ENVIRON International Corporation, Life Sciences, 4350 North Fairfax Drive, Suite 300, Arlington, Virginia 22203
ABSTRACT
The article highlighted in this issue is "The Hormetic Dose-Response Model is More Common than the Threshold Model in Toxicology" by E. J. Calabrese and L. A. Baldwin (pp. 246250).
Calabrese and Baldwin examine 664 dose-response relationships from 195 publications, selected because they provide information for at least two doses between the control and the no observed adverse effect level (NOAEL), and also include a lowest observed adverse effect dose (LOAEL). The database examined is highly diverse, and includes studies in plants, animals (both vertebrates and invertebrates), microbes and protozoa; endpoints ("responses") are equally diverse, and are described in the paper in broad terms only (the authors provide more specificity regarding endpoints in earlier publications, see below). Responses include both adverse and nonadverse effects and (though it is nowhere stated) are presumably identical in nature whether occurring above or below the NOAEL. The various data analyses presented provide evidence that, as the authors conclude, "responses at doses below the NOAEL, ...are nonrandomly distributed, with the strong majority having values greater than the control response" (p. 248). This finding is offered as a challenge to the long-standing view among toxicologists that treatment-related effects below a NOAEL are unexpectedthe so-called threshold model of dose-response relationships. Instead, the authors suggest that the "hormetic-like biphasic dose-response model" more accurately reflects biological reality.
The hormetic model of dose response is usually described as U-shaped (although a "J" seems more accurate, because a "U" suggests that as doses approach zero, responses found at high doses recur at increasingly greater rates, a phenomenon for which there is no evidence). As Calabrese and Baldwin present this phenomenon, the U is inverted when the response represents some normal function (e.g. growth) that is stimulated at low doses (i.e., it exceeds the control), then approaches the control as dose increases, and then becomes inhibited above the NOAEL. A noninverted U describes the hormesis phenomenon when the response represents some dysfunction, such that low doses produce improvement relative to controls, with doses above the NOAEL creating excessive dysfunction (Calabrese and Baldwin 2001). The inverted U is apparently more commonlow-dose stimulation, high-dose inhibition.
The paper highlighted in this volume adds to a long and increasingly discussed body of extraordinarily interesting work that has emerged in recent years, primarily from the University of Massachusetts, the home of Calabrese and Baldwin and of the decade-old Biological Effects of Low Level Exposures (BELLE) initiative. Through conferences, newsletters, and publications BELLE has inspired a fresh look at low-dose phenomena, and the revitalization of hormesis, a fairly ancient notion that has been on the margins of science, has been a major consequence of the BELLE effort (Calabrese, 2001a,b
). Calabrese has not only provided the major share of the modern effort to document the phenomenon, but has also argued forcefully for moving to center stage its further study by toxicologists (and scientists in other disciplines, including biomedical scientists investigating prevention and treatment of diseases). The implications of hormesis for toxicological risk assessment appear, on the surface, to be profound, and suggestions that this may be the case now appear in at least two major toxicology reference works (Eaton and Klaassen, 2001
; Beck et al., 2001
); the second of these can be fairly described as enthusiastic. Indeed, a paper in Risk Analysis attempts to document its highly suggestive title: "Hormesis: a highly generalizable and reproducible phenomenon with important implications for risk assessment" (Calabrese et al., 1999
).
The phenomenon certainly appears real, and the analysis presented in this volume does much to diminish further the usual criticism that responses different from control at sub-NOAEL doses represent only normal variation, and are therefore not truly stimulatory. But its potential importance in toxicology and toxicological risk assessment is not readily grasped, at least by this author. Several questions arise.
First, it cannot be discerned from the highlighted paper, and from those publications I have reviewed, how many of the hormetic-like dose-response relationships that have been uncovered represent the type of toxicological effects and experimental models that are typically used in risk assessments, especially those conducted for regulatory purposes; and, among these, the number that represent the "critical effect" for the chemical under study. Moreover, it would seem important to understand the extent to which the phenomenon is reproducible in different mammalian species, and across endpoints for a specific chemical. Perhaps there are discussions of these matters in some of the literature I have not reviewed, but they are not mentioned in the highlighted paper and are important if the consequences of this work for risk assessment are to be understood. Calabrese et al. (1999) provides several examples of experimental carcinogenesis data that illustrate hormesis, and a few are impressive. The well-known example of dioxin-induced and -inhibited tumorigenesis seems to represent a hormetic-like dose-response relationship when total tumors are combined. If all of the several tumor rate decreases and increases that comprise this "composite" dose response are the result of a common biological mechanism, then perhaps this represents an example of true hormesis; but if mechanisms differ, then perhaps calling this dose-response relationship a hormetic one stretches the definition too far. In any case, most of the many impressive cases of true (apparently single end-point) hormesis in the database reviewed by Calabrese and Baldwin are effects in nonmammalian species. Whether this supports strong generalizations regarding mammalian toxicity is uncertain.
As a second matter, it may be asked whether the threshold model, as used by toxicologists, is truly threatened by the hormetic model. Thus, toxicologists usually define a threshold dose as one below which no adverse effects occur (Eaton and Klaassen, 2001). If hormesis occurs, the effects observed below the NOAEL are (presumably) not adverse to health, indeed may be beneficial; this would not seem to deny the existence of a threshold dose for the high dose toxicity. Furthermore, the typical risk assessment approach involves dividing a NOAEL by uncertainty factors to deal with inter- and intraspecies variability, to arrive at EPAs Toxicity Reference Dose (RfD) or FDAs Acceptable Daily Intake (ADI), measures of daily intake likely to be protective for a large and diverse population. As long as this method protects against the high dose adverse effects, does the existence of hormesis matter?
One response to this question is that it certainly could matter. If the low-dose response is truly protective, and attempts to protect against high-dose toxicity resulted in reducing exposures into that protective range of doses, the extent of benefit (risk reduction) achieved will be less than expected. If the effect is to reduce exposures below the protective range, that protection will be lost. If the cost of such reductions is large, then money and resources will be wasted; more importantly, public health may not be maximally protected. These arguments have much merit, but it would seem that the understanding necessary to derive risk assessment models to deal with them is highly incomplete. If, for example, it is accepted that low doses are adaptive or protective, should we not seek to protect individuals who are least sensitive to these effects, in the same way we seek to protect individuals most sensitive to high-dose toxicity? Are the sources and magnitude of variability identical for both sides of the U? Do we multiply the maximum protective dose by an uncertainty factor to account for variability? How far into the adverse effect zone will doing so bring us?
The problem of taking into account hormesis to arrive at some estimate of the dose that provides an optimum balance between preventing high-dose toxicity and enhancing the low-dose hormetic effects seems not only methodologically difficult, but runs up against the problem of whether those low-dose adaptive effects, as observed experimentally, are truly beneficial to humans.
Thus, a regulatory or public health official is likely to require far more convincing evidence that the low-dose, hormetic (adaptive?) response is truly expected to provide human health benefits than can be discerned from an experimental model. In this context, it might be said that, from a public health perspective, the two arms of the biphasic dose-response relationship are not symmetrical; it will be argued that far greater evidence is needed to support the human benefits of exposures in the low-dose region than is needed to support a concern for high-dose toxicity, perhaps extending to the type of clinical evidence the FDA ordinarily seeks to support claims of disease or symptom prevention for pharmaceuticals and nutritional supplements. The issue that has to be confronted is whether, for example, decreases below control rates of certain tumors seen in animals receiving dioxin provides convincing evidence that the compound offers a protective effect in humans. There is scientific support for treating both arms of the U equally, but it is likely public health advocates and officials will find this difficult to accept.
This last issue brings forward the question of evidence of hormesis from human studies. In a thoughtful and useful recent review, Mundt and May (2001) note that although hormetic-like (biphasic) dose-response relationships have been observed in some epidemiological studies, it is generally not possible to discern whether they reflect true hormesis, or instead the combined effect of several different biological or disease processes that result in a hormetic-like dose-response relationship. This describes, for example, the well-established dose-response relationships between alcohol consumption and all-cause mortality, which appears hormetic-like (J-shaped) but is a true example of hormesis only if the definition of this phenomenon is enlarged to include dose-response relationships that incorporate diverse causes of mortality and disease mechanisms (similar to, but more complex than, the case of dioxin). After describing the serious limits of observational studies for identifying a hormetic response, these same authors provide evidence suggesting that if "hormesis in the analysis of epidemiological data containing mild hormesis [is ignored]...relative risks at exposure levels above the hormetic region are systematically overestimated" (Mundt and May, 2001
, p. 795). If correct, this finding (which would not hold for controlled experimental studies) strengthens the case for the importance of hormesis in understanding the human health risks of chemical exposures. Judging the importance of hormesis in toxicological risk assessment would, of course, greatly benefit from the availability of at least some comparative data from human studies.
Although hormesis may be a widespread phenomenon in biological science, and may have diverse ramifications, it remains unclear whether the toxicology community would regard it as such. As Calabrese has noted, the type of experimental models and designs necessary to detect hormesis are substantially different, and probably more costly to implement, than those now in common use. Their wholesale adoption without several stages of multilaboratory study and refinement is unlikely. Moreover, until a path toward the use of data on hormesis in risk assessment is well defined, and shows that its understanding is critical to public health decision-making, it is hard to envision the diversion of significant institutional investments from the current massive programs of toxicological testing now underway in the United States and abroad. Nevertheless, the challenges to current thinking that the work of Calabrese and his associates represents (and they are to be highly commended for their efforts) should not go unexplored. Review by those institutions that drive risk-based decision-making, perhaps through bodies such as the National Research Council, would seem to be in order.
NOTES
1 For correspondence via fax: (703) 516-2393. E-mail: jrodricks{at}environcorp.com.
REFERENCES
Beck, B. D., Slayton, T. M., Calabrese, E. J., Baldwin, L., and Rudel, R. (2001). The Use of Toxicology in the Regulatory Process. In Principles and Methods of Toxicology (A. W. Hayes, Ed.), 4th ed. pp. 6164. Taylor and Francis, Philadelphia.
Calabrese, E. J. (2001a). The future of hormesis: Where do we go from here? Crit. Rev. Toxicol. 31, 637648.[ISI][Medline]
Calabrese, E. J. (2001b). Overcompensation stimulation: A mechanism for hormetic effects. Crit. Rev. Toxicol. 31, 425470.[ISI][Medline]
Calabrese, E. J., and Baldwin, L. A. (2001). Hormesis: U-shaped dose responses and their centrality in toxicology. Trends Pharmacol. Sci. 22, 285291.[ISI][Medline]
Calabrese, E. J., Baldwin, L. A., and Holland, C. D. (1999). Hormesis: A highly generalizable and reproducible phenomenon with important implications for risk assessment. Risk Anal. 19, 261281.[ISI][Medline]
Eaton, D. L. and Klaassen, C. D. (2001). Principles of Toxicology. In Cassarett and Doulls Toxicology (C. D. Klaassen, Ed.), pp. 1134. McGraw-Hill, New York.
Mundt, K. A. and May, S. (2001). Epidemiological assessment of hormesis in studies with low-level exposure. Human Ecol. Risk Assess. 7, 795809.[ISI]
|