Legal Implications of Hormesis
Frank B. Cross, Professor
University of Texas
Dept. of Management Science & Information Systems,
Campus Mail Code: B6500
Austin, TX 78712
University of Texas
Dept. of Management Science & Information Systems,
Campus Mail Code: B6500
Austin, TX 78712
Readers of this publication are surely aware that the evidence for beneficial hormetic effects of low level exposures to hazardous substances continues to mount. The translation of the concept of hormesis into legal regulation of hazardous substances, however, is still largely unexplored. This article provides a tentative and preliminary examination of how existing legal structures can be deployed to acknowledge the reality of hormetic effects and how those structures might best be altered to deal with recognition of hormesis. Such structures can be adapted to an assumption of hormesis, but such recognition may provoke statutory change as well.
The first section of this article contains a model of hormetic effects in the context of legal regulation. The second section consists of a brief review of prevailing legal standards for regulation of hazardous substances, particularly carcinogens. The third and final section addresses how hormetic effects can be integrated into these legal standards. Recognition of hormesis will require administrative changes but not necessarily statutory ones.
Hormesis, as used in this article, means a dose-response relationship in which exposures to substances at some low level are actually beneficial to human health, even though higher level exposures to those substances may be harmful or even extremely hazardous. I will assume that the benefit at low levels is typically the reciprocal of the risk at higher levels (e.g., lower/higher risk of cancer), though this need not be the case. This assumption helps simplify a legal analysis that will be quite complicated enough already.
Scientific research on hormesis has often
displayed its effects as an "inverted U-shaped" or ",-shaped"
curve. In this depiction, the y axis depicts healthiness, with
a hormetic zone of beneficial exposure to a
substance rising to a tipping point where defense mechanisms
are overwhelmed, followed by a zone of declining health,
as exposures increase beyond that point. For
analytical legal purposes, it is more helpful to change the
y curve to represent adverse health consequences, because
are aimed at avoiding adverse health effects, rather
than maximizing health. This yields a "J-shaped" curve,
as depicted in Figure 1.
The curve declines in the hormetic zone, as greater exposures help produce fewer adverse health consequences. After exposures reach the tipping point or trough of the curve, additional exposures cause an increase in net health harms and the curve turns upward. The curve of hormetic effects differs significantly from the traditional dose-effects curves, which often assume a linear dose-response relationship often, but not necessarily, starting from the origin.
The United States has a great variety of laws dealing with health hazards, and a concomitant variety of regulatory standards. These prevailing laws and standards are not well integrated logically but nevertheless provide the structure that must be used for regulation in at least the immediate future. My analysis will focus on the sections of those laws dealing with carcinogenic exposures, because this is the area where recognition of hormetic effects would mean the greatest change from current practice. Traditionally, agencies charged with controlling human exposures to hazardous substances have assumed that the dose-response curve for carcinogens is a linear one, with no safe threshold level of exposure (above zero) to substances that may cause cancer. Under this assumption, reducing exposure levels is always beneficial, which is a presumption of much of today's regulation.
Some of the legal standards for regulation are difficult to adapt to the no-threshold linear risk
assumption for carcinogens. Early environmental statutes,
for example, called for standards to be set with a "margin
of safety," presuming that there is some apparently
safe level of exposure and that standards should be
set somewhere below that level and provide the
necessary margin. Yet the concept of a margin of safety
seems incomprehensible under a theory in which there was
no safe level of exposure. The linear hypothesis allows
no room for such a margin. Other statutory standards
more flexible. Table 1 briefly summarizes the key
language of the regulatory standards for carcinogens
under the federal government's major public health
statutes and the scope of their application.
Table 1: Regulatory Standards for Carcinogens
Clean Air Act
|"ample margin of safety"||Air pollution generally|
|Clean Water Act||"ample margin of safety"||Water pollution generally|
Safe Drinking Water Act (SDWA)
|"adequate margin of safety"||Drinking water quality|
Toxic Substances Control Act
|"unreasonable risk"||All toxic substances|
Food Quality Protection Act (FQPA)
|"reasonable certainty no harm will result"||Pesticides in food|
Resource Conservation & Recovery Act (RCRA)
|"reasonably to protect" health||Hazardous wastes|
Occupational Safety & Health Act (OSHA)
|"reasonably necessary and appropriate"||Occupational exposures|
Federal Insecticide, Fungicide & Rodenticide Act (FIFRA)
|"unreasonable risk"||Pesticides generally|
One of the most significant environmental statutes, the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA) is not included in this list because it generally adopts the standards of the other acts for its cleanup standards. Each of the Acts in Table 1 has its own jurisdiction, though they often overlap. For example, airborne exposures to a substance are covered by the Clean Air Act but may also be addressed by the Toxic Substances Control Act and RCRA. Moreover, some the above regulatory standards are subject to a major qualification in the nature of their implementation. The standards calling for ambient standards to be set with a margin of safety are not self-executing but must be implemented by imposing control requirements on particular sources of pollution. The statutes authorizing those source category standards are typically limited by a feasibility standard. Under these source-specific emission standards, an agency can impose only those control measures that are found to be technologically (and sometimes economically) feasible. Hence, an agency could set a zero level ambient exposure standard for a substance but be unable to require its elimination from the ambient environment, because regulated sources were unable feasibly to reach zero emissions.
The vague regulatory standards set forth above have acquired an elaborating gloss from interpreting judicial opinions over the years. "Unreasonable risk," for example, has been interpreted to require a balancing of the economic costs of a regulation against the health benefits that it would provide.1 It is now regarded as a cost/benefit standard. OSHA's standard of "reasonably necessary and appropriate to provide safe or healthful employment" was interpreted in the context of a benzene regulation to require a finding that a risk was "significant" before regulatory controls were authorized. 13 Most other agencies have concluded that the Benzene decision applies to their organic statutes and likewise quantify risks and regulate only those deemed significant, though interpretations of significant risk vary.3 Significant risk may be a condition on all federal public health regulation.14 As a precondition, though, it is not in itself a statutory standard. Once risk significance is established, regulators must establish rules to conform to their standard (feasibility analysis in the case of OSHA).
Courts have struggled unsuccessfully with the margin of safety standards and have yet to reconcile them with the no-threshold hypothesis of carcinogen risk assessment. An early decision by the D.C. Circuit Court of Appeals suggested that the "margin of safety" could be found in conservative assumptions employed in agency risk assessment procedures.9 A subsequent decision by Judge Bork of the D.C. Circuit suggested that cost could be considered in determining what level of safety is acceptable,10 but that opinion was promptly reversed by the full panel of the court, which held that EPA had to establish safety based only on health considerations.11 Most recently, the court has suggested that the "margin of safety" standard is so vague that it may be unconstitutional.12 That controversial decision is presently under review by the United States Supreme Court. Despite years of analytical effort, neither EPA nor the courts have been able to produce a sound theoretical basis for regulating no-threshold pollutants under a margin of safety standard.14 Congressional amendments have ameliorated but not cured this problem, and the agency continues to muddle through on a case-by-case basis.
Whatever the applicable statutory standard, regulators must develop methods for identifying risk and measuring its magnitude. In applying the standards under all of these authorities, the federal agencies generally, and EPA particularly, have established a strong default process for low level exposures that assumes there is no threshold below which exposures are safe and extrapolates estimated risks at those levels through a linear model. Given the general inability to prove conclusively that any exposure to a carcinogen is safe and the presence of the "one hit" model of carcinogenesis, agencies have precautionarily presumed that all exposures should be eliminated, insofar as possible. This assumption has been at the heart of the agency's problems, at least under margin of safety standards, and it is the assumption that hormesis calls into question.
The integration of hormesis into carcinogen regulation must consider both the regulatory decisionmaking process of the Executive Branch agencies, such as EPA, and the statutory constraints of the relevant enabling legislation. Hormesis cannot be reconciled with prevailing regulatory practice, but such practices can be changed with relative ease. An agency can conduct a rulemaking and alter its practices, though such alteration may be challenged in court. Statutory language is more difficult to change, as it requires legislative action.
Regulatory agencies have to date failed to consider hormetic effects. There is but a lone reference to hormesis in the entire history of the Federal Register and that was in connection with a grant proposal, not a regulatory determination. Agencies have considered something vaguely like hormesis. They had to recognize the potential value of low level exposures when essential nutrients have been shown to produce carcinogenic effects. Selenium, for example, is both a carcinogen and an essential nutrient and the agency adopted a standard that exceeded the recommended dietary allowance for the mineral, though it did not explain its analysis of the low-level beneficial effects of exposure.2
Recently, the Environmental Protection Agency has had a difficult effort regulating chloroform under the Safe Drinking Water Act. Chloroform, an established carcinogen, is a byproduct of chlorine used as a water disinfectant for microbial pathogens. The SDWA provides for enforceable standards called maximum contaminant levels (MCLs) plus aspirational maximum contaminant level goals (MCLGs), which represent the level "at which no known or anticipated adverse effects on the health of persons occur and which allows an adequate margin of safety." EPA assessed the data and rejected its assumption of a linear non-threshold carcinogenic response after considering chloroform's mode of action and concluded that there was a safe threshold. However, when setting its MCLG for chloroform, the agency chose a zero level, holding that more evidence was required for the agency to abandon the use of no-threshold models. An industry group sued, and the court vacated the zero level MCLG as contrary to the best scientific evidence.8 While the chloroform regulation took place in the shadow of obviously health benefits from chlorination, the EPA still failed to consider this sort of hormetic effect in its rulemaking.
EPA has historically been wedded to the linear,
no threshold hypothesis for its risk assessments of
carcinogens. Risks are assessed by identifying a substance
as carcinogenic at very high dose levels in
laboratory testing and then extrapolating risk at lower
levels through a biomathematical model. The method
has employed a variety of conservative assumptions
to ensure that risk is not underestimated,
assumptions including the use of highly susceptible subjects, use
of maximum tolerated doses, assumptions that the
mode of administration does not matter, and
others.5 The theory has been that while such conservative
assessment may cause actual risk to be overstated, that
consequence is untroubling as it errs on the side of public safety.
The precautionary nature of conservative linear risk
extrapolation, though, disappears in the presence of
hormesis. The contrast between current linear models and
the hormetic risk curve is shown in Figure 2.
There is obviously a substantial disparity at low exposure levels between the risks projected by hormetic models (curve a) and those projected by linear models (curve b). It is likewise obvious that there is a region where the linear model projects that greater exposures mean greater risks though the hormetic model projects that greater exposures mean lesser risks.
EPA has historically adhered to the linear model, though this application has been relaxed slightly in recent years. The most recent proposed agency guidelines for carcinogen risk assessment were published in 1996.6 The 1996 guidelines permit consideration of different dose-response models for regulated substances. While they have a default presumption of linearity, the guidelines allow for departures based on scientific evidence. The guidelines make clear that the default assumption is a powerful one, so the agency's assessments rarely use non-linear models.
I believe that the evidence for hormesis is sufficiently strong that the government should alter its default presumption of linear dose-response models. Readers of this journal are well aware of the magnitude of evidence that has accumulated in support of a hormetic dose-response function, evidence that surpasses the data suggesting a linear function at very low levels of exposure. Government policy should be grounded in the best scientific evidence (though that is not always the case in practice).9
B. Statutory Requirements
While such recognition of hormesis requires a change in agency risk assessment methods, those methods are not compelled by statute and could be administratively altered without legislative action. Hormesis does not fit readily into the statutory structure of public health and environmental regulation, but neither did the no-threshold linear hypothesis for carcinogenic risk. The margin of safety standard better suits hormesis than it does linear dose-response extrapolation. One can envision having a margin of safety with a hormetic curve, though its nature is not immediately obvious (as I address in the following section).
Risk assessment employing hormesis fits readily into the other statutory standards as well. Such assessment would alter benefits assessment but not alter the basic structure of the unreasonable risk standard. Similarly, hormesis will alter what regulation is "reasonably" necessary to protect health but in no way contradict that standard.
C. Using Hormesis in Regulation
How could hormesis be incorporated into rulemaking? Where the regulatory standard is one of unreasonable risk, such incorporation is simple. The hormetic model simply calls for a different assessment of health benefits to be compared with regulatory costs. If the proposed exposure reduction were in the zone of hormetic effects, there would be no benefits and hence no regulation. If the proposed exposure reduction were above the tipping point, the curve projected by hormesis would alter the measure of those benefits.
It is commonly assumed that consideration of hormetic effects would be used by regulated entities
to weakening regulations, at least for carcinogens.
The hormetic dose-response model clearly predicts less risk
at low exposure levels than does the linear model.
Hence, use of hormesis could dejustify some regulatory
actions. Yet in other circumstances, the hormetic risk curve
could have the effect of justifying more environmental
regulation. Consider the possible circumstances modeled
in Figure 3.
Suppose that a regulator is considering a reduction in exposure for a substance from E to E*. Under linear models this would yield a health benefit amounting to H H1. Suppose that the reduction involves certain compliance cost, called C. The regulation will be warranted if the H H1 benefit is greater than the C cost, according to whatever principles were chosen by the regulator for valuing risk reductions. Under a hormetic model, the benefit of the regulations would be H H2, which is noticeably greater than H H1. The choice of risk assessment models produces no change in compliance cost C. Under a cost-benefit standard, therefore, when hormetic models are used and exposures exceed the tipping point, regulation should generally be stronger than under traditional linear models. The same analysis applies under a significant risk standard, so long as both the current exposure and the post-regulation exposure are greater than the tipping point. The situation becomes more complex when that is not the case.
In a world where exposures are relatively high (with respect to the hormetic tipping point), the acknowledgement of a hormetic risk curve will often call for more controls. Moreover, despite the symbolic boasts of environmental legislation and regulation, government action rarely forces ambient exposures close to zero. Hence, with marginal, incremental regulation, hormesis would not infrequently help justify additional pollution control rules.
So long as exposures are well above the tipping point, hormesis should tend to justify stricter
regulation under any of the legal standards. The analysis of
lower exposures, under prevailing legal standards, is
more difficult. Consider the hormetic curve framework
of Figure 4.
In this figure Et represents the tipping point, where greater exposures mean greater risk. Eo is the point at which those greater exposures exceed the risk associated with zero exposure, absent any of the hormetic benefits of low level exposures.
As environmental regulations have become increasingly stringent over the years, the salient part of the hormetic curve (below Eo) may come into play. The environmental laws have not addressed themselves to the difficult questions associated with this portion of the curve. The issue raises fundamental questions about the objectives of our environmental public health statutes. Should the Et point be the goal of regulation? If low-level exposures are in fact beneficial, should regulators strive to ensure that populations receive these exposures? Should they set standards so as to compel a minimum level of exposure? Government entities have occasionally acted to ensure public exposure to desirable substances (fluoridation in public water supplies, folic acid in foods), but they have not used environmental laws to do so. When EPA has recognized beneficial consequences from exposure to essential nutrients, such as selenium, it has taken no action to ensure that the population received such exposures. Our laws are written from the perspective of avoiding exposure to harms, not increasing exposure to benefits. It seems unlikely that the environmental laws would be interpreted so as to authorize regulation increasing exposure when ambient levels are below Et.
Assuming that regulators are concerned only with the portion of the curve above Et, tricky questions remain to be answered. Should regulators aim for the level that corresponds with zero exposure (Eo) or should they aim for the bottom of the curve (Et) where human health is maximized? This is a more difficult legal question, as both would involve prevention of adverse health effects. Suppose that the statute in question calls for a "margin of safety." If the objective is set at Et, erring in either direction would be "unsafe." Setting the objective at Eo, however, would give real meaning to the concept of a margin of safety. If the regulatory agency set the standard somewhere around Em the margin of safety standard makes sense. The net adverse effects of the cumulative exposure would be fewer than those with zero exposure to the substance.
Given scientific uncertainty, we can never be sure exactly where Et and Eo are located. We can only estimate their location based upon the best available scientific data. Setting the standard midway, at Em, preserves a margin of safety in both directions it helps ensure that risks remain less than zero exposure and that the regulation does not cause more harm than good by reducing risks below Et. The regulatory goal thus preserves a margin of safety on both ends of the continuum of risk. This intermediate standard also accounts somewhat for uncertainty about the actual levels of exposure for individuals.
The recognition of hormesis will call for reconsideration of other risk assessment issues as well. Source-specific feasibility standards are, by their nature, not health based. Inevitably, they reflect a policy decision that risks will be suffered in order to avoid putting emitters out of business with infeasible control requirements. While recognition of hormetic effects does not intrinsically alter this decision, it could provoke a desirable change in the use of feasibility standards.
The most recently enacted regulatory standard is that of the FQPA. In addition to its new regulatory standard, that law established the use of a "risk cup" for cumulative exposures within categories of pesticides.4 The provision means that EPA must add up exposures from various pathways, including diet, air, and water. The risk cup also groups categories of pesticides together and imposes a maximum exposure permissible for all the constituents of the group. Rather than assessing the risk from individual pesticide applications in isolation, the agency must consider existing exposures before permitting a use. When the risk cup is filled, no further sources of exposure will be permitted, unless some existing sources are eliminated. While the risk cup actually makes little logical sense under linear risk extrapolation models, it is a key step toward regulating with hormesis in mind.
Recognition of hormesis might spark reconsideration of feasibility analysis and source-specific emission standards and replacement with an ambient risk cup. Under an unreasonable risk standard, a product or industry at least theoretically might be banned, if their overall risks exceeded their overall benefits. No such radical regulation has been undertaken, though EPA tried with certain asbestos products and was reversed in court. It should be troubling that public health could be compromised, even significantly, in circumstances where necessary emission controls are not technologically feasible.
Ambient levels of pollution cannot be directly regulated but must be translated into source-specific emission standards. That translation is conducted by government agencies and suffers the flaws associated with bureaucratic action. The strictness of emission controls may have more to do with an industry's effectiveness in lobbying than with scientific or engineering realities. The hard political choices associated with setting source-specific standards may force agencies to fail to set sufficiently strict ones. And the result may be, as under the Clean Air Act's ambient standards, numerous regions that fail to attain the ambient standards for many years. With a binding risk cup, ambient standards can be met and met efficiently, as free market transactions will ensure that the most valuable sources buy up pollution rights, consistently with a general trend toward market-based environmental regulation. Recognition of hormesis is not a prerequisite to using a risk cup for regulation, but the theory of hormesis would promote consideration of such an alternative.
The approach of a broad-based risk cup for environmental contaminants has the benefit of (a) ensuring that ambient exposures are not unduly hazardous and (b) providing for an economically efficient means of reducing exposures. For this desirable outcome to be realized, though, probably requires a rewriting of the environmental laws and a more global approach to regulation. In the interim, the recognition of hormesis could offer an incremental step in the right direction.
This article's discussion of the regulatory
implications of hormesis is necessarily tentative and
preliminary, as those implications have not been ventilated in
the literature nor considered by the government
agencies charged with implementation of environmental
and public health statutes. It is time to begin exploring
the legal implications of recognition of a paradigm
of hormesis. Consideration of hormetic effects
could distinctly enhance the benefits of public health
regulation and might even provide the sort of dramatic
change that could provoke a legislative breakthrough or
other action to restructure environmental regulation and
offer additional benefits.
1. Applegate, JS. The Perils of Unreasonable Risk: Information, Regulatory Policy, and Toxic Substances Control. Columbia Law Review 1991; 91: 261.
2. Poirier, KA & Dourson, ML. Is the Current Risk Assessment Paradigm Used by U.S. EPA and Others Compatible with the Concept of Hormesis? BELLE Newsletter 1999; 8: 22.
3. Cross FB. Beyond Benzene: Establishing Principles for a Significance Threshold on Regulatable Risks of Cancer. Emory Law Journal 1986: 35: 1-57.
4. Cross FB. The Consequences of Consensus: Dangerous Compromises of the Food Quality Protection Act. Washington University Law Quarterly 1997; 75: 1155-1206.
5. Abelson, PH. Commentary: Exaggerated Risks of Chemicals, Journal of Clinical Epidemiology 1995; 48: 173.
6. Environmental Protection Agency. Proposed Guidelines for Carcinogen Risk Assessment Federal Register 1996; 61: 17960.
7. Wagner, WE. The Science Charade in Toxic Risk Regulation. Columbia Law Review 1995: 95: 1613.
8. Chlorine Chemistry Council v. Environmental Protection Agency. 2000; 206 F.3d 1286.
9. Lead Industries Association v. Environmental Protection Agency. 1980; 647 F.2d 1130.
10. Natural Resources Defense Council v. Environmental Protection Agency. 1986; 804 F.2d 710.
11. Natural Resources Defense Council v. Environmental Protection Agency. 1987; 824 F.2d 1146.
12. American Trucking Associations v. Environmental Protection Agency. 1999; 175 F.3d 1027.
13. Industrial Union Department, AFL-CIO v. American Petroleum Institute. 1980; 448 U.S. 607.
14. Goldsmith, R. Nuclear Power Meets the 101st Congress, A "One-Act" Comedy: Regulation of Nuclear Regulatory Commission Licensees Under the Clean Air Act. Virginia Environmental Law Journal 1992; 12: 103.