Full opinion text
PER CURIAM: Respirable crystalline forms of silica, a compound made of silicon and oxygen, are commonly found in workplaces with rock, sand, gravel, concrete, and brick. Exposure to silica is one of the oldest known occupational hazards. And the health effects of exposure to silica—most commonly silicosis, a .progressive and irreversible lung disease caused by the inflammatory effects of silica—are not a thing of the past. “Currently, silicosis is the most prevalent chronic occupational disease in the world.” Robbins & Cotran, Pathologic Basis Of Disease 690 (9th ed. 2015). In the United States, more than two million workers are currently exposed to some level of silica. In 2016, the Occupational Safety and Health Administration (OSHA), an agency within the United States Department of Labor, published a final rule regulating workplace exposure to silica. Occupational Exposure to Respira-ble Crystalline Silica, 81 Fed. Reg. 16,285 (Mar. 25, 2016) (codified at 29 C.F.R. Pts. 1910, 1915, and 1926) (Silica Rule or Rule). Petitions to review the Rule came from both sides; a collection of industry petitioners (Industry) believes OSHA imper-missibly made the Rule too stringent and several union petitioners (Unions) believe OSHA improperly failed to make the Rule stringent enough. Industry petitioned for review of five issues: (1) - whether substantial evidence supports OSHA’s finding that limiting workers’ silica exposure to the level set by the Rule reduces a significant risk of material health impairment; (2) whether substantial evidence supports OSHA’s finding that the Rule is technologically feasible for the foundry, hydraulic fracturing, and construction industries; (3) whether substantial evidence supports OSHA’s finding that the Rule is economically feasible for the foundry, hydraulic fracturing, and construction industries; (4) whether OSHA violated the Administrative Procedure Act (APA) in promulgating the Rule; and (5) whether substantial evidence supports two ancillary provisions of the Rule—one that allows workers who undergo medical examinations to keep the results confidential from their employers and one that prohibits employers from using dry cleaning methods unless doing so is infeasible. We reject all of Industry’s challenges. The Unions petitioned for review of two parts of the Rule: (1) the requirement that medical surveillance for construction workers be provided only if the employee has to wear a respirator for 30 days for one employer in a one-year period; and (2) the absence of medical removal protections. We reject the Unions’ challenge to the construction standard’s 30-day trigger for medical surveillance. We conclude that OSHA failed to adequately explain its decision to omit medical removal protections from the Rule and remand for further consideration of the issue. I. BACKGROUND The Occupational Safety and Health Act (OSH Act) authorizes the Secretary of Labor (Secretary) to “promulgate, modify, or revoke any occupational safety or health standard,” 29 U.S.C. § 655(b), by requiring conditions or the adoption of practices, means, or methods “reasonably necessary or appropriate to provide safe or healthful employment and places of employment,” id. § 652(8). If the standard applies to toxic materials or harmful physical agents, 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 or functional capacity even if such employee has regular exposure to the hazard” regulated by the standard “for the period of- his working life.” Id. § 655(b)(5). The Secretary has delegated his authority to OSHA. See 72 Fed. Reg. 31,160 (June 5,2007). In 1971, OSHA adopted a standard regulating exposure to a variety of substances, including silica. Occupational Safety and Health Standards; National Consensus Standards and Established Federal Standards, 36 Fed. Reg. 10,466 (May 29, 1971). The 1971 rule established a permissible exposure limit (PEL)—a time-weighted average of a worker’s exposure during a workday—of 100 micrograms per cubic meter (pg/m) in general industry and 250 pg/m in the construction industry. See 81 Fed. Reg. at 16,294. In the 1990s, OSHA studied the efficacy of the 1971 rule regarding silica-related health effects in the workplace and concluded a new rule was needed. See id. at 16,295. In 2016, OSHA promulgated its final Silica Rule. 81 Fed. Reg. 16,285. The Rule lowers the PEL to 50 pg/m for all covered industries, including as particularly relevant here, the foundry, hydraulic fracturing, brick, and construction industries. See 29 C.F.R. §§ 1910.1053(c), 1926.1153(d)(1). Employers must assess silica exposure levels in the workplace (or, for certain construction industry tasks, adopt specific “safe-harbor” practices) and, if necessary, must implement engineering and work practice controls to keep exposures below the PEL. Id. §§ 1910.1053(f)(1), 1926.1153(c)(1), 1926.1153(d)(3)©. If engineering and work practice controls cannot reduce exposures to the PEL, the employer must use controls to the extent feasible and provide supplementary respirator protections. Id. The Silica Rule also establishes various ancillary provisions including, again, as relevant here, housekeeping requirements and medical surveillance requirements. Under the challenged housekeeping provision, employers are prohibited from using dry sweeping methods to clean worksites if doing so could contribute to employee exposure to silica unless wet cleaning methods are infeasible. Id. §§ 1910.1053(h)(1), 1926.1153(f)(1). Under the challenged medical surveillance provisions, employers must provide medical screening to silica-exposed workers if certain conditions are met. Most of the information from the medical examinations, including medical professionals’ recommendations limiting the employee’s exposure to silica, are confidential and cannot be released to- the employer unless the employee authorizes disclosure. Id. §§ 1910.1053(i)(6), 1926.1153(h)(6). Finally, the Rule provides no medical removal protections to workers whose doctors recommend either permanent or temporary removal from silica exposure on the job. Different compliance dates were established for each industry: June 23, 2017 for the construction industry, id. § 1926.1153(k); June 23,- 2018 for the foundry industry, id. § 1910.1053(0; and June 23, 2021 for the hydraulic fracturing industry, id. II. ANALYSIS We first decide Industry’s challenges. In order, we address OSHA’s significant risk findings, its technological feasibility findings, its economic feasibility findings, the procedural regularity of the Rule, and the challenged ancillary provisions. The substantive issues are governed by the “substantial evidence” standard, 29 U.S.C. § 655(f), under which we require OSHA to “identify relevant factual evidence, to explain the logic and the policies underlying any legislative choice, to state candidly, any assumptions on which it re-: lies, and to present its reasons for rejecting significant contrary evidence and argu-. ment,” United Steelworkers of America v. Marshall (Lead I), 647 F.2d 1189, 1207 (D.C. Cir. 1980). The APA governs the procedural challenge to ensure the Rule is not promulgated “without observance of procedure required by law.” 5 U.S.C. § 706(2)(D). We then turn to the Unions’ challenges and address the "30-day medical surveillance trigger in the construction standard and the lack of medical removal protections in the general industry standard. Where the Unions have failed to identify evidence that their proposals would be feasible and generate more than a de min-imis benefit to worker health, we reject them. See Building & Construction Trades Department, AFL-CIO v. Brock (Asbestos), F.2d 1258, 1271 (D.C. Cir. 1988). Where the Unions have met this initial burden, we ask whether OSHA has supported its decision with substantial evidence and otherwise engaged in reasoned decisionmaking.. A. Significant Risk Before OSHA promulgates any permanent health or safety standard, it must make a “threshold finding” that “it is at least more likely than 'not that' long-term expo'sure” to the regulated substance at current exposure levels “presents a significant risk of material impairment” that “can be eliminated or lessened by a change' in practices.” Industrial Union Department, AFL-CIO v. American Petroleum Institute (Benzene), 448 U.S. 607, 642, 653, 100 S.Ct. 2844, 65 L.Ed.2d 1010 (1980) (plurality).. The Supreme Court has provided the guidepost that OSHA follows: a one-in-a-thousand risk that exposure to the regulated substance will be fatal can reasonably be considered significant but a one-in-a-billion risk is likely not significant. Id. at 655-56,100 S.Ct. 2844. OSHA must support its significant .risk finding with substantial evidence. Id, at 653,100 S.Ct. 2844. Although it must rely on a “body of reputable scientific thought” when assessing risk, id. at 656, 100 S.Ct. 2844, OSHA does not have to “calculate the exact probability of harm” or support its finding “with anything approaching scientific certainty,” id. at 655-56, 100 S.Ct. 2844, OSHA is entitled to “some leeway” when its “findings must be made on the frontiers of scientific knowledge.” Id. at 656, 100 S.Ct. 2844, We “do not reweigh the evidence and come to our own conclusion[s]; rather, we assess the reasonableness of OSHA’s conclusion.” Public Citizen Health Research Group v. Tyson (Ethylene Oxide), 796 F.2d 1479, 1495 (D.C. Cir. 1986). In promulgating the Silica Rule, OSHA conducted a Quantitative Risk Assessment in which it reviewed toxicological, epidemiological, and experimental studies about the adverse health effects of silica exposure. 81 Fed. Reg. at 16,380. OSHA quantified the excess risk of silica-related health effects assuming exposure over a working life (45 years) to various levels of silica, including the original general industry PEL of 100 (xg/m3, the original construction PEL of 250 (xg/m3, and the new PEL of 50 (xg/m3. Id. at 16,300. OSHA concluded that silica exposure significantly “increases the risk of’ four adverse health effects:, silicosis and other non-malignant respiratory disease (NMRD) mortality, lung cancer mortality, silicosis morbidity, and renal disease mortality. Id: at 16,300, 16,386-87. OSHA also concluded that the risks at 50 (xg/m3—the new PEL—are lower than the risks at the original PELs of 100 (xg/m3 and 250 (xg/m3.- Id. at 16,300. In total, OSHA estimated that the Silica Rule will prevent 642 deaths and 918 cases of silica-related disease each year. Id. at 16,-399. Industry challenges OSHA’s significant risk findings in three ways. First, Industry attack's two parts of OSHA’s risk-assessment methodology. Second, it challenges OSHA’s findings on each of the four individual health risks. Finally, Industry challenges, OSHA’s decision to include the brick industry within the scope of the Rule. We reject each challenge. 1. OSHA’s Methodology Industry challenges two components ^of OSHA’s risk-assessment methodology: its no-threshold assumption and its failure to account for a dose-rate effect; We uphold OSHA’s decisions on both. First, Industry challenges OSHA’s use of no-threshold exposure-response models in its risk assessments for - silicosis and lung cancer. 81 Fed. Reg. at 16,351. The no-threshold concept means there is no exposure-level below which workers would not be expected to develop adverse health effects. Id. OSHA did not definitively find that no threshold exists. Instead, it found' that if a threshold exists it does so below the PEL, which justified its use of a no-threshold model OSHA supported its selection of the PEL with studies showing that risks of lung cancer exist at 36 (xg/m3 and 10 pg/m3, levels lower than the PEL. Id. at 16,351,16,356. To OSHA, the studies showing risks below the PEL support its conclusion that any threshold, if. it exists, does so below the PEL. See id. at-16,351 (“As 36 pg/m3 is well below the previous industry PEL of 100 pg/m3 and below the final PEL of 50 pg/m3, the ... study showed no evidence of an exposure-response threshold high enough to impact OSHA’s choice of PEL.”). Industry, in contrast, points to studies it claims not only show a threshold exists but also show a threshold exists above the PEL. OSHA rejected Industry’s argument because the contrary studies used non-reactive and poorly soluble particles—which silica is not—and therefore the “findings regarding” the particles “[cannot] be extrapolated to crystalline silica.” Id. at 16,349. OSHA acknowledged “there is considerable uncertainty” about whether a threshold exists but found that “the weight of evidence supports the view that, if there is a threshold,” it is “likely lower than the” PEL. Id. at 16,351. OSHA’s no-threshold assumption is supported by substantial evidence. Although Industry claims OSHA’s position is inconsistent with common sense and “mounting judicial skepticism” of no-threshold models, citing to several district court and state court cases disapproving-a no-threshold approach, Industry Br. at 28-29, OSHA’s position is in line with our precedent. In Ethylene Oxide, we upheld a no-threshold model based on OSHA’s having found evidence of adverse health effects at levels of exposure to ethylene oxide below the established PEL, then extrapolating that evidence to assume no threshold of ethylene oxide exposure existed below which risks did not exist and rejecting two contrary comments that purportedly showed a threshold did exist. 796 F.2d at 1500. As in Ethylene Oxide, Industry presents, and urges us to adopt, “one side of the debate.” Id. But OSHA has explained why it rejected Industry’s side of the debate, presented the other side of the debate, and supported it with evidence from which a reasonable conclusion could be made, as OSHA did here, that no threshold of safe exposure to silica exists. We cannot “choose a particular side as the ‘right’ one” in a scientific dispute. Id. Accordingly, OSHA’s no-threshold assumption satisfies our substantial evidence test. Second, Industry challenges OSHA’s decision not to include a- dose-rate effect in the model, which means OSHA assessed health risks based on the cumulative amount of silica exposure without accounting for the intensity of exposures. 81 Fed. Reg. at 16,375. OSHA took its position “because each of the key ... studies” OSHA relied on used cumulative exposure as the only metric. Id. at 16,374-75. Multiple commenters supported the notion that “cumulative exposure is a reasonable and practical choice” and that cumulative exposure “is often the best predictor of chronic disease.” Id. at 16,375. Competing com-menters argued that OSHA’s risk assessment should account for the intensity of exposures. Id. Industry relied on studies showing that , not accounting for a dose-rate effect “could overestimate risk at lower concentrations.” Id. The studies supporting ^Industry’s position, however, largely observed an intensity-based effect at 500 pg/m3 and 2,000 pg/m3, exposure levels so “far above the previous PEL,” id. at 16,395, that OSHA determined the studies were of little use to the “exposure range of interest”—25 to 500 pg/m3, id. at 16,376. In Ethylene Oxide, we upheld OSHA’s decision not to include a dose-rate effect in its model when faced with “competing technical opinions” about whether the amount or the intensity of ethylene oxide exposure mattered more. Ethylene Oxide, 796 F.2d at 1604. OSHA did the same in its Silica Rule: it took competing evidence, favored one side, and explained the reasons for its decision. We “cannot expect OSHA to [locate and use] absolutely conclusive studies on these difficult medical issues” and we must uphold OSHA’s choice, even in the face of “controverted” evidence, if it falls within a “zone of reasonableness.” Lead I, 647 F.2d at 1253 (quoting Hercules, Inc. v. EPA, 598 F.2d 91, 107 (D.C. Cir. 1978)). We believe OSHA’s conclusions on handling the purported dose-rate effect are reasonable. “[Cjourts cannot interfere with reasonable interpretations of equivocal evidence,” Ethylene Oxide, 796 F.2d at 1505, and therefore we do not interfere here. 2. Adverse Health Effects As noted earlier, OSHA concluded that long-term silica exposure above the PEL presents a significant risk of four discrete adverse health effects: (1) silicosis and NMRD mortality; (2) lung cancer mortality; (3) silicosis morbidity; and (4) renal disease mortality. 81 Fed. Reg. at 16,300, 16,386-87. Industry challenges OSHA’s findings as to all four. Industry acknowledged at oral argument that, to prevail, it would have to show none of the discrete findings is supported by substantial evidence. We address each in turn. We conclude OSHA’s significant risk findings as to the first three adverse health effects are supported by substantial evidence, which supports OSHA’s overall finding of a significant risk. We do not reach OSHA’s finding with respect to renal disease mortality. i. Silicosis or Non-Malignant Respiratory Disease Mortality Silicosis is a progressive, irreversible lung disease caused by the inflammatory effects of silica in the lungs. OSHA found that silica exposure at the original PEL of 100 ¡xg/m3 created an excess risk of silicosis mortality for 11 in 1,000 workers that would be reduced to 7 in 1,000 workers at the Rule’s PEL of 50 ¡xg/m3. 81 Fed. Reg. at 16,303, 16,312. Other NMRD caused by silica exposure include emphysema, chronic obstructive pulmonary disease, and chronic bronchitis. Id. at 16,304. OSHA found that silica exposure at the 100 ¡xg/m3 PEL created an excess risk of NMRD mortality (including silicosis mortality) for 85 in 1,000 workers that would be reduced to 44 in 1,000 workers at the Rule’s PEL of 50 ¡xg/m3. Id. at 16,303. Both Industry and the Chambers Intervenors challenge OSHA’s findings on silicosis and NMRD mortality. To support its findings on silicosis and NMRD mortality, OSHA relied on two studies: the Mannetje study, which showed a statistically significant association between silicosis mortality and cumulative exposure to silica, and the Park study, which quantified the relationship between silica exposure and NMRD mortality. Id. at 16,317. Industry’s objections to OSHA’s conclusions primarily attack the reliability of the Park study. Industry claims the Park study (1) focused on workers with cumulative exposure levels far above what workers typically faced under the original PEL and (2) produced results that were likely skewed by smoking because the study had smoking data for only one-half of the studied workers. In its rulemaking, OSHA addressed both criticisms. On the first point, OSHA acknowledged “some uncertainty in using models heavily influenced by exposures above the previous PEL” but noted that the average cumulative exposure of the studied workers was “lower than what the final rule would permit'over 45 years of exposure.” Id. at 16,318. Accordingly, OSHA “[dis]agree[d] that the Park study should be discounted” and instead concluded that the study was both relevant and appropriate to rely on. Id. On the second point, OSHA" acknowledged that “comprehensive smoking data would be ideal” but assessed the Park study’s mechanics in detail and concluded that the risk estimates were “not likely to be exaggerated due to [studied workers’] smoking habits.” Id. Under our substantial evidence standard, OSHA has a duty to “present its reasons for rejecting significant contrary evidence and argument.” Lead I, 647 F.2d at 1207. OSHA acknowledged and adequately responded to Industry’s criticisms of the Park study. Even if the Park study was “flawed in some way,” OSHA is not precluded from relying on- imperfect evidence so long as it “recognize[s] and accounts] for the methodological weaknesses” of. the evidence. Ethylene Oxide, 796 F.2d at 1487; see id. at 1495 (“While some of OSHA’s evidence suffers from shortcomings, such incomplete proof is inevitable when the Agency regulates on the frontiers of scientific knowledge.”). OSHA did recognize. and account for the weaknesses of the two studies it. relied on here. The Chambers, meanwhile, present a record of death certificates and their listed cause- of death that shows silicosis-attributed deaths dropped from 1,065 in 1968 (three years before the 1971 PEL was implemented) to 123 in 2007. The decline, according to the Chambers, shows that the current risks are due not to exposure levels at the 1971 PEL but instead are due to pre-1971 exposures or exposures occurring in violation of the 1971 PEL. Thus, the Chambers argue, the 1971 rule is working and there is no need for a new one. But here again, OSHA adequately explained why it rejected this evidénce. First, OSHA concluded that the death certificate data underreported risks after one commenter found that silicosis was listed as the cause of death for only 14 percent of people with confirmed -silicosis. 81 Fed. Reg. at 16,328. Second, the death certificate data “d[id] not include information about exposure[]” levels for those who died as a result of silicosis, which made the data “inadequate and inappropriate for” setting a standard regulating silica at particular exposure levels. Id. at 16,326. Indeed, the agency that compiled and analyzed the death certificate data testified that relying on the death certificates to show no significant risk exists- would be a “misuse” of the data. Id. Moreover, OSHA responded directly to the Chambers’ arguments that the death certificate data showed the risks of silica exposure are no longer significant. OSHA acknowledged that silicosis-related deaths have dropped since 1968 but pointed to evidence showing that the decline leveled off at approximately 90 to 180 deaths per year since 2000. Id: at 16,324. This evidence “suggests] that the number of silicosis deaths ... may be stabilizing,” id:, which also suggests that the significant risk of silicosis mortality would not disappear if OSHA simply let the 1971 PEL. run its course, as Industry argued, id. at 16,-325. OSHA also pointed to evidence showing that the decline in silicosis-related deaths tracks the decline in high-exposure jobs as much as it tracks improved working conditions, further 'suggesting that OSHA “still h[as] work to do” to make silica exposure safe. Id. at 16,325-26. Thus, although OSHA agreed that the death certificate data was “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,” the “limited and incomplete”- data made reliance on the death certificates “inappropriate.” Id. at 16,330.. OSHA “described in some detail [its] reasons for choosing between competing alternatives.” Asbestos, 838 F.2d at 1266. Accordingly, OSHA has met its burden to identify the evidence it relied on and explain why it rejected contrary evidence. ii. Lung Cancer Mortality OSHA found that silica exposure at the 100 (Tg/m3 PEL created an excess risk of lung* cancer mortality equal to 11 to 54 deaths per 1,000 workers that would be reduced to an excess risk of 5 to 23 deaths per 1,000 workers at the 50 |xg/m3 PEL. 81 Fed. Reg. at 16,338. Industry argues the conclusion hinges on OSHA’s unsupported .assumption that silica exposure directly increases the risk of lung cancer in the absence of silicosis. That is, if the risk of lung cancer depends on pre-existing silicosis, then silica exposure alone does not create an independent risk of lung cancer. Industry points to evidence that asserts the association between silicosis and lung cancer is “more compelling” than the association between silica exposure and lung cancer. Joint Appendix (J.A.) 3027. But the mere suggestion in some evidence that silicosis is a necessary precursor of lung cancer does not bind the agency. See Ethylene Oxide, 796 F.2d at 1504 (noting that suggestive statements “do' not amount to a scientific certainty binding on the agency”). Meanwhile, OSHA also cites to numerous studies that show silica exposure can lead directly to lung cancer. 81 Fed. Reg. at 16,309 (recapping and summarizing findings). As one commenter put it, the literature OSHA relied on shows “silica has been established as a cause of lung cancer.” J.A,' 7815. We lack the technical expertise to. second-guess OSHA’s judgment when it “reviewed] all sides of the issue and reasonably resolvefd] the matter.” Ethylene Oxide, 796 F.2d at 1500. We do not second-guess OSHA’s conclusions here. Industry specifically challenges OSHA’s decision to give weight to a 2004 Attfield and Costello study, which showed there is an association between silica exposure and lung cancer, instead of a 2011 Vacek study showing there is no such association. 81 Fed. Reg. at 16,338. Industry provides a laundry list of reasons why it believes the Vacek study is better: it is more recent, covered more workers, covered more years, and used more detailed information. But OSHA explained its reasons for rejecting the Vacek study. Among them: the Vacek study found an unexplained significant excess risk of lung cancer that called ■into question all of its results and had a low risk' estimate for a particular type of worker (channel bar operators) that OSHA concluded had major consequences for the entire exposure analysis. Id. at 16,335-37. Moreover, OSHA provided affirmative reasons for choosing the Attfield and Costello study. Most importantly, OSHA reasoned, that study accounted for a healthy worker survivor effect—the tendency of healthy workers to remain in the workforce longer than ill workers and therefore face more exposure than ill workers, which- “may” make the “risk of disease at higher exposures” improperly “appear to be constant or decrease”—but the Vacek study did not assess the healthy worker survivor effect. Id. at 16,336. “We have then, at worst, the ordinary situation of controverted evidence, in which we must defer to the reasonable and conscientious interpretations of the agency.” Lead I, 647 F.2d at 1258. Hi. Silicosis Morbidity To support its finding of a significant risk of silicosis morbidity, OSHA relied on five studies that showed an excess risk between 60 and 773 cases of’silicosis morbidity per 1,000 workers at a level of 100 fxg/m3 that would be reduced to an excess risk between 20 and 170 cases of silicosis morbidity per 1,000 workers at a level of 50 |xg/m3. 81 Fed. Reg. at 16,317. The variance among studies, according to Industry, “suggests that none of [the studies] is a reliable guide to a correct quantification” of exposures and therefore none of the studies can support a finding of a significant risk of silicosis morbidity. J.A. 3368. OSHA concluded the results of the five studies did not “differ remarkably,” 81 Fed. Reg. at 16,321, which Industry asserts is “arbitrary and capricious reasoning,” Industry Br. at 39. In the rulemaking OSHA responded to critiques against the individual studies upon which OSHA relied. See 81 Fed. Reg. at 16,320-22. OSHA also responded to critiques of the variance among the studies, albeit in less detailed fashion, and concluded that the risk estimates among the studies “are in reasonable agreement.” Id. at 16,322. OSHA’s reconciliation of the data’s variance is not airtight. A more important question, however, is whether the studies constitute substantial evidence supporting OSHA’s finding of a significant risk of silicosis morbidity at the initial PEL that is reduced at the Rule’s PEL. They do. The variance in results may show uncertainty as to the precise amount of the risk of silicosis morbidity. Maybe it falls closer to 60 cases per 1,000 workers at 100 ¡xg/m3 ; maybe it falls closer to 773 per 1,000. Maybe it falls closer to 20 cases per 1,000 workers at 50 |xg/m3; maybe it falls closer to 170. “While each study individually may not be a model of textbook scientific inquiry,” we assess, again, the “cumulative evidence” OSHA relied on. Ethylene Oxide, 796 F.2d at 1489. Even assuming the actual amount of risk is closer to the low end, a “reasonable person could draw from this evidence the conclusion that exposure to” silica presents a significant risk of silicosis morbidity. Id. “Even if a reasonable person could also draw the opposite conclusion, we must uphold the agency’s findings.” Id. We conclude, then, that OSHA’s conclusion that exposure to silica presents a significant risk of silicosis morbidity is supported by substantial evidence. iv. Renal Disease Mortality OSHA concluded that the excess risk of renal disease mortality would drop from 39 deaths per 1,000 workers at the 100 |xg/m3 PEL to 32 deaths per 1,000 workers at the 50 |xg/m3 PEL. 81 Fed. Reg. at 16,342. OSHA relied on a single pooled study that provided “considerably less data” compared to the studies of the other disease endpoints. Id. at 16,345. OSHA rejected numerous other studies that showed no risk of renal disease. Id. at 16,344-45. Industry argues that OSHA lacked substantial evidence to support a finding of a significant risk of renal disease mortality and failed to explain its resolution, of conflicting evidence. OSHA acknowledged in the rulemaking that the evidence supporting its finding regarding renal disease mortality was “less robust” than the evidence supporting its findings for other silica-related health effects. Id. at 16,345. OSHA defended its position with a single footnote in its brief. We note OSHA’s concession that the evidence is weak; if OSHA had relied solely on the risk of' renal disease mortality to support the Silica Rule, its decision may well have been unsupported by substantial evidence. But we need not and do not decide whether OSHA supported its renal disease findings with substantial evidence because OSHA’s findings with respect to silicosis and NMRD mortality, lung cancer mortality, and silicosis morbidity are sufficient to uphold the requisite threshold finding of a significant risk of material health impairment at the 100 |xg/m3 PEL that will be reduced at the new PEL. See National Maritime Safety Association v. OSHA, 649 F.3d 743, 752 n.11 (D.C. Cir. 2011) (upholding OSHA’s significant risk finding where OSHA relied On four contributors to the risk but one was flawed; if OSHA “relied on [the flawed] factor alone, its significant risk determination might well have been arbitrary and capricious” but the presence of the other substantiated factors sufficiently supported OSHA’s significant risk finding). And Industry does not show that any weakness with respect to OSHA’s renal disease findings infected OSHA’s findings regarding the other adverse health effects such that the entire significant risk conclusion is undermined. Accordingly, we do not decide the renal disease issue because OSHA, through its supported findings on the other three adverse health effects, has met its burden to show that the Silica Rule regulates a significant risk of material harm. 3. Brick Industry Industry argues OSHA should have excluded the brick industry from the scope of the Silica Rule because OSHA did not have substantial evidence to find a significant risk of material harm in the brick industry. OSHA pins its findings on one study (the Love study) that surveyed brick plant workers. 81 Fed. Reg. at 16,-378. The Love study reported that 1.4 percent—a rate below OSHA’s risk estimates in other industries but exceeding Benzene’s general 0.1 percent benchmark—of brick workers had small abnormalities in x-rays, which the authors said were “most likely” silicosis. Id. Industry makes three arguments to contest the findings based on the Love study. First, Industry argues that OSHA used the Love study when it wanted to and did not use the Love study when it did not want to. Specifically, OSHA found the Love study showed a significant risk of silicosis but declined to include the Love study in the group of studies that formed the basis of OSHA’s silicosis morbidity quantitative risk assessment. See id. at 16,377-78. If OSHA exhibits “apparently inconsistent handling of the evidence available to it,” OSHA cannot be said to have relied on the best available evidence. See American Iron & Steel Institute v. OSHA (Lead II), 939 F.2d 975, 1009 (D.C. Cir. 1991) (per curiam) (rejecting OSHA’s conclusions when it criticized one industry study yet relied on another study with the same flaws). But OSHA explained its rationale. The Love study excluded retired workers and had little follow-up data on the workers it included. 81 Fed. Reg. at 16,378. These two data pieces are “extremely important” to fully quantify risks of silicosis morbidity because “silicosis typically develops slowly and becomes detectable [decades after] a worker’s first exposure.” Id. at 16,377-78. Without the two data pieces, the Love study did not meet OSHA’s “rigorous standards used in the studies on which OSHA’s [silicosis morbidity] risk assessment relies” and therefore eould not be included. Id. at 16,377. But the lack of the two data points did not render the Love study meaningless—if anything, OSHA reasoned, the failure to study workers at- later stages of their career, when the latent effects of silica exposure are more likely to manifest, meant the Love study “underestimated” the risk of silicosis to brick industry workers. Id. at 16,378. Moreover, the Love study was the only study specific to the brick industry that used exposure-response information, making it the “highest-quality” 1 study for ascertaining risks. Id. As one commenter testified, the Love study was the “only sensible study to be used for setting an exposure limit ... in brick manufacturing.” Id. OSHA, then, explained its reasoning and . supported it with substantial evidence. Second, Industry argues that even if the Love study is a credible source, OSHA’s risk estimates in other industries and for other disease endpoints (between 2 and 17 percent) at the PEL are greater than the risk estimates for the brick industry (1,4 percent), and therefore OSHA should have let the brick industry’s risks remain unaltered. But Industry misunderstands the legal standard. The mere fact that the brick industry faces a lower risk than other industries does not mean the brick industry’s risks are not significant. And the 1.4 percent risk quantified by -the Love study surpasses the Supreme Court’s 0.1 percent benchmark. Benzene, 448 U.S. at 665-56,100 S.Ct. 2844. Industry finally argues that OSHA’s different treatment of the brick industry and the sorptive minerals industry is arbitrary and capricious. According to Industry, the substances in both industries have chemical properties that reduce the toxicity of silica (which would reduce the health risks of exposure to silica) yet the Rule includes the■ brick'industry• but not-the sorptive-minerals industry. OSHA explained its decision in the preamble to the rule. The evidence for the sorptive minerals industry was unclear and thus insufficient to conclude a significant risk exists. 81 Fed. Reg. at 16,379-80. In contrast, the evidence in the brick industry—the Love study, ■ primarily—showed there is a significant risk; Id. at 16,377-78. Even if brick clay and sorptive minerals have similar chemical properties that reduce the toxicity of silica within those compounds, OSHA found the evidence as it existed in the record was not similar enough to treat them similarly. Id. OSHA’s position is supported by substantial evidence and a reasonable explanation, and therefore we, uphold the inclusion of the brick industry in the Süica Rule. , B. Technological Feasibility This court has interpreted the OSH Act’s requirement that OSHA health standards protect workers “to the extent feasible,” 29 U.S.C. § 655(b)(5), to include “two types of feasibility,” namely, “technological. and economic.” Lead I, 647 F.2d at 1264. Our .standard of review narrowly cabins our consideration of OSHA’s finding of technological feasibility. Specifically, we must ensure, only that OSHA found its standard feasible and supported that finding with substantial evidence. “To establish technological feasibility, OSHA, after consulting the ‘best available ‘evidence,’ 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 [standard] in most of its operations.’ ” Lead II, 939 F.2d at 980 (quoting Lead I, 647 F.2d at 1272). OSHA need not show with certainty that all firms will be able to meet the new standard in all operations. If “‘only the most technologically advanced plants in an industry have been able to achieve [the standard]—even if only in some of their operations some of the time,’ then the standard is considered feasible for the entire industry.” Id. (alteration in original) (quoting Lead I, 647 F.2d at 1264). As with its finding of significant risk, OSHA must support its finding of technological feasibility with substantial evidence. Substantial evidence does not require absolute “certainty.” Id. Where OSHA regulates on the frontiers of scientific knowledge, it is bound to confront inconsistency and uncertainty. But the mere “possibility of drawing two inconsistent conclusions from the evidence does not prevent [the] agency’s finding from being supported by substantial evidence.” American Textile Manufacturers Institute, Inc. v. Donovan (Cotton Dust), 452 U.S. 490, 523, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981) (quoting Consolo v. FMC, 383 U.S. 607, 620, 86 S.Ct. 1018, 16 L.Ed.2d 131 (1966)). So long as “OSHA makes reasonable predictions based on ‘credible sources.of information’, (e.g., data from existing plants and expert testimony), then the court should defer to OSHA’s feasibility determinations.” Lead II, 939 F.2d at 980. Where OSHA has demonstrated technological feasibility for the typical firm in most operations and has supported that finding with substantial evidence, it has satisfied its burden and we must defer to its conclusions. To mount a successful attack on OSHA’s feasibility finding, then, challengers must do more than suggest that compliance will be infeasible for some firms or in “a few isolated operations.” Id. In the robust process leading up to the promulgation of the silica rule, OSHA found that the rule would be technologically feasible based on a thorough consideration of available sources of information. For general industry and construction, OSHA identified job categories that involve silica exposure and developed profiles showing current exposure- levels. OSHA then identified the individual jobs for which additional controls are required to comply with the new PEL, and identified available Controls that would reduce exposure below- the new PEL. 81 Fed. Reg. at 16,433-34. OSHA concluded that achieving the new PEL is technologically feasible for 87 out of 90 job categories considered in general industry—including 36 categories in the foundry industry, all of which were deemed feasible—and 19 of 23 tasks con-' sidered in construction. Id. at 16,454-55, 16,459. On this basis, OSHA found that there was a reasonable possibility that the new standard could-be achieved by the typical employer in.most operations and was thus technologically feasible. In performing its analysis, OSHA relied on data from a variety of sources, including reports from OSHA inspections, National Institute for Occupational Safety and Health (NIOSH) reports, site visits conducted by a contractor, data from external stakeholders, and a variety of studies looking at the effectiveness of various controls. OSHA also considered and responded to testimony and comments submitted to the rulemaking record. With our highly deferential standard of review and OSHA’s process in mind, we now turn to Industry’s objections. Industry challenges OSHA’s feasibility findings in only three industries: foundries, hydraulic fracturing, and construction. While Industry identifies sundry examples of infeasibility for certain firms or in certain operations, their objections do not collectively undermine OSHA’s overall finding of feasibility for the typical firm in most operations nor do they meaningfully call into question the evidence on which OSHA relied. 1. Foundries Industry disputes OSHA’s finding of technological feasibility on two grounds: that variability in exposure levels makes compliance infeasible; and that OSHA did not rely on the best available evidence. On the issue of exposure variability, Industry contends that because of the dynamic and unpredictable nature of silica exposure, firms must strive to attain an exposure level well below the new PEL to ensure compliance with certainty. This argument runs headlong into our standard of review: “Feasibility of compliance turns on whether exposure levels ... can be met in most operations most of the time; therefore, it is the routine exposure levels that determine feasibility, and atypical outliers cannot invalidate a feasibility finding.” Lead II, 939 F.2d at 990. Industry’s focus on whether all foundries can always meet the new standard with certainty is thus beside the point. The relevant question is whether OSHA has shown that the typical firm can meet the standard in most operations. OSHA has done just that. It pointed to data—including over 1,000 samples from nearly 100 foundries—supporting its feasibility finding. Indeed, a study by the American Foundry Society, which Industry itself relies on, shows that the new PEL is already being met in most foundry job categories. OSHA further recognized that variability can be smoothed through consistent use of engineering controls. And OSHA expressly contemplates flexible enforcement to accommodate unexpected swings in exposure levels, an approach this court has approved in prior feasibility determinations. 81 Fed. Reg. at 16,459; see Lead II, 939 F.2d at 991. Industry may well be right that exposure levels vary uncontrollably and unpredictably across the foundry industry and within individual firms. That, however, is exactly why our standard of review does not require compliance from all firms in all operations all of the time; it is designed to permit OSHA to regulate in the face of variability and uncertainty. And Industry has failed to show that variability in the foundry industry undermines OSHA’s finding of feasibility for the typical firm in most operations most of the time. Industry also challenges the foundry-industry evidence on which OSHA relied. The data OSHA considered came from a variety of sources including its own visits to worksites, enforcement data, and other inspection reports, as well as NIOSH reports, state program reports, industrial hygiene literature, and survey data from the American Foundry Society, all of which supported OSHA’s feasibility finding. Industry, insisting that “no two foundries are alike,” contends that OSHA ignored the best available evidence, namely, the experiences of foundries attempting and failing to comply with the prior standard. In particular, Industry singles out sand system operators and finishers as two job categories in which compliance is infeasible. Industry Br. at 63-65. But Industry’s evidence suggests, at most, that compliance will be infeasible for some foundries or in some operations. And OSHA identifies controls that might be able to achieve compliance in the specific foundries and operations that Industry identifies. Even assuming that Industry is correct that compliance is unachievable in the foundries and operations it identifies, such isolated examples of infeasibility are, under our standard of review, insufficient to defeat OSHA’s finding of feasibility for “the typical” foundry in “most ... operations.” Lead I, 647 F.2d at 1272. 2. Hydraulic Fracturing Because OSHA only recently recognized the risk of silica exposure in the hydraulic fracturing industry, available data is limited and what data is available shows, unsurprisingly, that the vast majority of firms are not yet in compliance with the new standard. According to Industry, this evidence shows that the new standard is unattainable as there is no evidence of any controls reducing exposure below the new PEL. But even if sufficient controls do not yet exist, Industry’s challenge to OSHA’s feasibility finding nonetheless fails. In considering which controls can feasibly be implemented, OSHA “is not bound to the technological status quo.” Lead I, 647 F.2d at 1264. “Because the OSH Act is a ‘technology-forcing’ statute, OSHA can also ‘force industry to develop and diffuse new technology’ ” to meet its standard. Lead II, 939 F.2d at 980 (quoting Lead I, 647 F.2d at 1264). So long as OSHA “gives industry a reasonable time to develop new technology” and “presents substantial evidence that companies acting vigorously and in good faith can develop the technology,” it can “require industry to meet PELs never attained anywhere.” Lead I, 647 F.2d at 1264-65. Given the nascent state of silica-control technology in the hydraulic fracturing industry, OSHA gave firms five years to comply with the new. standard. Acknowledging that controls have yet to be widely implemented in the industry, OSHA identified controls, some currently available and others under development, that promise to sufficiently reduce exposure, citing to comments from several vendors. 81 Fed. Reg. at 16,455. In support of the five-year grace period, OSHA relied on an industry expert who described significant progress made over the prior five years and an inventor of one silica-control technology who explained that the technology took only three years to develop. Id. at 16,457. Though Industry disagrees with OSHA’s forecast of future silica-control developments in hydraulic fracturing, the agency’s evidence is more than sufficient “to show that modern technology has at least conceived some industrial strategies or devices which are likely to be capable of meeting the PEL and which the industries are generally capable of adopting” in the extended time horizon OSHA provided. Lead II, 939 F.2d at 1006 (quoting Lead I, 647 F.2d at 1266). 3. Construction In assessing the technological feasibility of its rule in the construction indusr try, OSHA relied on the Table 1 safe harbor. Under the new rule, if a construction employer implements the controls listed on Table 1—applicable to nineteen of twenty-three construction tasks—it is freed from its obligation to achieve the new PEL. OSHA determined not only that most employers would follow Table 1 for most tasks, but also that it would be technologically feasible for them to do so given the ready availability of .Table 1 controls. OSHA also found the rule to be technologically feasible for tasks not appearing on Table 1.81 Fed. Reg. at 16,458. Industry’s primary challenge to OSHA’s feasibility finding is that the Table 1 controls cannot always be implemented and sometimes require respiratory protection. But even were we to accept Industry’s arguments, these isolated exceptions hardly undermine OSHA’s finding of feasibility for the typical firm in most operations. As to situations where Table 1 controls cannot be implemented, Industry focuses on six tasks for which wet methods are prescribed, arguing that it is sometimes infeasible to introduce water to the work environment, such as for some indoor work or in cold-weather environments. Industry Br. at 98-99. But OSHA adduced evidence showing that employers can overcome many of the barriers identified by Industry, for example; by using heated water in cold-weather environments. 81 Fed. Reg. at 16,460. Moreover, even where wet methods cannot be implemented, Table 1 functions as , just one of two paths to compliance: where an employer cannot or elects not to follow Table 1, it is free to take the traditional path to ■ compliance by implementing controls of its choice to reduce exposures below the new PEL. OSHA acknowledged in the rulemaking record that such situations may arise and contemplated alternative- controls that might be implemented. Id. at 16,460-61, Even accepting Industry’s arguments that compliance for some tasks is infeasible under certain work conditions does not overcome OSHA’s finding of feasibility for the typical employer in most operations. Because Industry argues neither that the typical employer cannot implement wet methods nor that such methods are required in most operations, it has failed to carry its burden of showing that the use of wet methods renders the rule infeasible. On the issue of respiratory protection, OSHA assesses technological feasibility based on whether firms can “develop and install engineering and work practice controls" to meet the standard “without relying on respirators.” Lead I, 647 F.2d at 1272. The fact that “respirators -will be necessary in a few ... operations, will not undermine th[e] general presumption in favor of feasibility/’ Id. Thus the question for our review remains whether the need for respirators is so widespread as to undermine OSHA’s finding of feasibility for the typical firm in most operations. OSHA, however, contemplates only limited respirator use. Industry argues that “one-third- of [Table 1 tasks] require some form of respiratory protection when the task is performed for just over four hours,” which is “significant and .completely undercuts OSHA’s claim of technological feasibility.” Industry Br. at 95-96. Table 1, however, includes nineteen construction tasks, thirteen of which require no respiratory protection at all. OSHA Br. at. 92 n.56 (explaining that though only eighteen tasks are listed, a nineteenth task, performed by ground crew assisting equipment operators, is covered by Table 1). Certain others, the one-third of tasks to which Industry refers, require respirators under only certain circumstances, such as when the task is performed indoors or for over four hours. See 29 C.F.R. § 1926.1153(c)(3). And OSHA credibly found that most tasks would be performed for four hours or less and/or outdoors. 81 Fed. Reg. at 16,724. Again, the fact that respiratory protection will be required in some operations some of the time fails to satisfy Industry’s burden to rebut OSHA’s feasibility finding for the typical firm in most operations. Industry points to OSHA’s finding that it expects 13% of all workers to need some amount of respiratory protection as an indication that the rule is infeasible. Specifically, Industry argues that in litigation related to OSHA’s Hexavalent Chromium rule, the agency rejected respirator use by 9.5% of employees as unacceptably high. Industry Reply Br. at 51-52. The very language Industry relies on, however, defeats its claim. In the Hexavalent Chromium litigation, OSHA stated: “While the agency estimated that a total of 9.5% of all employees in all application groups would need respirators ,.. that overall figure did not factor into OSHA’s technological feasibility findings .... ” Public Citizen Health Research Group v. OSHA, Nos. 06-1818 and 06-2604, Final Brief for Respondents, at 45 (3d Cir. Dec. 14, 2007). There, OSHA noted that “réspirator use was more than ‘isolated’ where almost one third or more of the exposed employees in the affected groups would have to use respirators.” Id. Here, OSHA’s conclusion that 13% of workers using respirators amounts to only “isolated” respirator use neither overwhelms its finding of technological-feasibility nor conflicts with its position in the Hexavalent Chromium litigation. As we have explained, OSHA must show only that compliance is feasible for the typical firm in most operations—that some respirator .use may sometimes be needed is not enough to defeat OSHA’s finding. Even combining the effects of these two issues—the sometimes need for respiratory protection and the occasional situations where wet methods are infeasible—Industry has failed to show that it is infeasible for the typical employer to meet the standard in most operations. Some employers may be unable to implement the Table 1 controls in all operations—though OSHA reasonably explains why there are fewer such situations than Industry suggests. And some may have to resort to respiratory protection for certain tasks, though, as OSHA points out, only for a minority of tasks and only under certain circumstances. But Industry’s identification of atypical circumstances in a minority of operations where compliance with Table 1 is infeasible falls far short of rebutting OSHA’s well-supported finding of feasibility for the typical firm in most operations. Industry mounts a handful of additional challenges. None has merit. First, Industry again raises the issue of exposure variability. But this argument fails in construction just as it failed for the foundry industry: OSHA provided eyi-dence suggesting that variability is controllable and, in any event, our standard of review is designed to accommodate just such variability. Moreover, exposure variability—to the extent it presents a problem—is further mitigated in construction, where Table 1 provides a path to compliance without any need for exposure testing. Next, Industry criticizes the evidence upon which OSHA relied in determining how the PEL could feasibly be met. Specifically, Industry takes issue with OSHA’s reliance on short-duration exposure samples and its calculation of an eight-hour average, assuming no additional exposure during the un-sampled portion of the eight-hour period. Industry Br. at 90-94. Although few of OSHA’s exposure samples were eight hours long, the vast majority (70%) were four hours or longer and nearly half. (43%) were more than six hours long, 81 Fed. Reg. at 16,435. And OSHA considered Industry’s objection and adequately justified the no-further-exposure assumption by adducing evidence of the intermittent and short-duration.nature of silica exposure in construction tasks. Id. Moreover, OSHA’s assumption aligns with its enforcement practice. When OSHA compliance officers collect partial-shift samples during an inspection, they calculate eight-hour time-weighted average exposures using the same assumption of no further exposure during the un-sampled period. The alignment between OSHA’s evidence and its enforcement practice confirms that any harm to Industry from this assumption is more semantic than substantive. Finally, Industry disputes OSHA’s finding of feasibility for four particular tasks: hole drillers using handheld or stand-mounted drills, jackhammering and using other powered handheld- chipping tools, masonry cutters using stationary saws, and mobile crushing machine operators and tenders. Industry Br. at 99-105., We have no need to address Industry’s arguments as to these tasks, for even were we to accept them, Industry would still have failed to rebut OSHA’s finding of feasibility in “most operations.” Lead II, 939 F.2d at 990. In any event, OSHA cited evidence that employers could reduce exposure levels for each task using available controls. In response, Industry recites a number of by-now familiar arguments: that OSHA’s data was inadequate, that the tasks are sometimes performed for longer than OSHA assumes, and that particular controls (again, wet methods) sometimes cannot be implemented. But OSHA considered and responded to each of these objections, making “reasonable predictions based on ‘credible sources of information.’” Lead II, 939 F.2d at 980 (quoting Lead I, 647 F.2d at 1266). Once again, Industry’s insistence that compliance is infeasible for some firms in some operations some of the time cannot upend our deference to OSHA’s well-supported finding that compliance is feasible for the typical firm in most operations. C. Economic Feasibility The OSH Act’s requirement that OSHA health standards protect workers “to the extent feasible,” 29 U.S.C. § 665(b)(6), also requires OSHA to show that its rule is economically feasible, Lead I, 647 F.2d at 1264. As with technological feasibility, the scope of our review of OSHA’s economic feasibility finding is narrowly circumscribed. A rule is economically feasible in a particular industry so long as it does not “threaten massive dislocation to, or imperil the existence of, the industry.” Id. at 1265. Thus, “[a] standard is not infeasible simply because it is financially burdensome or even because it threatens the survival of some companies within an industry.” Id. (citation omitted). “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 980-81. OSHA must also provide “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. “Courts, [moreover], ‘cannot expect hard and precise estimates of costs.’ ” Lead II, 939 F.2d at 1006 (quoting Lead I, 647 F.2d at 1266). As before, the mere “possibility of drawing two inconsistent conclusions from the evidence” or deriving two divergent cost models from the data “does not prevent [the] agency’s finding from being supported by substantial evidence.” Cotton Dust, 452 U.S. at 523, 101 S.Ct. 2478 (quoting Consolo, 383 U.S. at 620, 86 S.Ct. 1018). Industry does not challenge OSHA’s overall methodology for assessing economic feasibility. Instead, it questions the evidence on which OSHA relied in the foundry, hydraulic fracturing, and construction industries. Industry also gestures towards a challenge to OSHA’s findings on the brick industry, claiming only that OSHA “cannot adopt a standard that imposes very large costs on an industry without producing any quantifiable health benefit.” Industry Br. at 130. But because OSHA found significant risk in the brick industry, as we explained above, and Industry does not otherwise claim economic infeasibility, this argument is foreclosed. Industry’s economic feasibility arguments, like its technological feasibility arguments, raise a host of claims about OSHA’s sources that do not collectively undermine the evidence OSHA relied on and the conclusions it reached, especially in light of our standard of review and the narrow scope of Industry’s challenge. In its economic feasibility analysis, OSHA developed estimates of the annualized cost of compliance for each affected industry— and for small and very small employers within each industry—and compared those costs against industry revenues and profits. See 81 Fed. Reg. at 16,462-582 (describing OSHA’s economic feasibility methodology). OSHA explained that “while there is no hard and fast rule,” it “generally considers a standard to be economically feasible” for an industry where annualized costs of compliance are less than one percent of revenue or ten percent of profit. Id. at 16,533. OSHA considers this benchmark to be “fairly modest,” so costs exceeding the threshold do not imply per se infeasibility, but rather serve as a trigger for further analysis. Id. For each of the industries at issue here—foundries, hydraulic fracturing, and construction—OSHA determined that costs as a percentage of revenues' and profits were below the one percent and ten percent thresholds. Id. at 16,536, 16,538, 16,573. For foundries and construction, these costs were well below these benchmarks for all industry subgroups considered: even doubling OSHA’s cost estimates in foundries and tripling them in construction would only barely trigger the thresholds for further inquiry. Id. at 16,538 (showing, among subgroups within the foundry industry, costs as a percentage of profits of 5.62% at the greatest); id. at 16,573 (showing, among subgroups in construction, costs as a percentage of profits of 3.66% at the greatest). For hydraulic fracturing, compliance costs were somewhat nearer OSHA’s thresholds, though still below, with costs as a percentage of revenues of 0.56% and costs as a