OPINION OF THE COURT

WEIS, Circuit Judge.

The Environmental Protection Agency adopted a rule requiring petitioner manufacturers to conduct extensive testing of certain chemicals, fluoroalkenes, to determine their potential for producing adverse health effects. The manufacturers have petitioned for judicial review, asserting that the rule is contrary to law and should be set aside. An examination of the record persuades us that the agency action is supported by substantial evidence, and accordingly, we deny review.

The Toxic Substances Control Act, 15 U.S.C. §§ 2601-29, authorizes EPA to promulgate rules directing manufacturers to arrange for testing of chemical substances that "may present an unreasonable risk of injury to health." 15 U.S.C. § 2603(a)(1)(A)(i). Petitioners are the manufacturers of four chemicals, collectively known as fluoroalkenes: vinyl fluoride, vinylidene fluoride, hexafluoropropene, and tetrafluoroethene.

Fluoroalkene monomers, precursors in the preparation of specialty fluorocarbon polymers, are incorporated into various end products. Vinyl fluoride (VF) is processed to form a high-strength film used for wall coverings, lighting panels and automobile trim, as well as for laminating aluminum, steel, wood and plastics for industrial buildings. Vinylidene fluoride (VDF) is employed in the manufacture of resins for computer panel boards, tank linings, pumps, valves and pipes. Hexafluoropropene (HFP) is used in the production of fluorinated oils and greases for the electrical and chemical industries. Tetrafluoroethene (TFE) is converted to make resins, elastomers and fluids for electrical insulation, chemical processing equipment and medical devices.

The manufacturing and processing of these fluoroalkenes takes place within closed pipes, reaction vessels and other equipment located behind sealed and barricaded areas of the factory. At room temperature the substances are odorless gases. in the course of converting the monomers into polymers, fluoroalkenes are chemically altered and cease to exist in their original form. Several of the fluoroalkenes are highly flammable. The danger of explosion combined with the economic cost of inefficient production act as incentives to guard against release of the gases into the environment.

Five companies produce one hundred percent of the fluoroalkenes: DuPont, Pennwalt, ICI Americas, Allied and American Hoechst. Currently no commercially available substitutes for these monomers exist. Consequently, the inelastic demand permits manufacturers to pass along price increases and testing costs. The major end-use markets include the automotive, chemical-petrochemical, industrial pollution control, hydraulic-pneumatic, and aerospace industries. Growth rates in these markets are favorable.

The parties disagree on the number of workers who may be affected by contact with the chemicals. In their brief, petitioners assert that only fifty workers are "potentially exposed" to vinyl fluoride and that the number potentially exposed to the other fluoroalkenes is less than five hundred. EPA estimates that "between eleven hundred fifty and two thousand ninety-five workers" may be exposed to these substances during their production and use. This numerical disparity may have been caused partly by recounting workers exposed to more than one of the chemicals, and at oral argument EPA conceded that its estimates may have involved some double counting.

Nearly seven million pounds of vinyl fluoride are transported annually by rail in closed tanks from DuPont's production plant in Louisville, Kentucky to its facilities in Buffalo, New York. Although in the event of derailment these movements carry a possibility of leaks and a threat of chemical escape into the atmosphere, EPA believes that human exposure to these chemicals is unlikely to occur outside the workplace. See 46 Fed.Reg. 53,706 (1981).

Petitioners assert that exposure above two parts per million (2 ppm) has occurred only during non-recurring short-term events — operational upsets, atypical work day exposures, reactor plugging problems, and unusual work practices. Petitioners characterize these incidents as "fugitive emissions." Although some workers wear air-supplied respirators, these devices are not used generally where industry considers engineering controls adequate to prevent emissions. An industrial hygiene study by the Fluoroalkene Industry Group (Industry Group) revealed, however, that in the instances where exposures exceeded 2 ppm, workers were not wearing respirators.

The Toxic Substances Control Act created the Interagency Testing Committee, an entity consisting of eight members from various federal agencies whose work focuses on various concerns associated with toxic substances. Represented are such agencies as EPA, the National Cancer Institute, the National Science Foundation and the National Institute for Occupational Safety and Health. See 15 U.S.C. § 2603(e)(2)(A). The Committee is charged with preparing a list limited to fifty chemicals to which EPA must give first order scrutiny. In establishing such a list, "the committee shall give priority attention to those chemical substances and mixtures which are known to cause or contribute to or which are suspected of causing or contributing to cancer, gene mutations, or birth defects." Id. § 2603(e)(1)(A).

In November 1980 the Committee listed the fluoroalkenes for priority testing. After a number of conferences with industry groups and an aborted "negotiated testing proposal," see Natural Resources Defense Council v. EPA, 595 F. Supp. 1255, 1261 (S.D.N.Y. 1984), EPA issued a notice of proposed rulemaking to require testing. Written and oral comments were submitted. The agency held a public meeting, examined scientific data, and then published a final test rule on June 8, 1987. 52 Fed.Reg. 21,516 (codified at 40 C.F.R. § 799.1700 (1987)).

The rule requires reproductive effects testing for VDF, subchronic toxicity testing for HFP, chronic oncogenicity bioassays for VF and VDF, mutagenicity testing for VF, VDF, HFP, and TFE. Depending on the outcome of the mutagenicity testing, chronic oncogenicity bioassays for HFP and TFE may also be ordered. EPA estimates total testing costs to range from $4,783,500 to $6,196,200, for a unit cost of about 2.6 to 3.3 cents per pound of the listed substances. Petitioners expect the expense to be from $5,000,000 to $9,000,000.

In asking for review, petitioners contend that before issuing a testing rule, EPA must demonstrate that humans are actually exposed to the chemicals to such a degree that serious harm could result if the substances are toxic. Petitioners assert that EPA may not base a rule on speculation but must found it on evidence establishing a reasonable likelihood of significant risk. The agency insists that the Act authorizes it to issue a rule based on potential exposure, and that evidence in the rulemaking record supports the testing order here. EPA also contends that scientific uncertainty over the possible harmful effects of the chemicals provides the justification for testing.

In enacting the Toxic Substances Control Act, Congress made a number of preliminary findings. 15 U.S.C. § 2601(a). It indicated concern that the manufacture and processing of some chemicals "may present an unreasonable risk of injury to health." Id. § 2601(a)(2). To meet this threat, the government declared its policy that the manufacturers and processors of these substances must develop "adequate data" with respect to "the effect of chemical substances and mixtures on health and the environment." Id. § 2601(b)(1).

Agency authority, however, should be exercised "in such a manner as not to impede unduly or create unnecessary economic barriers to technological innovation", while assuring that those activities do not "present an unreasonable risk of injury to health." Id. § 2601(b)(3). EPA is directed to carry out the provisions of the Act in a "reasonable and prudent manner" and to "consider the environmental, economic, and social impact of any action" taken. Id. § 2601(c).

These themes underlie section 2603(a)(1)(A), which directs that testing be conducted when the EPA Administrator "finds" that "the manufacture, distribution in commerce, processing . . . of a chemical substance . . . may present an unreasonable risk of injury to health or the environment" and

"(ii) there are insufficient data and experience upon which the effects of such manufacture . . . or . . . of such activities on health . . . can reasonably be determined or predicted, and

(iii) testing . . . is necessary to develop such data. . . ."

Judicial review, governed by section 2618, requires the court of appeals to "hold unlawful and set aside" EPA action if the court "finds that the rule is not supported by substantial evidence in the rulemaking record . . . taken as a whole." 15 U.S.C. § 2618(c)(1)(B)(i). We read this standard of review as more demanding than the arbitrary and capricious test often applied to administrative rulemaking. See Charlton v. United States, 412 F.2d 390, 398 (3d Cir. 1969) (Stahl, J., concurring). See also Note, Convergence of the Substantial Evidence and Arbitrary and Capricious Standards of Review During Informal Rulemaking, 54 Geo.Wash.L.Rev. 541 (1986).

Nevertheless, the statutory language "unreasonable risk of injury" and "may present" are not completely specific. We can agree with petitioners that Congress did not intend EPA to direct expensive experimentation based on mere speculation. We appreciate, too, that with thousands of substances in the marketplace, the agency must be reasonably discriminate in selecting subjects for testing. But section 2603 focuses on investigating areas of uncertainty as a prelude to regulating harmful substances.

Although mere scientific curiosity does not form an adequate basis for a rule, as the seriousness of risk becomes known and the extent of exposure increases, the need for testing fades into the necessity for regulatory safeguards. The issue presented here is where in the spectrum from screening to regulation this rule falls. In most administrative proceedings, we examine the record to see if there is a foundation for an agency determination of fact; however, here we look to see if the Administrator produced substantial evidence to demonstrate not fact, but doubt and uncertainty.

Both parties agree that risk is a critical factor. Risk implicates two concepts — toxicity and exposure. In each instance, quantity is important. Essentially, petitioners argue that neither factor has been shown to exist in sufficiently threatening amounts to justify promulgation of a rule. Petitioners argue that EPA has not pointed to any evidence showing the fluoroalkenes to be "unusually toxic" and has never posited that long-term health effects have been established.

To bolster their argument, petitioners question the necessity of oncological testing for VDF because it has not been shown to cause cancer. The record, however, does refer to published work by scientists Maltoni and Tovoli showing that VDF is a carcinogen. 52 Fed.Reg. 21,522-523 (1987). Although skeptical about the scientific conclusions of that work, the Industry Group has not produced evidence strong enough to fully discredit the Maltoni and Tovoli data.

Petitioners also criticize EPA's practice of relying on its information about one harmful substance to assess the danger from another of similar molecular structure. In this case, EPA has identified such a "structure-activity relationship" (SAR) between VDF and vinylidene chloride, which has been suspected, but not definitely established, as a carcinogen. 52 Fed.Reg. 21,519 (1987). See Natural Resources Defense Council, Inc. v. EPA, 824 F.2d 1146, 1148 (D.C.Cir. 1987) (court noted that vinyl chloride is a "strong carcinogen"). Again, we believe that the petitioners' attempt to transform EPA's concerns about the lack of scientific certainty into mere speculative scouting for data actually strengthens the government's position. These questions broaching the frontiers of scientific knowledge highlight the need for testing.

We have examined the other instances in which EPA did not accept the Industry Group's reservations about the validity and strength of scientific data. We have also scrutinized theories that raise concern about health hazards from the fluoroalkenes. The petitioners' submissions may weigh against a finding that regulation is required; but they do not demonstrate that testing to reduce or dispel doubt is unnecessary.

Petitioners principally attack the exposure aspect. They interpret their own study as showing minimal quantities of the chemicals in the plant atmosphere and note the small number of workers exposed to the substance. Although the data support these claims to some extent, EPA calls attention to the results of the Industry Group's industrial hygiene survey indicating that "worker exposure to fluoroalkenes typically takes the form of brief episodes of relatively high exposure followed by extended periods of little or no exposure." 52 Fed.Reg. 21,518 (1987). For purposes of risk assessment, EPA assumes that a high dose of a carcinogen received over a short period is equivalent to a corresponding low-dose spread over a lifetime. 51 Fed.Reg. 33,998 (1986).

Moreover, as the agency argues, certain workers because of their jobs are more apt to be present during periods when gases become "fugitive" and thus may experience substantial concentrations of fluoroalkenes over the long term. Id. As we observed above, not all workers who may inhale gases while performing tasks wear respirators routinely.

Petitioners rely heavily on Shell Chemical Co. v. EPA, 826 F.2d 295, 297-98 (5th Cir. 1987), in which the manufacturers of mesityl oxide, a chemical intermediate, challenged an EPA testing order. The court observed that the degree of risk was disputed because production of the chemical in an enclosed, controlled environment presented only a limited opportunity for exposure, and only "a few hundred" workers were involved. At argument in that case, the manufacturers asserted that use of the chemical had declined so that "only about 100" workers were engaged in production. See id. at 298. To the contrary, EPA argued that use of the chemical was increasing and had expanded to include additional end products. Viewing the case as a "close" one, the court remanded for supplemental findings on the extent of usage since promulgation of the test rule. Id.

The court in Shell Chemical recognized the problem of drawing a line between testing based on simply speculation and on scientific uncertainty when risk was not merely illusory. Because substantial developments in use and exposure could be relevant in that determination, the court properly remanded for further findings. The holding in Shell Chemical therefore is inconclusive and of no real assistance in the dispute before us.

The manufacturers here do not question the agency's finding that prospects are good for continued manufacturing and marketing of the chemicals. We have no indication of diminished use or of a reduced number of persons potentially affected. Even the limited holding of Shell Chemical requiring more factfinding is not applicable here, nor do petitioners argue that as few as one hundred workers may be affected. Consequently, these circumstances more strongly support approval of a testing rule than did those in Shell Chemical.

We are not prepared to say, as petitioners urge, that the number of persons potentially exposed to the fluoroalkenes, or the evidence of incidents which have occurred, is such that the element of risk is insignificant. Ultimately the question in this case is one of degree, and on that issue there is substantial evidence in the record to justify a finding of unreasonable risk.

In construing the statutory phrase "may present," we agree to some extent with both parties. The necessity for testing depends on lack of knowledge. If no doubt existed, testing would not be required and the "actual" risk concept espoused by petitioners would have little meaning. Likewise, if EPA views "potential" exposure as merely a remote possibility founded on theoretical factual situations, the statutory directive for "reasonable and prudent" agency action would counsel against expensive testing.

Our reading of the statutory phrase falls between the poles staked out by the parties. We believe that this interpretation of the statute prevents a testing rule based on little more than scientific curiosity, yet allows the agency to act when an existing possibility of harm raises reasonable and legitimate cause for concern. So read, the statute expresses the intention of Congress.

The congressional conference committee report on the Act stated that the purpose of the testing provision is to

"focus the Administrator's attention on those chemical substances and mixtures about which there is a basis for concern, but about which there is inadequate information to reasonably predict or determine their effects on health or the environment. The Administrator need not show that the substance or mixture does or will present a risk."

H.R.Conf.Rep. No. 1679, 94th Cong., 2d Sess. 61, reprinted in 1976 U.S. Code Cong. Admin. News 4491, 4539, 4546.

Congress obviously had serious concerns about toxic substances and made a policy judgment to place the expense of testing on manufacturers who ultimately can transfer the cost to consumers. Although cautioning that the agency must act reasonably and prudently, and take into consideration the economic impact of any action, of necessity Congress granted EPA fairly broad discretion in exercising its expertise to determine when data must be produced.

In Natural Resources Defense Council, 824 F.2d at 1153, the court addressed the agency's mandate to regulate hazardous substances in language pertinent here: "Congress chose . . . to deal with the pervasive nature of scientific uncertainty and the inherent limitations of scientific knowledge by vesting in the Administrator the discretion to deal with uncertainty in each case."

We conclude that the rule promulgated by EPA falls within the statutory authorization and is supported by substantial evidence. Accordingly, the petition for review will be denied.