven Johnson, of the Operative Plasterers' and Cement Masons' International Association, testified that he thought 45 years was relevant, as many members of his union had received gold cards for 50 and 60 years of membership; he also noted that there was a 75-year member in his own local union (Document ID 3581, Tr. 1625-1626). Similarly, Sarah Coyne, representing the International Union of Painters and Allied Trades, testified that 45 years was adequate, as “we have many, many members who continue to work out in the field with the 45 years” (Document ID 3581, Tr. 1626). Charles Austin, of the International Association of Sheet Metal, Air, Rail and Transportation Workers, added that thousands of workers in the union's dust screening program have been in the field for 20 to 30 years (Document ID 3581, Tr. 1628-1629).

In its post-hearing comment, the BCTD submitted evidence on behalf of the United Association of Plumbers, Fitters, Welders and HVAC Service Techs, which represents a portion of the workers in the construction industry. A review of membership records for this association revealed 35,649 active members with 45 years or more of service as a member of the union. Laurie Shadrick, Safety and Health National Coordinator for the United Association, indicated that this membership figure is considered an underestimate, as many members had previous work experience in the construction industry prior to joining the union, or were not tracked by the union after transitioning to other construction trades (Document ID 4073, Attachment 1b). The post-hearing comment of the BCTD also indicated a trend of an aging workforce in the construction industry, with workers 65 years of age and older predicted to increase from 5 percent in 2012 to 8.3 percent in 2022 (Document ID 4073, Attachment 1a, p. 1). This age increase is likely due to the fact that few construction workers have a defined benefit pension plan, and the age for collecting Social Security retirement benefits has been increasing; as a result, many construction workers are staying employed for longer in the industry (Document ID 4073, Attachment 1a, p. 1). Thus, the BCTD expressed its support for using a 45-year working life in the construction industry for risk assessment purposes (Document ID 4073, Attachment 1a, p. 1).

In addition to BAC and BCTD, OSHA received post-hearing comments on the 45-year working life from the International Union of Operating Engineers (IUOE) and the American Federation of State, County and Municipal Employees (AFSCME). The IUOE reviewed records of the Central Pension Fund, in which IUOE construction and stationary local unions participate, and determined that the average years of service amongst all retirees (75,877 participants) was 21.34 years, with a maximum of 49.93 years of active service. Of these retirees, 15,836 participants recorded over 30 years of service, and 1,957 participants recorded over 40 years of service (Document ID 4025, pp. 6-7). The IUOE also pointed to the testimony of Anthony Bodway, Special Projects Manager at Payne & Dolan, Inc. and appearing with the National Asphalt Pavement Association (NAPA), who indicated that some workers in his company's milling division had been with the company anywhere from 35 to 40 years (Document ID 3583, Tr. 2227, 2228). Similarly, the AFSCME reported that, according to its 2011 poll, 49 percent of its membership had over 10 years of experience, and 21 percent had over 20 years (Document ID 3760, p. 2).

The rulemaking record on this topic of the working life thus factually refutes the Chamber's assertion that “no such 45-year career silica exposures exist in today's working world, particularly in construction, energy production, and other short term work-site industries” (Document ID 4194, p. 11; 2288, p. 11). Instead, OSHA concludes that the rulemaking record demonstrates that the Agency's use of a 45-year working life as a basis for estimating risk is legally justified and factually appropriate.

2. Life Tables

Dr. Cox, on behalf of the ACC, commented that OSHA should use “modern methods,” such as Bayesian competing-risks analyses, expectation-maximization (EM) methods, and copula-based approaches that account for subdistributions and interdependencies among competing risks (Document ID 2307, Attachment 4, p. 61). Such methods, according to Dr. Cox, are needed “[t]o obtain risk estimates . . . that have some resemblance to reality, and that overcome known biases in the naïve life table method used by OSHA” (Document ID 2307, Attachment 4, p. 61). Dr. Cox then asserted that the life table method used in the following studies to estimate mortality risks is also incorrect: Steenland et al. (2001a, Document ID 0452), Rice et al. (2001, Document ID 1118), and Attfield and Costello (2004, Document ID 0285) (Document ID 2307, Attachment 4, pp. 61-63).

OSHA does not agree that the life table method it used to estimate mortality risks is incorrect or inappropriate. Indeed, the Agency's life table approach is a standard method commonly used to estimate the quantitative risks of mortality. As pointed out by Rice et al. (2001), the life table method was developed by the National Research Council's BEIR IV Committee on the Biological Effects of Ionizing Radiations (BEIR), Board of Radiation Effects Research, in its 1988 publication on radon (Document ID 1118, p. 40). OSHA notes that the National Research Council is the operating arm of the National Academy of Sciences and the National Academy of Engineering, and is highly respected in the scientific community. As further described by Rice et al., an “advantage of this [actuarial] method is that it accounts for competing causes of death which act to remove a fraction of the population each year from the risk of death from lung cancer so that it is not necessary to assume that all workers would survive these competing causes to a given age” (Document ID 1118, p. 40). Because this life table method is generally accepted in the scientific community and has been used in a variety of peer-reviewed, published journal articles, including some of the key studies relied upon by the Agency in its Preliminary QRA (e.g., Rice et al., 2001, Document ID 1118, p. 40; Park et al., 2002, 0405, p. 38), OSHA believes it is appropriate here.

Regarding the alternative methods proposed by Dr. Cox, OSHA believes that these methods are not widely used in the occupational epidemiology community. In addition, OSHA notes that Dr. Cox did not provide any alternate risk estimates to support the use of his proposed alternative methods, despite the fact that the Agency made its life table data available in the Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 360-378). Thus, for these reasons, OSHA disagrees with Dr. Cox's claim that the life table method used by the Agency to estimate quantitative risks was inappropriate.

3. Exposure Metric

In its risk assessment, OSHA uses cumulative exposure, i.e., average exposure concentration multiplied by duration of exposure, as the exposure metric to quantify exposure-response relationships. It uses this metric because each of the key epidemiological studies on which the Agency relied to estimate risks used cumulative exposure as the exposure metric to quantify exposure-response relationships, although some Start Printed Page 16375also reported significant relationships based on exposure intensity (Document ID 1711, p. 342). As noted in the Review of Health Effects Literature, the majority of studies for lung cancer and silicosis morbidity and mortality have consistently found significant positive relationships between risk and cumulative exposure (Document ID 1711, p. 343). For example, nine of the ten epidemiological studies included in the pooled analysis by Steenland et al. (2001a, Document ID 0452) showed positive exposure coefficients when exposure was expressed as cumulative exposure (Document ID 1711, p. 343).

Commenting on this exposure metric, the ACC argued that cumulative exposure undervalues the role of exposure intensity, as some studies of silicosis have indicated a dose-rate effect, i.e., short-term exposure to high concentrations results in greater risk than longer-term exposure to lower concentrations at an equivalent cumulative exposure level (Document ID 4209, p. 58; 2307, Attachment A, pp. 93-94). The ACC added that, given that silica-related lung cancer and silicosis may both involve an inflammation-mediated mechanism, a dose-rate effect would also be expected for lung cancer (Document ID 4209, p. 58). It concluded that “assessments of risk based solely on cumulative exposure do not account adequately for the role played by intensity of exposure and, accordingly, do not yield reliable estimates of risk” (Document ID 4209, p. 68). Patrick Hessel, Ph.D., representing the Chamber, pointed to the initial comments of OSHA peer reviewer Kenneth Crump, Ph.D., who stated that “[n]ot accounting for a dose-rate effect, if one exists, could overestimate risk at lower concentrations” (Document ID 4016, p. 2, citing 1716, pp. 165-167).

OSHA acknowledges these concerns regarding the exposure metric and finds them to have some merit. However, it notes that the best available studies use cumulative exposure as the exposure metric, as in common in occupational epidemiological studies. As discussed below, there is also substantial good evidence in the record supporting the use of cumulative exposure as the exposure metric for crystalline silica risk assessment.

Paul Schulte, Ph.D., of NIOSH testified that “cumulative exposure is a standard and appropriate metric for irreversible effects that occur soon after actual exposure is experienced. For lung cancer and nonmalignant respiratory disease, NMRD mortality, cumulative exposure lagged for cancer is fully justified . . . For silicosis risk assessment purposes, cumulative exposure is a reasonable and practical choice” (Document ID 3579, Tr. 127). NIOSH also conducted a simulated dose rate analysis for silicosis incidence with data from a Chinese tin miners cohort and, in comparing exposure metrics, concluded that the best fit to the data was cumulative exposure with no dose-rate effect (Document ID 4233, pp. 36-39). This finding is consistent with the testimony of Dr. Steenland, who stated, “Cumulative exposure, I might say, is often the best predictor of chronic disease in general, in epidemiology” (Document ID 3580, Tr. 1227). OSHA also notes that using a cumulative exposure metric (e.g., mg/m3-yrs) factors in both exposure intensity and duration, while using only an exposure intensity metric (e.g., μg/m3) ignores the influence of exposure duration. Dr. Crump's comment that “[e]stimating risk based on an `incomplete' exposure metric like average exposure is not recommended . . . . [E]xposure to a particular air concentration for one week is unlikely to carry the same risk as exposure to that concentration for 20 years, although the average exposures are the same” also supports the use of a cumulative exposure metric (Document ID 1716, p. 166).

With regard to a possible dose-rate effect, OSHA agrees with Dr. Crump that if one exists and is unaccounted for, the result could be an overestimation of risks at lower concentrations (Document ID 1716, pp. 165-167). OSHA is aware of two studies discussed in its Review of Health Effects Literature and Preliminary QRA that examined dose-rate effects on silicosis exposure-response (Document ID 1711, pp. 342-344). Neither study found a dose-rate effect relative to cumulative exposure at silica concentrations near the previous OSHA PEL (Document ID 1711, pp. 342-344). However, they did observe a dose-rate effect in instances where workers were exposed to crystalline silica concentrations far above the previous PEL (i.e., several-fold to orders of magnitude above 100 μg/m3) (Buchanan et al., 2003, Document ID 0306; Hughes et al., 1998, 1059). For example, the Hughes et al. (1998) study of diatomaceous earth workers found that the relationship between cumulative silica exposure and risk of silicosis was steeper for workers hired prior to 1950 and exposed to average concentrations above 500 µg/m3 compared to workers hired after 1950 and exposed to lower average concentrations (Document ID 1059). Similarly, the Buchanan et al. (2003) study of Scottish coal miners adjusted the cumulative exposure metric in the risk model to account for the effects of exposures to high concentrations where the investigators found that, at concentrations above 2000 µg/m3, the risk of silicosis was about three times higher than the risk associated with exposure to lower concentrations but at the same cumulative exposure (Document ID 0306, p. 162). OSHA concluded that there is little evidence that a dose-rate effect exists at concentrations in the range of the previous PEL (100 µg/m3) (Document ID 1711, p. 344). However, at the suggestion of Dr. Crump, OSHA used the model from the Buchanan et al. study in its silicosis morbidity risk assessment to account for possible dose-rate effects at high average concentrations (Document ID 1711, pp. 335-342). OSHA notes that the risk estimates in the exposure range of interest (25-500 μg/m3) derived from the Buchanan et al. (2003) study were not appreciably different from those derived from the other studies of silicosis morbidity (see Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1.).

In its post-hearing brief, NIOSH also added that a “detailed examination of dose rate would require extensive and real time exposure history which does not exist for silica (or almost any other agent)” (Document ID 4233, p. 36). Similarly, Dr. Crump wrote, “Having noted that there is evidence for a dose-rate effect for silicosis, it may be difficult to account for it quantitatively. The data are likely to be limited by uncertainty in exposures at earlier times, which were likely to be higher” (Document ID 1716, p. 167). OSHA agrees with Dr. Crump, and believes that it has used the best available evidence to estimate risks of silicosis morbidity and sufficiently accounted for any dose-rate effect at high silica average concentrations by using the Buchanan et al. (2003) study.

For silicosis/NMRD mortality, the ACC noted that Vacek et al. (2009, Document ID 2307, Attachment 6) reported that, in their categorical analysis of the years worked at various levels of exposure intensity, only years worked at >200 µg/m3 for silicosis and >300 µg/m3 for NMRD were associated with increased mortality (Document ID 2307, Attachment A, p. 93, citing 2307, Attachment 6, pp. 21, 23). However, OSHA believes it to be inappropriate to consider these results in isolation from the other study findings, and notes that Vacek et al. (2009) also reported statistically significant associations of silicosis mortality with cumulative exposure, exposure duration, and average exposure intensity in their Start Printed Page 16376continuous analyses with univariate models; for NMRD mortality, there were statistically significant associations with cumulative exposure and average exposure intensity (Document ID 2307, Attachment 6, pp. 21, 23).

In addition, OSHA notes that Vacek et al. (2009) did not include both an exposure intensity term and a cumulative exposure term in the multivariate model, after testing for correlation between cumulative exposure and years at particular exposure intensity; such a model would indicate how exposure intensity affects any relationship with cumulative exposure. As Dr. Crump stated in his comments:

To demonstrate evidence for a dose-rate effect that is not captured by cumulative exposure, it would be most convincing to show some effect of dose rate that is in addition to the effect of cumulative exposure. To demonstrate such an effect one would need to model both cumulative exposure and some effect of dose rate, and show that adding the effect of dose rate makes a statistically significant improvement to the model over that predicted by cumulative exposure alone (Document ID 1716, p. 166).

Indeed, both Buchanan et al. (2003, Document ID 0306) and Hughes et al. (1998, Document ID 1059), when examining possible dose-rate effects for silicosis morbidity, specifically included both cumulative exposure and exposure intensity in their multivariate models. Additionally, as described in the lung cancer section of this preamble, the Vacek et al. study may be affected by both exposure misclassification and the healthy worker survivor effect. Both of these biases may flatten an exposure-response relationship, obscuring the relationship at lower exposure levels, which could be the reason why a significant effect was not found at the lower exposure levels in the Vacek et al. (2009, Document ID 2307, Attachment 6) multivariate analysis.

Regarding lung cancer mortality, the ACC pointed out that Steenland et al. (2001a, Document ID 0452) acknowledged that duration of exposure did not fit the data well in their pooled lung cancer study. The ACC indicated that exposure intensity should be considered (Document ID 2307, Attachment A, p. 93; 4209, p. 58, citing 0452, p. 779). OSHA interpreted the results of the Steenland et al. (2001, Document ID 0452) study to simply mean that duration of exposure alone was not a good predictor for lung cancer mortality, where a lag period may be important between the exposure and the development of disease. Indeed, Steenland et al. found the model with logged cumulative exposure, with a 15-year lag, to be a strong predictor of lung cancer (Document ID 0452, p. 779). Additionally, no new evidence of a dose-rate effect in lung cancer studies was submitted to the record.

For these reasons, OSHA does not believe there to be any persuasive data in the record that supports a dose-rate effect at exposure concentrations near the revised or previous PELs. OSHA concludes that cumulative exposure is a reasonable exposure metric on which to base estimates of risk to workers exposed to crystalline silica in the exposure range of interest (25 to 500 μg/m3).

N. Comments and Responses Concerning Physico-Chemical and Toxicological Properties of Respirable Crystalline Silica

As discussed in the Review of Health Effects Literature and Preliminary Quantitative Risk Assessment (Document ID 1711, pp. 344-350), the toxicological potency of crystalline silica is influenced by a number of physical and chemical factors that affect the biological activity of the silica particles inhaled in the lung. The toxicological potency of crystalline silica is largely influenced by the presence of oxygen free radicals on the surfaces of respirable particles; these chemically-reactive oxygen species interact with cellular components in the lung to promote and sustain the inflammatory reaction responsible for the lung damage associated with exposure to crystalline silica. The reactivity of particle surfaces is greatest when crystalline silica has been freshly fractured by high-energy work processes such as abrasive blasting, rock drilling, or sawing concrete materials. As particles age in the air, the surface reactivity decreases and exhibits lower toxicologic potency (Porter et al., 2002, Document ID 1114; Shoemaker et al., 1995, 0437; Vallyathan et al., 1995, 1128). In addition, surface impurities have been shown to alter silica toxicity. For example, aluminum and aluminosilicate clay on silica particles has been shown to decrease toxicity (Castranova et al., 1997, Document ID 0978; Donaldson and Borm, 1998, 1004; Fubini, 1998, 1016; Donaldson and Borm, 1998, Document ID 1004; Fubini, 1998, 1016).

In the preamble to the proposed standard, OSHA preliminarily concluded that although there is evidence that several environmental influences can modify surface activity to either enhance or diminish the toxicity of silica, the available information was insufficient to determine to what extent these influences may affect risk to workers in any particular workplace setting (Document 1711, p. 350). NIOSH affirmed OSHA's preliminary conclusion regarding the silica-related risks of exposure to clay-occluded quartz particles, which was based on what OSHA believed to be the best available evidence. NIOSH stated:

NIOSH concurs with this assessment by OSHA. Currently available information is not adequate to inform differential quantitative risk management approaches for crystalline silica that are based on surface property measurements. Thus, NIOSH recommends a single PEL for respirable crystalline silica without consideration of surface properties (Document ID 4233, p. 44).

Two rulemaking participants, the Brick Industry Association (BIA), which represents distributors and manufacturers of clay brick, and the Sorptive Minerals Institute (SMI), which represents many industries that process and mine sorptive clays for consumer products and commercial and industrial applications, provided comment and supporting evidence that the crystalline silica encountered in their workplace environments presents a substantially lower risk of silica-related disease than that reflected in the Agency's Preliminary QRA.

BIA argued that the quartz particles found in clays and shales used in clay brick are occluded in aluminum-rich clay coatings. BIA submitted to the record several studies indicating reduced toxicity and fibrogenicity from exposure to quartz in aluminum-rich clays (Document ID 2343, Attachment 2, p. 2). It purported that “OSHA lacks the statutory authority to impose the proposed rule upon the brick and structural clay manufacturing industry because employees in that industry do not face a significant risk of material impairment of health or functional capacity” (Document ID 2242, pp. 2-3). BIA concluded that its industry should be exempted from the rule, stating: “OSHA should exercise its discretion to exempt the brickmaking industry from compliance with the proposed rule unless and until it determines how best to take into account the industry's low incidence of adverse health effects from silica toxicity” (Document ID 2242, p. 11).

SMI argued that silica in sorptive clays exists as either amorphous silica or as geologically ancient, occluded quartz, “neither of which pose the health risk identified and studied in OSHA's risk assessment” (Document ID 4230, p. 2). SMI further contended that OSHA's discussion of aged silica “does not accurately reflect the risk of geologically ancient, (occluded) silica formed millions of years ago found in Start Printed Page 16377sorptive clays” (Document ID 4230, p. 2). Additionally, SMI noted that clay products produced by the sorptive minerals industry are not heated to high temperatures or fractured, making them different from brick and pottery clays (Document ID 2377, p. 7). In support of its position, SMI submitted to the record several toxicity studies of silica in sorptive clays. It stated that the evidence does not provide the basis for a finding of a significant risk of material impairment of health from exposure to silica in sorptive clays (Document ID 4230, p. 2). Consequently, SMI concluded that the application of a reduced PEL and comprehensive standard is not warranted.

Having considered the evidence SMI submitted to the record, OSHA finds that although quartz originating from bentonite deposits exhibits some biological activity, it is clear that it is considerably less toxic than unoccluded quartz. Moreover, evidence does not exist that would permit the Agency to evaluate the magnitude of the lifetime risk resulting from exposure to quartz in bentonite-containing materials and similar sorptive clays. This finding does not extend to the brick industry, where workers are exposed to silica through occluded quartz in aluminum rich clays. The Love et al. study (1999, Document ID 0369), which BIA claimed would be of useful quality for OSHA's risk assessment, shows sufficient cases of silicosis to demonstrate significant risk within the meaning used by OSHA for regulatory purposes. In addition, OSHA found a reduced, although still significant, risk of silicosis morbidity in the study of pottery workers (Chen et al., 2005, Document ID 0985) that BIA put forth as being representative of mortality in the brick industry (Document ID 3577, Tr. 674). These findings are discussed in detail below.

1. The Clay Brick Industry

BIA did not support a reduction in the PEL because although brick industry employees are exposed to crystalline silica-bearing materials, BIA believes silicosis is virtually non-existent in that industry. It contended that silica exposure in the brick industry does not cause similar rates of disease as in other industries because brick industry workers are exposed to quartz occluded in aluminum-rich layers, reducing the silica's toxicity. BIA concluded that “no significant workplace risk for brick workers from crystalline silica exposure exists at the current exposure limit” (Document ID 3577, Tr. 654) and that reducing the PEL would have no benefit to workers in the brick industry (Document ID 2300, p. 2). These concerns were also echoed by individual companies in the brick industry, such as Acme Brick (Document ID 2085, Attachment 1), Belden Brick Company (Document ID 2378), and Riverside Brick & Supply Company, Inc. (Document ID 2346, Attachment 1). In addition, OSHA received over 50 letters as part of a letter campaign from brick industry representatives referring to BIA's comments on the lack of silicosis in the brick industry (e.g., Document ID 2004).

The Tile Council of North America, Inc., also noted that “[c]lay raw materials used in tile manufacturing are similar to those used in brick and sanitary ware manufacturing” and also suggested that aluminosilicates decrease toxicity (Document ID 3528, p. 1). OSHA agrees with the Tile Council of North America, Inc., that their concerns mirror those of the BIA and, therefore, the Agency's consideration and response to BIA also applies to the tile industry.

a. Evidence on the Toxicity of Silica in Clay Brick.

On behalf of BIA, Mr. Robert Glenn presented a series of published and unpublished studies (Document ID 3418), also summarized by BIA (Document ID 2300, Attachment 1) as evidence that “no significant workplace risk for brick workers from crystalline silica exposure exists at the current exposure limit” (Document ID 3577, Tr. 654). Most of these studies, including an unpublished report on West Virginia brick workers (West Virginia State Health Department, 1939), a study of North Carolina brick workers (Trice, 1941), a study of brick workers in England (Keatinge and Potter, 1949), a study of Canadian brick workers (Ontario Health Department, 1972), two studies of North Carolina brick workers (NIOSH, 1978 and NIOSH, 1980), a study of English and Scottish brick workers (Love et al., 1999, Document ID 0369), and an unpublished study commissioned by BIA of workers at 13 of its member companies (BIA, 2006), reported little or no silicosis among the workers examined (Document ID 3418; 3577, Tr. 655-669).

Based on its review of the record evidence, OSHA finds that there are many silica-containing materials (e.g., other clays, sand, etc.) in brick and concludes that BIA's position is not supported by the best available evidence. The analysis contained in the studies Mr. Glenn presents does not meet the rigorous standards used in the studies on which OSHA's risk assessment relies. Indeed the studies cited by Mr. Glenn and BIA do not adequately support their contention that silicosis is “essentially non-existent.” Several studies were poorly designed and applied inappropriate procedures for evaluating chest X-rays (Document ID 3577, Tr. 682-685). Dr. David Weissman of NIOSH underscored the significance of such issues, stating: “It's very important, for example, to use multiple [B] readers [to evaluate chest X-rays] and medians of readings, and it is very important for people to be blinded to how readings are done” (Document ID 3577, Tr. 682). Also problematic was Mr. Glenn's failure to provide key information on the length of exposure or time since the first exposure in any of the studies he presented, which examined only currently employed workers. Information on duration of exposure or time since first exposure is essential to evaluating risk of silicosis because silicosis typically develops slowly and becomes detectable between 10 years and several decades following a worker's first exposure. In the hearing, Dr. Ken Rosenman also noted inadequacies related to silicosis latency, testifying that “we know that silicosis occurs 20, 30 years after . . . first exposure . . . if people have high exposure but short duration, short latency, you are not going to see positive x-rays [even if silicosis is developing] and so it's not going to be useful” (Document ID 3577, Tr. 688-689).

Mr. Glenn acknowledged shortcomings in the studies he submitted for OSHA's consideration, agreeing with Dr. Weissman's points about quality assurance for X-ray interpretation and study design (e.g., Document ID 3577, Tr. 683). In response to Dr. Rosenman's concerns about silicosis latency, he reported that no information on worker tenure or time since first exposure was presented in Trice (1941), Keatings and Potter (1949), Rajhans and Buldovsky (1972), the NIOSH studies (1978, 1980), or Love et al. (1999), and that more than half of the West Virginia brick workers studied by NIOSH (1939) had a tenure of less than 10 years (Document ID 4021, pp. 5-6), a time period that OSHA believes is too short to see development of most forms of silicosis. He suggested that high exposures in two areas of the West Virginia facilities could trigger accelerated or acute silicosis, which could be observed in less than 10 years, if the toxicity of the silica in clay brick was comparable to silica found in other industries (post-hearing comments, p. 5). However, OSHA notes that a cross-sectional report on actively employed workers would not necessarily capture cases of accelerated or acute silicosis, Start Printed Page 16378which are associated with severe symptoms that compromise individuals' ability to continue work, and therefore would result in a survivor effect where only unaffected workers remain at the time of study.

Mr. Glenn further argued that the Agency should assess risk to brick workers based on studies from that industry because the incidence of silicosis among brick workers appears to be lower than among workers in other industries (Document ID 3577, Tr. 670). For the reasons discussed above, OSHA does not believe the studies submitted by Mr. Glenn provide an adequate basis for risk assessment. In addition, studies presented did not: (1) Include retired workers; (2) report the duration of workers' exposure to silica; (3) employ, in most cases, quality-assurance practices for interpreting workers' medical exams; or (4) include estimates of workers' silica exposures. Furthermore, Mr. Glenn acknowledged in the informal public hearing that the Love et al. (1999, Document ID 0369) study of 1,925 workers employed at brick plants in England and Scotland in 1990-1991 is the only available study of brick workers that presented exposure-response information (Document ID 3577, Tr. 692). He characterized the results of that study as contradictory to OSHA's risk assessment for silicosis morbidity because the authors concluded that frequency of pneumoconiosis is low in comparison to other quartz-exposed workers (Document ID 4021, p. 2). He also cited an analysis by Miller and Soutar (Document ID 1098) (Dr. Soutar is a co-author of the Love et al. study) that compared silicosis risk estimates derived from Love et al. and those from Buchanan et al.'s study of Scottish coal workers exposed to silica, and concluded that silicosis risk among the coal workers far exceeded that among brick workers (Document ID 3577, Tr. 671). He furthermore concluded that the Love et al. study is “the only sensible study to be used for setting an exposure limit for quartz in brick manufacturing.” (Document ID 3577, Tr. 679).

Based on review of the Love et al. study (Document ID 0369), OSHA agrees with Mr. Glenn's claim that the silicosis risk among workers in clay brick industries appears to be somewhat lower than might be expected in other industries. However, OSHA is unconvinced by Mr. Glenn's argument that risk to workers exposed at the previous PEL is not significant because the cases of silicosis reported in this study are sufficient to show significant risk within the meaning used by OSHA for regulatory purposes (1 in 1,000 workers exposed for a working lifetime).

Love et al. reported that 3.7 percent of workers with radiographs were classified as ILO Category 0/1 (any signs of small opacities) and 1.4 percent of workers were classified as ILO Category 1/0 (small radiographic opacities) or greater. Furthermore, among workers aged 55 and older, the age category most likely to have had sufficient time since first exposure to develop detectable lung abnormalities from silicosis exposure, Love et al. reported prevalences of abnormal radiographs ranging from 2.9 percent (cumulative exposure below 0.5 mg/yr-m3) to 16.4 percent (exposure at least 4 mg/yr-m3) (Love et al. 1999, Document ID 0369, Table 4, p. 129). According to the study authors, these abnormalities “are the most likely dust related pathology—namely, silicosis” (Document ID 0369, p. 132). Given that OSHA considers a lifetime risk of 0.1 percent (1 in 1,000) to clearly represent a significant risk, OSHA considers the Love et al. study to have demonstrated a significant risk to brick workers even if only a tiny fraction of the abnormalities observed in the study population represent developing silicosis (see Benzene, 448 U.S. 607, 655 n. 2). According to the study authors, “the estimated exposure-response relation for quartz suggests considerable risks of radiological abnormality even at concentrations of 0.1 mg/m3 [100 μg/m3] of quartz” (Document ID 0369, p. 132).

OSHA concludes that, despite the possibly lower toxicity of silica in the clay brick industry compared to other forms, and despite the Love et al. study's likely underestimation of risk due to exclusion of retired workers, the study demonstrates significant risk among brick workers exposed at the previous general industry PEL. It also suggests that the silicosis risk among brick workers would remain significant even at the new PEL. Furthermore, OSHA is unconvinced by Mr. Glenn's argument that the Agency should develop a quantitative risk assessment based on the Love et al. study, because that study excluded retired workers and had inadequate worker follow-up. As explained earlier in this section, adequate follow-up time and inclusion of retired workers is extremely important to allow for latency in the development of silicosis. Therefore, OSHA relied on studies including retired workers in its QRA for silicosis morbidity.

Mr. Glenn additionally argued that the risk of lung cancer from silica exposure among brick workers is likely to be lower than among workers exposed to silica in other work settings. Mr. Glenn acknowledged that “there are no published mortality studies of brick workers that look at cause of death or lung cancer death” (Document ID 3577, Tr. 674). However, he stated that “pottery clays are similar to the structural clays used in brickmaking in that the quartz is occluded in aluminum-rich layers of bentonite, kaolinite, and illite,” and that OSHA should consider studies of mortality among pottery workers as representative of the brick industry (Tr. 674). Mr. Glenn cited the Chen et al. (2005) study of Chinese pottery workers, which reported a weak exposure-response relationship between silica exposure and lung cancer mortality, and which appeared to be affected by PAH-related confounding. He concluded that the Chen et al. study “provides strong evidence for aluminum-rich clays suppressing any potential carcinogenesis from quartz” (Document ID 3577, Tr. 675).

OSHA acknowledges that occlusion may weaken the carcinogenicity of silica in the brick clay industry, but does not believe that the Chen et al. study provides conclusive evidence of such an effect. This is because of the relatively low carcinogenic potential of silica and the difficulty involved in interpreting one cohort with known issues of confounding (see Section V.F, Comments and Responses Concerning Lung Cancer Mortality). OSHA also notes, however, that it estimated risks of silicosis morbidity from the cited Chen et al. (2005, Document ID 0985) study, and found the risk among pottery workers to be significant, with 60 deaths per 1,000 workers at the previous PEL of 100 μg/m3 and 20 deaths per 1,000 workers at the revised PEL of 50 μg/m3 (as indicated in Section VI, Final Quantitative Risk Assessment and Significance of Risk, Table VI-1). Thus, given Mr. Glenn's assertion that pottery clays are similar to the clays used in brickmaking, OSHA believes that while the risk of silicosis morbidity may be lower than that seen in other industry sectors, it is likely to still be significant in the brickmaking industry.

Thus, OSHA concludes that the BIA's position is not supported by the best available evidence. The studies cited by Mr. Glenn to support his contention that brick workers are not at significant risk of silica-related disease do not have the same standards as those studies used by OSHA in its quantitative risk assessment. Furthermore, in the highest-quality study brought forward by Mr. Glenn (Love et al. 1999, Document ID 0369), there are sufficient cases of silicosis to demonstrate significant risk within the meaning used by OSHA for Start Printed Page 16379regulatory purposes. Even if the commenters' arguments that silica in clay brick is less toxic were, to some extent, legitimate, this would not significantly affect OSHA's own estimates from the epidemiological evidence of the risks of silicosis.

2. Sorptive Minerals (Bentonite Clay) Processing

SMI asserted that the physico-chemical form of respirable crystalline silica in sorptive clays reduces the toxicologic potency of crystalline silica relative to the forms of silica common to most studies relied on in OSHA's Preliminary QRA. In other words, the risk associated with exposure to silica in sorptive clays is assertedly lower than the risk associated with exposure to silica in other materials. SMI based this view on what it deemed the “best available scientific literature,” epidemiological, in vitro, and animal evidence OSHA had not previously considered. It believed the evidence showed reduced risk from exposure to occluded quartz found in the sorptive clays and that occluded quartz does not create a risk similar to that posed by freshly fractured quartz (Document ID 2377, p. 7). Based on this, SMI contended that the results of OSHA's Preliminary QRA were not applicable to the sorptive minerals industry, and a more stringent standard for crystalline silica is “neither warranted nor legally permissible” (Document ID 4230, p. 1). As discussed below, OSHA reviewed the evidence submitted by SMI and finds that although the studies provide evidence of some biological activity in quartz originating from bentonite deposits, there is not quantitative evidence that would permit the Agency to evaluate the magnitude of the lifetime risk resulting from exposure to quartz in bentonite-containing materials and similar sorptive clays.

a. Evidence on the Toxicity of Silica in Sorptive Minerals

SMI submitted a number of studies to the rulemaking record. First, it summarized a retrospective study by Waxweiler et al. (Document ID 3998, Attachment 18e) of attapulgite clay workers in Georgia in which the authors concluded that there was a significant deficit of non-malignant respiratory disease mortality and no clear excess of lung cancer mortality among these workers. It used the study as the basis for its recommendation to OSHA that the study “be cited and that exposures in the industry be recognized in the final rule as not posing the same hazard as those in industries with reactive crystalline silica” (Document ID 2377, p. 10).

Based on its review of the rulemaking record, OSHA concludes that the Waxweiler et al. study is of limited value for assessing the hazard potential of quartz in bentonite clay because of the low airborne levels of silica to which the workers were exposed. The Agency's conclusion is supported by NIOSH's summary of the time-weighted average (TWA) exposures calculated for each job category in Waxweiler et al. (1988, Document ID 3998, Attachment 18e), which were found to be “within the acceptable limits as recommended by NIOSH (i.e., <0.05 mg/m3 [50 μg/m3]) . . . and most were substantially lower” (Document ID 4233, p. 41). It cannot be known to what extent the low toxicity of the dust or the low exposures experienced by the workers each contributed to the lack of observed disease.

SMI also presented a World Health Organization (WHO) document (2005, Document ID 3929), which recognized that “studies of workers exposed to sorptive clays have not identified significant silicosis risk” (Document ID 2377, p. 10). However, although WHO did find that there were no reported cases of fibrotic reaction in humans exposed to montmorillonite minerals in the absence of crystalline silica (Document ID 3929, p. 130), the WHO report does discuss the long-term effects from exposure to crystalline silica, including silicosis and lung cancer. In fact, with respect to evaluating the hazards associated with exposure to bentonite clay, WHO regarded silica as a potential confounder (Document ID 3929, p. 136). Thus, WHO did not specifically make any findings with respect to the hazard potential of quartz in the bentonite clay mineral matrix but instead recognized the hazard presented by exposure to crystalline silica generally.

Additionally, the WHO (Document ID 3929, pp. 114, 118) cited two case/case series reports of bentonite-exposed workers, one demonstrating increasing prevalence of silicosis with increasing exposure to bentonite dust (Rombola and Guardascione, 1955, Document ID 3998, Attachment 18) and another describing cases of silicosis among workers exposed to bentonite dust (Phibbs et al. 1971, Document ID 3998, Attachment 18b). Rombola and Guardascione (1955) found silicosis prevalences of 35.5 and 12.8 percent in two bentonite processing factories, and 6 percent in a bentonite mine. In the factory where the highest exposures occurred, 10 of the 26 cases found were severe and all cases developed with seven or fewer years of exposure, indicating that exposure levels were extremely high (Document ID 4233, p. 42, citing 3998, Attachment 18). Phibbs et al. (1971) reviewed chest x-rays of 32 workers in two bentonite plants, of which x-ray films for 14 indicated silicosis ranging from minimal to advanced. Although the exposure of affected workers to respirable dust or quartz is not known, industrial hygiene surveys conducted in four bentonite plants showed some areas having particle counts in excess of 3 to 11 times the ACGIH particle count limit (Document ID 3998, Attachment 18b, p. 4). This is roughly equivalent to exposure levels between 8 and 28 times OSHA's former general industry PEL of 100 μg/m3 (given that the particle count limit is about 2.5 or more times higher than the gravimetric limit for respirable quartz (see Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA). Exposures of this magnitude are considerably higher than those experienced by worker cohorts of the studies relied on by OSHA in its Final Risk Assessment and discussed in Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA. For example, the median of average exposures reported in the ten cohort studies used by Steenland et al. (2001, Document ID 0684, p. 775) ranged from about one-half to six times the former general industry PEL.

The lack of specific exposure information on bentonite workers found with silicosis, combined with the extraordinary exposures experienced by workers in the bentonite plants studied by Phibbs et al. (1971), make this study, while concerning, unsuitable for evaluating risks in the range of the former and final rule PELs. OSHA notes that the WHO report also concluded that available data were inadequate to conclusively establish a dose-response relationship or even a cause-and-effect relationship for bentonite dust, and that its role in inducing pneumoconiosis remains uncertain.

SMI also presented evidence from animal and in vitro studies that it believes shows that respirable crystalline quartz present in sorptive clays exists in a distinct occluded form, which significantly mitigates adverse health effects due to the physico-chemical characteristics of the occluded quartz. As discussed below, based on careful review of the studies SMI cited, OSHA believes these studies indicate that silica in bentonite clay is of lower toxicologic potency than that found in other industry sectors.

SMI submitted two studies: an animal study (Creutzenberg et al. 2008, Start Printed Page 16380Document ID 3891) and a study of the characteristics of quartz samples isolated from bentonite (Miles et al. 2008, Document ID 4173). SMI contended that these studies demonstrate the low toxicity potential of geologically ancient occluded quartz found in sorptive clays (Document ID 2377, pp. 8-9).

Creutzenberg et al. (2008) summarized the findings from a rat study aimed at “characterizing the differences in biological activity between crystalline ground reference quartz (DQ12) and a quartz with occluded surfaces (quartz isolate) obtained from a clay deposit formed 110-112 million years ago” (Document ID 3891, p. 995). Based on histopathological assessment of the lungs in each treatment group, Creutzenberg et al. (2008, Document ID 3891) found that the DQ12 reference quartz group exhibited a significantly stronger inflammatory reaction than the quartz isolate, which showed a slight but still statistically significant inflammatory response compared to the control group. The increased inflammatory response was observed at day 3 but not at 28 or 90 days. Thus, reaction elicited by the quartz isolate, thought to have similar properties to bentonite, was considered by the investigators to represent a moderate effect that did not progress. In light of this, the implications of this study for development of silicosis are unclear.

SMI also cited Miles et al. (2008, Document ID 4173), who studied the mineralogical and chemical characteristics of quartz samples isolated from bentonite, including the quartz isolate used by Creutzenberg et al. (2008) in their animal study. Their evaluation identified several differences in the chemical and physical properties of the quartz isolates and unoccluded quartz that could help explain the observed differences in toxicity (Document ID 4173); these included differences in crystal structure, electrical potential of particle surfaces, and, possibly, differences in the reactivity of surface-free radicals owing to the presence of iron ions in the residual clay material associated with the quartz isolates.

With respect to the two studies just discussed, animal evidence cited by SMI demonstrates that quartz in bentonite induces a modest inflammatory reaction in the lung that does not persist (Creutzenberg et al., 2008, Document ID 3891). Such a reaction is notably different from the persistent and stronger response seen with standard experimental quartz material without surface occlusion (Creutzenberg et al., 2008, Document ID 3891). Physical and chemical characteristics of quartz from bentonite deposits have been shown to differ from standard experimental quartz in ways that can explain its reduced toxicity (Miles et al., 2008, Document ID 4173). However, the animal studies cited by SMI are not suitable for risk assessment since they were short-term (90 days), single-dose experiments.

In sum, human evidence on the toxicity of quartz in bentonite clay includes one study cited by SMI that did not find an excess risk of respiratory disease (Waxweiller et al., Document ID 3998, Attachment 18e). However, because exposures experienced by the workers were low with most less than that of the final rule PEL, the lack of an observed effect cannot be solely attributed to the nature of the quartz particles. Two studies of bentonite workers found a high prevalence of silicosis based on x-ray findings (Rombola and Guardascione, 1955, Document ID 3998, Attachment 18; Phibbs et al., 1971, Document ID 3998, Attachment 18b). Limited exposure data provided in the studies as well as the relatively short latencies seen among cases of severe silicosis make it clear that the bentonite workers were exposed to extremely high dust levels. Neither of these studies can be relied on to evaluate disease risk in the exposure range of the former and revised respirable crystalline silica PELs.

OSHA finds that the evidence for quartz originating from bentonite deposits indicates some biological activity, but also indicates lower toxicity than standard experimental quartz (which has similar characteristics to quartz encountered in most workplaces where exposures occur). For regulatory purposes, however, OSHA finds that the evidence does not exist that would permit the Agency to evaluate the magnitude of the lifetime risk resulting from exposure to quartz in sorptive clays at the 100 μg/m3 PEL. Instead, OSHA finds that the record provides no sound basis for determining the significance of risk for exposure to sorptive clays containing respirable quartz. Thus, OSHA is excluding sorptive clays (as described specifically in the Scope part of Section XV, Summary and Explanation) from the scope of the rule, until such time that sufficient science has been developed to permit evaluation of the significance of the risk. However, in excluding sorptive clays from the rule, the general industry PEL, as described in 29 CFR 1910.1000 Table Z-3, will continue to apply.

VI. Final Quantitative Risk Assessment and Significance of Risk

A. Introduction

To promulgate a standard that regulates workplace exposure to toxic materials or harmful physical agents, OSHA must first determine that the standard reduces a “significant risk” of “material impairment.” Section 6(b)(5) of the OSH Act, 29 U.S.C. 655(b). The first part of this requirement, “significant risk,” refers to the likelihood of harm, whereas the second part, “material impairment,” refers to the severity of the consequences of exposure. Section II, Pertinent Legal Authority, of this preamble addresses the statutory bases for these requirements and how they have been construed by the Supreme Court and federal courts of appeals.

It is the Agency's practice to estimate risk to workers by using quantitative risk assessment and determining the significance of that risk based on the best available evidence. Using that evidence, OSHA identifies material health impairments associated with potentially hazardous occupational exposures, and, when possible, provides a quantitative assessment of exposed workers' risk of these impairments. The Agency then evaluates whether these risks are severe enough to warrant regulatory action and determines whether a new or revised rule will substantially reduce these risks. For single-substance standards governed by section 6(b)(5) of the OSH Act, 29 U.S.C. 655(b)(5), OSHA sets a permissible exposure limit (PEL) based on that risk assessment as well as feasibility considerations. These health and risk determinations are made in the context of a rulemaking record in which the body of evidence used to establish material impairment, assess risks, and identify affected worker population, as well as the Agency's preliminary risk assessment, are placed in a public rulemaking record and subject to public comment. Final determinations regarding the standard, including final determinations of material impairment and risk, are thus based on consideration of the entire rulemaking record.

In this case, OSHA reviewed extensive toxicological, epidemiological, and experimental research pertaining to the adverse health effects of occupational exposure to respirable crystalline silica, including silicosis, other non-malignant respiratory disease (NMRD), lung cancer, and autoimmune and renal diseases. Using the information collected during this review, the Agency Start Printed Page 16381developed quantitative estimates of the excess risk of mortality and morbidity attributable to the previously allowed and revised respirable crystalline silica PELs; these estimates were published with the proposed rule. The Agency subsequently reexamined these estimates in light of the rulemaking record as a whole, including comments, testimony, data, and other information, and has determined that long-term exposure at and above the previous PELs would pose a significant risk to workers' health, and that adoption of the new PEL and other provisions of the final rule will substantially reduce this risk. Based on these findings, the Agency is adopting a new PEL of 50 μg/m3.

Even though OSHA's risk assessment indicates that a significant risk also exists at the revised action level of 25 μg/m3, the Agency is not adopting a PEL below the revised 50 μg/m3 limit because OSHA must also consider the technological and economic feasibility of the standard in determining exposure limits. As explained in the Summary and Explanation for paragraph (c), Permissible Exposure Limit (PEL), of the general industry/maritime standard (paragraph (d) for construction), OSHA has determined that, with the adoption of additional engineering and work practice controls, the revised PEL of 50 μg/m3 is technologically and economically feasible in most operations in the affected general industrial and maritime sectors and in the construction industry, but that a lower PEL of 25 μg/m3 is not technologically feasible for most of these operations (see Section VII, Summary of the Final Economic Analysis and Final Regulatory Flexibility Analysis (FEA) and Chapter IV, Technological Feasibility, of the FEA). Therefore, OSHA concludes that by establishing the 50 μg/m3 PEL, the Agency has reduced significant risk to the extent feasible.

B. OSHA's Findings of Material Impairments of Health

As discussed below and in OSHA's Review of Health Effects Literature and Preliminary QRA (Document ID 1711, pp. 7-229), there is convincing evidence that inhalation exposure to respirable crystalline silica increases the risk of a variety of adverse health effects, including silicosis, NMRD (such as chronic bronchitis and emphysema), lung cancer, kidney disease, immunological effects, and infectious tuberculosis (TB). OSHA considers each of these conditions to be a material impairment of health. These diseases make it difficult or impossible to work and result in significant and permanent functional limitations, reduced quality of life, and sometimes death. When these diseases coexist, as is common, the effects are particularly debilitating (Rice and Stayner, 1995, Document ID 0418; Rosenman et al., 1999, 0421). Based on these findings and on the scientific evidence that respirable crystalline silica substantially increases the risk of each of these conditions, OSHA has determined that exposure to respirable crystalline silica increases the risk of “material impairment of health or functional capacity” within the meaning of the Occupational Safety and Health Act.

1. Silicosis

OSHA considers silicosis, an irreversible and potentially fatal disease, to be a clear material impairment of health. The term “silicosis” refers to a spectrum of lung diseases attributable to the inhalation of respirable crystalline silica. As described more fully in the Review of Health Effects Literature (Document ID 1711, pp. 16-71), the three types of silicosis are acute, accelerated, and chronic. Acute silicosis can occur within a few weeks to months after inhalation exposure to extremely high levels of respirable crystalline silica. Death from acute silicosis can occur within months to a few years of disease onset, with the affected person drowning in his or her own lung fluid (NIOSH, 1996, Document ID 0840). Accelerated silicosis results from exposure to high levels of airborne respirable crystalline silica, and disease usually occurs within 5 to 10 years of initial exposure (NIOSH, 1996, Document ID 0840). Both acute and accelerated silicosis are associated with exposures that are substantially above the previous general industry PEL, although no precise information on the relationships between exposure and occurrence of disease exists.

Chronic silicosis is the most common form of silicosis seen today, and is a progressive and irreversible condition characterized as a diffuse nodular pulmonary fibrosis (NIOSH, 1996, Document ID 0840). Chronic silicosis generally occurs after 10 years or more of inhalation exposure to respirable crystalline silica at levels below those associated with acute and accelerated silicosis. Affected workers may have a dry chronic cough, sputum production, shortness of breath, and reduced pulmonary function. These symptoms result from airway restriction caused by the development of fibrotic scarring in the lower regions of the lungs. The scarring can be detected in chest x-ray films when the lesions become large enough to appear as visible opacities. The result is a restriction of lung volumes and decreased pulmonary compliance with concomitant reduced gas transfer. Chronic silicosis is characterized by small, rounded opacities that are symmetrically distributed in the upper lung zones on chest radiograph (Balaan and Banks, 1992, Document ID 0289, pp. 347, 350-351).

The diagnosis of silicosis is based on a history of exposure to respirable crystalline silica, chest radiograph findings, and the exclusion of other conditions that appear similar. Because workers affected by early stages of chronic silicosis are often asymptomatic, the finding of opacities in the lung is key to detecting silicosis and characterizing its severity. The International Labour Organization (ILO) International Classification of Radiographs of Pneumoconioses (ILO, 1980, Document ID 1063; 2002, 1064) is the currently accepted standard against which chest radiographs are evaluated for use in epidemiological studies, medical surveillance, and clinical evaluation. The ILO system standardizes the description of chest x-rays, and is based on a 12-step scale of severity and extent of silicosis as evidenced by the size, shape, and density of opacities seen on the x-ray film. Profusion (frequency) of small opacities is classified on a 4-point major category scale (0-3), with each major category divided into three, giving a 12-point scale between 0/− and 3/+. Large opacities are defined as any opacity greater than 1 cm that is present in a film (ILO, 1980, Document ID 1063; 2002, 1064, p. 6).

The small rounded opacities seen in early stage chronic silicosis (ILO major category 1 profusion) may progress (through ILO major categories 2 and/or 3) and develop into large fibrotic masses that destroy the lung architecture, resulting in progressive massive fibrosis (PMF). This stage of advanced silicosis is usually characterized by impaired pulmonary function, permanent disability, and premature death. In cases involving PMF, death is commonly attributable to progressive respiratory insufficiency (Balaan and Banks, 1992, Document ID 0289).

Patients with ILO category 2 or 3 background profusion of small opacities are at increased risk, compared to those with category 1 profusion, of developing the large opacities characteristic of PMF. In one study of silicosis patients in Hong Kong, Ng and Chan (1991, Document ID 1106, p. 231) found the risk of PMF increased by 42 and 64 percent among patients whose chest x-Start Printed Page 16382ray films were classified as ILO major category 2 or 3, respectively. Research has shown that people with silicosis advanced beyond ILO major category 1 have reduced life expectancy compared to the general population (Infante-Rivard et al., 1991, Document ID 1065; Ng et al., 1992a, 0383; Westerholm, 1980, 0484).

Silicosis is the oldest known occupational lung disease and is still today the cause of significant premature mortality. As discussed further in Section V.E, Comments and Responses Concerning Surveillance Data on Silicosis Morbidity and Mortality, in 2013, there were 111 deaths in the U.S. where silicosis was recorded as an underlying or contributing cause of death on a death certificate (NCHS data). Between 1996 and 2005, deaths attributed to silicosis resulted in an average of 11.6 years of life lost by affected workers (NIOSH, 2007, Document ID 1362). In addition, exposure to respirable crystalline silica remains an important cause of morbidity and hospitalizations. National inpatient hospitalization data show that in the year 2011, 2,082 silicosis-related hospitalizations occurred, indicating that silicosis continues to be a significant health issue in the U.S. (Document ID 3577, Tr. 854-855). Although there is no national silicosis disease surveillance system in the U.S., a published analysis of state-based surveillance data from the time period 1987-1996 estimated that between 3,600-7,000 new cases of silicosis occurred in the U.S. each year (Rosenman et al., 2003, Document ID 1166).

It has been widely reported that available statistics on silicosis-related mortality and morbidity are likely to be understated due to misclassification of causes of death (for example, as tuberculosis, chronic bronchitis, emphysema, or cor pulmonale), lack of occupational information on death certificates, or misdiagnosis of disease by health care providers (Goodwin et al., 2003, Document ID 1030; Windau et al., 1991, 0487; Rosenman et al., 2003, 1166). Furthermore, reliance on chest x-ray findings may miss cases of silicosis because fibrotic changes in the lung may not be visible on chest radiograph; thus, silicosis may be present absent x-ray signs or may be more severe than indicated by x-ray (Hnizdo et al., 1993, Document ID 1050; Craighhead and Vallyahan, 1980, 0995; Rosenman et al., 1997, 4181).

Although most workers with early-stage silicosis (ILO categories 0/1 or 1/0) typically do not experience respiratory symptoms, the primary risk to the affected worker is progression of disease with progressive decline of lung function. Several studies of workers exposed to crystalline silica have shown that, once silicosis is detected by x-ray, a substantial proportion of affected workers can progress beyond ILO category 1 silicosis, even after exposure has ceased (e.g., Hughes, 1982, Document ID 0362; Hessel et al., 1988, 1042; Miller et al., 1998, 0374; Ng et al., 1987a, 1108; Yang et al., 2006, 1134). In a population of coal miners whose last chest x-ray while employed was classified as major category 0, and who were examined again 10 years after the mine had closed, 20 percent had developed opacities consistent with a classification of at least 1/0, and 4 percent progressed further to at least 2/1 (Miller et al., 1998, Document ID 0374). Although there were periods of extremely high exposure to respirable quartz in the mine (greater than 2,000 μg/m3 in some jobs between 1972 and 1976, and more than 10 percent of exposures between 1969 and 1977 were greater than 1,000 μg/m3), the mean cumulative exposure for the cohort over the period 1964-1978 was 1.8 mg/m3-yrs, corresponding to an average silica concentration of 120 μg/m3. In a population of granite quarry workers exposed to an average respirable silica concentration of 480 μg/m3 (mean length of employment was 23.4 years), 45 percent of those diagnosed with simple silicosis (i.e., presence of small opacities only on chest x-ray films) showed radiological progression of disease after 2 to 10 years of follow up (Ng et al., 1987a, Document ID 1108). Among a population of gold miners, 92 percent progressed in 14 years; exposures of high-, medium-, and low-exposure groups were 970, 450, and 240 μg/m3, respectively (Hessel et al., 1988, Document ID 1042). Chinese mine and factory workers categorized under the Chinese system of x-ray classification as “suspected” silicosis cases (analogous to ILO 0/1) had a progression rate to stage I (analogous to ILO major category 1) of 48.7 percent, and the average interval was about 5.1 years (Yang et al., 2006, Document ID 1134).

The risk of silicosis carries with it an increased risk of reduced lung function as the disease irreversibly progresses. There is strong evidence in the literature for the finding that lung function deteriorates more rapidly in workers exposed to silica, especially those with silicosis, than what is expected from a normal aging process (Cowie, 1988, Document ID 0993; Hughes et al., 1982, 0362; Malmberg et al., 1993, 0370; Ng and Chan, 1992, 1107). The rates of decline in lung function are greater in those whose disease showed evidence of radiologic progression (Begin et al., 1987, Document ID 0295; Cowie, 1988, 0993; Ng and Chan, 1992, 1107; Ng et al., 1987a, 1108). Additionally, the average deterioration of lung function exceeds that in smokers (Hughes et al., 1982, Document ID 0362).

Several studies have reported no decrease in pulmonary function with an ILO category 1 level of profusion of small opacities but found declines in pulmonary function with categories 2 and 3 (Ng et al., 1987a, Document ID 1108; Begin et al., 1988, 0296; Moore et al., 1988, 1099). However, one study found a statistically significantly greater annual loss in forced vital capacity (FVC) and forced expiratory volume in one second (FEV1) among those with category 1 profusion compared to category 0 (Cowie, 1988, Document ID 0993). In another study, the degree of profusion of opacities was associated with reductions in several pulmonary function metrics (Cowie and Mabena, 1991, Document ID 0342). Some studies have reported no associations between radiographic silicosis and decreases in pulmonary function (Ng et al., 1987a, Document ID 1108; Wiles et al., 1972, 0485; Hnizdo, 1992, 1046), while other studies (Ng et al., 1987a, Document ID 1108; Wang et al., 1997, 0478) have found that measurable changes in pulmonary function are evident well before the changes seen on chest x-ray. Findings of pulmonary function decrements absent radiologic signs of silicosis may reflect the general insensitivity of chest radiography in detecting lung fibrosis, or may also reflect that exposure to respirable silica has been shown to increase the risk of non-malignant respiratory disease (NMRD) and its attendant pulmonary function losses (see Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA).

Moreover, exposure to respirable crystalline silica in and of itself, with or without silicosis, increases the risk that latent tuberculosis infection can convert to active disease. Early descriptions of dust diseases of the lung did not distinguish between TB and silicosis, and most fatal cases described in the first half of this century were a combination of silicosis and TB (Castranova et al., 1996, Document ID 0314). More recent findings demonstrate that exposure to silica, even without silicosis, increases the risk of infectious (i.e., active) pulmonary TB (Sherson and Lander, 1990, Document ID 0434; Cowie, 1994, 0992; Hnizdo and Murray, 1998, 0360; teWaterNaude et al., 2006, 0465). Both conditions together can Start Printed Page 16383hasten the development of respiratory impairment and increase mortality risk even beyond that experienced by persons with active TB who have not been exposed to respirable crystalline silica (Banks, 2005, Document ID 0291).

Based on the information presented above and in its review of the health literature, OSHA concludes that silicosis remains a significant cause of early death and of serious illness, despite the existence of an enforceable exposure limit over the past 40 years. Silicosis in its later stages of progression (i.e., with chest x-ray findings of ILO category 2 or 3 profusion of small opacities, or the presence of large opacities) is characterized by the likely appearance of respiratory symptoms and decreased pulmonary function, as well as increased risk of progression to PMF, disability, and early mortality. Early-stage silicosis, although without symptoms among many who are affected, nevertheless reflects the formation of fibrotic lesions in the lung and increases the risk of progression to later stages, even after exposure to respirable crystalline silica ceases. In addition, the presence of silicosis increases the risk of pulmonary infections, including conversion of latent TB infection to active TB. Silicosis is not a reversible condition, and there is no specific treatment for the disease, other than administration of drugs to alleviate inflammation and maintain open airways, or administration of oxygen therapy in severe cases. Based on these considerations, OSHA finds that silicosis of any form, and at any stage of progression, is a material impairment of health and that fibrotic scarring of the lungs represents loss of functional respiratory capacity.

2. Lung Cancer

OSHA considers lung cancer, an irreversible and frequently fatal disease, to be a clear material impairment of health (see Homer et al., 2009, Document ID 1343). According to the National Cancer Institute (SEER Cancer Statistics Review, 2006, Document ID 1343), the five-year survival rate for all forms of lung cancer is only 15.6 percent, a rate that has not improved in nearly two decades. After reviewing the record as a whole, OSHA finds that respirable crystalline silica exposure substantially increases the risk of lung cancer. This finding is based on the best available toxicological and epidemiological data, reflects substantial supportive evidence from animal and mechanistic research, and is consistent with the conclusions of other government and public health organizations, including the International Agency for Research on Cancer (1997, Document ID 1062; 2012, Document ID 1473), the HHS National Toxicology Program (2000, Document ID 1417), the CDC's National Institute for Occupational Safety and Health (2002, Document ID 1110), the American Thoracic Society (1997, Document ID 0283), and the American Conference of Governmental Industrial Hygienists (2010, Document ID 0515).

The Agency's primary evidence comes from evaluation of more than 50 studies of occupational cohorts from many different industry sectors in which exposure to respirable crystalline silica occurs, including: Granite and stone quarrying; the refractory brick industry; gold, tin, and tungsten mining; the diatomaceous earth industry; the industrial sand industry; and construction. In addition, the association between exposure to respirable crystalline silica and lung cancer risk was reported in a national mortality surveillance study (Calvert et al., 2003, Document ID 0309) and in two community-based studies (Pukkala et al., 2005, Document ID 0412; Cassidy et al., 2007, 0313), as well as in a pooled analysis of 10 occupational cohort studies (Steenland et al., 2001a, Document ID 0452). Toxicity studies provide supportive evidence of the carcinogenicity of crystalline silica, in that they demonstrate biologically plausible mechanisms by which crystalline silica in the deep lung can give rise to biochemical and cellular events leading to tumor development (see Section V.H, Mechanisms of Silica-Induced Adverse Health Effects).

3. Non-Malignant Respiratory Disease (NMRD) (Other Than Silicosis)

Although many of the stakeholders in this rule have focused their attention on the evidence related to silicosis and lung cancer, the available evidence shows that exposure to respirable crystalline silica also increases the risk of developing NMRD, in particular chronic bronchitis and emphysema. OSHA has determined that NMRD, which results in loss of pulmonary function that restricts normal activity in individuals afflicted with these conditions (see American Thoracic Society, 2003, Document ID 1332), constitutes a material impairment of health. Both chronic bronchitis and emphysema can occur in conjunction with the development of silicosis. Several studies have documented increased prevalence of chronic bronchitis and emphysema among silica-exposed workers even absent evidence of silicosis (see Document ID 1711, pp. 182-192; NIOSH, 2002, 1110; American Thoracic Society, 2003, 1332). There is also evidence that smoking may have an additive or synergistic effect on silica-related NMRD morbidity or mortality (Hnizdo, 1990, Document ID 1045; Hnizdo et al., 1990, 1047; Wyndham et al., 1986, 0490; NIOSH, 2002, 1110). In a study of diatomaceous earth workers, Park et al. (2002, Document ID 0405) found a positive exposure-response relationship between exposure to respirable cristobalite (a form of silica) and increased mortality from NMRD.

Decrements in pulmonary function have often been found among workers exposed to respirable crystalline silica absent radiologic evidence of silicosis. Several cross-sectional studies have reported such findings among granite workers (Theriault et al., 1974a, Document ID 0466; Wallsh, 1997, 0477; Ng et al., 1992b, 0387; Montes II et al., 2004b, 0377), gold miners (Irwig and Rocks, 1978, Document ID 1067; Hnizdo et al., 1990, 1047; Cowie and Mabena, 1991, 0342), gemstone cutters (Ng et al., 1987b, Document ID 1113), concrete workers (Meijer et al., 2001, Document ID 1243), refractory brick workers (Wang et al., 1997, Document ID 0478), hard rock miners (Manfreda et al., 1982, Document ID 1094; Kreiss et al., 1989, 1079), pottery workers (Neukirk et al., 1994, Document ID 0381), slate workers (Surh, 2003, Document ID 0462), and potato sorters exposed to silica in diatomaceous earth (Jorna et al, 1994, Document ID 1071).

OSHA also evaluated several longitudinal studies where exposed workers were examined over a period of time to track changes in pulmonary function. Among both active and retired granite workers exposed to an average of 60 μg/m 3, Graham et al. did not find exposure-related decrements in pulmonary function (1981, Document ID 1280; 1984, 0354). However, Eisen et al. (1995, Document ID 1010) did find significant pulmonary decrements among a subset of granite workers (termed “dropouts”) who left work and consequently did not voluntarily participate in the last of a series of annual pulmonary function tests. This group of workers experienced steeper declines in FEV1 compared to the subset of workers who remained at work and participated in all tests (termed “survivors”), and these declines were significantly related to dust exposure. Thus, in this study, workers who had left work had exposure-related declines in pulmonary function to a greater extent than did workers who remained on the job, clearly demonstrating a survivor effect among the active Start Printed Page 16384workers. Exposure-related changes in lung function were also reported in a 12-year study of granite workers (Malmberg, 1993, Document ID 0370), in two 5-year studies of South African miners (Hnizdo, 1992, Document ID 1046; Cowie, 1988, 0993), and in a study of foundry workers whose lung function was assessed between 1978 and 1992 (Hertzberg et al., 2002, Document ID 0358).

Each of these studies reported their findings in terms of rates of decline in any of several pulmonary function measures, such as FVC, FEV1, and FEV1/FVC. To put these declines in perspective, Eisen et al. (1995, Document ID 1010) reported that the rate of decline in FEV1 seen among the dropout subgroup of Vermont granite workers was 4 ml per mg/m3-yrs of exposure to respirable granite dust; by comparison, FEV1 declines at a rate of 10 ml/year from smoking one pack of cigarettes daily. From their study of foundry workers, Hertzberg et al., reported finding a 1.1 ml/year decline in FEV1 and a 1.6 ml/year decline in FVC for each mg/m3-yrs of respirable silica exposure after controlling for ethnicity and smoking (2002, Document ID 0358, p. 725). From these rates of decline, they estimated that exposure to the previous OSHA general industry quartz standard of 100 µg/m3 for 40 years would result in a total loss of FEV1 and FVC that is less than but still comparable to smoking a pack of cigarettes daily for 40 years. Hertzberg et al. also estimated that exposure to the current standard for 40 years would increase the risk of developing abnormal FEV1 or FVC by factors of 1.68 and 1.42, respectively (2002, Document ID 0358, pp. 725-726). OSHA believes that this magnitude of reduced pulmonary function, as well as the increased morbidity and mortality from non-malignant respiratory disease (NMRD) that has been documented in the studies summarized above, constitute material impairments of health and loss of functional respiratory capacity.

4. Renal and Autoimmune Effects

Finally, OSHA's review of the literature reflects substantial evidence that exposure to crystalline silica increases the risk of renal and autoimmune diseases, both of which OSHA considers to be material impairments of health (see Section V.C, Summary of the Review of Health Effects Literature and Preliminary QRA). Epidemiological studies have found statistically significant associations between occupational exposure to silica dust and chronic renal disease (e.g., Calvert et al., 1997, Document ID 0976), subclinical renal changes including proteinurea and elevated serum creatinine (e.g., Ng et al., 1992c, Document ID 0386; Rosenman et al., 2000, 1120; Hotz, et al., 1995, 0361), end-stage renal disease morbidity (e.g., Steenland et al., 1990, Document ID 1125), chronic renal disease mortality (Steenland et al., 2001b, Document ID 0456; 2002a, 0448), and granulomatosis with polyangitis (Nuyts et al., 1995, Document ID 0397). Granulomatosis with polyangitis is characterized by inflammation of blood vessels, leading to damaging granulomatous formation in the lung and damage to the glomeruli of the kidneys, a network of capillaries responsible for the first stage of blood filtration. If untreated, this condition often leads to renal failure (Nuyts et al., 1995, Document ID 0397, p. 1162). Possible mechanisms for silica-induced renal disease include a direct toxic effect on the kidney and an autoimmune mechanism (see Section V.H, Mechanisms of Silica-Induced Adverse Health Effects; Calvert et al., 1997, Document ID 0976; Gregorini et al., 1993, 1032). Steenland et al. (2002a, Document ID 0448) demonstrated a positive exposure-response relationship between exposure to respirable crystalline silica and end-stage renal disease mortality.

In addition, there are a number of studies that show exposure to be related to increased risks of autoimmune disease, including scleroderma (e.g., Sluis-Cremer et al., 1985, Document ID 0439), rheumatoid arthritis (e.g., Klockars et al., 1987, Document ID 1075; Rosenman and Zhu, 1995, 0424), and systemic lupus erythematosus (e.g., Brown et al., 1997, Document ID 0974). Scleroderma is a degenerative disorder that leads to over-production of collagen in connective tissue that can cause a wide variety of symptoms including skin discoloration and ulceration, joint pain, swelling and discomfort in the extremities, breathing problems, and digestive problems. Rheumatoid arthritis is characterized by joint pain and tenderness, fatigue, fever, and weight loss. Systemic lupus erythematosus is a chronic disease of connective tissue that can present a wide range of symptoms including skin rash, fever, malaise, joint pain, and, in many cases, anemia and iron deficiency. OSHA considers chronic renal disease, end-stage renal disease mortality, granulomatosis with polyangitis, scleroderma, rheumatoid arthritis, and systemic lupus erythematosus clearly to be material impairments of health.

C. OSHA's Final Quantitative Risk Estimates

To evaluate the significance of the health risks that result from exposure to hazardous chemical agents, OSHA relies on epidemiological and experimental data, as well as statistical methods. The Agency uses these data and methods to characterize the risk of disease resulting from workers' exposure to a given hazard over a working lifetime at levels of exposure reflecting both compliance with previous standards and compliance with the new standard. In the case of respirable crystalline silica, the previous general industry, construction, and shipyard PELs were formulas that limit 8-hour TWA exposures to respirable dust; the limit on exposure decreased with increasing crystalline silica content of the dust. OSHA's previous general industry PEL for respirable quartz was expressed both in terms of a particle count and a gravimetric concentration, while the previous construction and shipyard employment PELs for respirable quartz were only expressed in terms of a particle count formula. For general industry, the gravimetric formula