Occupational Safety and Health Administration

OSHA was created by, and has its regulatory authority under, the Occupational Safety and Health Act of 1970. (Because the agency and the act share the same acronym, the act is typically abbreviated as "OSHAct" and the agency itself as "OSHA.") The act's stated purpose is "to assure so far as possible every working man and woman in the Nation safe and healthful working conditions" by several means, including "providing medical criteria which assure insofar as practicable that no employee will suffer diminished health, functional capacity, or life expectancy as a result of his work experience" (OSHAct §2). It was passed during the heyday of public concern about environmental health that also saw the founding of the Environmental Protection Agency. Regulatory decision-making under the OSHAct is formally invested in the Secretary of Labor.

The act mandates in §5(a) that "Each employer...shall furnish to each of his employees employment and a place of employment which are free from recognized hazards that are causing or are likely to cause death or serious physical harm." The regulatory authority of OSHA is provided by §6 of the act, which sets out methods and criteria for issuance of occupational safety and health standards. In particular, §6(b)(5) states that "The Secretary, in promulgating standards dealing with toxic materials or harmful physical agents...,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...for the period of his working life." This paragraph further states that "In addition to the attainment of the highest degree of health and safety protection for the employee," the Secretary must consider "the feasibility of the standard" and that "Whenever practicable, the standard promulgated shall be expressed in terms of objective criteria and of the performance desired."

In other words, the achievement of safe and healthful workplaces is to be brought about by the setting of enforceable workplace standards, in practice framed primarily in terms of allowable limits to employee exposure. For a workplace to be considered healthful, the limits to exposure are to be set so that an employee could be exposed at the limit for an entire working life without suffering harm. The authority is over the exposure limits, not over how they are achieved. In practice, engineering controls are preferred to respirators, where feasible. In 6(b)(7), however, it is stated that "Where appropriate, such standards shall also prescribe suitable protective equipment and control or technological procedures to be used in connection with such hazards." (This paragraph goes on to prescribe labels, warnings, and provisions for ongoing monitoring of employee exposure.)

The OSHAct does not mention risk assessment as such, nor does it say much about the establishment of safe exposures. It is more explicit than some other laws about what constitutes an adverse health effect, however. In §2 it refers to "diminished health, functional capacity, or life expectancy" while §6 mentions "material impairment of health or functional capacity" as outcomes to be avoided. The mandated focus is on individual risk to a hypothetical employee experiencing an agent at the permissible exposure limit for a working lifetime, with regulation set "to the extent feasible" so that such an employee will suffer no impairment. The interpretation of these provisions has undergone considerable evolution as the result of some key judicial challenges. A full account is beyond the scope of this report, but the history and issues are reviewed by Graham et al. (1988).

Initially, the mandate was interpreted as essentially a health-based standard with an added proviso that health-based regulations could not be set so low as to be infeasible, interpreted as meaning having significant financial impact on the industry. For carcinogens, the lack of demonstrable exposure thresholds for toxic effect was interpreted to mean that no workplace exposure standard, however low, could assure that "no employee will suffer material impairment of health." Accordingly, the "feasibility" provision becomes the limiting factor, and workplace standards for carcinogens were set as low as was deemed to be technically feasible at reasonable cost. Under this interpretation, in a proposed "carcinogen policy" (42 FR 54148, 1977), risk assessment for carcinogens played a rather minor role in OSHA's setting of workplace standards, and OSHA staff generally argued that the uncertainties of quantitative cancer risk assessment precluded its use as a basis for regulation.

A proposed 1 ppm standard for workplace benzene exposure set under this interpretation was challenged in court, eventually leading to a 5-4 Supreme Court decision [Industrial Union Department v. American Petroleum Institute, 448 U.S. 607 (1980)], commonly known as the "benzene decision," which imposed fundamental changes in the interpretation of the OSHAct mandate. The court ruled that, before issuing a standard, OSHA must first demonstrate that the chemical posed a "significant risk." Unless the risk is significant, the material does not become a "toxic material" or "harmful physical agent" controllable under the act, and its presence cannot be said to meaningfully lead to an unhealthy workplace. A key part of this finding was that the §3(8) definition of a standard as a "reasonably necessary or appropriate" action was taken as grounds that action under §6(b)(5) must be shown to be necessary in some quantitative sense. While stating that "OSHA is not required to support its finding that a significant risk exists with anything approaching scientific certainty," the court ruled that the case for significant risk could in principle be made using quantitative risk analysis. On the question of how large a cancer risk is "significant," Justice Stevens, in his opinion, stated that this was OSHA's responsibility, conceded to be a matter of policy, but that "If, for example, the odds are one in a billion..., the risk clearly could not be considered significant. On the other hand, if the odds are one in a thousand..., a reasonable person might well consider the risk significant and take appropriate steps to decrease or eliminate it."

In effect, the benzene decision prompts OSHA to conduct quantitative risk assessment in order to set standards for carcinogens. The court declined to address the related question about whether the "feasibility" and "reasonably required" standard-setting issues should be interpreted to require cost-benefit analysis of proposed standards. In a later supreme court decision, the "cotton dust decision" [American Textile Manufacturers Institute v. Donovan, 452 U.S. (1981)], the court ruled that OSHA may set a level as protective of health as feasible, even if a less stringent one has a more favorable cost-benefit ratio.

One further court case of note is the recent ruling [AFL-CIO v. OSHA, 965 F.2d. 962 (1992)] that OSHA must make its risk case for each chemical according to its own analysis. The practice of adopting outside standards, and of setting standards based on general risk arguments rather than case-by-case demonstration of significant risks, was struck down, invalidating 428 OSHA permissible exposure limits.

Since the benzene decision, risk assessment at OSHA has been dominated by the question of showing "significant" risk from exposure to workplace carcinogens. The question that Justice Stevens threw back to OSHA in his benzene opinion--what constitutes a "significant" risk (within the limits he set)?--has never been fully answered. Justice Stevens' statement that a lifetime risk of one in a thousand is clearly significant has served as something of a benchmark; in practice risks below 10^-5 are rarely given much significance, but the lower bound on risks considered significant is hard to define because it is difficult to show. There is no real case to date where OSHA did not pursue a standard because cancer risks were calculated to be low. In this case, the "significance" question is one of individual risk (rather than of public health impact on the whole exposed population), because the question is still posed in terms of the hypothetical worker exposed at the permitted limit. (OSHA has a policy of forbidding rotation of employees through jobs with high carcinogen exposure as a work practice to ensure no employee experiences a PEL for a 45 year working life. The grounds are that this strategy would only increase the number of workers exposed. In essence, this is a population risk argument.)

In practice, the technical and financial feasibility of achieving a standard is usually the limiting factor in choosing a permissible exposure level (P. Infante, personal communication). That is, limits are usually proposed under which a worker exposed to that limit would be calculated to experience risk in the upper end of Justice Stevens' range. (This is not to say that real workers with their actual exposures are necessarily suffering significant risk.) Under these conditions, the particular numerical estimate of risk level is not the driving issue in regulation, only the more general argument that "significant" risks could be generated. OSHA is able to entertain a variety of risk analyses based on somewhat different data sets and assumptions without muddying the regulatory decision with questions about which single analysis is the "right" one to choose to set a standard.

In the analyses that in practice drive the permissible levels specified in standards--that is, the determination of what levels are feasible to achieve--the costs and performances of various technical control options are considered. In these analyses, actual worker exposure levels and durations of exposure can be considered, including the resulting changes in residual risk to be expected after various regulatory options. Thus, there is opportunity, albeit indirect, for information on distributions of actual exposure to come into play in determining OSHA regulations. Nonetheless, the key consideration in feasibility is not risk, but rather the costs and technical ability needed to reach various ambient concentration levels.

Although the benzene decision has profoundly affected OSHA's approach to the analysis of risk, the practical result is that decisions are not very different from what would have been done under the pre-1980 carcinogen policy. The benzene decision stated that OSHA could not simply limit exposures according to feasibility of control without first showing that lack of control leads to significant risk. In practice, this is usually shown, at least for the standards that OSHA has pursued since 1980, so controls are set primarily on feasibility all the same. The role of risk assessment in this process is largely to establish (1) that significant risks exist under current exposures, and (2) that reducing the exposure as proposed in the standard will reduce the risk. The major practical impact is that the case for significant risk must be made for each compound, focusing the agency's activities and resources to pursue regulation on those compounds where risk can be clearly shown. Feasibility is a particular problem for OSHA because the characteristics of the indoor environment make it very difficult to control exposures to levels that other agencies might seek.

The principal notable features of risk assessment at OSHA are that the size of the risks in question are a good deal larger than those encountered in other regulatory programs. Frequently, risks may be assessed on human data directly relevant to the regulatory interest; in recent years about one-half of OSHA PELs have been based primarily on human data. Even when animal data are used, human exposures of interest are often not far below the tested levels. Real, directly relevant exposure data are often available, and they are often quite defined and less variable compared to environmental exposures for the general population. As a consequence, OSHA risk assessments have to grapple much less with extrapolation questions, and OSHA's methods have less built-in conservatism (for example, use of maximum likelihood estimates instead of upper bounds). Since PELs are in practice set by feasibility, with risk assessment determining the need for controls, OSHA is able to entertain a variety of risk analyses without settling on a single "number" as the canonical one for its regulatory activities. The regulatory focus is on the risk to a worker exposed to the permitted level for a full working life; although in practice and for a variety of reasons, this hypothetical exposure may not be much higher than that actually experienced by many workers, and indeed some workers (those doing overtime or previously exposed under a higher standard, for example) may exceed this theoretical "maximum."