Table of Contents
Hormesis As A Default Parameter in Red Derivation

Edward J. Calabrese, Ph.D. and Linda A. Baldwin

School of Public Health, Environmental Health Sciences, University of Massachusetts, Amherst, MA 01003

Tel: (413) 545-3164, Fax: (413) 545-4692, email:


This essay addresses the question of how an assumption of hormesis could affect the process of hazard assessment, as well as NOAEL (no observable adverse effect level) and Reference Dose (RfD) derivation. The selection of this topic is designed to extend the work of Calabrese (1996) and Gaylor (1998) (see the BELLE website at who presented RfD derivation methods for optimizing theoretical public health benefits due to hormesis. At the onset it is important to note that the US EPA has acknowledged the functional equivalent of the hormesis concept in their Federal Register notification of Proposed Guidelines for Neurotoxicity Risk Assessment (Oct. 4, 1995, p. 52049) which stated that "dose-response curves may exhibit not only monotonic but also U-shaped or inverted U-shaped functions1 (Davis and Svendsgaard, 1990). Such curves are hypothesized to reflect multiple mechanisms of action, the presence of homeostatic mechanisms, and/or activation of compensatory or protective mechanisms." Moreover, this same Federal Register document (p. 52051) articulated a role for hormetic responses in the RfD derivation noting that U-shaped dose-response relationships may be useful in the estimation of NOAEL/LOAEL (lowest observed adverse effect level) or the benchmark dose (BMD).


The principal goal of the hazard assessment process is the estimation of a reliable value for the NOAEL. The reason for the adherence to the derivation of a highly reliable value for the NOAEL is that it is the crucial starting point for the RfD derivation process. This goal would not change whether hormesis is assumed to exist, were reliably demonstrated to exist or were assumed not to exist.

The current methods for estimating the NOAEL employ a testing strategy that first establishes highly toxic responses and then backs off on the dose with respect to duration of exposure until a reasonable approximation of the chronic NOAEL is obtained. That is, in order to obtain a chronic NOAEL the testing process is designed to define a dose-response relationship that encompasses the NOAEL, LOAEL and toxic responses greater than the LOAEL. This process permits the direct estimation of the NOAEL as defined by the highest dosage not significantly different from the control group, its estimation via traditional uncertainty factor (UF) applications (i.e., LOAEL to NOAEL UF application) if no NOAEL were observed experimentally, and/or its derivation via a benchmark dose (BMD)2 process.

While the NOAEL may be derived from any of the above three procedures, the estimation of the NOAEL may be affected by the presence of U-shaped dose-responses in two ways: (1) doses within the hormetic range could influence the BMD derivation process or (2) a dose within the hormetic range could be selected as the NOAEL. In the case of the BMD derivation, the presence of apparent hormetic responses would flatten the model-based dose-response relationship resulting in a higher exposure to achieve the BMD05/10 response. In the case of the traditional NOAEL methodology (i.e., where the highest dose not different from the control is designated the NOAEL), a dose with a value less than the Zero Equivalent Point (ZEP) (i.e., value equal to the control value) would be the NOAEL. For example, Figure 1 depicts a stylized dose-response relationship in which a U-shaped dose-response is illustrated. In this case, the highest dose represents a significant toxic response and is deemed the LOAEL. The next highest dose displays a response lower than the control. By definition (i.e., the highest dose not significantly different with respect to an increased adverse response incidence), this dose would serve as the highest NOAEL for RfD derivation purposes and yet would be in the hormetic zone. Consequently, in these two cases at least, the concept of NOAEL as determined by the BMD is affected by the presence of hormetic-like responses (U-shaped data) while being part of the hormetic response curve (i.e., traditional NOAEL derivation process), respectively. However, in numerous and perhaps the overwhelming majority of cases where the number of doses is limited and emphasizes the high dose response, the dose-response will not have U-shaped data and therefore will not be influenced by an hormetic response during BMD derivation. Likewise, a dose greater than the ZEP value is often selected as the NOAEL and therefore would not be in the hormetic zone. Nonetheless, it is important to recognize the relationship of hormesis to current NOAEL derivation processes since it may impact methods for RfD derivation that attempt to optimize population-based hormetic responses.

The process of starting at the upper dose range in short-term toxicological studies and gradually defining a chronic NOAEL is clearly designed to assess hazard potential across a broad spectrum of dosages and exposure durations. This traditional approach to defining the NOAEL has been built into the vast majority of hazardous assessment schemes. The high to low dose testing strategy process is not the only way that the NOAEL may be derived. In fact, one could start at the opposite end of the dose spectrum and work up to higher doses until the NOAEL and LOAEL are estimated. Therefore, it would be toxicologically possible to arrive at the NOAEL from either side of the dose testing spectrum (i.e., from very low or very high dosages). However, starting at the very low dosage end to proceed to the NOAEL is not a particularly attractive notion since it is in society's best interest to define the major potential threats an agent could present rather than the more modest and theoretical benefits of a rather limited nature (i.e., hormetic responses) as well as the NOAEL. These responses are determined in the domain of the traditional hazard assessment testing protocol. While hormesis-oriented testing schemes which ensure the incorporation of a wide dose range and large number of doses could be used to derive an even more reliable NOAEL than with high-to-low dose testing, the hazard assessment process demands that the full range toxic potential be reasonably defined. This clearly favors the traditional testing protocol where dose number is limited and high doses emphasized. Thus, hormesis is not expected to play a role in altering the principal goals of defining toxic potential.

As reasonable as the above argument for high-to-low end hazard testing would appear, it is not without its biases or limitations. By deriving a NOAEL from the traditional high-to-low testing strategy it is more likely to miss possible hormetic responses. This would have implications for the derivation of a BMD since some precision would be lost in defining the low dose zone. However, regardless of how the NOAEL is derived, the concept of hormesis can be a principal component in the RfD derivation.


Assuming that hormesis exists and follows the quantitative scheme as presented in the Introduction of this issue of the BELLE Newsletter, the critical issue for hormesis is whether it should become a DEFAULT parameter or be required to be demonstrated for each agent. If it is a DEFAULT parameter then the current hazard assessment process could continue as is. One would need to correct the NOAEL value to equal the ZEP dose by a model-based procedure, thereby obtaining an adjusted ZEP-based NOAEL, and then to employ an hormesis optimization scheme to facilitate the RfD derivation (Calabrese, 1996; Gaylor, 1998).


If hormesis were required to be "proven" for each agent, this most likely would require a hazard assessment to obtain the NOAEL and then a separate set of experiments to explore the NOAEL-subNOAEL dose-response area (i.e., hormetic dose-response zone). This would place such an excessive demand on the sponsoring group that the specific integration of hormesis in any RfD derivation would be relegated to very, very rare exceptions, much like raising the high jump bar two feet above the world record and telling other high jumpers this is the new standard. Consider the massive experimental effort launched by industry concerning the capacity of selected hydrocarbons in petroleum and other products to cause renal cancer in male rats. After many millions of dollars of testing it was finally judged by EPA that the observed phenomenon (i.e., enhanced -2µ-globulin synthesis in male rat liver and the subsequent induced renal pathology, (Swenberg et al., 1992)) was a species-specific, gender-specific response with very limited relevance to humans. Requiring that hormesis be established for each agent, both sexes of each species tested, and possibly the optimum hormetic response, would likewise create an overwhelming financial burden and would surely be the deathknell of any practical significance for hormesis in the regulatory world. While it may appear rational to say that hormesis would be considered on a case-by-case basis, the implications of this statement are clearly apparent. It is for this reason that hormesis needs to be considered within the context of incorporation into the risk assessment process as a DEFAULT assumption.


If hormesis were considered as a DEFAULT parameter in the RfD process what type of evidence would be required for such a consensus toxicological judgment? How can decisions be made about whether to accept hormesis as a DEFAULT assumption? An EPA-supported major research effort on this topic, while possible, is unlikely and unwise since it would concentrate limited resources on only one issue. In addition, many hundreds of studies in the peer-reviewed literature already exist that are extremely pertinent to the issue of hormesis as a biological hypothesis. Such data which were derived without consideration of this rather challenging public policy (i.e., hormesis as a DEFAULT value in risk assessment) could be seen as essentially unbiased data.

In order to establish a decision making framework for assessing whether hormesis should achieve DEFAULT status the following considerations are offered. The concept of DEFAULT suggests a highly generalizable phenomenon with respect to chemical class, biological model and endpoint of concern which it must be reliably and practically linked to risk assessment processes. Hormesis therefore would need to be present for broad classes of inorganic and organic contaminations including those containing agents subjected to regulatory action. In fact, its relevance would be heightened considerably if it were shown for agents currently regulated by federal and state environmental and food and drug safety agencies. As with the case of contaminants, the capacity for generalization across species is critical. In this case, the question posed is whether hormesis is a general biological phenomenon or a species- or strain-specific response that can be generalized on a limited basis. It would also be important to establish if hormesis is likely to occur in humans even though establishing cause and effect relationships in humans for low dose environmental exposures is often beyond the power of even extraordinarily large epidemiological studies. The third issue is that of endpoint. Traditional environmental regulatory actions typically emphasize preventing organ-specific toxicity, reducing biomarkers of exposure such as blood lead levels and serum cholinesterase activity, and preventing teratogenic and other reproductive effects, as well as tumor formation. In the case of hormesis, its impact is not only on the above noted harmful effects and biomarker responses, but also on its capacity for enhancing more positive outcomes such as increased longevity and improved performance. Thus, for hormesis to be considered for DEFAULT status it must score sufficiently high in the three areas of generalizability by chemical class, biological model and biological endpoint.

The recently developed chemical hormesis database (Calabrese and Baldwin, 1997, 1997a) evaluates toxicological/pharmacological investigations for their capacity to demonstrate evidence of hormesis based on study design features, response criteria, statistical power, and reproducibility within the context of consistency with the - (or U-shaped) curve. While it is anticipated that a detailed description of the hormesis database will be presented in a future BELLE Newsletter, work to date clearly supports the conclusion that hormesis is highly generalizable across the three critical areas and that its relationship to the NOAEL is a fixed reference point (e.g., a factor 4-5 below the NOAEL), thereby making its relationship to the risk assessment process very practical (see Calabrese, 1996; Gaylor, 1998). In addition, many toxic substances subject to regulatory action have been shown to induce hormetic responses in various species for a wide variety of endpoints, although the endpoint of concern from a public health perspective may not have been the object of the investigation in which the hormetic response was shown. Nonetheless, there are other instances in which hormesis has been shown in which the intent was to directly confront the endpoint that was the object of public health concern (Calabrese and Baldwin, 1998). Based on the above information it would appear that the concept of hormesis is a good candidate for further consideration for the status of DEFAULT parameter.

What would it mean if hormesis were employed as a DEFAULT assumption for the RfD process? If hormesis received unrestricted DEFAULT status the RfD process would assume that hormesis exists for all agents, animal models and endpoints. It would utilize an hormesis optimizing procedure that would attempt to estimate the dosage that strikes the best balance between preventing the agent-induced adverse effects and enhancing hormetic/adaptive responses within the highly heterogeneous human population. As with other DEFAULT assumptions it would place the burden on the process to prove that hormesis did not exist in order to deviate from the DEFAULT hormetic assumption. However, it is possible that DEFAULT status could be acknowledged in a restrictive manner, for example, applying only to certain classes of chemicals or various biological endpoints. While the impact of hormesis would be significant for the non-carcinogen risk assessment process in terms of final RfD values (Calabrese, 1996; Gaylor, 1998), its capacity to affect the carcinogen risk assessment process would be enormous given the current assumptions and practices in this area. However, while the argument offered today focuses only on the issue of non-carcinogen risk assessment, there is no theoretical or practical reason why this concept cannot (and will not) be applied to the issue of cancer risk assessment (Sielken and Stevenson, 1998).


1 The terms hormesis and U-shaped dose-response relationship have been used interchangeably in the toxicological literature. However, as non-monotonic dose-responses become the object of more careful mechanistic evaluation, it is anticipated that a more uniform and mechanistically-based terminology will be developed (Calabrese and Baldwin, 1998a).

2 The BMD has been defined by EPA as a lower confidence limit on the effective dose associated with some defined level of effect [e.g., a 5 percent (BMD05) or 10 percent (BMD10) increase in response for a particular effect]. Since the biostatistical model is only employed to interpolate within the dose range of the study, no assumptions about the presence (or absence) of a threshold are required. This suggests that any model fitting the data well may be able to offer a reasonable estimate of the BMD.


Calabrese, E.J. 1996. Expanding the RfD concept to incorporate and optimize beneficial effects while preventing toxic responses from non-essential toxicants. BELLE Newsletter, 4(2):1-10.

Calabrese, E.J. and Baldwin, L.A. 1997. The dose determines the stimulation (and poison): development of a chemical hormesis database. Int. J. Toxicol. 15:545-559.

Calabrese, E.J. and Baldwin, L.A. 1997a. A quantitatively-based methodology for the evaluation of chemical hormesis. Hum. Ecol. Risk Assess. 3:545-554.

Calabrese, E.J. and Baldwin, L.A. 1998. Can the concept of hormesis be generalized to carcinogensis? Submitted.

Calabrese, E.J. and Baldwin, L.A. 1998a. A general classification of U-shaped dose-response relationships in toxicology and their mechanistic foundations. Submitted.

Davis, J.M. and Svendsgaard, D.J. 1990. U-shaped dose-response curves: their occurrence and implication for risk assessment. J. Toxicol. Environ. Health 30:71-83.

Gaylor, D. 1998. Safety assessment with hormetic effects. BELLE Newsletter 6(3):6-8.

Sielken, R.L., Jr. and Stevenson, D.E. 1998. Some implications for quantitative risk assessment if hormesis exists. BELLE Newsletter 6(3):13-17.

Swenberg, J.A., Dietrich, D.R., McClain, R.M., and Cohen, S.M. 1992. Species-specific mechanisms of carcinogenesis. In: Mechanisms of Carcinogenesis in Risk Assessment (H. Vainio et al, eds.), IARC, Lyon, France, pp.477-500.

US EPA. 1995. Proposed Guidelines for Neurotoxicity Risk Assessment. Federal Register (Oct. 4) 60:52032-52056.