The Hormetic Dose-Response Model Is More Common than the Threshold Model in Toxicology

Edward J. Calabrese1 and Linda A. Baldwin

Department of Environmental Health Sciences, University of Massachusetts, Amherst, Massachusetts 01003

Received September 12, 2002; accepted October 15, 2002

ABSTRACT

The threshold dose-response model is widely viewed as the most dominant model in toxicology. The present study was designed to test the validity of the threshold model by assessing the responses of doses below the toxicological NOAEL (no observed adverse effect level) in relationship to the control response (i.e., unexposed group). Nearly 1800 doses below the NOAEL, from 664 dose-response relationships derived from a previously published database that satisfied a priori entry criteria, were evaluated. While the threshold model predicts a 1:1 ratio of responses "greater than" to "less than" the control response (i.e., a random distribution), a 2.5:1 ratio (i.e., 1171:464) was observed, reflecting 31% more responses above the control value than expected (p < 0.0001). The mean response (calculated as % control response) of doses below the NOAEL was 115.0% ± 1.5 standard error of the mean (SEM). These findings challenge the long-standing belief in the primacy of the threshold model in toxicology (and other areas of biology involving dose-response relationships) and provide strong support for the hormetic-like biphasic dose-response model characterized by a low-dose stimulation and a high-dose inhibition. These findings may affect numerous aspects of toxicological and biological/biomedical research related to dose-response relationships, including study design, risk assessment, as well as chemotherapeutic strategies.

Key Words: hormesis; biphasic; risk assessment; dose response; linear; threshold.

It is widely accepted in essentially all disciplines dealing with dose-response relationships that the threshold model is the overwhelmingly dominant paradigm (Hayes, 2001Go; Klaassen, 2001Go). This model can affect numerous aspects of research activities including biological model selection, endpoint measured, and study design. It may also affect the interpretation and modeling of dose-response relationships. The threshold model has long been used by regulatory agencies such as the FDA and EPA in establishing acceptable exposures to non-carcinogens.

Despite its clear dominance, the threshold model has been receiving strong challenges over the past decade. Perhaps the most notable challenge has been from the hormetic dose-response model (i.e., the biphasic model characterized by a low-dose stimulation and a high-dose inhibition). This model appears to be quite common in the biomedical and toxicological literature, with responses highly generalizable according to biological model, endpoint measured, and chemical and physical stressor agents tested (Calabrese and Baldwin, 2001aGo,bGo,cGo,dGo, 2002aGo,bGo; Calabrese and Baldwin, in pressGo; Calabrese et al., 1999Go).

The present study, which was designed to assess the capacity of the threshold model to predict responses of doses below apparent toxicological thresholds, demonstrated that not only was the threshold model unable to adequately account for the data, but also that the responses were consistent with the hormetic model.

METHODS

A previously described database created from the published toxicological literature, using rigorous a priori entry criteria (Calabrese and Baldwin, 2001aGo), was employed for the evaluation. The a priori entry criteria required the dose-response relationships to have a lowest observed adverse effect level (LOAEL), a NOAEL, at least two doses below the NOAEL, and a concurrent control. The database is comprised of 664 dose-response relationships from 195 articles and contains 1791 doses below the NOAEL. All responses were converted to values representing a percentage of their respective control response. In the calculation of responses as percentage of control, fractions were rounded to whole numbers (e.g., the category 100% contains values ranging from 99.6% to 100.4%.

RESULTS

For simplification purposes, the ratio of responses "greater than" (or above) the control response to responses "less than" (or below) the control response are referred to as an A/B ratio (above/below).

Doses below NOAEL.
The mean response (calculated as the percentage of control) of the 1791 doses below the NOAEL was 115.0% ± 1.5 (SEM) and the median response was 105.0. Eleven hundred seventy-one responses were greater than the control response, 156 were equal to (i.e., rounded to) the control response, and 464 responses were less than the control response. The A/B ratios were 2.5:1 (1171/464). Table 1Go summarizes the response distribution by experimental model, endpoint, and agent. The A/B ratios were 3.4:1 (557/162) for plant and 2.1:1 (487/229) for vertebrate models. The mean (± SEM) and median responses for plant models were 118.4% ± 1.2 and 110.0%, respectively. The mean (± SEM) and median responses for vertebrate models were 109.0% ± 0.8 and 103.0%, respectively.


View this table:
[in this window]
[in a new window]
 
TABLE 1 Responses (Represented as % Control Response) of 1791 Doses below the NOAEL from 664 Dose-Response Relationships Evaluated by Model, Endpoint, and Agent
 
NOAEL doses.
The mean response (calculated as % control) of the 664 NOAEL doses was 107.1% ± 2.1 (SEM) and the median response was 98.0. Two hundred seventy-three NOAEL responses were greater than the control response, 37 NOAEL responses were equal to the control response, and 354 NOAEL responses were less than the control response. The A/B ratio was 0.8/1 (273/354). Table 2Go summarizes the NOAEL response distribution by experimental model, endpoint, and agent. The A/B ratios were 1.1:1 (157/140) for plant and 0.5:1 (84/163) for vertebrate models. The mean (± SEM) and median NOAEL responses for plant models were 111.5% ± 1.7 and 101.0%, respectively. The mean (± SEM) and median NOAEL responses for vertebrate models were 98.0% ± 1.3 and 95.0%, respectively.


View this table:
[in this window]
[in a new window]
 
TABLE 2 Responses (Calculated as % Control) of 664 NOAEL Doses from 664 Dose-Response Relationships Evaluated by Model, Endpoint, and Agent
 
Effect of NOAEL response on the A/B ratio of responses of doses below the NOAEL.
The A/B response ratio for doses below the NOAEL was markedly affected by the value of the NOAEL (Table 3Go). When the NOAEL response was greater than the control response the A/B response ratio of doses below the NOAEL was 6.7:1 (572/85); when the NOAEL response was equal to the control response the A/B response ratio was 2.5:1 (56/22), and when the NOAEL response was less than the control response the A/B response ratio was 1.5:1 (549/356).


View this table:
[in this window]
[in a new window]
 
TABLE 3 Relationship of the NOAEL Response (Calculated as % Control) to the A/B Ratio of Responses below the NOAEL
 
Effect of position of dose less than the NOAEL on response and A/B ratio.
Table 4Go summarizes the responses and A/B ratios of doses below the NOAEL, based on the position of the dose from the NOAEL (i.e., first dose, second dose, or third and greater doses less than the NOAEL). Although the mean and median responses are similar as the position of the dose relative to the NOAEL decreases, the response decreases (i.e., the highest response was observed in the dose closest to the NOAEL, the next highest in the second dose, and the lowest in the third and greater doses from the NOAEL). In contrast, the A/B ratio was observed to increase as the position of the dose from the NOAEL increased.


View this table:
[in this window]
[in a new window]
 
TABLE 4 Effect of Position of Dose below the NOAEL Dose on Response (Calculated as % Control) and A/B Ratio Position below NOAEL
 
DISCUSSION

The findings indicate that the responses of doses below the NOAEL, where there is a transition between adaptation and toxicity, are nonrandomly distributed, with the strong majority having values greater than the control response. This observation is further supported by the mean response of 115.0% ± 1.5 (SEM) of below NOAEL responses. These findings challenge the long-held belief that treatment-related effects below the NOAEL are unexpected and, if observed, are simply normal variation.

Despite the fact that the data are at variance with predictions of the threshold model, it is important to emphasize that there is considerable overlap in the low-dose stimulatory zone between the threshold and hormetic model predictions. That is, for those doses with responses greater than the control value, it is not possible to distinguish between the two models. This is because the magnitude of the low-dose stimulation of the hormetic model is quite modest (i.e., 30–60% greater than controls at maximum) and consistent with a response often regarded as normal variability depending on the model employed and endpoint measured. Thus, the principal manner by which the two models can be differentiated is the nonrandom distribution of responses greater than the control response as seen in this study. In fact, it is the high degree of overlap between the two models that makes it very difficult to differentiate the threshold and hormetic models when only data from a single dose response are considered.

An assessment of dose-response relationships without evidence of hormesis suggests that this may be due, at least in part, to responses occurring at doses less than the identified NOAEL. Even though the NOAEL, by definition, does not differ in a statistically significant manner from the control, it is still quite possible that doses less than the NOAEL, especially that dose closest to the NOAEL, may display a distinct but lesser degree of toxicity than the NOAEL. If this were true, it would affect the capacity to detect hormetic responses in such dose-response relationships. This concept was evaluated by assessing the responses of doses from dose responses not satisfying our functional definition of hormesis, in which the NOAEL was equal to or less than 95% of the control. In such circumstances it was hypothesized that the "residual (i.e., carryover) toxicity" response would progressively diminish with the lower doses. This suggests that the likelihood of observing a response greater than the control would increase as the dose decreased (at least within the optimal hormetic response zone). As seen in Table 5Go, this is the trend that is observed. There was a 25–30% increase (i.e., 121 vs. 94) in the number of responses equal to or greater than the control response when comparing the responses of the first and second doses less than the NOAEL. This observation suggests the occurrence of residual toxicity in such dose-response relationships. Since nearly 70% of vertebrate toxicology studies assessed here had NOAELs less than the control, it suggests the possibility of residual toxicity in a certain percentage of such dose-response relationships, a factor that could significantly underestimate the frequency of hormesis in the vertebrate toxicological literature. This observation also suggests why toxicologists may have mistakenly dismissed a hormetic hypothesis in favor of a threshold hypothesis.


View this table:
[in this window]
[in a new window]
 
TABLE 5 Evidence Supporting Residual Toxicity in Dose-Response Relationships Not Showing Evidence of Hormesis When the NOAEL Response Is Equal To or Less Than 95% of the Control Response NOAEL
 
An alternative interpretation to the residual toxicity hypothesis could involve the assumption that the population studied is highly heterogeneous, being comprised of a variety of subgroups with differential susceptibility. Under such a scenario, it is possible that one could account for such responses of doses less than the NOAEL based on subgroup-specific responses rather than the residual toxicity hypothesis. Numerous examples exist within the hormesis database in which hormetic responses occurred when the NOAEL was equal to or less than 95% of the control response. This suggests the possibility of population heterogeneity as a factor explaining the dose-response relationship (Calabrese and Baldwin, 2002aGo). In addition, it is possible that differences in study design, especially those relating to dose spacing below the NOAEL, could also be a critical determinant affecting the interpretation of the dose response. Further assessment in this area is necessary to clarify the factors affecting the nature of the dose response in the sub-NOAEL zone.

The threshold model is not only challenged by the nonrandom distribution of responses of doses below the NOAEL, but is further weakened by the data supporting the residual toxicity hypothesis in circumstances in which hormesis was not even observed. These findings raise the critical question of how the field of toxicology could have accepted the threshold model over the past century (Klaassen, 2001Go). This is particularly important since the concept of the dose response is the most central feature in toxicology.

The use of the NOAEL to provide a quasi-estimate of the threshold has both strengths and limitations. Most notably, it provides a statistically based framework by which a consistent comparison across the large number of dose responses may be evaluated. However, the precision by which the NOAEL provides a close approximation of the actual threshold is affected by the quality of the study design, especially with respect to the number of doses used and the nature of the dose spacing. Since the a priori entry criteria for the present study required a LOAEL, NOAEL, at least two doses below the NOAEL, and a concurrent control, the study designs were generally quite robust with respect to the number of doses, such that concerns dealing with adequacy of the NOAEL to provide a reasonable estimate of the threshold were minimized.

The findings are broadly generalizable according to endpoint measured, since the below NOAEL stimulatory response was the most dominant response for all general endpoint response categories and chemical classes (Table 1Go). These findings, along with the average magnitude of stimulation, are consistent with the published literature dealing with hormesis (Calabrese and Baldwin, 2001aGo,bGo,cGo,dGo, 2002bGo; Calabrese and Baldwin, in pressGo; Calabrese et al., 1999Go).

The implications of the findings are striking and challenge the fundamental teachings of the dose response and textbook treatment of this concept. They may affect study designs assessing dose-response relationships, risk assessment procedures for carcinogens and noncarcinogens, strategies for chemotherapeutic applications, and the selection of biological model and endpoints measured.

ACKNOWLEDGMENTS

We thank Ralph Kodell for his thoughtful and constructive review of the manuscript.

NOTES

This study was sponsored by the Air Force Office of Scientific Research, Air Force Material Command, USAF, under grant number F49620-98-1-0091. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the Air Force Office of Scientific Research or the U.S. Government. The U.S. Government is authorized to reproduce and distribute for Governmental purposes, notwithstanding any copyright notation thereon.

1 To whom correspondence should be addressed as above at the Morrill I Science Center, N344. Fax: (413) 545-4692. E-mail: edwardc{at}schoolph.umass.edu. Back

REFERENCES

Calabrese, E. J., and Baldwin, L. A. (2002a). Hormesis and high-risk groups. Reg. Toxicol. Pharmacol. 35, 414–428.[ISI][Medline]

Calabrese, E. J., and Baldwin, L. A. (in press). Hormesis: The dose-response revolution. Annu. Rev. Pharmacol. Toxicol.

Calabrese, E. J., and Baldwin, L. A. (2002b). Radiation hormesis and cancer. Hum. Ecol. Risk Assess. 8, 327–353.[ISI]

Calabrese, E. J., and Baldwin, L. A. (2001a). The frequency of U-shaped dose responses in the toxicological literature. Toxicol. Sci. 62, 330–338.[Abstract/Free Full Text]

Calabrese, E. J., and Baldwin, L. A. (2001b). Hormesis: U-shaped dose responses and their centrality in toxicology. Trends Pharmacol. Sci. 22, 285–291.[ISI][Medline]

Calabrese, E. J., and Baldwin, L. A. (Eds.) (2001c). Scientific foundations of hormesis. Crit. Rev. Toxicol. 31, 351–695[ISI][Medline]

Calabrese, E. J., and Baldwin, L. A. (2001d). U-shaped dose responses in biology, toxicology, and public health. Annu. Rev. Public Health 22, 15–33.[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, 261–281.[ISI][Medline]

Hayes, A. W. (Ed.) (2001). Principles and Methods of Toxicology, 4th ed. Taylor and Francis, Philadelphia.

Klaassen, C. D. (Ed.) (2001). Casarett and Doull’s Toxicology, 6th ed. McGraw-Hill, New York.