Views& Commentaries

:.. .. i} ;:...

.%siX: views & Commentaries |T____._-----i------.--r...... -,f,P.g....f.wRt.,4.....He .;8|S .

A Toxicological Basis to Derive a Generic Interspecies Uncertainty Factor Edward J. Calabrese and Linda A. Baldwin School of Public Health, University of Massachusetts, Amherst, MA 01003 USA

The use of an uncertainty factor (UF) to on experience, and a sense that it achieves account for interspecies variation in risk its goal of protecting human health. The assessment procedures for noncarcinogens present paper offers what the authors we is well known and implemented by regula- believe to be a toxicological and statistically tory agencies at the federal and state levels. defensible foundation for deriving the The approach that has been widely adopt- interspecies UF, its database requirements, ed is to assume that humans may be 10- and statistical procedures for its derivation. fold more sensitive than the animal model. In brief, the recommended interspecies UF This factor of 10 has become routinely is defined as the 95% of the population of adopted in essentially all risk assessment 95% prediction intervals (PI) for binary procedures involving animal model data interspecies comparisons based on phyloge- for extrapolation. netic relatedness. More specifically, the UF Despite the long-standing use of the is derived by determining the minimum interspecies UF of 10, only limited biologi- ratio of the estimated toxicity value and its cal and/or toxicological justification for the 95% upper or lower PI after back-transfor- interspecies UF has ever been put forth by mation from the logarithmic expression. any regulatory agency (1) or national advi- This paper presents the toxicological sory committee (e.g., National Academy of and statistical basis for this proposal and its Sciences Safe Drinking Committee). The implications for judging the reliability of adoption of the 10-fold factor appears to current regulatory interspecies UF proce- erous binary interspecies comparisons and have been based on a combination of pub- dures as well as offering a fundamentally organized them via phylogenetic related- lic health protection philosophy, prac- novel approach to deriving an interspecies ness. For example, interspecies compar- tical/intuitive toxicological insights based UF. isons were provided when the comparisons An extensive database on interspecies represented species-within-genus, genera- variation in susceptibility to toxic agents within-family, families-within-order, and twos Jlog.RlaB;2oroi~~~~~~~~~~~~~ali~~~nbrJB ~ ...... ].... exists in the aquatic toxicology area. The orders-within-class comparisons. For toxicity data are principally, though not example, in Figure 2 a species-within- 14 exclusively, based on acutely toxic respons- genus comparison would represent a bina- es. The data are arranged in the form of ry comparison of species 1 with species 2. binary interspecies comparisons with A genera-within-family binary comparison respect to toxicity from dozens to over 500 would be represented by a comparison of agents depending on the specific binary species 1 with species 3. The reason for comparison. A binary comparison in the organizing the comparisons in this phylo- .v... present context involves comparing the genetic manner is the assumption that responses of two species to agents that were interspecies variation in susceptibility tested in both species. For example, two would increase as the phylogenetic dis- J, species of fish (e.g., smallmouth bass and tance increased. perch) have been used to test over 500 of Table 1 provides a summary of the the same toxicants (Fig. 1). A binary com- database of phylogenetically based inter- parison of these two species would include species binary comparisons. The 95% PI more than 500 agents. These data have for each binary comparison is provided, An been organized to assess whether a mathe- along with the number of different chemi- matical relationship exists such that the cal agents tested for each binary compari- -5 0 5 10 15 LC50 of one species may be a useful predic- son. The weighted mean value indicates In Salmonifonnes Wm5 tor of the LC50 in the other species via the that in general the closer the animal use of were the smaller the Figure 1. Natural logarithms of LC50 values for regression modeling. species related, 95% Perciformes plotted against Salmoniformes The above binary comparison method- PI. The range of weighted means of 95% (orders of the same class, Osteichthyes). The ology has been used by various authors PI is from a low of 6.0 (species within solid line represents the least-squares linear (2-4) to estimate the LC50 for any new genus) to a high of 26.0 for the orders- regression of the natural logarithm of LC50 values chemical in an untested species (e.g., small- within-class grouping. for Perciformes species on the natural logarithm mouth bass) if the LC50 were known. for Slooff et al. (4) transformed the con- of values for Salmoniformes species. Each LC50 the perch. The estimate is made by calcu- cept of the 95% PI into a 95% UF. Figure circle represents the LC5, value of a specific chemical for both species. $he number of chemi- latng a prediction interval (PI) for the cals represented in the figure is 503. Data from unknown chemical. Barnthouse et al. (3) Address correspondence to E. J. Calabrese. Johnson and Finley (9). have provided 95% PI estimates for num- Received 21 June 1993; accepted 2 October 1993.

14 Environmental Health Perspectives I - - -1MLA - - - I - -

3 presents a graphic foundation of the PI as well as statistical definition and relationship to the UF concept. Thus, the species- within-genus 95% UF, as anticipated, is considerably smaller than the 95% UF for orders within class. The magnitude of interspecies variation in 95% PI values fol- lows fairly closely with phylogenetic relatedness, as expected. Inconsistencies such as the similar estimates for species within genus and genera within family are likely Figure Z Interspecies comparisons based on phylogenetic relatedness. S1 represents a species for which related to issues concerning representative- data are available. S, and S2 represent a species-within-genus comparison; S1 and S3 represent a genera- ness, number of binary comparisons, and within-family comparison; S1 and S5 represent a families-within-order comparison; and S1 and S9 represent number and nature ofchemical agents test- an orders-within-class comparison. ed. The binary comparison values do not Table 1. Taxonomic extrapolation: means and weighted means calculated for the 95% and 99% prediction represent the population (or universe) of intervals (PI) for uncertainty factors calculated from regression models (3) such values but must be considered a sample Uncertainty factor of the population. No knowledge exists X variable Yvariable n 95% PI 99% P1 concerning how representative this sample Taxonomic extrapolation: species within genera of values would be of the population. For Salmo clarkii S. gairdneri 18 9 13 the sake of argument, the samples of each S. clarkii S. salar 6 6 10 phylogenetic subgroup are considered repre- S. clarkii S. trutta 8 6 8 sentative of their respective population val- S. gairdneri S. salar 10 7 11 S. gairdneri S. trutta 15 4 5 ues. Table 2 provides an estimate of upper S. salar S. trutta 7 5 8 95% (using logistic regression modeling) of Ictalurus melas I. punctatus 12 5 7 the population of 95% PI values (see Figure Lepomis cyanellus L. macrochirus 14 6 9 3 for derivation of95% PI values) according Fundulus heteroclitus F majalis 12 6 8 to phylogenetic relatedness. The unexpect- Mean 6.1 10.1 Weighted mean 6.0 7.4 edly high value from the families-within- Taxonomic extrapolation: genera within families order extrapolation group is partially incon- Oncorynchus Salmo 56 5 6 sistent with the proposed phylogenetic rela- Oncorynchus Salvelinus 13 4 5 tionship. This inconsistency is principally a Salmo Salvelinus 56 5 7 result of the low number of binary compar- Carassius Cyprinus 8 4 6 Carassius Pimephales 19 7 9 isons (N= 7) and high variability ofindivid- Cyprinus Pimephales 10 7 10 ual estimates in the families-within-order Lepomis Micropterus 30 8 11 comparison group. This value is less stable Lepomis Pomoxis 8 9 13 than the orders-within-class grouping. Cyprinodon Fundulus 12 6 8 Given the amount of data, the orders-with- Mean 6.1 8.3 Weighted mean 5.8 7.7 in-class comparison offers the most stable Taxonomic extrapolation: families within orders and reliable perspective. We propose that Centrarchidae Percidae 47 10 14 these values can be used to provide a toxico- Centrarchidae Cichlidae 6 4 6 logically and statistically based foundation Percidae Cichlidae 5 13 24 for generic interspecies UFs when normal- Salmonidae Esocidae 11 9 13 Atherinidae Cyprinodontidae 32 7 9 ized for phylogenetic relatedness. The data Mugilidae Labridae 12 55 78 suggest that four different UFs be adopted Cyprinodontidae Poecillidae 12 3 5 according to phylogenetic relatedness. The Mean 14.4 21.3 choice of 95% UFs would range from a low Weighted mean 12.6 17.9 of 10 for the species within genus to a high Taxonomic extrapolation: orders within classes within The Salmoniformes Cypriniformes 225 20 27 of 65 for the orders class. gen- Salmoniformes Siluriformes 203 39 51 era-within-family and families-within-order Salmoniformes Perciformes 443 12 16 groupings are more difficult to determine. Cypriniformes Siluriformes 111 11 15 Based on the phylogenetic relatedness con- Cypriniformes Perciformes 219 32 43 cept, these two groups are estimated to be Siluriformes Perciformes 190 63 83 intermediary between the boundary values Anguiliformes Tetraodontiformes 12 13 18 Anguiliformes Perciformes 34 25 34 (i.e., species within genus, orders within Anguiliformes Gasterosteiformes 8 16 24 class), approximating 25 and 50. Anguiliformes Atheriniformes 46 9 12 The proposed methodology approach Atheriniformes Cypriniformes 7 5018 786° takes into account two critical components Atheriniformes Tetraodontiformes 46 13 17 in any interspecies UF estimation process: Atheriniformes Perciformes 148 25 33 Atheriniformes Gasterosteiformes 36 20 27 the need to address the universe of species Gasterosteiformes Tetraodontiformes 8 20 30 (as is done via the use of logistic regression) Gasterosteiformes Perciformes 33 32 43 and the need to incorporate the new chem- Perciformes Tetraodontiformes 34 25 34 icals (as is accomplished via the use of the PT Mean 23.5 31.7 approach). These findings and interpreta- Weighted mean 26.0 34.5 tions are based directly on data derived from "Not included in calculations.

Volume 102, Number 1, January 1994 15 __ 9- . -. .

Table 2. Upper 95% uncertainty factors calculated for the 95% and 99% prediction intervals (Table 1) based on the scheme of Van Straalen and Denneman (10) Prediction interval Regression model 95% 99% Species within genera 10.0 16.3 Genera within families 11.7 16.9 Families within orders 99.5 145.0 Orders within classes 64.8 87.5

mammalian/human responses. Rather, the U . ,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ...... ----- issue is whether the variation in response among species at the various levels of phylogenetic relatedness for the aquatic models is predictive of the mammalian phyloge- 1. netic variability that would be seen among _l LI pee A lo mammalian models and humans for the same chemical contaminants. On a con- ceptual level, the trend in increased variability in susceptibility as seen in fish as the phylogenetic relatedness decreases would be expected to occur with mammalian sys- tems. How quantitatively similar the weighted mean 95% PIs of the fish comparisons would be for the various phylogenetic relatedness comparisons in mammals T...... c...tainty------is unknown. However, the use of biological systematics to provide a common measure of evolutionary/biological relatedness among the various animal classifications (e.g., fish and mammals) is a valuable and powerful tool that has rarely been applied varanc numbr o oseratins knwnlog LC50species A an_ siae lgL5 pce to the field of toxicology/risk assessment. BigWhenx.D therpreditiononunetintevlbcmsy+ts+(/f)i2Teueranyfactors isFdefne)a For example, the basic unit of comparison, the species, is similarly defined in fish as well as mammals. Although less precise theretminmu praetiolfte estiaStedentoxiistyrvaueiandwits 95% uppegree or lowrepedicetionlmitafterbeidack than the species concept, the same concep- tual definitions proceeds to broader cate- transformation: UF =10 S[1 + (1 + (1/n))1t2. In applied terms: If the toxicity of a given compound for A is gories (genus to class) across the animal known, the value for B is in the range of N/UF< B< A* UF with a probability of 95%. kingdom. Thus, the trend of interspecies variability observed in various phylogenetic related categories in fish would be expected acute toxicology experiments in fish. It database considers a large number of species to be qualitatively similar in mammals as assumes that the concept of phylogenetic representing different sizes, various biologi- well. relatedness in relationship to toxicity that is cal adaptations, and variation in susceptibil- Another area of possible concern is that seen within fish species would apply to ities; 4) the database is composed of assess- the database uses acute rather than chronic mammals and that the magnitude of the ments of more than 400 different chemical toxic responses. This does not appear to be phylogenetic differences observed among agents representing several dozen chemical a serious concern because acute responses fish species would be quantitatively compa- classes (e.g., pesticides, metals, PAHs, etc.). have been shown to be effective predictors rable to mammalian toxicology. The database has the capacity to provide of chronic effects of both a carcinogenic The proposed methodology offers a strong generalizations to account for both (5) and noncarcinogenic nature (3,4,6,7). number of important strengths in providing inherent species variation and large num- In fact, the chronic no-observed adverse a foundation for the interspecies UF deriva- bers of chemical agents; 5) the database per- effect level in mammalian models and the tion: 1) it represents an extensive database mits the application of statistical evaluation chronic maximum acceptable toxicant con- obtained via a standardized testing protocol to describe the distribution of responses centration in fish have been similarly esti- with respect to a critical integrative end- with respect to both PIs for specific chemi- mated by dividing the acutely lethal dose point (i.e., LC50); 2) it has the capability to cal responses and species variation in (LD50/LC50) by approximately 50-75 incorporate phylogenetic relatedness to the responses. (4,6,7,8). These data show a high degree of predictive endpoint, which represents a sig- An area of potential concern with the fundamental concordance between fish nificant advance and is entirely consistent present proposal is that the database is and mammalian responses with respect to with the biologically persuasive evolutionary drawn entirely from aquatic models and is the capacity of acute doses to estimate paradigm of modern molecular biology being generalized to mammalian phyloge- chronic responses. relating genetic factors to susceptibility ny. The issue is not whether fish are effec- There is a need to define the biological and/or resistance to chemical insults; 3) the tive qualitative/quantitative predictors of and statistical meaning of the interspecies

16 Environmental Health Perspectives rPfiftl;-4y: 098oilA 9- 9 nations: influence of life history, data uncer- UF. The 95% UF as described here repre- final selection of which range of UF values tainty and exploitation intensity. Environ sents the upper 95% of the distribution of to select would be based on value judg- Toxicol Chem 9:297-311(1990). binary interspecies comparison 95% PI ments. 4. Slooff W, Van Oers JA, DeZwart D. Margins values. This is interpreted as 95% of exper- The field of mammalian toxicology in of uncertainty in ecotoxicological hazard iments in which a chemical is tested would which mice, rats, gerbils, guinea pigs, cats, assessment. Environ Toxicol Chem 5:84 1- respond within the given PI (i.e., 95% PI). and dogs are used as models to estimate 852(1986). 5. Zeise L, Crouch EAC, Wilson R. A possible This also is interpreted to mean that 95% human responses represents orders-with- relationship between toxicity and carcino- of every 100 unknown chemicals tested class comparisons. Using the scheme out- genicity. J Am Coll Toxicol 5(2):137-151 would display a response within the calcu- lined above suggests that the UF for such (1986). lated PI. The 95% PI can, therefore, be comparisons could range from 65 to 87 6. Layton DW, Mallon BJ, Rosenblatt DH, used as a measure of interspecies variation. (possibly rounded to 50-100) rather than Small MJ. Deriving allowable daily intakes for We then estimate the upper 95% of these the 10-fold value currently used, depend- systemic toxicants lacking chronic toxicity data. Regul Toxicol Pharmacol 7:96-112 "individual measures of interspecies varia- ing on which quantitative estimate for UF (1987). tion" (i.e., the distribution of the 95% PI). derivation were selected. 7. Kenaga EE. Predictability of chronic toxicity This is then collectively interpreted as the from acute toxicity of chemicals in fish and following: 95% of chemicals would not REFERENCES aquatic invertebrates. Environ Toxicol Chem exceed a given PI in 95% of species tested. 1:347-358(1982). The risk assessor has the flexibility to 1. Dourson ML, Stara JF. Regulatory history 8. Calabrese EJ, Baldwin L. Performing ecologi- and experimental support of uncertainty (safe- cal risk assessments. Chelsea. MI:Lewis change the size of the PI as well as that ty) factors. Regul Toxicol Pharmacol 3:224- Publishers, 1993. portion of the logistic distribution deemed 238(1983). 9. Johnson WW, Finley MT. Handbook of suitable for UF selection. For example, if 2. Suter GW II, Vaughan DS, Gardner RH. acute toxicity of chemicals to fish and aquatic the 99th percentile of the population of Risk assessment by analysis of extrapolation invertebrates. Washington, DC:U.S. Fish and 95% PI were selected for the UF, then the error, a demonstration for effects of pollutants Wildlife Service Resource Publication 137. range of phylogenetic UFs would be in- on fish. Environ Toxicol Chem 2:369-378 U.S. Department of the Interior, 1980. (1983). 10. Van Straalen NM, Denneman CAJ. Ecotox- creased from the 10- to 65-fold range to 3. Barnthouse LW, Suter GW, Rosen AE. Risks icological evaluation of soil quality criteria. the 16- to 87-fold range (Table 2). The of toxic contaminants to exploited fish popu- Ecotoxicol Environ Safety 18:241-251(1989).

OF GOVERNMENT B The U.S. Government Printing all about. Send for your free Office has a free catalog of new catalog. and popular books sold by the Government. Books about Free Catalog agriculture, energy, children, P.O. Box 37000 space, health, history, business, Washington, DC 20013-7000 vacations, and much more. Find out what Government books are

01"~~ ~ ~ ~

Volume 102, Number 1, January 1994 17