Table of Contents
Hormesis - Implications for Risk Assessment Caloric Intake (Body Weight) as an Exemplar

Angelo Turturro, Ph.D., D.A.B.T.a,1, Bruce Hass, Ph.D.b, Ronald W. Hart, Ph.D.c

a Division of Biometry and Risk Assessment
b Division of Genetic Toxicology
c Office of the Director

National Center for Toxicological Research, 3900 N.C.T.R. Road, Jefferson, AR 72079

1 Corresponding Author Tel: 870-543-7581; Fax: 870-543-7332

Hormesis can be considered a parameter which has a non-monotonic relationship with some endpoint. Since caloric intake is such a parameter, and the impact of this parameter on risk assessment has been fairly well characterized, it can provide clues as to how to integrate the information from a hormetic parameter into risk assessments for toxicants. Based on the work with caloric intake, one could: a)define a biomarker for hormetic effect; b) integrate specific information on when in the animals lifespan the parameter is active to influence parameters such as survival; c) evaluate component effects of the overall hormetic response; and d) address the consequences of a non-monotonic relationship between the hormetic parameter and endpoints critical for risk assessment.

These impacts on risk assessments have been best characterized for chronic tests, but are also true for short-term tests. A priority is the characterization of the dose-response curves for hormetic parameters. This quantification will be critical in utilizing them in risk assessment. With this information, one could better quantitatively address the changes one expects to result from the hormetic parameter, and limit the uncertainty and variability which occurs in toxicity testing.


Hormesis, in a simplified description the observation that small doses of a toxicant can have salutary effects, has been characterized for some time (1-4). One way to address this phenomenon is to consider agents that induce hormesis as having non-monotonic dose-response curves. Traditionally, the term is applied to compounds in which there is a positive effect at low-dose. However a negative effect at low dose followed by a positive effect is also possible. The implications of this phenomenon for risk assessment has been the object of some discussion (5-8), mostly of a theoretical nature. Recently, it has been shown that caloric intake, and its concomitant impact on body weight (BW), can have a significant impact on the assessment of risk, especially when the results of chronic tests are being evaluated (9-13). Since caloric intake appears to be hormetic for survival (14), the impact caloric intake has had on risk assessment can be used as a model for the impact of other hormetic processes on risk assessment.


The best characterized effect of caloric intake on risk assessment is on the interpretation of long-term toxicity tests. In vivo chronic bioassays in a variety of animal species are used as part of the process of estimating risk in humans (15-17). Because of practical considerations, many chronic tests are done in rodent species (17), such as mouse and rat, with some of the best examples the 500 or so compounds evaluated by the National Toxicology Program (NTP). For directly studying the effects of hormesis on critical parameters such as growth, survival and chronic disease, the NTP studies are of limited value because there are usually only three doses, and the lowest dose, often one-half the maximum tolerated dose (MTD), is usually quite high. However, the NTP has attempted to minimize variability in environmental conditions surrounding chronic bioassays and variations in chemical dosing (18), reducing the impact of variation in many environmental parameters, but almost never controlling the one significant environmental variable, caloric intake (and its concomitant effects on BW). Thus, the NTP dataset, supplemented by information from a long-term collaboration between the National Center for Toxicological Research and the National Institute of Aging on the modulation of risk by modulation of diet (19), has provided a unique opportunity to derive detailed information on the impact of a hormetic parameter on long-term toxicity tests and the consequences for their interpretation.


1) Biomarker

One of the first issues that arose was the derivation of a convenient biomarker for the effect of caloric intake. Caloric intake is not often measured in chronic tests, and when it is it is often problematic. Direct measurement of food consumption is complicated by spillage, coprophagy, feeder design, cage type, and behavioral factors. Few of these parameters are easily measured for individual animals. When accurately measured, however, food consumption is found to be quite variable from week-to-week (20), and the amount of food consumed can vary as a function of animal activity (21), food absorption, adequate access to food dispensers, types of diets (22), and environmental conditions (20,23). Estimates of caloric consumption based upon the disappearance of the amount of food have faced similar difficulties. Fortunately, for the effect of caloric intake on critical parameters in toxicity tests such as survival, body weight (BW) at various times-on-test is a good biomarker. In addition, this biomarker appears to be proximal to the effect of caloric intake to its effect. This is suggested by four lines of evidence:

  1. the best correlation of beneficial health effects with caloric restriction (CR) is with the BW changes that CR induces at certain ages (14,24,25);
  2. the demonstration that altering BW, but not caloric consumption, by altering diet composition, results in increased mortality when BW is increased (25);
  3. the demonstration of relationships between survival and BW at specific times-on-test in chronic bioassays for mouse (11) and rat (12);
  4. treatment by compounds which lower BW but do not decrease food consumption seem to be correlated with BW changes and not food consumption per se. An example is derived from the NTP bioassay of amphetamine (26) (Table 1), in which changes in pathological endpoints such as liver tumors in mice and adrenal pheochromatomas in rats are predicted from BW changes (since food consumption is unchanged or slightly increased by this agent).

Since many of the endpoints which are associated with hormesis are inconvenient to measure (such as lifespan), or difficult to directly use in risk assessment calculations, a biomarker which is relatively easy to measure and simple to use would be useful. For instance, survival (percentage of a cohort) at 26 months of age (often the age of animals after a long-term chronic test) may be quicker to measure than lifespan, have more information in the literature about it in different experiments, as well as provide a parameter whose use is fairly well defined in risk assessment.

For risk assessment, these markers can serve as easily manipulable surrogates that can be used in the risk analysis.

Additionally, use of the biomarker may result in some new ways of looking at the risk evaluation. For instance, use of the BW biomarker results in a prediction that any procedure which effects BW can effect survival and pathological endpoints, a prediction which has been extremely fruitful in interpretation of survival information in toxicity test, aging, etc.

2. Time (Age) is an Important Variable

Another suggestion derived from the work evaluating caloric intake is that the time (age) at which a phenomenon occurs can be an important parameter. For instance, although average BW is of very limited value in predicting survival, the BW at a particular time-on test (age) can be very good (11,14). For BW biomarkers this appears to occur because the different mechanisms which make up the multiple response of an organism to altered BW's can be prominent at different stages in life. For instance, elevated BW's at relatively young ages (approximately 3-4 months of age) seem to stimulate disease in rapidly dividing and growing compartments, such as blood dyscrasias (e.g., lymphoma), while elevated BW's at later ages seem to be associated with disease in organs which proliferate relatively slowly but follow the steady increase in BW seen in later life (13-14 months of age), such as liver (11,12). Equally, the effect of an agent may be totally different at different times. Caloric restriction, which extends life and function when started at 6 weeks of age, can inhibit brain development in growing, developing animals.

Other hormetic biomarkers may be similar. For instance, an agent which improves growth of a particular part of a plant at one developmental stage influence another at another stage. Additionally, a compound could stimulate growth during the growth phase of an animal or plant while becoming ineffective or inhibitory if given later in life.

This has direct impact on risk assessment using this agent since this time variable should be taken into account to interpret the study. In addition, the final analysis should suggest that inhibition of risk will only occur within a “window” in the development life of the organism.

3. It Can Be Useful to Evaluate Effects That Contribute to an Overall Hormetic Response.

The BW markers are useful in predicting survival, but become more useful when they are used to predict the onset and progress of chronic diseases, which are components of the survival. In a similar fashion, other hormetic markers become much more useful when they are used to predict the individual components that make up the hormetic endpoint. For instance, when evaluating survival, the incidences of chronic diseases become a series of multiple tests of the hormesis, as well as point to potential mechanisms of action. For example the impact of BW on liver tumors, coupled with data on the relationship of cellular proliferation/apoptosis and BW, has given insight into the mechanisms by which BW impacts on liver tumors and other chronic diseases (27). This mechanism of action has contributed to the ability to extrapolate the effects of toxicants on liver tumors in mice to man, as well as quantitative adjustments for this extrapolation.

Similarly, hormetic effects on a tumor may complement information on the effects of hormesis on survival in order to investigate how an agent could induce inhibition of tumors (or other endpoints).

4. Non-Monotonic Dose-Response Curves

As an hormetic parameter, BW at various times-on-test can provide mathematical models for addressing hormesis. Intuitively, too light BW's can indicate starvation and early death, while we have shown the effect of increasing BW on survival (14). To quantitate this, we can use the relationship of BW at 2 months-on-test (BW2) to survival. This is illustrated in Figure 1, which is an updated version of work previously published for male B6C3F1 mice (14). The data which are the bases of the curve post 100% survival (at 24 months-on-test) is derived from previous work (11). The data which are bases of the curve with BW's less than those required for 100% survival is derived from subchronic (3-month) studies reported by the NTP in which survival in the mice is measured at 3 months-on-test (see table footnote). Thus, this curve can be thought to underestimate the detrimental effects of low BW2 (i.e., it is probable that the BW2 in this range are the upper reaches of what is compatible with long-term survival in these mice). It can be seen that the relationship is non-monotonic.

Based on the non-monotonic relationship, a number of issues relevant to risk assessment have been characterized.

1) The control survival for any drug treatment is a function of BW2.

Therefore, the proper background survival incidence as a control needs to be adjusted by the BW2. Simple comparison of control to treated is inadequate since the drug treatment may result in changes in BW2, which impacts on survival (see 14), unless dietary control (DC) is instituted (10,11), which provides an appropriate untreated control for the treated animals.

Similarly, the biomarker for the hormetic effect needs either to be controlled (such as DC does for BW biomarkers) or a mathematical adjustment needs to be made which corrects for the true background incidence.

2) Any toxic response could be a non-linear function of the background incidence.

The response to a toxicant can be a non-linear function of the BW, as indicated for liver tumor response (12). Therefore, not only the control, but also any response in dosed animals could be a function of the BW. This issue is very difficult to address without DC, which is the recommended approach to addressing it.

Similarly, it will probably be hard to address this issue, modification of a toxic response by the interaction of the toxic response and background incidence without some form of control of the hormesis biomarker for control and treated group.

3) A significant portion of the variability in critical parameters, such as survival, may be as a result of variability in the hormesis biomarker.

For the BW biomarkers, up to 95% of the variability in critical parameters such as tumor incidences may be a result of variability in different BW's at different times-on-test. Thus much of the “biologic variability” which complicates test interpretation may be related to the hormetic biomarker. Addressing this, using the hormetic biomarker as a guide, may reduce the uncertainty associated with a test result for risk assessment, and improves its utility, especially if the test value for the biomarker falls into the observable range.

4) Evaluation of the biomarker allows use of historical controls to evaluate tests.

One aspect of risk assessment is to evaluate the significance of safety tests that are conducted. The present variability in chronic tests is so great that it is presently almost impossible to obtain a result which is not in the historical range for a number of parameters. With DC, one can create a BW-stratified approach to test evaluation which limits the historical range so that it again becomes useful.

Similarly, a hormetic biomarker may be associated with a wide historical range, and may explain much of the variance in that range. If so, evaluation of the impact of the variable may allow better interpretation of tests which are problematic.

5) Stratification may be possible on an individual animal basis.

Evaluating BW markers for individual animals, as opposed to mean values in an experiment of 50 animals, as noted above, has been shown to be effective in predicting the risk of certain chronic diseases (28). Thus, the risk of animals developing spontaneous disease, and the potential to develop tumors after toxic insult, can be estimated. This is useful in evaluating the effects of BW distributions in experiments (12). For instance, based on Figure 2, a female mouse with a BW12 of 60 grams has better than a 70% chance of developing a liver tumor spontaneously so if the study arm with treated animals have animals in that weight range the significance of a liver tumor is problematic.

Similarly, an hormetic biomarker may indicate very poor or very good survival in a particular individual, so the effect of treatment can be blunted or artificially enhanced by the action of the uncontrolled hormetic parameter. This complicates study interpretation, and again indicates some form of control of the hormetic biomarker.

6. Low-Dose Extrapolation Becomes Complicated.

The impact of the BW biomarker complicates the high- to low-dose extrapolation of chemical effects. One reason is a consequence of the often-seen loss of BW that occurs with chemical exposure which (within a range) result in improved parameters such as survival. This effect can result in treated animals with improved survival compared to controls complicating interpretation, as well as mitigation of the adverse toxic effects seen at the higher doses. Since these changes are occurring in the observable range of doses, these modified toxic responses are used in setting the initial point for the low-dose extrapolation procedure used, and thus BW effects will directly impact the extrapolation. This is especially true for low-dose extrapolation procedures which use the information from all the doses, e.g., a technique which uses a certain percentile response derived from the entire dose-response curve. The situation is further complicated by the non-linear aspect of the dose-response curve, e.g., weight loss when the animals are in the higher weight ranges generally has more of an impact than an equal weight loss at average weights. This non-linear aspect has another problem associated with it in that either a weight loss from initially low BW's, or a dramatic weight loss, can result in animals falling in a BW range where they are almost refractory to toxic insult. Similarly, a non-monotonic impact of an hormetic parameter can complicate the interpretation of a toxic effect, especially if the hormetic biomarker is tied to the toxic effect as BW changes often are.


The impact of BW is less well characterized in these assays than in chronic tests. Acute toxicity (29,30) and subchronic toxicity (31) is inhibited, sometimes extinguished, in very low BW animals. This suggests that the endpoints are sensitive to caloric intake, but the actual dose-response curves are not known. Similarly, hormetic parameters may influence the results of short-term toxicity tests, effecting a number of aspects of risk assessment, especially when short-term tests are used as adjuncts to long-term tests (set doses, look at pharmacokinetics, etc.). Similar to BW, the influence of hormetic parameters should be controlled or, at a minimum, evaluated in order to properly interpret the results of short-term toxicity tests.


As an hormetic parameter which is known to influence risk assessments, caloric intake can be useful in defining aspects of hormetic parameters which influence risk assessments in myriad ways. Characterizing the dose-response curves of these hormetic parameters will be critical in utilizing them in risk assessment. With this information, one could better quantitatively address the changes one expects to result from the hormetic parameter, and limit the uncertainty and variability which occurs in toxicity testing.


1. Calabrese, E., McCarthy, M., and Kenyon, E., The occurrence of chemically induced hormesis, Health Physics, 52, 531, 1987.

2. Sagan, L., What is hormesis?, Health Physics, 52, 521, 1987.

3. Fritz-Niggli, H. 100 years of radiobiology: Implications for biomedicine and future perspectives. Experientia 51:652-664, 1995.

4. Roth, J., Schweitzer, P., and Guckel, C. Grundlagen des strahlenschutzes. Schweiz. Med. Wochenschr., 126:1157-1171, 1996.

5. Lukey, T. Radiation hormesis in cancer mortality. Chin. Med. J. Engl. 107:627-630, 1994.

6. Van Wyngaarden, K. and Pauwels, E. Hormesis: Are low doses of ionizing radiation harmful or beneficial. Eur. J. Nuc. Med 22:399-401, 1995

7. Kitchin, K.T., and Brown, J. Dose-response relationship for rat liver DNA damage caused by 1,2-dimethylhydrazine. Toxicol. 114:113-124, 1996.

8. Hart, R.W. and Turturro, A. Is a new cancer risk assessment paradigm needed? Belle Newsletter 5:14-18, 1996.

9. Turturro, A., P. Duffy and Hart, R.. Modulation of toxicity by diet and dietary macronutrient restriction. Mutation Res., 295, 151-164, 1993.

10. Turturro, A., Duffy, P. and Hart, R. The effect of caloric modulation on toxicity studies. In Dietary Restriction: Implications for the Design and Interpretation of Toxicity and Carcinogenicity Studies (R. Hart, D. Neuman, and R. Robertson, Eds.) ILSI Press, Wash., DC. 1995,pp. 79-98.

11. Turturro, A., Duffy, P., Hart, R., and Allaben, W. Rationale for the use of dietary control in toxicity studies - B6C3F1 mouse. Toxicologic Pathology 24:769-775, 1996

12. Turturro, A., Duffy, P., Hart, R., and Allaben, W. Body weight impact on spontaneous and agent-induced diseases in chronic bioassays. Inter. J. Toxicol. In press.

13. Turturro, A., Leakey, J., Allaben, W., and Hart, R. Response to Michael Festing's “Fat Rats”. Nature 389:326, 1997.

14. Turturro, A. and Hart, R. Modulation of toxicity by diet: Implications for response at low-level exposures. In: Biological Effects of Low Level Exposures: Dose-Response Relationships (Edward Calabrese, Ed.) Lewis Publishers, Chelsea, MI 48118, Chapter 9, pp. 143-152, 1994.

15. FDA, (Food and Drug Administration) Toxicological principles for the safety assessment of direct food additives and color additives used in food. USFDA, Center for Food Safety and Applied Nutrition, Wash., D.C. (1986).

16. FDA, (Food and Drug Administration) Advisory Committee for Protocols for Safety Evaluation, Panel on Carcinogenesis: Report on cancer testing in the safety of food additives and pesticides. Toxicol. Appl. Pharmacol. 20:419-438 (1993).

17. Interagency Staff Group. Chemical carcinogens: A review of the science and associated principles. Environ. Health Perspect. 67:201-282 (1986).

18. NTP, (National Toxicology Program). Specifications for the Conduct of Studies to Evaluate the Toxic and Carcinogenic Potential of Chemical, Biological, and Physical Agents in Laboratory Animals for the National Toxicology Program, National Institutes for Environmental Health Sciences, Research Triangle Park, NC, August, (1992).

19. Lewis, S., Leard, B.L., Turturro, A., and Hart, R. Long-term housing of rodents under specific pathogen-free barrier conditions. In: (B.P. Yu, Ed.). Methods in Aging Research, Section C: Vertebrate Models in Aging Research, CRC Press, Boca Raton, FL, in press.

20. Witt, W., Sheldon, W., and Thurman, J. Pathological endpoints in dietary restricted rodents - Fischer 344 rats and B6C3F1 mice. In: Biological Effects of Dietary Restriction, (Fishbein, L., Ed.) Springer-Verlag, New York, pp. 73-86, (1991).

21. Holloszy, J. Exercise increases average longevity of female rats despite increased food intake and no growth retardation. J. Gerontol. 48:B97-100, (1993).

22. Su, W., and Jones, P. Dietary fatty acid composition influences energy accretion in rats. J. Nutr. 123:2109-2114 (1993).

23. Keenan, K., Smith, P., Ballam, G., Soper, K., and Bokelman, D. The effect of diet and dietary optimization (caloric restriction) on survival in carcinogenicity studies: an industry viewpoint. In: The Carcinogenicity Debate. (J. McAuslane, C. Lumley and S. Walker, Eds.) Quay Publishing, London, pp. 77-102 (1992).

24. Turturro, A. Blank, K., Murasko, D., and Hart, R.. Mechanisms of caloric restriction effecting aging and disease. Ann. N. Y. Acad. Sci. 719:159-170, 1994.

25. Turturro A. and Hart, R. Dietary alteration in the rate of cancer and aging. Experimental Gerontology 27:583-592, 1992.

26. NTP, (National Toxicology Program) Toxicology and carcinogenesis studies of d+-amphetamine sulfate in F344/N rats and B6C3F1 mice. (Feed Studies) Technical Report Series Number 387, Dept. of Health and Human Services, NIH, NIEHS (1991).

27. James-Gaylor, J., Muskhelishvili, L., Gaylor, D., Turturro, A., and Hart, R. Upregulation of apoptosis with dietary restriction: Implications for carcinogenesis and aging. Environ. Health Perspect., in press.

28. Seilkop, S. The effect of body weight on tumor incidence and carcinogenicity testing in B6C3F1 mice and F-344 rats. Fund. Appl. Toxicol. 24:247-259 (1995).

29. Duffy, P., Feuers, R., Pipkin, J. Berg, T.F., Divine, B., Leakey, J., Turturro, A., and Hart, R. The effect of caloric modulation and aging on the physiological response of rodents to drug toxicity. In Dietary Restriction: Implications for the Design and Interpretation of Toxicity and Carcinogenicity Studies (R. Hart, D. Neuman, and R. Robertson, Eds.) ILSI Press, Wash., DC. ppg. 127-140 (1995).

30. Berg, T., Breen, P., Feuers, R., and Hart, R. Acute toxicity of Ganciclovir: Effect of dietary restriction and chronobiology. Food Chem. Toxicol. 32:45-50 (1994).

31. Keenan, K., Laroque, P., Ballam, G., Soper, K., Dixit, R., Mattson, B., Adams, S., and Coleman, J. The effects of diet, ad libitum overfeeding, and moderate dietary restriction on the rodent bioassay: The uncontrolled variable in safety assessment. Toxicol Pathol. 24:757-768, 1996.