Members of the smelting industry petition for review of regulations promulgated by the Occupational Safety and Health Administration (OSHA or Secretary). 1 The regulations establish maximum permissible exposure levels (PELs) of airborne arsenic to which workers may be exposed. Arsenic is a by-product of nonferrous metal (e.g., copper, lead, and zinc) smelting. Because the regulations are supported by substantial evidence, we deny the petition for review and uphold the challenged regulations. We also hold that the Secretary did not abuse his discretion in refusing to reopen the record to reconsider the feasibility of the standard.

FACTUAL AND PROCEDURAL BACKGROUND

For a detailed account of the procedural history, rulemaking, and underlying scientific data in this case see 48 Fed.Reg. 1869-1903 (Jan. 14, 1983), 43 Fed.Reg. 19584-631 (May 5, 1978). Briefly stated, arsenic 2 is a by-product of non-ferrous metal smelting processes. Although sometimes an unwanted by-product, arsenic is marketed for many uses, including use as an insecticide, pesticide, or clarifying agent in glass production. It may also be further processed into “pentavalent arsenic” which is used, for example, as a herbicide, desiccant (drying agent), or wood preservative. See 43 Fed.Reg. at 19584-85.

Following the receipt of studies indicating that arsenic was a human carcinogen, OSHA proposed a reduction of the existing standard for occupational exposure to arsenic from 500 ug/m3 to 4 ug/m3. 3 40 Fed.Reg. 3392 (Jan. 21, 1975). Thereafter, OSHA held public hearings and received and considered epidemiological studies of the relationship between arsenic exposure and cancer. OSHA also received and considered studies evaluating the technological and economic feasibility of possible measures to reduce occupational exposure to arsenic. On May 5, 1978, OSHA published a final standard limiting occupational exposure to arsenic to 10 ug/m3. OSHA also found that the 10 ug/m3 PEL was both technologically and economically feasible. 43 Fed.Reg. 19584-631 (May 5, 1978) (codified at 29 C.F.R. § 1910.1018 (1983)).

The final standard regulates employee exposure to arsenic and applies to most workplaces. Exemptions are granted for pesticide application, agriculture, and the treatment and use of arsenically preserved woods. 29 C.F.R. 1910.1018(a). The standard went into effect August 1, 1978, and lowered the permissible exposure level (PEL) from 500 to 10 ug/m3 averaged over an eight-hour period. 4 Id. at 1910 .1018(c). The standard includes provisions designed to assist in the reduction of employee exposure to arsenic and to aid the detection of arsenic-induced disease (for example, the regulations require employers to provide changing rooms, filtered air lunch rooms, medical examinations, and exposure moni *488 taring). 29 C.F.R. § 1910.1018 (m), (n), (e). Several affected companies challenged the standard and the cases were consolidated in this court. 5 Before we had ruled on those challenges, the Supreme Court decided Industrial Union Dept. v. American Petroleum Institute (IUD v. API), 448 U.S. 607 , 100 S.Ct. 2844 , 65 L.Ed.2d 1010 (1980) (reviewing OSHA’s proposed benzene standard). In IUD v. API, the Court held that the Secretary must make a finding that a toxic substance poses a “significant risk” to workers’ health before he may regulate that substance under § 6(b)(5) of the Occupational Safety and Health Act (Act), 29 U.S.C. §§ 651-678 , § 655(b)(5) (1982). 448 U.S. at 642 , 100 S.Ct. at 2864 . Because the Secretary had not specifically made such findings with respect to the proposed arsenic standard, on April 7,1981, we remanded the record to the Secretary to make the required findings. We did, however, keep the 10 ug/m3 PEL in effect during the remand because OSHA had evidence demonstrating that arsenic did indeed pose a significant risk at the then-existing 500 ug/m3 level. ASARCO conceded that such a risk existed. ASA RCO Inc. v. OSHA, 647 F.2d 1 , 2 (9th Cir.1981). Our remand was limited to the “significant risk” issue although we did leave open the possibility that the remand could be modified upon petition by the parties in light of an expected opinion by the Supreme Court that would focus on feasibility issues. 6 ASARCO, Inc., et. al v. OSHA, Nos. 78-1959, 78-2764, 78-3039, 78-2477, 78-2478 (Order of April 7, 1981).

On remand, the Secretary reopened the record for the limited purpose of receiving evidence and making the required findings on the degree and significance of risk posed by occupational exposure to arsenic and to determine whether further adjustments in the final standard were warranted. 47 Fed.Reg. 15358 (April 9, 1982). On January 14, 1983, after receiving and considering additional evidence and public comments on the significant risk issue, the Secretary published his final statement justifying the 10 ug/m3 arsenic standard. 48 Fed.Reg. 1864-1903 (January 14, 1983), as corrected by 48 Fed.Reg. 24068-73 (May 31, 1983). Petitioners requested the Secretary to reopen the record also on the question whether the 10 ug/m3 PEL was tech-' nologically and economically feasible. The Secretary did not do so, treating petitioners’ request to reopen the record as a proposal to amend the standard. The Secretary (through OSHA’s Director of Health Standards) rejected the proposed amendment. Letter of June 7, 1983. See also 48 Fed.Reg. at 1869.

After considering epidemiological studies on the relationship of arsenic exposure to lung and other cancers, the Secretary found that arsenic was a human carcinogen that posed a significant risk to workers at the pre-existing 500 ug/m3 PEL. 7 The Secretary also found that a 10 ug/m3 PEL would significantly reduce that risk. The Secretary estimated that the “excess risk” 8 to workers at the 500 ug/m3 PEL was 400 excess deaths per 1000 employees, while the excess risk to workers at the 10 ug/m3 PEL would be 8 excess deaths per 1000 employees. The Secretary therefore concluded that reducing the PEL from 500 ug/m3 to 10 ug/m3 was likely to reduce significantly the cancer risk.

Prior to our remand, the Secretary had found the 10 ug/m3 PEL standard to be *489 feasible — both technologically and economically — in all affected industries. 43 Fed. Reg. at 19601-12. The Secretary based his overall feasibility findings on OSHA and industry studies which agreed that the 10 ug/m3 PEL could be obtained through the use of existing engineering controls in combination with sound work practices and supplementary respirator use. The Secretary determined that engineering and work practice controls would bring virtually all areas of the majority of copper smelters and most areas of even those smelters refining high-arsenic-content ore into compliance with the 10 ug/m3 PEL. 9 43 Fed.Reg. at 19601-03, 19605. The Secretary also found that the 10 ug/m3 PEL could be met in high-emission areas through supplemental use of respirators. 43 Fed.Reg. at 19616-17.

The Secretary found that the 10 ug/m3 standard was also economically feasible for all affected industries. In reaching this conclusion, he paid particularly close attention to the copper smelting industry, the industry most affected by the standard. 43 Fed.Reg. at 19604-07.

Once the Secretary made his significant risk findings on remand, the matter was resubmitted to this court and rebriefed. Industry petitioners ASARCO, Inc., Kennecott (formerly Kennecott Copper Corporation), and Anaconda Minerals Company 10 again challenge the standard. They contend that the Secretary’s risk and feasibility findings are not supported by substantial evidence as required by the Occupational Safety and Health Act, 29 U.S.C. § 655 (f) (1982). Petitioners also contend that the Secretary abused his discretion by not reopening the record to consider additional evidence on the feasibility issues. The Chemical Manufacturer’s Association (CMA) intervened in support of petitioners; the United Steelworkers of America, AFL-CIO (Steelworkers) intervened on behalf of OSHA. Representatives of the arsenical chemicals industry (Koppers, et al.) filed an amicus brief challenging the Secretary’s feasibility findings as applied to the arsenical chemicals industry.

ISSUES

1. Does substantial evidence support the Secretary’s finding that a 500 ug/m3 PEL poses a “significant risk” to workers’ health and that a 10 ug/m3 PEL would “significantly reduce” that risk?

2. Does substantial evidence support the Secretary’s finding that the 10 ug/m3 PEL is technologically feasible?

3. Does substantial evidence support the Secretary’s finding that the 10 ug/m3 PEL is economically feasible?

4. Did the Secretary abuse his discretion by not reopening the record on feasibility?

DISCUSSION

The challenged standard appears in part at 29 C.F.R. § 1910.1018 (g)(1) (1983):

*490 The permissible exposure limit (PEL) is defined in § 1910.1018(c): “The employer shall assure that no employee is exposed to inorganic arsenic at concentrations greater than 10 micrograms per cubic meter of air (10 ug/m3), averaged over any 8-hour period.”

I. Significant Health Risk

Before the Secretary promulgates a new standard, he is required by § 3(8) of the Occupational Safety and Health Act (Act), 29 U.S.C. § 652 (8) (1982), to make a threshold determination that (1) the current standard governing the toxin poses a significant health risk, and (2) the risk can be diminished by lowering the standard. Industrial Union Dep’t v. American Petroleum Inst. (IUD v. API), 448 U.S. 607, 642 , 100 S.Ct. 2844, 2864 , 65 L.Ed.2d 1010 (1980) (plurality); United Steelworkers of America v. Marshall, 647 F.2d 1189, 1246 (D.C. Cir.1980), cert. denied, 453 U.S. 913 , 101 S.Ct. 3148 , 69 L.Ed.2d 997 (1981). A majority of the Court has endorsed this requirement. See American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490 , 505-06 n. 25, 101 S.Ct. 2478 , 2488-89 n. 25, 69 L.Ed.2d 185 (1981).

Petitioners and intervenor CMA divided the arguments. CMA argues petitioners’ challenges to the “significant risk” determinations required under IUD v. API and made by OSHA on remand. CMA also challenges OSHA’s decision to include pentavalent arsenic within the arsenic standard. Petitioners ASARCO, Inc. and Kennecott argue petitioners’ challenges to the Secretary’s feasibility findings.

(a) Standard of Review

Section 6(f) of the Act, 29 U.S.C. § 655 (f), provides that “[t]he determination of the Secretary shall be conclusive if supported by substantial evidence in the record as a whole.” Substantial evidence is “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” American Textile, 452 U.S. at 522 , 101 S.Ct. at 2497 (quoting Universal Camera Corp. v. NLRB, 340 U.S. 474, 477 , 71 S.Ct. 456, 459 , 95 L.Ed. 456 (1951)). Because study of OSHA’s attempts to safeguard worker health demands that we consider complex and often conflicting scientific evidence, we elaborate below some of the considerations that must guide our review.

Although we must review contradictory evidence in the record, “the possibility of drawing two inconsistent conclusions from the evidence does not prevent an administrative agency’s finding from being supported by substantial evidence.” American Textile, 452 U.S. at 523 , 101 S.Ct. at 2497 (quoting Consolo v. FMC, 383 U.S. 607, 620 , 86 S.Ct. 1018, 1026 , 16 L.Ed.2d 131 (1966)). Similarly, in United Steelworkers, the D.C. Circuit held,

647 F.2d at 1263 .

In the context of significant risk determinations, “OSHA is not required to support its finding ... with anything approaching scientific certainty.” IUD v. API, 448 U.S. at 656 , 100 S.Ct. at 2871 (plurality). Moreover, we “give OSHA some leeway where its findings must be made on the frontiers of scientific knowledge.” Id. (citations omitted). And,

IUD v. API, 448 U.S. at 656 , 100 S.Ct. at 2871 (emphasis added) (footnote omitted). See also Ethyl Corp. v. EPA, 541 F.2d 1 , 22-33 (D.C. Cir.) (en banc), cert. denied, 426 U.S. 941 , 96 S.Ct. 2662 , 49 L.Ed.2d 394 (1976) (for eloquent discussion of problems associated with scientific certainty in environmental decisionmaking).

It is the Secretary’s responsibility to determine what risk he considers to be *491 “significant,” but in doing so he need not be forced into a “mathematical straightjacket.” IUD v. API, 448 U.S. at 655 , 100 S.Ct. at 2870 . He can be reasonable in determining that a one in a thousand risk is significant, id., and be justified in regulating such a risk. We turn now to the facts of this ease.

(b) The Arsenic Standard and Petitioners’ Challenges

The epidemiological studies and methodologies used by the Secretary to determine that a significant health risk exists at and below the pre-existing 500 ug/m3 PEL and that the new 10 ug/m3 PEL will reduce that risk are discussed in the summary of the regulations. See 48 Fed.Reg. 1895-1903. 11 Although a familiarity with these studies is necessary to resolve whether the standard is supported by substantial evidence, we shall not resummarize them-here. Instead, for brevity’s sake, we respond primarily to CMA’s contentions after we summarize OSHA’s findings.

On the basis of the studies, OSHA determined that a range of 148-767 excess deaths per 1000 employees was expected at 500 ug/m3 PEL, and a range of 2.2-29 excess deaths per 1000 employees was expected at 10 ug/m3 PEL. 48 Fed.Reg. at 1866. OSHA’s best estimate of risks in those ranges was 400 excess deaths per 1000 employees at 500 ug/m3 and 8 excess deaths per 1000 employees at 10 ug/m3. 12 Id. Thus, OSHA concluded that the 500 ug/m3 PEL posed a significant risk, and that the 10 ug/m3 PEL would substantially reduce that risk. Id. 13

CMA’s primary argument is that the epidemiological studies of trivalent arsenic, including the Lee-Feldstein, Lee & Fraumeni, Pinto, and Marsh & Enterline studies, relied upon by OSHA to make its risk assessments are flawed and hence cannot constitute “substantial evidence.” CMA also contends that another study, by Higgins, which found no excess risk of death below the pre-existing standard, is more reliable and relevant than the studies relied on by OSHA. We find these contentions unpersuasive. In fact, we find just the opposite to be true. The studies relied on by OSHA constitute' substantial evidence and the Higgins study is critically flawed. Moreover, CMA asks us to hold OSHA to a degree of scientific certainty far more exacting than the Act requires. 14

First, as we have already held in maintaining the 10 ug/m3 PEL during the remand, “it is undisputed that exposure to inorganic arsenic at the level of 500 ug/m3 — which would be permitted were we to vacate the current standard — poses a ‘significant’ health risk, and ASARCO concedes that exposure at a level above 50 ug/m3 may present some danger.” ASARCO, 647 F.2d at 2 . Second, even assuming that the Higgins study is not defective, see infra slip op. at pp. 4022-4024, 746 F.2d at pp. 493-494, most of CMA’s methodology challenges would require us to evaluate conflicting inferences drawn from comparable or identical evidence. In such a case, we must defer to OSHA. American Textile, 452 U.S. at 523 , 101 S.Ct. at 2497 ; see supra slip op. at p. 4019, 746 F.2d at p. 490. Third, even if we were to evaluate comparable evidence and second-guess OSHA, we would reject CMA’s challenges. We now turn to CMA’s specific contentions.

*492 Exposure Levels. We are not persuaded by CMA’s attack on the quantification of arsenic exposure levels in the Anaconda smelter studies of Lee & Fraumeni and Lee-Feldstein. 15 Morris subsequently supplied reasonable quantifications in support of those studies. 48 Fed.Reg. at 1870. After reviewing the record, none of CMA’s other challenges to the studies that OSHA relies on persuade us that those studies have defects that should cause us to question their validity. They thus remain as valid evidence supporting OSHA’s 10 ug/m3 standard.

Urinary Measurements. CMA also attacks OSHA’s reliance on the Enterline & Marsh studies. Enterline and Marsh compared worker urinary arsenic levels at ASARCO’s Tacoma smelter with worker incidence of lung cancer. Those studies found an excess risk of cancer at low arsenic exposure levels. OSHA supports its reliance on Enterline & Marsh with Pinto’s correlation between airborne arsenic and urinary arsenic levels. 48 Fed.Reg. at 1878-79. CMA contends that such correlations are invalid. Pinto found a correlation, however, and ASARCO itself presented his data as the “best available data” on the subject at the earlier proceedings. Id. Similar correlations have been accepted by reviewing courts. See, e.g., United Steelworkers, 647 F.2d at 1259-63 (air/lead; blood/lead); cf. Lead Industries Ass’n. v. EPA, 647 F.2d 1130,1166-67 (D.C.Cir.1980) (discussing airborne lead’s eventual arrival in bloodstream); Ethyl Corp. v. EPA, 541 F.2d 1 at 7-9, 30-31 (discussing blood lead levels). After reviewing the record, we also accept the correlations. The Pinto and Enterline & Marsh studies of the Tacoma smelter therefore constitute evidence in support of OSHA’s standard.

Risk Assessments. CMA also contends that OSHA impermissibly relied on a linear, no-threshold, cumulative dosage model that CMA argues incorrectly determines that there is no safe level of exposure to arsenic. 16 First, CMA contends that OSHA’s use of the “no safe level” (i.e., linear) model is impermissible because the model was not derived through scientific reasoning — but was used as a matter of administrative fiat. CMA’s contention is *493 not supported by the record. OSHA’s experts in fact adopted the linear assumption of “zero risk at zero exposure” on the basis of “best available evidence,” not by administrative fiat as CMA contends. 48 Fed. Reg. at 1886-90. Moreover, the linear model “fits” the epidemiological data obtained by OSHA’s experts far better than the threshold model 17 advocated by CMA does. 48 Fed.Reg. at 1886; see also 48 Fed.Reg. at 1878 (expert opinion of Dr. C. H. Crump).

Second, with respect to the cumulative dosage model employed by OSHA, CMA argues that exposure to high levels of arsenic (exposure intensity), rather than long-term exposure at low levels (cumulative exposure), causes cancer. The record, however, supports OSHA’s risk estimation method. Contrary to CMA’s allegations, the cumulative exposure model is not an unsubstantiated assumption. Instead, the model is supported by direct evidence and expert opinion. After examining many factors including duration of exposure, age, and exposure intensity, OSHA’s experts (Brown and Chu), found that duration was “the most important single factor in the excess lung cancer risk.” 48 Fed.Reg. at 1889. They also acknowledged that exposure intensity was important. Id. CMA’s argument, however, that exposure intensity, rather than duration (cumulative exposure at lower exposure levels), is the only critical variable is unpersuasive. OSHA considered the evidence supporting exposure intensity and adequately stated its reasons for preferring duration. 48 Fed. Reg. at 1890. Finally, the “no safe level” model easily falls within the class of “conservative assumptions” approved by the IUD v. API plurality. 448 U.S. at 656 , 100 S.Ct. at 2871 . See supra 746 F.2d at p. 490. We therefore defer to the agency’s expertise on this point. See American Textile, 452 U.S. at 523 ; 101 S.Ct. at 2497 ; United Steelworkers, 647 F.2d at 1248 .

Higgins Study. CMA seeks to supplant OSHA’s studies and conclusions with those of Higgins. CMA argues that OSHA’s rejection of Higgins was unreasonable and an abuse of discretion. We disagree.

CMA alleges that the Higgins study reflects most accurately the risks associated with arsenic exposures below 500 ug/m3. Higgins hypothesized that no excess risk would occur to workers whose ceiling exposure level was 500 ug/m3 or below. The record, however, shows that the Higgins study is flawed. 18 Foremost of those flaws is the study’s low statistical power (ability to predict). 19 Higgins examined only a limited number of individuals (only 22% of the workers studied by Lee & Fraumeni and Lee-Feldstein at the same plant). Despite CMA’s attempts to introduce extra-record evidence now to rebut OSHA’s conclusion that the Higgins study was unreliable, we are firmly convinced that the Higgins study does not warrant re-evaluation of the standard. 20

*494 In sum, CMA’s attempt to discredit and/or supplant OSHA’s studies with the Higgins study does not fail just because CMA asks us to choose one comparable study over another (in effect, to “second-guess” OSHA). See United Steelworkers, 647 F.2d at 1263 (when faced with comparable evidence on both sides, court should defer to OSHA). This case is far easier. CMA’s challenge fails because OSHA supports its rule with several reliable studies and CMA asks us to reject those studies in favor of one that is speculative and of questionable validity. 21

Here, OSHA considered the Higgins study and rejected it on reasonable grounds, and OSHA’s reliance on “a reputable body of scientific thought” — e.g., the Lee & Fraumeni, Lee-Feldstein, Pinto, and Enterline & Marsh studies — was reasonable. 22 Given the generous standard of review with which we review OSHA regulations — “risking error on the side of over-protection rather than underproteetion” 23 — we hold that OSHA’s finding that arsenic poses a significant risk to worker health at the pre-existing 500 ug/m3 PEL and that the new standard of 10 ug/m3 will signifieantly reduce that risk is amply supported by substantial evidence. 24

CMA also challenges OSHA’s inclusion of pentavalent arsenic in the arsenic standard. The Secretary concluded that pentavalent arsenic was a carcinogen and hence poses a “significant risk” to worker health. 48 Fed.Reg. at 1893-1894. CMA contends that OSHA did not rely on substantial evidence to show that pentavalent arsenic is carcinogenic. Although the question whether substantial evidence supports the Secretary’s finding of significant risk is closer here than with trivalent arsenic, we find that substantial evidence supports the Secretary’s finding.

OSHA initially concluded that pentavalent arsenic was carcinogenic on the basis of a study by Ott. 48 Fed.Reg. at 1893. CMA points out, however, that OSHA itself at one time concluded that the Ott study’s “methodological limitations” prevented it from being used to show adequate dose-response levels. See 43 Fed.Reg. at 19596. Nonetheless, OSHA now reaffirms its inclusion of pentavalent arsenic on the basis *495 of the Ott study and on the subsequent opinions of several additional experts indicating that Ott’s study can indeed be used for risk assessment purposes. 48 Fed.Reg. at 1890, 1897-98. OSHA also explains that the Supreme Court’s intervening decision in IUD v. API clarified that OSHA was free to use “conservative assumptions” and err on the side of overprotection. 448 U.S. at 655 , 100 S.Ct. at 2870 ; see supra 746 F.2d at p. 490. Given this clarification, the Secretary decided that the Ott study was sufficiently reliable to support OSHA’s regulation of pentavalent arsenic. 48 Fed. Reg. at 1894.

We defer to the Secretary on his decision to include pentavalent arsenic in the standard. CMA’s contentions on this point do not compel a different result. OSHA states adequate reasons for rejecting the rival Nelson study proffered by CMA. 48 Fed.Reg. at 1893. 25 OSHA also found that new evidence indicates that the Ott study is more reliable than originally thought. 48 Fed.Reg. at 1893-94.

As with our consideration of the arsenic standard in general, the deferential standard of review enunciated by the court in IUD v. API, 448 U.S. at 656 , 101 S.Ct. at 2871 (scientific certainty not required), also leads us to the conclusion that the Ott study, along with the other evidence, constitutes sufficiently substantial evidence to support OSHA’s inclusion of pentavalent arsenic in its regulations. Similarly, the Court’s directive in IUD v. API that the Secretary may rely on “conservative assumptions” to safeguard worker health, 448 U.S. at 656 , 101 S.Ct. at 2871 , supports the conclusion that, along with Ott, OSHA may use trivalent arsenic studies and conclusions to support inclusion of pentavalent arsenic in the standard. See, e.g., Environmental Defense Fund v. EPA, 598 F.2d 62 , 83-85 (D.C.Cir.1978) (studies of highly chlorinated PCBs could be used to extrapolate knowledge about less chlorinated PCBs).

II. Feasibility Issues

Section 6(b)(5) of the Act, 29 U.S.C. § 655 (b)(5) (1982), states in pertinent part:

(emphasis added). The section requires that an OSHA standard be both technologically and economically feasible. United Steelworkers, 647 F.2d at 1264 . See also Donovan v. Castle & Cooke Foods, 692 F.2d 641, 647 (9th Cir.1982) (reviewing noise standard promulgated under § 6(a) of the Act). “Feasible” is defined as “capable of being done,” American Textile, 452 U.S. at 508-09 , 101 S.Ct. at 2490 ; Castle & Cooke, 692 F.2d at 648 . We review the Secretary’s feasibility determinations for substantial evidence. Boise Cascade Corp. v. Secretary of Labor, 694 F.2d 584, 589 (9th Cir.1982); Castle & Cooke, 692 F.2d at 647 n. 8.

A. Technological Feasibility

In finding that a standard is “technologically feasible,” the Secretary is not restricted to the state of the art in the regulated industry. Boise Cascade, 694 F.2d at 589 . The Act is designed to be “technology-forcing.” United Steelworkers, 647 F.2d at 1264 . Hence, “[s]o long as [the Secretary] presents substantial evidence that companies acting vigorously and in good faith can develop the technology, OSHA can require industry to meet [standards] never attained anywhere.” Boise Cascade, 694 F.2d at 589 -90 (quoting United Steelworkers, 647 F.2d at 1264 ) (emphasis added). In fact, for our purposes *496 here — reviewing OSHA’s general rulemaking for the purpose of protecting worker health — all OSHA need demonstrate is 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.” United Steelworkers, 647 F.2d at 1266 . Thus all OSHA need demonstrate is “a general presumption of feasibility for an industry.” United Steelworkers, 647 F.2d at 1269 (original emphasis). 26

We find that substantial evidence supports the Secretary’s finding that the 10 ug/m3 PEL is technologically feasible.

(a) OSHA’s findings.

The Secretary found that the standard is technologically feasible through engineering and work practice controls for most of the affected industry. 43 Fed.Reg. 19601. For the rest of the industry,

43 Fed.Reg. at 19616. See 29 C.F.R. § 1910.1018 (g). In high-exposure areas of some plants, the standard would have to be met through the supplemental use of respirators. 43 Fed.Reg. at 19601, 19617. The Secretary also specifically found that respirators alone were not an adequate substitute for engineering and work practice controls. 43 Fed.Reg. at 19616-19619. 27 Because conditions at each plant may vary significantly, see 43 Fed.Reg. at 19601-02; see also supra note 9, the Secretary provided that the standard would be met through the development of individual compliance plans. 43 Fed.Reg. at 19601. The Secretary found that this approach would provide “maximum protection without heavy respirator use.” Id.

(b) Petitioners’ challenges.

ASARCO and Kennecott argue petitioners’ challenges to the regulations on both technological and economic feasibility grounds. Their challenges on technological grounds fall into three categories — (1) the use of respirators, (2) allegations that OSHA has impermissibly deferred its feasibility findings to the compliance stage, and (3) the alleged vagueness of the standard. Koppers, et al. (amicus curiae) challenge the feasibility of the standard as applied to the arsenical chemicals industry. We briefly address these challenges below.

(1) Use of respirators. Petitioners claim that the standard is “admittedly infeasible” because OSHA found that the five domestic smelters with the highest arsenic intake could not meet the standard through engineering and work practice controls alone and that supplemental respirator use would be required in high exposure areas in those plants. 43 Fed.Reg. at 19600. These smelters comprise 39% of the nation’s smelting capacity. 43 Fed.Reg. at 19602 (Table I). Kennecott contends that the two other high-intake smelters should be added to this list, bringing the total number of smelters in which respirators would be used up to half of the country’s smelting capacity.

OSHA demonstrates, however, that engineering and work practice controls, within *497 the industry’s present grasp, actually can achieve compliance for 11 of the 16 smelters “with engineering and work practices and only very limited use of respirators.” 43 Fed.Reg. at 19603. At the other smelters, the evidence shows that compliance is possible with only “limited” to “moderate” respirator use. 43 Fed.Reg. at 19603. 28

Petitioners argue that what OSHA terms “limited-to-moderate” use really means that respirators will be used extensively in the production areas, the heart of the smelting operation, and not used outside the production areas. Thus, they argue that the standard actually requires “extensive” respirator use in all areas central to the smelting process. Because the standard requires such use of respirators, petitioners argue that the engineering and work practice controls must therefore be “infeasible.” Petitioners greatly overstate the use of respirators. See 43 Fed.Reg. at 19600-04. We find the respirator use envisioned in the arsenic standard to be merely supplemental to engineering and work practice controls. The limited respirator use that the standard requires does not in any way render the standard infeasible, let alone “admittedly infeasible.” Cf. American Textile, 452 U.S. 490 , 101 S.Ct. 2478 , 69 L.Ed.2d 185 (where Court approved OSHA standard that relied on heavy respirator use until engineering and work practice controls would be technologically feasible); 29 C.F.R. § 1910.1043 (e)(1) (cotton dust standard).

Petitioners also challenge the standard’s inclusion of respirators on statutory grounds. Petitioners offer some unpersuasive contentions concerning the interaction of § 3(8) and § 6(b)(5) of the Act. Section 3(8) of the Act defines an OSHA standard. An OSHA standard is one which requires “adoption or use of one or more practices ... reasonably necessary or appropriate to provide safe or healthful employment____” 29 U.S.C. § 652 (8). Petitioners’ argument here is that although “feasibility” for § 6(b)(5) purposes may mean “capable of being done,” § 3(8) limits the means of compliance to those that are “reasonably necessary or appropriate.” ASARCO claims that the standard, as applied to the five high-intake smelters is “redundant,” and hence not “reasonably necessary or appropriate” because it requires engineering and work practice controls even though only respirators will be able to reduce worker exposure to the required 10 ug/m.3 In support of this argument, ASARCO cites American Textile, 452 U.S. at 513 -14 n. 32, 101 S.Ct. at 2492 -93 n. 32, where the Supreme Court generally reserved the question of the applicability of § 3(8) in § 6(b)(5) feasibility determinations.

We reject petitioners’ contention that the limited supplemental use of respirators renders the standard “unreasonable” or “inappropriate.” Their argument rests on the notion that respirator use is “redundant” and somehow unreasonably superfluous. As we have already explained, see supra note 27, the many problems associated with exclusive respirator use render respirators woefully inadequate, standing alone, to protect worker health. Hence it is incorrect to characterize them as redundant rather than as supplemental. Cf. United Steelworkers, 647 F.2d 1189, 1269 (where supplemental respirator use did not render standard unreasonable).

Moreover, petitioners misconstrue the extent of § 3(8)’s modification of § 6(b)(5) as set forth in American Textile . There, the Court noted that in IUD v. API it had used § 3(8)’s “reasonably necessary” language to support the requirement that the Secre *498 tary find a significant risk before he could regulate a toxin. The Court also noted that a standard that required the use of five respirators where one would suffice could conceivably be “redundant” and hence not “reasonably necessary or appropriate.” The situation here is completely different from the situation noted by the Court — respirator use under the arsenic standard is merely supplemental, and not redundant because respirators alone are inadequate. See supra note 27. Furthermore, the Court in American Textile rather clearly rejected an argument that § 3(8) had the kind of impact that petitioners suggest. In rejecting an argument that § 3(8)’s “reasonably necessary” language required the Secretary to undertake a cost-benefit analysis of any proposed standard, the Court indicated that although § 3(8) might modify § 6(b)(5), the former section was not intended to supplant the latter. 452 U.S. at 513 , 101 S.Ct. at 2492 ; cf. Castle & Cooke, 692 F.2d at 648 .

Petitioners also contend that § 3(8)’s language requires OSHA to set conditions and requirements more specific than those set forth in the arsenic standard. Petitioners focus on that part of the new standard which requires the high-intake smelters to use work practice and engineering controls as much as possible and then to rely on respirators. 29 C.F.R. § 1910.1018 (g)(l)(ii). OSHA’s standard requires the smelters to devise their own compliance programs which OSHA will then review. 43 Fed.Reg. at 19616-19617. Petitioners argue that the standard is not “reasonable” or “appropriate” because OSHA has not told them specifically what technology to employ. They assert that requiring them to devise their own compliance plans is “unreasonable,” thereby violating § 3(8).

Petitioners’ “specificity” contention lacks merit largely because § 6(b)(5) states that “performance desired” and “objective criteria” are required only “whenever practicable.” 29 U.S.C. § 655 (b)(5). Because of the individualized operations of most smelters, OSHA’s explanation that such specificity is not practicable here is “reasonable.” See 43 Fed.Reg. at 19601-19604. OSHA’s approach has been accepted by other courts and we also accept it. See United Steelworkers, 647 F.2d at 1274-77 ; AFL-CIO v. Marshall, 617 F.2d 636 (D.C.Cir.1979), aff'd in part and rev’d in part on other grounds sub nom. American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490 , 101 S.Ct. 2478 , 69 L.Ed.2d 185 (1981) (cotton dust); American Iron & Steel Inst. v. OSHA, 577 F.2d 825 (3d Cir.1978), cert. dismissed, 448 U.S. 917 , 101 S.Ct. 38 , 65 L.Ed.2d 1180 (1980) (coke oven emissions); Society of Plastics Industry, Inc. v. OSHA, 509 F.2d 1301 , 1310 (2d Cir.), cert. denied, 421 U.S. 992 , 95 S.Ct. 1998 , 44 L.Ed.2d 482 (1975) (vinyl chloride); cf. Industrial Union Dept. v. Hodgson, 499 F.2d 467 (D.C.Cir. 1974) (asbestos).

Thus, we find that OSHA easily satisfies the test for general feasibility: “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.” United Steelworkers, 647 F.2d at 1272 (technological feasibility not found). 29 See also id. at 1269-70 (explaining how OSHA general feasibility requirement may be met).

(2) Deferral to Compliance Stage. Petitioners make two related contentions with regard to OSHA’s approach that we have just approved. They argue that by requiring industries to adopt all compliance measures that are feasible through consultation with the Secretary: (1) OSHA has impermissibly deferred its rulemaking duty to the compliance stage, and (2) by so doing, OSHA has impermissibly shifted the burden of proving feasibility to the industry. Both contentions are unpersuasive.

ASARCO supports these contentions by citing an apparent conflict between Boise *499 Cascade in this circuit and United Steelworkers in the D.C. Circuit. Boise Cascade requires the Secretary to bear the initial burden of proof as to feasibility in both rulemaking (pre-enforcement) and compliance (enforcement) proceedings. 694 F.2d at 589 . United Steelworkers indicates that, once the presumption of general feasibility has been proved in rulemaking proceedings, the burden of proof shifts to the industry in enforcement proceedings. 647 F.2d at 1269-70.

We need not concern ourselves with the potential intercircuit conflict as to the burden of proof in enforcement proceedings because neither this case nor United Steelworkers concerns a review of an enforcement proceeding. United Steelworkers is a rulemaking feasibility case. Boise Cascade is an enforcement case. We here only adopt the United Steelworkers rule that only general feasibility need be proved in rulemaking/pre-enforcement proceedings. This does not conflict in any way with Boise Cascade. See Boise Cascade, 694 F.2d at 590 (nature of burden at enforcement stage is different from burden at rulemaking stage). 30

(3) Vagueness. ASARCO’s contention that the standard is unconstitutionally vague is without merit. We have rejected a similar challenge against a noise standard of virtually identical specificity. See Plum Creek Lumber Co. v. Hutton, 608 F.2d 1283, 1288-89 (9th Cir.1979). 31

(4) Arsenical chemical industry challenges. Amici allege that the Secretary has not found the 10 ug/m3 PEL technologically feasible for their industry. See United Steelworkers, 647 F.2d at 1277 (where court reviewed Secretary’s feasibility findings for each affected industry challenging the standard). Their only support for this contention is a statement by one of OSHA’s experts that the arsenical chemicals industry could not reduce the PEL to 4 ug/m3 through engineering and work practices alone. OSHA’s experts, however, actually found that even a 4 ug/m3 PEL could be reached by the glass and desiccant manufacturing industries without respirator use, but that herbicide, pesticide, wood preservative, and lead arsenical manufacturers would require some respirator use to reach 4 ug/m3. OSHA determined that those industries’ manufacturing processes could be “enclosed” and that engineering and work practice controls alone could indeed meet the higher 10 ug/m3 PEL. Only the lead arsenicals industry would require “limited” respirator use. 43 Fed.Reg. at 19604. OSHA then found that the standard was feasible for the arsenical chemicals industry. OSHA’s findings are supported by substantial evidence. 32

B. Economic Feasibility

A standard is not economically infeasible because it is “financially burdensome” or even if it “threatens the survival of some companies within an industry.” United Steelworkers, 647 F.2d at 1265 (citation omitted). Similarly, “[a] standard is feasible if it does not threaten ‘massive dislocation’ to, ... or imperil the existence of ... the industry.” Id. (citations omitted). See also American Textile, 452 U.S. at 514 , 101 S.Ct. at 2493 (“Congress was *500 aware that the Act would impose real and substantial costs of compliance on industry”). We review the Secretary’s economic feasibility findings for substantial evidence. Boise Cascade, 694 F.2d at 589 .

(a) OSHA’s findings

OSHA’s economic feasibility findings, although open to criticism, are supported by substantial evidence. Moreover, OSHA’s methodology complies with the guidelines for proving economic feasibility set out in United Steelworkers, 647 F.2d at 1266 -67: “[T]he agency must ... provide a reasonable assessment of the likely range of costs of its standard, and the likely effects of those costs on the industry.” Id. at 1266 . We adopt those guidelines.

In making its economic feasibility finding, OSHA relied on four studies by consultants, one of which projected $32 million in annualized costs of compliance. 43 Fed. Reg. at 19604. Giving specific reasons, OSHA adjusted industry projections of $94.7 million down to derive a $37-63 million range for annualized compliance costs (low estimate/high estimate). 43 Fed.Reg. at 19605-06. OSHA then estimated the industry’s ability to pay by averaging annual pretax profits in the years 1966-75. 43 Fed.Reg. at 19606. That amount totaled $525 million. Id. OSHA determined that $30 million of the costs could be passed back to the industry’s customers, the copper mines. Id. OSHA estimated that the “passback” could account for all of the low-estimate costs, and half of the high estimate. Id. OSHA then estimated the cost of compliance with the standard (using the high estimate) as 6% of the industry’s average pretax profits. Id.

(b) Petitioners’ Contentions

Petitioners challenge OSHA’s economic feasibility findings on several grounds, including: (1) OSHA did not assess the impact on the competitive stability of the industry should ASARCO’s Tacoma plant close due to high costs of compliance with the arsenic standard, (2) OSHA did not take into consideration regulatory costs imposed by other statutes such as the Clean Air Act, (3) OSHA incorrectly assumed that industry could “passback” costs to the mines, and (4) OSHA should have assessed profits solely from smelting operations and not from an entire company’s profits. We address these arguments in turn.

Tacoma. Compliance with the arsenic standard would indisputably fall most heavily on ASARCO’s Tacoma smelter. ASARCO’s arguments centering on the Tacoma smelter, however, are now moot. After briefing in this ease, ASARCO informed us that it planned to cease copper smelting operations at its Tacoma plant by June 30, 1985. The Tacoma plant will remain open to continue processing smelting by-products from ASARCO’s other plants. There is no indication that ASARCO decided to cease copper smelting operations as a result of the OSHA arsenic standard. In its letter to the court, ASARCO cited the depressed state of the copper smelting industry, the shortage of copper ore concentrates that can be profitably processed at Tacoma, and other environmental regulations whose costs were many times higher than those envisioned in the occupational arsenic standard. 33 We therefore need no longer consider ASARCO’s economic feasibility challenges as they pertain to the Tacoma plant.

Other Regulatory Costs. Petitioners contend that OSHA’s economic cost estimates are skewed because OSHA failed to take into account the substantial costs associated with regulations promulgated under other health and pollution laws, such as the Clean Air Act (codified at scattered sections of 42 U.S.C.) (1982). Petitioners contend, and OSHA agrees, that these costs were not felt until after OSHA made its feasibility findings. Petitioners expect these non-OSHA-imposed costs to reach *501 $389.3 million per year by 1987 — 86% of the industry’s annual pretax profits. 34

Petitioners’ argument fails here largely because OSHA is not required to be prescient. OSHA is only required to use the best evidence available at the time it reaches its decision. United Steelworkers, 647 F.2d at 1267 . OSHA’s estimates of pretax profits reflect all regulatory costs for the years considered. 43 Fed.Reg. at 19606. Petitioners’ contention here is relevant, not to the adequacy of OSHA’s initial feasibility finding, but instead to the reasonableness of OSHA’s refusal to reopen the record on feasibility. We consider this argument infra, 746 F.2d at pp. 501-502.

Passback. ASARCO also contends that OSHA’s passback plan is not supported by substantial evidence. Petitioners argue that because of the structure of the smelting industry, a passback may drive business to foreign smelters. It appears, oddly enough, that OSHA relied on an ASARCO annual report to support its findings. 43 Fed.Reg. at 19606. Moreover, petitioners again rely primarily on evidence collected after OSHA’s determination was completed. See infra, 746 F.2d at pp. 501-502. Given conflicting evidence and OSHA’s use of industry-supplied evidence, we defer to OSHA on this point.

Profits. ASARCO contends that OSHA’s profit estimates for ASARCO should have been drawn solely from ASARCO’s smelting business, rather than from the entire company. OSHA counters with evidence that ASARCO’s operations are highly integrated. (Ex. 111-7, Appendix B). Relatedly, ASARCO argues that OSHA failed to consider adequately the industry’s debt capability. OSHA points to evidence to rebut that contention. (Ex. 135A, p. V-13). ASARCO points to contrary evidence (Ex. 202-7H, p. 2-35). We find the evidence to be “comparable.” On both points we defer to agency expertise. See American Textile, 452 U.S. at 523 , 101 S.Ct. at 2497 .

In sum, OSHA’s economic feasibility determinations are supported by substantial evidence. OSHA is not required to write an economic treatise on the smelting industry. See American Textile, 452 U.S. at 530 n. 55, 101 S.Ct. at 2501 n. 55. It need only show that the “long-term profitability and competitiveness” of the industry will not be threatened by the standard. Id. We find that OSHA’s calculations were reasonably made and were based on the best available evidence. We also note that the Steelworkers support imposition of the standards. See AFL-CIO v. Marshall, 617 F.2d 636, 665 (D.C.Cir.1979), aff'd in part and rev’d in part on other grounds sub. nom., American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490 , 101 S.Ct. 2478 , 69 L.Ed.2d 185 (1981). The Steelworkers’ support suggests that continued existence of the industry is not imperiled. Id.

III. Refusal to Reofen the Record

Petitioners contend that the Secretary abused his discretion by refusing to reopen the record on remand to consider new evidence on economic and technological feasibility. The procedural history of the request complicates the issue. On balance, we find that OSHA did not abuse its discretion in denying the request.

We will not set aside the agency’s refusal to reopen an administrative hearing absent a showing of abuse of discretion. United States v. Pierce Auto Freight Lines, Inc., 327 U.S. 515, 535 , 66 S.Ct. 687, 697 , 90 L.Ed. 821 (1946); Duval Corf. v. Donovan, 650 F.2d 1051, 1054 (9th Cir. 1981).

*502 During the remand, petitioners requested that OSHA reconsider its feasibility determinations in light of new evidence about both technological and economic feasibility. 48 Fed.Reg. at 1869. On January 14, 1983, OSHA declined and noted that it would treat the requests as petitions to amend the standard. Id. By letter dated June 7, 1983, OSHA denied the petition. The denial occurred several days prior to the deadline for submission of petitioners’ opening briefs in this matter. OSHA stated that the order of April 7, 1981 from this court limited the remand to the significant risk determinations required under IUD v. API. In its denial of petitioners’ request, OSHA also stated that, in apy event, the record on feasibility is sufficient to support the standard. We agree. 35

Our April 7 order did not necessarily limit the remand of this matter as strictly and as explicitly as OSHA contends. Order at 111 and If 2. Our order did, however, authorize the parties to petition this court to modify the scope of the remand in light of the Supreme Court’s then-upcoming decision in American Textile (which directly addressed feasibility issues). Order at H 2. Petitioners offer inadequate explanation for their failure to petition for modification of the remand. 36

In support of its argument for reopening, Kennecott seeks to introduce evidence showing the lack of technological feasibility to meet the arsenic standard at its Utah smelter, which was under reconstruction in 1978 when OSHA first issued the final standard. ASARCO seeks to offer evidence collected since 1978, purportedly demonstrating the current depressed economic state of the smelting industry as well as of the magnitude of compliance costs associated with environmental regulations. 37 Thus, petitioners appear to have had reasons to petition this court to expand the remand. Our order gave petitioners until 28 days after the Supreme Court issued its decision in American Textile to petition for such modification of the remand. They did not do so. 38

We are concerned that agency consideration of new evidence relevant to agency decisionmaking should not be excluded merely because the agency wishes to bring a long rulemaking or other administrative proceeding to a close. Nonetheless, we are also mindful of the limitations that we specifically placed on this remand as well as of petitioners’ failure to ask us to modify the scope of that remand. The Secretary would have arguably violated the terms of our remand had he reopened the record on feasibility issues without our consent. We therefore find that the Secretary did not abuse his discretion in refusing to reopen the record. 39

PETITION DENIED.

. This opinion will use the terms "OSHA” and “the Secretary” interchangeably when referring to the agency, the Secretary of Labor, or the Assistant Secretary for Occupational Safety and Health. The Secretary of Labor has delegated the authority to promulgate occupational safety and health standards to the Assistant Secretary. See 29 C.F.R. § 1910.4 (1983).

. Unless otherwise noted, the term "arsenic" refers to inorganic arsenic. Approximately 97 per cent of the arsenic at issue here is arsenic trioxide (AS2O3) (trivalent arsenic).

. The symbol “ug/ms" refers to micrograms per cubic meter of air.

. The 500 ug/m3 limit was originally set by the American Conference of Governmental Industrial Hygienists in 1947, apparently to protect against dermatitis and without regard to carcinogenesis. 43 Fed.Reg. at 19585. The 500 ug/m3 level was subsequently adopted by the Secretary under the Walsh Healey Public Contracts Act. 41 U.S.C. §§ 35-45 (1982). 34 Fed. Reg. 788 -96 (1969). Pursuant to § 6(a) of the OSH Act, the standard was repromulgated as an OSHA standard. 36 Fed.Reg. 10466 (May 29, 1971).

. Those consolidated cases were ASARCO, Inc., et al v. OSHA, No. 78-1959; The Anaconda Co. & Kennecott Copper Corp. v. OSHA, Nos. 78-2764, 78-3038; and General Motors Corp. & Chrysler Corp. v. OSHA, Nos. 78-2477, 78-2478.

. The expected opinion was American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490 , 101 S.Ct. 2478 , 69 L.Ed.2d 185 (1981) (reviewing OSHA’s "cotton dust’’ standard).

. For the Secretary’s detailed scientific and other findings see 48 Fed.Reg. 1864-903 (January 14, 1983), as corrected by 48 Fed.Reg. 24068-73 (May 31, 1983); 43 Fed.Reg. 19584-631 (May 5, 1978), as corrected by 43 Fed.Reg. 28472-73 (June 30, 1978).

. "Excess risk" is one method for describing the results of an epidemiological study. It estimates the number of additional deaths that will occur in a given group of people because of their exposure to a harmful substance. See 48 Fed.Reg. 1868 (describing risk statistics).

. Some smelters may have especially high arsenic emission levels because the smelter may refine ores (feed) with particularly high arsenic content. See 43 Fed.Reg. at 19602 (Table I).

. The petitioners are in the smelting industry. Pursuant to stipulation, we dismissed the petitions for review that had been filed by General Motors and Chrysler, Nos. 78-2477, 78-2478 (Order of April 8, 1983). A petition filed in the 3rd Circuit by ASARCO and Koppers Company, Inc. and transferred here (No. 78-2583) was dismissed without prejudice. (Order of June 9, 1983).

. OSHA relies on quite a few studies, and we refer the reader to 48 Fed.Reg. 1895-1903 for brief descriptions of those studies. Throughout this opinion we shall refer to them by their author or authors’ name(s).

. Note that even the lowest estimate of 2.2 excess deaths per 1000 easily exceeds the excess risk of one per thousand that the Supreme Court stated "a reasonable person might well consider ... significant and take appropriate steps to decrease or eliminate... ”. Industrial Union Dept. v. American Petroleum Inst., 448 U.S. 607, 655 , 100 S.Ct. 2844, 2870 , 65 L.Ed.2d 1010 (1980).

. The Secretary concluded that there was no safe level of exposure to arsenic, but only lowered the standard to 10 ug/m3 because of feasibility limitations. 43 Fed.Reg. at 19600.

. See supra 746 F.2d at pp. 489-490. See also IUD v. API, 448 U.S. at 655-56, 100 S.Ct. at 2870-71 ; American Textile Mfrs. Inst. v. Donovan, 452 U.S. 490, 523 , 101 S.Ct. 2478, 2497 , 69 L.Ed.2d 185 (1981); and United Steelworkers, 647 F.2d at 1248 .

. CMA’s contentions center on the quantification of exposure levels employed by the Lee & Fraumeni and Lee-Féldstein studies. The stud-, ies quantify exposure levels into "heavy," “medium,” and "light” exposures. The “light” category included workers whose exposure levels were 500 ug/m3 or below. CMA complains that employees in the "light” exposure groups in fact had arsenic exposures well over 10 ug/m3 and some of them had been exposed at some time to greater than 500 ug/m3. Morris, however, a consulting engineer at the Anaconda smelter, quantified the "light” group at a mean of 290 ug/m3. CMA contends that only Higgins’ study actually measured excess risk for workers who had only been exposed to very low arsenic levels. Thus, they argue, Higgins must be accepted, and all other studies rejected.

CMA’s challenges do not succeed. We reject CMA’s implication that the Secretary could not set a 10 ug/m3 PEL unless he has evidence measuring worker exposures close to that level. IUD v. API clearly explains that the Secretary need not "wait for deaths to occur," 448 U.S. at 655, 100 S.Ct. at 2871 , nor need he "calculate the exact probability of harm.” Id. Prior to the Secretary’s issuing the 10 ug/m3 PEL, the PEL was 500 ug/m3. Because of this, it is likely that the pool of workers exposed to only 10 or even under 100 ug/m3 is quite small. We reject the notion that the Secretary must have studies at or quite near a PEL that may not exist before he may set such a PEL. Moreover, the Secretary's studies are not invalidated because of a possibility that some employees in the "light” group may have had some higher exposures. The law is clear that the Secretary need only find that the risk to worker health is "significant” at a certain level and show a likelihood that reducing the standard to a given level will significantly reduce that risk. IUD v. API, 448 U.S. at 655, 100 S.Ct. at 2870 . The Supreme Court has expressly declined to put the Secretary into the "mathematical straightjacket” that CMA would have him wear. Id. See also infra 746 F.2d at pp. 493-494 (discussing OSHA’s determination of excess risk at 10 ug/m3). We find CMA’s other criticisms of the Lee & Fraumeni and Lee-Feldstein studies unpersuasive.

. This model posits that cancer risk diminishes as arsenic exposure approaches zero and that, therefore, there is no safe threshold exposure other than zero. It thus accepts that cumulative exposure to even low levels of arsenic is likely to cause cancer. CMA, on the other hand, argues that the intensity of a worker’s exposure to arsenic is the critical variable in arsenic’s carcinogenicity and that there is a threshold exposure level under which no cancer risk occurs. CMA relies on Higgins to support this theory.

. See supra note 16.

. For example, Higgins’ hypothesis, unlike the hypotheses entertained by the other studies, was not based on a well-established biological model. Despite CMA's criticism of OSHA’s quantifications, Higgins’ quantifications are also open to question. See 48 Fed.Reg. at 1866, 1873-75, 1899 for OSHA’s criticisms of Higgins’ work.

. An epidemiological study can fail to detect a true excess risk because of low statistical power, low significance, poor sampling, or other types of errors. “Power” is a statistical concept which quantifies the ability of a study to detect an excess risk that truly exists. 48 Fed.Reg. at 1875. Larger excess risks are more readily detectable than excess risks of a small magnitude. An increase in the amount of data increases the chance or "power" of observing a given risk. For example, Enterline and Marsh expanded a Pinto study of 527 retirees to 2802 workers. Based on the retiree study alone, Pinto initially suggested that there might be a threshold for arsenic-induced lung cancer. 48 Fed.Reg. 1900. However, the expanded study showed a risk well below the threshold initially suggested by Pinto — even among workers with less than 19 years of employment. Expanding the study and increasing the statistical power detected an increased risk at the exposure level that would otherwise erroneously have appeared to be a threshold.

. CMA also relies on two Fifth Circuit decisions to challenge OSHA’s trivalent arsenic (arsenic trioxide) risk assessments. This reliance is misplaced and merely highlights the weakness of CMA’s challenge to the Secretary’s risk *494 assessments. The cases are distinguishable because of the glaring lack of substantial evidence found in each.

In Gulf South Insulation v. Consumer Product Safety Comm’n, 701 F.2d 1137 , 1146 (5th Cir. 1983), the court held that no substantial evidence existed to show carcinogenicity of urea-formaldehyde foam used in home insulation because the sole empirical datum consisted of one 240-rat study.

Similarly, in Texas Independent Ginners Ass’n v. Marshall, 630 F.2d 398, 407-08 (5th Cir. 1980), OSHA inferred a correlation between an American study, which showed no evidence of “brown lung” in American ginners, and foreign studies that did find it in their workers. The court held that the putative correlation could not constitute substantial evidence because ginning conditions in the two study areas were substantially different. Id.

In contrast, OSHA’s evidence here consists of large scale studies of American workers.

. As we have noted, OSHA need not make its significant risk determinations with the degree of "scientific certainty” that petitioners would impose upon them. See supra 746 F.2d at pp. 490-492 n. 15.

. OSHA also had corroboration from 21 other experts. See, e.g., 43 Fed.Reg. at 19600 (listing studies).

. Obviously, our review would be somewhat less generous were this a challenge that OSHA had "underprotected” the public.

. CMA also suggests that we remand the record once again because Drs. Higgins, Lee-Feldstein, and Enterline are reexamining their data. Dr. Higgins is working with Dr. Lee-Feldstein to reexamine the entire Anaconda smelter workforce. Dr. Enterline is examining airborne, as opposed to urinary, arsenic levels at the Tacoma smelter. We decline to remand the record on the basis of studies now in progress. The earlier studies are adequate to support the Secretary’s findings. Should the studies, once completed, reach significant new findings that would support amendment of the standard, petitioners will then be free to petition the Secretary to amend the standard on the basis of those studies.

. See American Textile, 452 U.S. at 528-29 , 101 S.Ct. at 2500-01 , where the Court assessed Secretary's “reasonableness” in accepting or rejecting scientific studies.

. As with the Secretary’s significant risk determination: "OSHA cannot let workers suffer while it awaits the Godot of scientific certainty. It can and must make reasonable predictions on the basis of 'credible sources of information,’ whether data from existing plants or expert testimony.” United Steelworkers, 647 F.2d at 1266 (citation omitted).

The D.C. Circuit’s excellent opinion in United Steelworkers comprehensively discusses the feasibility issue. We therefore do not duplicate its analysis here — we adopt it by reference.

. The problems associated with respirators include problems with adequate facial fit, increased heat stress, reduced vision, increased breathing resistance, speech limitation, limited mobility, and excess weight. 43 Fed.Reg. at 19617.

. Petitioners’ technological feasibility challenges rely heavily on the fact, conceded by OSHA, 43 Fed.Reg. at 19603, that ASARCO’s Tacoma smelter would require particularly pervasive respirator use to meet the 10 ug/m8 PEL. Since this case was briefed, however, ASARCO has decided to cease most of its smelting operations at the Tacoma plant by mid-1985. ASARCO and OSHA have also reached agreement on Tacoma’s compliance plan until that time. The specific problems associated with meeting the arsenic standard at the smelter, therefore, no longer require our consideration. The plant will continue to process arsenic from other facilities. The limited respirator use that will be required for these remaining operations does not defeat the standard for the reasons stated immediately infra in text.

. We note also that OSHA's rulemaking here is very different from its rulemaking in setting the lead standard under review in United Steelworkers , where the D.C. Circuit found that OSHA had failed to prove feasibility for any affected industrial operation. Id. at 1298 .

. Boise Cascade's holding is not so broad as petitioners suggest. In that case, the Secretary had promulgated a noise standard under § 6(a) of the Act. Relying on earlier Ninth Circuit cases, we held that where the standard makes feasibility an element of the violation, the burden of proving that controls are feasible is on the Secretary. 694 F.2d at 589 (citing Castle & Cooke, 692 F.2d 801 , 803 (9th Cir.1981)). Boise Cascade does not indicate that the Secretary bears the burden of proving feasibility in every enforcement proceeding.

. Kennecott also contends that OSHA has failed to use the "best available evidence” required under § 6(b)(5) because the agency refused to consider data collected at Kennecott’s Utah smelter after OSHA published its findings. This contention actually contests OSHA’s refusal to reopen the record to admit more evidence on feasibility. This contention is considered infra 746 F.2d at pp. 501-502.

. The Secretary also found the standard economically feasible for the arsenical chemicals industry. Amici do not contest the standard on economic feasibility grounds.

. In particular, ASARCO cited a projected cost of $150 million to meet air quality standards for sulphur dioxide (SO2) emissions.

. In its letter notifying the court of its intention to cease copper smelting operations at the Tacoma smelter, ASARCO indicated that its action "forcefully supports" ASARCO's complaint that OSHA gave inadequate weight to the impact of other regulatory costs on the economic viability of the Tacoma smelter. While our decision here must rest on the record before the Agency at the time it made its decision rather than upon post-argument letters from parties, we note here that ASARCO’s argument is without merit for the reasons stated immediately infra (OSHA need not be prescient) and infra text at 746 F.2d at p. 502 (petitioners should have petitioned this court for expansion of the remand as instructed by our order of April 7, 1981).

. See infra notes 38 & 39.

. Petitioners contend that they did not think our remand was so limited. Had our remand not been limited, however, we would obviously have had no need to structure our order to specifically provide for a petition for modification.

. In support of reopening the record, ASARCO cites Atchison, Topeka & Santa Fe R.R. Co. v. United States, 284 U.S. 248 , 52 S.Ct. 146 , 76 L.Ed. 273 (1932), in which the court ordered the ICC to reopen hearings to readjust rates, set prior to the Depression, to Depression realities. The Supreme Court has said that Atchison is unique, ICC v. Jersey City, 322 U.S. 503, 515 , 64 S.Ct. 1129, 1135 , 88 L.Ed. 1420 (1944), and, therefore, the decision is inapposite here.

Moreover, OSHA took market fluctuations in the copper industry into account in averaging the industry’s profits over a ten year period. Thus, the drop in copper prices over the past few years is a factor OSHA has already accounted for. See 43 Fed.Reg. at 19606.

. Kennecott argues that its experience with its new Utah smelter is especially relevant to the technological feasibility of the standard because the Utah smelter is the most technologically advanced smelter in the nation. We note, however, that this showing would not be relevant to our pre-enforcement review because all the Secretary need show here is general- feasibility for the industry. The fact that the Utah smelter will require some supplemental respirator use does not defeat the standard.

. Moreover, even had our remand not been so limited, we would find that the Secretary did not abuse his discretion in refusing to reopen the record. As to technological feasibility, see immediately supra note 38. As to economic *503 feasibility, petitioners’ most plausible argument pertains to the Secretary's failure to consider the magnitude of the industry’s projected costs of compliance with Clean Air Act regulations. That argument fails, however, for a number of reasons. For example, the Clean Air Act directs the Administrator of the Environmental Protection Agency to adjust compliance requirements for smelters that encounter economic or other difficulty in meeting the standard. 42 U.S.C. § 7419 (b)(3) (1982). Petitioners gave no indication that such adjustments would not be forthcoming. Second, even if an economic bind were to occur as a direct result of OSHA’s regulation, OSHA will adjust the affected company’s compliance plan. See 43 Fed.Reg. at 19606. Thus, the speculative nature of industry contentions on this point amply justifies the Secretary’s refusal to reopen the record. It also justifies his refusal to amend the standard.