Hormesis as a Biological HypothesisEdward J. Calabrese and Linda A. BaldwinDepartment of Environmental Health Sciences, School of Public Health,University of Massachusetts, Amherst, Massachusetts

A comprehensive effort was undertaken to identify articles demonstrating chemical hormesis.Nearly 4000 potentially relevant articles were retrieved from preliminary computer databasesearches by using various key word descriptors and extensive cross-referencing. A priorievaluation criteria were established including study design features (e.g., number of doses, doserange), statistical analysis, and reproducibility of results. Evidence of chemical hormesis was

judged to have occurred in approximately 350 of the 4000 studies evaluated. Chemical hormesiswas observed in a wide range of taxonomic groups and involved agents representing highlydiverse chemical classes, many of potential environmental relevance. Numerous biological endpoints were assessed; growth responses were the most prevalent, followed by metaboliceffects, longevity, reproductive responses, and survival. Hormetic responses were generallyobserved to be of limited magnitude. The average low-dose maximum stimulation wasapproximately 50% greater than controls. The hormetic dose-response range was generallylimited to about one order of magnitude, with the upper end of the hormetic curve approachingthe estimated no observable effect level for the particular end point. Based on the evaluationcriteria, high to moderate evidence of hormesis was observed in studies comprised of > 6 doses;with >3 doses in the hormetic zone. The present analysis suggests that chemical hormesis is a

reproducible and relatively common biological phenomenon. A quantitative scheme is presentedfor future application to the database. Environ Health Perspect 106(Suppl 1):357-362 (1998).http.//ehpnetl1.niehs.nih.gov/docs/1998/Suppl- 1/357-362calabrese/abstract.html

Key words: hormesis, P-curve, stimulation, low dose, U-shaped

Introduction

The concept of chemical hormesis has along history, originating over a century agofrom the research of Schulz (1), who notedthat many chemicals were able to stimulategrowth and respiration of yeast at low dosesbut were inhibitory at higher levels. Thisconcept of a generalized low-dose stimula-tion-high-dose inhibition was graduallysupported by similar observations withother chemicals and eventually becameknown as the Arndt-Schulz law. AlthoughSchulz (1) ushered in the so-called modern

This paper is based on a presentation at The ThirdBELLE Conference on Toxicological DefenseMechanisms and the Shape of Dose-ResponseRelationships held 12-14 November 1996 inResearch Triangle Park, NC. Manuscript received atEHP29 April 1997; accepted 17 July 1997.

This report was sponsored in part by an award tothe University of Massachusetts (E.J. Calabrese,Principal Investigator) by the Texas Institute forAdvancement of Chemical Technology, Inc.

Address correspondence to Dr. E.J. Calabrese,Department of Environmental Health Sciences,School of Public Health, University of Massachusetts,Amherst, MA 01003. Telephone: (413) 545-3164. Fax:(413) 545-4692. E-mail: edwardc@schoolph.umass.edu

Abbreviations used: LOEL, lowest observableeffect level; NOEL, no observable effect level.

concept of hormesis, Paracelsus (2), writingin the 16th century, likewise noted thatvarious toxic substances may be beneficialin small quantities.

Despite the widespread recognition ofapparent hormetic effects, which continuedinto the early decades of the 20th century,Stebbing (2) argues that the Arndt-Schulzlaw gradually fell into disuse because it didnot provide an adequate explanatory (i.e.,mechanism-based) capacity. Nonetheless,over the years a continuing stream ofobservations has been reported (2) intoxicological publications and the broaderbiological literature that documentlow-dose stimulations.

Although there has been long-standinginterest in the concept of chemical hormesisfew attempts have been made to summarizethe extent of its occurrence in biologicalsystems and its potential to generalize withrespect to animal models, biological endpoints, or chemical class. Previous limitedsummaries have been reported (2-6). Inaddition, Davis and Svendsgaard (7)attempted to assess the statistical likelihoodof low-dose stimulation among a randomsample of experimental studies published in

prominent toxicological journals. The goalof this research was to extend the findingsof these previous reports by attempting toevaluate in a comprehensive manner thosestudies that are believed or alleged to dis-play evidence of chemical hormesis. Thesefindings would then be employed to assessthe scientific basis of the hypothesis thathormesis is induced by chemical agents andis a generalized biological phenomenon.

Criteria for EvaluatingHormesisTo conduct this investigation it wasnecessary to define chemical hormesis anddevelop a priori criteria to evaluate its pos-sible occurrence in experimental or empiri-cal investigations. The definition derivedfrom Stebbing (2) is low-dose stimulationfollowed by higher-dose inhibition; themost common form of hormesis followsthe widely recognized P-curve (Figure 1).The use of the P-curve follows principallyfrom the widespread use of growth as aprincipal end point in hormesis research.However, the term U shaped, as empha-sized by Davis and Svendsgaard (8), wouldmost appropriately be applied when theend point relates to a traditional toxicologi-cally based health end point such as cancerincidence. The criteria applied in thepresent methodology were the same for thei-curve and U-shaped relationships.

Because hormesis is a scientific hypoth-esis the question of whether it is beneficial isoften contextual. To eliminate subjectivedecisions concerning beneficial versus harm-ful effects, the decision was made to evaluatemodel- and end point-specific responseswith respect to stimulation and inhibition.For example, stimulation of detoxifyingenzyme levels observed in the larval form ofa species would be evaluated for its hormeticpotential even though this increased meta-bolic activity, although beneficial in theshort-term, may have a detrimental effect on

0

0-curve\

Dose

Figure 1. The most common dose-response curveshowing hormesis-the P-curve.

Environmental Health Perspectives * Vol 106, Supplement 1 a February 1998

llimilinne

357

CALABRESE AND BALDWIN

other end points. Likewise the stimulationof microbial reproduction by antibiotics wasevaluated for its hormetic potential eventhough these low dose effects are harmful tothe host organism.

This assessment of chemical hormesishas been restricted to those dose-responserelationships most conforming to theP-curve and would be affected by themagnitude of the low-dose stimulatoryresponse, the number of doses establishingthe reliability of the P-curve, the presenceof statistical analysis, and the reproducibil-ity of the findings. Within this categoryonly the well-known types of dose-responserelationships exhibiting P- or U-shapedcurves (represented by nutritionally essen-tial substances, with the exception of cop-per) were exduded, as this phenomenon isgenerally accepted.

The capacity to evaluate high confor-mity to the P-curve ideally requires theestablishment of an end point-specific low-est observable effect level (LOEL) and noobservable effect level (NOEL), with mul-tiple doses within two orders of magnitudeimmediately below the NOEL. This sug-gests that to be a relevant study for theevaluation of chemical hormesis, an experi-ment would be expected to have four ormore doses distributed in a highly specificmanner relative to the NOEL. Therefore,highly restrictive study design requirementsmust be satisfied to adequately assess chem-ical hormesis. Most toxicological studies donot satisfy these design criteria and wouldbe classified as nonrelevant, as they areunable to demonstrate no evidence or someevidence (i.e., equal to or greater than lowevidence) of hormesis. Similarly, data fromepidemiological studies, with the exceptionof reports on ethanol and cardiovasculardisease, were difficult to conform to thesecriteria. Within this evaluative contextjudgment on the evidence supporting con-sistency of data from an individual experi-ment with the definition of chemicalhormesis was made by a weight-of-evidenceprocedure. It should be noted that upward-curving P- or U-shaped dose-responsecurves (e.g., characteristic of certain studieswhere low doses reduce mortality) wereinduded in the analysis. In addition manystudies contain multiple dose-responserelationships for the same or different endpoints. In these cases all end points withina study showing low-dose stimulationwere evaluated.

To facilitate an appreciation of weight-of-evidence evaluation methodology,several graphic examples are presented

that illustrate how such judgmentsconcerning chemical hormesis may bemade (Figure 2A-E).

Figure 2A depicts a hypothetical studywith a dose range of 10-fold that displays amodest statistically insignificant increase(i.e., stimulation) in response at the lowestdose followed by a more definitive decrease(i.e., inhibition). Using the criteria applied inthe present methodology this study would bejudged as a) displaying an extremely limiteddosage range probably inadequate for assess-ment of the dynamics of the dose-responsecontinuum, b) inadequately describing thedose-response relationship in the hypotheti-cal hormetic zone, and c) having inadequatestatistical power to conclude that the stimu-latory effort was treatment related. An initialscreen of such an experiment would mostlikely result in a designation of not highlyrelevant to assess the hormesis hypothesis.However, the study would be retained forfurther evaluation within a weight-of-evi-dence context, based on the observationof the low-dose stimulation. At presentthis experiment would be most consistentwith either a low or not relevant evidencedesignation of hormesis. Neither categoricalplacement is convincing.

Figure 2B likewise depicts a study witha limited dose range (10-fold) with a lim-ited number of doses. However in contrastto Figure 2A, a more striking stimulatoryresponse is seen at the lowest dose, which ishighly statistically significant. However,this study is limited by having only onedose showing a stimulatory response eventhough the response was impressive. A casecan be made for either a low or moderateevidence classification of hormesis.

Figure 2C depicts a study with amarkedly larger dose range (500-fold) andnumber of doses (seven), with three dosesin the hypothetical hormetic zone (i.e.,doses less than the NOEL). However, themagnitude of stimulation is very limitedand the observed increases are not statisti-cally significant. This experiment wouldprobably be considered as no or low evi-dence of hormesis. It would of course be arelevant study because of the wide dosagerange, the substantial number of doses, theinclusion of doses below the LOEL, andthe observation, although inconclusive, ofstimulation at low dose.

Figure 2D depicts a study with a verybroad dose range (> 1000-fold) and a largenumber of doses (1 1), with a low-dose stim-ulatory response observed in seven doses inthe hypothetical hormetic zone. In additionthe results display considerable statistical

power. This type of study would receive ahigh evidence ranking.

Figure 2E depicts a study with a widerange of doses (> 500) and a large number ofconcentrations with adequate statisticalpower. However the data do not show anyevidence of a low-dose stimulation. Thisstudy would receive a not relevant designa-tion because it does not establish a NOELnor does it have any doses below the NOEL.

A

0

0.(4amCC

B

a)(4

00.1CO

C

00.a,

D

a,0

Q(4

CO0

E

0)0.(4.CO

10Dose

Dose

DooDose

I-100

Dose>1000

>500Dose

Figure 2. Dose-response curves illustrating variousdata sets evaluated for chemical hormesis.

Environmental Health Perspectives * Vol 106, Supplement 1 * February 1998

[1

358

HORMESIS AS A BIOLOGICAL HYPOTHESIS

Nature and Yieldof Search StrategyTable 1 summarizes the search strategyand yield. Computerized library searcheswere conducted on BIOSIS (Knight-Ridder Information, Mountain View, CA;1969-1996), Chemical Abstracts (Knight-Ridder Information; 1967-1996), andMedline (Knight-Ridder Information;1966-1996) using hormesis and the com-bination ofU shaped plus dose response askey word descriptors. One hundred fifty-two publications were retrieved usinghormesis as the key word descriptor; 165publications were retrieved using the com-bination of U shaped plus dose response.Radiation hormesis was the subject of 104of the 317 publications.

Based on information obtained from theinitial searches described above, additionalsearch strategies were employed using thesame three databases and the following keyword descriptors: low dose plus stimulation,beta curve plus dose response, adaptationplus pollution, and homeopathy. A total of4058 articles was identified then reduced to3272 following elimination of databasereplication of articles. Manual review of the3272 abstracts revealed approximately172 potentially relevant publications, the

majority of which were chemically oriented.Radiation studies comprised approximately25% of the 172 articles.A computerized library search con-

ducted on the database Agricola (Knight-Ridder Information; 1970-September1995) using the key word descriptorshormesis, U shaped plus dose response,and low dose plus stimulation revealed veryfew articles not already identified in theprevious searches.

To include the most recent articles, com-puterized searches of the Current ContentsLife Sciences and Agricultural, Biological,and Environmental Sciences databases(Institute for Scientific Information,Philadelphia, PA) for the time period 15May 1995 through 5 June 1996 were con-ducted using the key word descriptorshormesis, U shaped plus dose response, lowdose plus stimulation, adaptation plus pollu-tion, and beta curve. Only a small numberof new articles not included in the priorsearches was identified.

Potentially relevant artides not includedin the computerized library searches wereobtained from extensive cross-referencing ofprimary sources.

Most recently additional searchstrategies were employed using BIOSIS,

Table 1. Summary of computerized search strategy.

Key words DatabasesHormesisU shaped + dose responseLow dose + stimulationBeta curve + dose responseAdaptation + pollutionHomeopathyHormesisU shaped + dose-responseLow dose + stimulationAdaptation + pollutionBeta curveDose response + sublethalDose response + subthresholdDose response + subtoxic

BIOSIS, Medline, Chemical AbstractsaBIOSIS, Medline, Chemical AbstractsBIOSIS, Medline, Chemical Abstracts

Agricola (1970-Sept 1995)Current Contents (15/5/95-6/5/96)

Abstractsretrieved, no

150165

3272

>200

Current Contents (15/5/95-6/5/96)

BIOSIS, Medline, Chemical Abstracts

Agricola (1970-Sept 1996)Current Contents (16/10/95-7/10/96)

Stebbing ARD Science Citations (1990-Mar 1996)Luckey TD Science Citations (1990-Sept 1996)8B10SIS (1969-1996), Medline (1966-1996), Chemical Abstracts (1967-1996).

3776

230214207200

Potentiallyrelevant, no

Medline, and Chemical Abstracts and thefollowing key word descriptors: doses plussublethal, doses plus subtoxic, doses plussubthreshold, responses plus sublethal,responses plus subtoxic, and responses plussubthreshold. A total of 5569 articles wasidentified then reduced to 3776 followingelimination of database replication of arti-cles. Manual review of the 3776 abstractsrevealed approximately 67 potentiallyrelevant publications.A computerized library search using

the same key word descriptors listed aboveand the database Agricola (1970-September1996) identified 230 articles. Whenreviewed approximately 25 were consid-ered potentially relevant. The same keyword descriptors were also used in com-puterized searches of the Current ContentsLife Sciences and Agricultural, Biological,and Environmental Sciences databases forthe time period 16 October 1995 through7 October 1996 to include the mostrecent articles. Of the 214 abstracts iden-tified, 27 new articles were consideredpotentially relevant.

Finally, computer searches of ScienceCitations (Institute for Scientific Infor-mation; 1990-1996) were conducted usingauthors' names Stebbing ARD and LuckeyTD. Approximately 400 articles wereidentified, of which 149 were consideredpotentially relevant.

Results of Article Evaluation90 Those studies placed within a high evidence31 category of chemical hormesis had the172 greatest number of total study doses (i.e.,

6.3 on average) and doses in the so-calledhormetic zone (i.e., 3.4 on average), fol-

24 lowed by studies demonstrating moderateevidence and more distantly by studiesdemonstrating low evidence (Table 2).

The types of chemicals that inducehormetic effects represent a broad range of

67 chemical classes (Table 3). The most studied

25278465

Table 2. Characteristics defining studies demonstrating high, moderate, low, and no evidence of hormesis.

Evidence of Mean Mean doses, Inadequate Inadequate Inadequate Lack ofhormesis doses, no no < NOAEL doses, no dose range end point statistical power

High 6.3 3.4 - - - -Moderate 5.5 2.8low 4.7 1.8 X X - XNo 4.7 < 1.8 X X X X

Abbreviations: -, characteristic was generally not evident; X, characteristic was generally evident.

Table 3. Results of initial screening organized by agent.a

Agent Percent

Alcohol and metabolites 6.2Antibiotics 7.9Auxin related 4.6Hydrocarbons 3.4Metals 29.6Herbicides 7.2Insecticides 6.1Fungicides 1.5Pesticides 2.9Miscellaneous 30.6

Total 100.0

°High, moderate, and low data are combined.

Environmental Health Perspectives * Vol 106, Supplement 1 * February 1998

1

359

CALABRESE AND BALDWIN

Table 4. Results of initial screen organized by end point.

High Moderate Low Number Percent

Growth 161 147 61 369 62.2Metabolic effects 38 27 25 90 15.2Longevity 21 5 5 31 5.2Survival 13 18 3 34 5.7Reproduction 7 18 9 34 5.7Miscellaneous 8 22 5 35 5.8

Total 248 237 108 593 99.8

agents were metals, followed by alcohols,antibiotics, auxin-related compounds, andnumerous biocidal agents. The range ofhormetic responses is listed in Table 4 andindicates that the principal end point isgrowth, followed by metabolic changes(e.g., enzyme activity), longevity, andvarious reproductive indices.

Characteristics of theChemical Hormetic ZoneTo assess the characteristics of the chemicalhormetic dose-response zone, experimentaldata were evaluated with respect to a) thedosage range of the hormetic zone (i.e.,from the estimated dosage where theresponse starts to deviate from the controlto the estimated dosage where the responsebegins to dip below the controls); b) themaximum stimulatory response (as a per-centage greater than the control response);and c) the magnitude of dosage differencefrom the maximum stimulatory responseand the estimated NOEL (Figure 3).

In general the hormetic dose-responserange is usually within a 10-fold range.Stimulatory effects, however, have beenreported over dosage ranges of two or moreorders of magnitude as well as over a morenarrow range of dosages depending on theagent, end point, and model assessed. Themagnitude of stimulatory responses hasbeen observed as high as several-fold butthe majority of low-dose stimulations are

Maximum stimulationaverages 30 to 60% greater than control

X, {4- to 5-fo Estimated high NOEL(ZEP)

Hormetic zone \ OELaverages 10-fold

Dose

30 to 60% greater than the controls. Thedistance from the maximum stimulatoryresponse to the NOEL is difficult todiscern as it is a function of the numberof doses employed, their variability inresponse, and the estimated value of theNOEL. Nonetheless, the distance betweenthe maximum stimulatory response andthe estimated NOEL is typically observedin the 3- to 6-fold range (i.e., the NOELis about 3- to 6-fold greater than themaximum stimulatory response).

Hormesis as a GeneralBiological PhenomenonHormetic responses are observed innumerous species from a broad range oftaxonomic groups including microbes,plants, and animals (Table 5). Theseresponses occur with a large number ofchemicals representing a broad range ofchemical classes (Table 3). AlthoughStebbing (2) focused principally on growthhormesis, the present report indicates thathormetic effects are observed in a broadrange of biological end points that involvenot only growth but survival, longevity,reproduction, and numerous metabolicand physiological responses (e.g., metal-lothionein synthesis, DNA synthesis, RNAsynthesis, mitosis, oxygen consumption,altered hepatic foci, photosynthesis rate,tissue regeneration, immune response,stress protein synthesis, germination ofseeds, etc.). Thus hormesis appears rela-tively common with respect to species,chemical, and biological end point.

The ability to generalize hormeticresponses also extends to the descriptivenature of the dose-response phenomenon

Table 5. Results of initial screen organized by test model.

itself. As Stebbing (2) noted earlier, whenthe data are precise and comprehensive,the points appear to fit a P-curve and haveremarkable similarity with respect to therange and amplitude of response. Howeverit should be emphasized that the develop-mental dynamics of the hormetic doseresponse over time have not been widelyor systematically studied. For example,while Stebbing (2) found that the form ofthe curve varied during the course of theexperiment with hydra, Calabrese andHowe (9) observed a consistent shape ofthe P-curve over 4 to 6 weeks in plantgrowth experiments.

Why is HormesisInfrequently Observed?If hormesis is believed to be relatively com-mon, questions arise as to why it is notreported more frequently and why theArndt-Schulz law failed to become estab-lished. The infrequent reports of hormesisare most likely attributable to a combina-tion of factors, predominantly the issue ofappropriate study design, along with theinfluence on safety evaluation, whichemphasizes the upper end of the dose-response continuum (i.e., where higherconcentrations establish toxic responsesthat can be used in chemical evaluationand risk assessment). The present conclu-sions support this assessment; a direct rela-tionship has been shown between thestrength of the evidence supporting horme-sis and the number of doses, includingboth overall experimental number of dosesand the number of doses in the hormeticzone. Furthermore, because the averagerange of the hormetic zone is about oneorder of magnitude this phenomenonis difficult to discern when wide doseintervals (e.g., > 10-fold) are used.

Predictive insight into the number ofpublished articles potentially displayinghormesis may be derived as follows.Assuming 500,000 toxicology studies havebeen published this century (based onsearches of Chemical Abstracts, IndexMedicus, and BIOSIS), it is estimated thatapproximately 350,000 toxicology articles

High Moderate Low Number PercentBacteria 17 27 11 55 9.3Protozoa 7 9 2 18 3.0Fungi 24 8 6 38 6.4Plants 112 63 32 207 34.9Animals 88 130 57 275 46.3

Total 248 237 108 593 99.9

Environmental Health Perspectives * Vol 106. Supplement 1 * February 1998

a005

Figure 3. Dose-response curve depicting charac-teristics of the chemical hormetic zone. ZEP, zeroequivalent point.

360

HORMESIS AS A BIOLOGICAL HYPOTHESIS

have been published since 1966. Thus weassume for the sake of argument that500,000 toxicology papers comprise theavailable pool for evaluation. If we alsoassume that 2% of these studies include sixor more doses (10) then 10,000 studiesmay contain dose ranges adequate forhormesis evaluation. Of these 10,000studies, based on the characteristics defin-ing hormetic studies, approximately 10%(i.e., 1000) have three or more doses belowthe estimated NOEL. Further refinementof this estimation can be made by assum-ing that 90% of these 1000 studies havedoses in the low-dose range within one totwo orders of magnitude and close to theestimated NOEL (10).

These figures suggest that mammaliantoxicologists may have had only limiteddirect interaction with the concept ofhormesis, as only an estimated 900 poten-tially relevant studies exist. Therefore it isnot surprising that toxicologists may viewhormesis more as a belief than a phenome-non and that the Arndt-Schulz law (i.e.,hormesis) fell into general disuse.

In addition to the low number ofhormetic observations reported, Stebbing(2) suggested that the Arndt-Schulz Lawfell into disuse because it lacked an explicitmechanism component. However, the con-cept of adaptation, a potentially importantexplanatory component for hormesis, hasevolved for the most part independent ofhormesis. Although numerous studies ofadaptation exist, only a limited number

address specific mechanisms applicable tochemically induced hormetic dose-responserelationships. Nonetheless there arestudies that have sought to mechanisticallyexplain specific hormetic dose-responserelationships.

Perhaps the most systematically assessedmechanism-based research has been in thearea of herbicide-induced stimulatoryeffects. Hormetic responses have long beenrecognized by herbicide researchers whohave conducted studies assessing not onlythe molecular basis for the response butalso the effect of the plant species and ageon the response.A growing number of mammalian

examples also exists where plausibleexplanatory mechanisms have been putforth to account for specific hormeticdose-response relationships (11-15).

The wide range of hormetic effects(e.g., increased growth, fecundity, lon-gevity, and decreased disease incidence)suggests that these changes are fundamen-tal and affect thousands of genes. Thisimplies that hormetic mechanisms arelikely to be operational in a very upstreamlocation. Nevertheless investigators oftenfocus on mechanisms more closely relatedto biological protection. For example, sub-stantial evidence exists in numerous speciesthat specific alterations in patterns of geneexpression occur in response to toxicantexposure. Such responses can be sortedinto two classes: those resulting in anenhanced metabolic capacity for detoxi-

Table 6. Summary of study design criteria with assigned point values used in the quantitative evaluation of arti-cles for evidence of chemical hormesis.

Number of dosesbelow NOEL Point value NOEL determined Point value

1 1 Yes 12 2 No 03 34 4

.5 5

Table 7. Summary of response criteria with assigned point values used in the quantitative evaluation of articlesfor evidence of chemical hormesis.

Number of doses Magnitude of response,statistically significant Point value % of control Point valuea

1 2 >110<125% 0.52 4 >125<150% 13 8 >150<200% 2

>4 16 >200<400% 3>400% 4

aThe point value is multiplied by the number of experimental doses falling within the corresponding percentagerange. For example, if an experiment has three doses exhibiting stimulatory responses within the 125 to 150%range, the total number of points will be 3 x 1.0=3.

fication (e.g., the cytochrome P450 genefamily) and those that offer a more generalprotection against cellular damage causedby a wide variety of agents (e.g., heat shockor stress proteins).

Proposed QuantitativeEvaluation SchemeA quantitative scheme has been developedto provide a more objective and repro-ducible methodology for ranking studieswith respect to hormetic potential. Criteriahave been established and assigned pointvalues based on: the number of experimen-tal doses below the NOEL, experimentaldetermination or estimation of the highNOEL, the statistical significance of thestimulatory response, the magnitude (per-centage of control value) of the stimulatoryresponse, and the reproducibility of data byother studies (Tables 6 and 7). Evidence ofhormesis will be assessed by comparing thesummation of point values to point rangesestablished for six evidence categories:high, moderate-high, moderate, low-mod-erate, low, and no-low (Table 8). Resultsof the application of this methodology andcomparison with the current qualitativefindings will be published elsewhere (16).

SummaryA goal of the present research is to create adatabase of studies demonstrating objectiveevidence of hormesis. It is hoped that thisdatabase, when complete, will enable thescientific community to evaluate more rig-orously and efficiently the concept ofhormesis with respect to its status as a bio-logical hypothesis, its potential to be gener-alized, and its impact on environmentaland human health.

The findings to date indicate thatexamples of low-dose stimulation consis-tent with P-curve characteristics are copi-ous, diverse, independently derived, andreproducible. Yet despite the large numberof such observations no long-term system-atic effort has been made to uncoverexplanatory mechanisms, except in limitedcases (e.g., herbicidal agents).

Table 8. Summary of total point ranges for hormesisevidence categories used in the quantitative evaluationof articles for evidence of chemical hormesis.

Total point range Hormesis evidence category

1-2 No to low>2-8 Low>8-12 Low to moderate>12-16 Moderate>16-20 Moderate to high>20 High

Environmental Health Perspectives * Vol 106, Supplement 1 * February 1998 361

CALABRESE AND BALDWIN

A more objective and reproducibleevaluation methodology for rankingstudies with respect to hormetic potentialis needed. The development of quantitativecriteria based on study design, response,and reproducibility of findings is proposedand will be applied to the current database.

Furthermore, statistical simulations ofdose-response relationships given varioustypes of variability in control groupscan provide important insight into theestablishment of more quantifiable criteriain the evaluation of possible hormeticfindings. The area of hormesis and its

evaluation as a biological hypothesishas striking similarities to the evolvingmathematical area of meta-analysis in epi-demiology. In fact the application of meta-analysis techniques to the evaluation ofhormetic response data is likely to yieldsignificant advances.

REFERENCES

1. Schulz H. Ueber Hefegifte. Pfluegers Arch Gesamte PhysiolMenschen Tiere 42:517 (1888).

2. Stebbing ARD. Hormesis-the stimulation of growth by lowlevels of inhibitors. Sci Total Environ 22:213-234 (1982).

3. Townsend JF, Luckey TD. Hormoligosis in pharmacology. JAm Med Assoc 173:44-48 (1960).

4. Luckey TD. Hormology with inorganic compounds. In: HeavyMetal Toxicity, Safety, and Hormology, Supplement Volume 1(Coulston F, Korte F, eds). Stuttgart:George Thieme,1975;83-118.

5. Calabrese EJ, McCarthy M, Kenyon E. The occurrence ofchemical hormesis. Health Phys 57:531-54 (1987).

6. Calabrese EJ. Primer on BELLE. In: Biological Effects of LowLevel Exposures: Dose-Response Relationships (Calabrese EJ,ed). Boca Raton, FL:CRC/Lewis Publishers, 1994;27-42.

7. Davis JM, Svendsgaard DJ. Nonmonotonic dose-response rela-tionships in toxicological studies. In: Biological Effects of LowLevel Exposures: Dose-Response Relationships (Calabrese EJ,ed). Boca Raton, FL:CRC/Lewis Publishers, 1994;67-86.

8. Davis JM, Svendsgaard DJ. U-shaped dose-response curves:their occurrence and implications for risk assessment. J ToxicolEnviron Health 30:71-83 (1990).

9. Calabrese EJ, Howe KJ. Stimulation of growth of peppermint

(Mentha piperita) by phosfon, a growth retardant. Physiol Plant37:163-165 (1976).

10. Calabrese EJ. Unpublished data.11. Calabrese EJ, Baldwin LA, Mehendale HM. Contemporary

issues in toxicology: G2 subpopulation in rat liver induced intomitosis by low level exposure to carbon tetrachloride: an adap-tive response. Toxicol Appl Pharmacol 121:1-7 (1993).

12. Calabrese EJ, Mehendale HM. A review of the role of tissuerepair as an adaptive strategy: why low doses are often non-toxic and why high doses can be fatal. Food Chem Toxicol34:301-311 (1996).

13. Hart RW, Frame LT. Toxicological defense mechanisms andhow they may affect the nature of dose-response relationships.BELLE Newslett 5:1-16 (1996).

14. Mukherjee SN, Rawal SK, Ghumare SS, Sharma RN.Hormetic concentrations of azadirachtin and isoesterase profilesin Tribolium castraneum (Herbst) (Coleoptera:Tenebrionidae).Experientia 49:557-560 (1993).

15. Vichi P, Tritton TR. Stimulation of growth in human andmurine cells by adriamycin. Cancer Res 49:2679-2682 (1989).

16. Calabrese EJ, Baldwin LA. Quantitatively-based methodologyfor the evaluation of chemical hormesis. Hum Ecol Risk Assess4:545-554 (1997).

362 Environmental Health Perspectives * Vol 106, Supplement 1 a February 1998