Mutation Research, 239 (1990) 83-115 Elsevier
83
MUTREV 07281
INTERNATIONAL
COMMISSION
ENVIRONMENTAL
FOR PROTECTION
AGAINST
MUTAGENS AND CARCINOGENS
ICPEMC Working Paper 1/2
A multi-factor ranking scheme for comparing the carcinogenic activity of chemicals Stephen Nesnow Carcinogenesis and Metabolism Branch, Health Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711 (U.S.A.) (Received 31 January 1990) (Accepted 2 February 1990)
Keywords: Cancer; Carcinogens; Potency; Chemicals; B6C3F1 mice; F344 rats; Carcinogen-ranking scheme; TDs0; Highest average daily dose (HADD); Historical control tumour incidence
Summary A scheme for ranking the quantitative activity of chemical carcinogens is described. This activity scheme uses as its base, dose potency measured as TDs0, the chronic dose rate that actuarially halves the adjusted percentage of tumor-free animals at the end of the study (Gold et al., Environ. Health Perspect., 67, 161-200, 1986). The TDs0 is converted into an inverse log scale, a decile scale, and then adjusted by weighting factors that describe other parameters of carcinogenic activity. These factors include positive or negative weightings for: the induction of tumors at tissues or organs associated with high historical control tumor incidences; the induction of tumors at multiple sites; the induction of tumors in both sexes of the species; and the induction of tumors in more than one species. These factors were chosen as they represented qualitative descriptions of the general specificity or non-specificity of chemicals with regard to the activity in r o d e n t s a n d h a v e s o m e b e a r i n g o n the p o t e n t i a l activity o f c h e m i c a l s in h u m a n s . I n o r d e r to
The research described in this article has been reviewed by the Health Effects Research Laboratory, U.S. Environmental Protection Agency and approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of the Agency and no official endorsement should be inferred. 0165-1110/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
84 construct a measure to express the inactivity of chemicals towards the induction of cancer, a measure analogous to the TDs0 has been developed: the highest average daily dose or H A D D . The H A D D is the highest average daily dose in mg chemical/kg body weight administered in a chronic cancer study and that did not induce a statistical increase in tumors. H A D D values were similarly converted to log decile units and adjusted by weighting factors according to lack of activity in both sexes of a species, and the lack of activity in more than one species. In order to explore the use of this multi-factor activity scheme for both carcinogens and non-carcinogens, a group of 142 chemicals was selected that had been tested according to an oral dosing protocol in two sexes of two species of rodents and whose data was peer-reviewed and available for this analysis. This data came from the National Toxicology Program/National Cancer Institute Bioassay Technical Reports. Three activity ranking schemes were developed: the Carcinogen Activity-F344 Rat, an activity scheme based on cancer data obtained with the F344 rat; the Carcinogen Activity-B6C3F1 Mouse, an activity scheme based on cancer data obtained with the B6C3F1 mouse, and the Carcinogen Activity-Combined, an activity scheme based on selecting data from both the F344 rat and the B6C3F1 mouse. This selection was based on the most sensitive rodent responding to the carcinogenic activity of active chemicals and the least sensitive rodent responding to the toxic effects of inactive chemicals. This paper discusses the construction and development of these ranking schemes and analyzes the results in terms of distributions of values within each ranking scheme, and the relative contributions of TDs0 (or H A D D ) and the weighting factors in the activity schemes.
The concept of carcinogenic activity as it relates to the carcinogenic effects of chemicals is one that has had much discussion and debate in the literature. Several models for estimating the carcinogenic activity of chemicals have been proposed in which activity or potency is described in combinations of terms of administered dose, tumor incidence or tumor latency. In 1939, Iball described a potency scheme for chemical carcinogens, the Iball Index: [tumor incidence/tumor latency] × [100] (Iball, 1939). Druckrey (1967) reported a potency scheme, that was termed "indices of carcinogenic dosage" and that related tumor latency (t) and the daily dose ( d ) by the equation: index = [d] × [t2"3]. Meselson and Russell (1977) suggested that carcinogenic potency could be defined as In 2/D1/2 where D1/2 is the daily dose that gives 50% cumulative single-risk incidence of induced cancer after 2 yr of exposure. Peto et al. (1984) and Gold et al. (1984, 1986a, 1987) have produced a cancer potency data base of 975 chemicals based on the TDs0 (tumorigenic dose rate 50), a value calculated from the tumor incidence data that estimates the chronic dose rate needed to halve the actuarially adjusted percentage of tumor-free animals at the end of the standard experiment time. Two additional models
of carcinogenic potency have recently appeared in the literature: Kaldor et al. (1988) have proposed an index of carcinogenic potency for human data, the 10 year cumulative tumor incidence per gram of total dose administered; and Pitot and Campbell (1987) have proposed indices of tumor initiation and tumor promotion. While each of these potency models is useful, there is a need to develop a cancer activity scheme that incorporates other factors in addition to dose in its construct. These factors should account for the specificity or non-specificity of chemicals with respect to the ability to induce cancer in more than one sex of an experimental test species as well as the ability of the chemical to induce tumors in more than one test species. These characteristics are particularly important if the endpoint of concern is the potential of that chemical to induce cancer in man. It is believed that a chemical which is non-specific with regard to tumor site, sex and species of experimental animal is more likely to be a potential hazard to man. This belief arises, in part, from the observations that the majority of chemicals or processes evaluated by the International Agency for Research on Cancer (IARC) and rated as Sufficient Evidence of Carcinogenicity based on human carcinogenicity data are also carcinogenic
85 in experimental animals (Wilbourn et al., 1986) and that most of these human carcinogens are also highly active as genetic toxins (Garrett et al., 1984; Shelby, 1988). Moreover, IARC in its evaluation criteria describing the strength of evidence for the carcinogenic activity of chemicals places weight on chemicals which induce cancer in more than one species of experimental animal. Therefore, the factors of importance for carcinogenic potency in addition to dose include the ability of the chemical to induce tumors in multiple organs, sexes and species of test animals and also at sites associated with low historical control incidences. The incorporation of a number of these factors into a semi-quantitative ranking scheme for animal carcinogens has been proposed by Squire (1981), while Gold et al. (1986b) have suggested that some of these factors be included as additional descriptors in summarizing the potential hazards of chemical carcinogens. This activity scheme would then allow quantitative comparisons to be made between chemicals possessing a broad range of carcinogenic activities. Therefore, in constructing this scheme to quantitate carcinogenic activity it was of interest to integrate dose-response relationships with factors that describe the generality or specificity of chemicals with regard to target-species, sex, organs and tissue sites. In addition to utilizing this scheme to compare the carcinogenic activities of chemicals, this cancer activity scheme could also be compared with similar activity schemes being developed for short-term tests for genotoxic activity. These short-term tests have the ability to detect many kinds of genotoxic damage including gene mutation, chromosomal aberration, morphological cell transformation, and other related endpoints. The development of this cancer activity scheme is a parallel effort with another ICPEMC Committee 1 effort to derive activity scheme for chemicals using combinations of short-term genetic toxicity tests (Brusick et al., in press). This paper describes the construction of such a scheme and applies it to a data set of 142 chemicals taken from the National Toxicology Program and National Cancer Institute Bioassay Technical Reports.
Methods
General considerations and philosophy In considering a cancer activity ranking scheme for chemicals, it would be advantageous to incorporate the following into its construct: (1) An estimate of dose potency as the major determinant. (2) A weighting for tumorigenic responses in organs or tissues that have a low historical control tumor incidence. (3) A weighting for tumorigenic responses in multiple tissues or organs. (4) A weighting for tumorigenic responses in both sexes of the same species. (5) A weighting for tumorigenic responses in multiple species. Since dose potency would be the major determinant, the most reasonable method for applying the weightings would be to directly adjust the dose potency value after converting it to a log-type numeric scale. In applying these weightings, it was necessary to select a base increment adjustment value and the concept of doubling dose was employed for this purpose (National Research Council, 1972). Therefore, weighting factors increasing or decreasing the dose potency value of chemicals by factors of 2 were established. In a relative weighting scheme, a determination must first be made as to the maximum change that will be acceptable when all weighting factors are operating. In the present case, the comparison would be between two chemicals, that both induced tumors at the same dose level (equivalent dose potency). However, one chemical was active in both sexes of one species and at least one sex of a second species. This chemical also produced tumors at multiple sites. The second chemical was active in only one sex of one species and produced only one tumor type. In this analysis, the maximum difference between these two extremes in activity was arbitrarily set to a factor of 2 l° or approximately 1000-fold. Another consideration is the relative weight placed on the presence or absence of data. If an activity scheme evaluates data from multiple sexes and species, a decision needs to be made to accommodate chemicals that have not been tested in all the components of the matrix test group. It was decided not to down-weight for the absence of data but to adjust the activities of chemicals that
86 had complete data sets by the concordance or discordance of results between sexes and between species. In order to construct an activity scheme that compares the responses of different chemicals, it would be desirable to compare data generated by the administration of chemicals by the same route of administration to the same sexes, strain, and species of test animal. Comparisons made between the activity of chemicals would therefore be made using a common set of variables and should be more meaningful. In addition to using the abovementioned variables as a common set, it would be desirable to utilize data derived from standardized protocols, with acceptable numbers of animals treated, sufficient numbers of tissues examined, and appropriate quality control procedures. Selection of N T P / N C I bioassay data as the standard data set
Data that fulfill the criteria stated above can be found in the reports of the National Toxicology Program ( N T P ) / N a t i o n a l Cancer Institute (NCI) Bioassay Program (Ashby and Tennant, 1988; Griesemer and Cueto, 1980; Haseman et al., 1984a, 1986). Generally, chemicals evaluated in that program are administered to both sexes of mice and rats in a two-dose feeding protocol under lifetime exposure and observation periods. The bioassays are performed under strict quality assurance procedures, the bioassay reports are subjected to peer review, and the results are reported in detail. It is for these reasons that the N T P / N C I bioassay data were selected as the standard data set to develop this concept. Selection of oral routes of administration as the standard routes and the F344 rat and the B6C3F1 mouse as the standard rodent strains
In its evolution the N T P / N C I bioassay testing program has used mainly two strains of rats, Osborne-Mendel and F344, one strain of mouse, B6C3F1, and a number of routes of administration. In the construction of this activity scheme, it was decided to utilize only data from the N T P / N C I bioassay reports that described the oral administration (diet, water or gavage) of chemicals to F344 rats and B6C3F1 mice. This would allow the creation of two independent cancer activity
rankings, one based on data generated in the F344 rat and one based on data generated in the B6C3F1 mouse, and each based on one generalized route of administration. The advantages of using this approach are that chemicals are being compared in the same strain of rodent using the same route of administration and using similar protocols for dosing, length of treatment, time of observation, number of animals, number of doses, tissue selection, and levels of histopathologic observation. The use of two separate activity schemes for two specific species and strain combinations circumvents some of the problems inherent with chemicals that have dissimilar potencies in different strains or species of rodent. The disadvantage of this approach is that chemicals that are only active in strains of rodents other than F344 rats or B6C3F1 mice will not be detected or compared. Selection of the TDso values as a measure of activity
In the search for a method to represent the carcinogenic activity of chemicals it was immediately recognized that the major publications of Peto et al. (1984) and Gold et al. (1984, 1986a, 1987) on the development of a carcinogen activity data base by use of the TDs0 concept would greatly aid in this effort. These reports present the calculated TDs0 potencies of 975 chemicals. Of the 975 chemicals, a significant number of chemicals were bioassayed by the N T P / N C I and could be used in this analysis. Selection of the H A D D value as a measure of inactivity
For those chemicals that were bioassayed and not found to induce significant numbers of tumors in specific sex-species combinations, the highest average daily dose ( H A D D ) administered to the test animals was used as a measure of inactivity. The H A D D values in mg chemical administered/ kg body weight/day were calculated according to the methods described by Gold et al. (1984) from the highest dose administered as described in the N T P / N C I Technical Reports. Selection of chemicals for inclusion into the standard data set
Chemicals (mixtures were not included) evaluated by the N T P / N C I in both F344 rats and
87 B6C3F1 mice in which the route of administration was in the diet, by gavage, or in the drinking water were selected for development of this concept and incorporated into this activity scheme according to the following criteria: (1) The summary evaluation for each sexspecies combination (male rat, female rat, male mouse, female mouse) given in the Technical Report or as found in the Gene-Tox Carcinogen Data Base (Nesnow et al., 1987) was used for the categorization of each study as positive, negative, inadequate or equivocal; and only studies that were rated positive or negative were utilized in this report. (2) The selection of individual TDs0 values found in Gold et al. (1984, 1986a) for each sexspecies combination of each chemical was predicated on three rules: (A) If several TDs0 values were available for a sex-species combination then the one with the lowest value was chosen (the most sensitive endpoint). (B) In the Gold et al. (1984, 1986a) reports, several TDs0 values are calculated for each chemical. Only TDs0 values with a probability of p < 0.05 were chosen. (C) TDs0 values were only used for those studies rated as positive in the N T P / N C I technical report. (3) Chemicals that were tested b y the N T P / N C I in the F344 rat and the B6C3F1 mouse by oral routes of administration and that were found not to induce tumors in all four sex-species combinations were also included with their H A D D values. In total, 142 chemicals were selected. Table 1 lists each chemical, the Chemical Abstract Registry (CAS) No., the N T P / N C I Technical Report No., the TDso values in mg chemical administered/kg body w e i g h t / d a y for those chemicals that induced tumorigenic effects in specific sexspecies combinations, and the H A D D values in mg chemical administered/kg body w e i g h t / d a y for the chemicals that were inactive in specific sex-species combinations.
Conversion of TDso values and HADD values into decile units In the use of the TDs0 data base, a method was
needed that could be applied to the TDs0 values and H A D D values to convert them into a simpler numeric scale for comparisons among chemicals. A log-type conversion was chosen to encompass the wide range (109 ) of values encountered in the data base. These converted TDs0 values are termed decile units and are inversely related to TDs0 values by the formula: decile value (active chemical) = log[107/TDs0] Decile values for active chemicals (carcinogens) are positive. It was decided to scale the active chemicals between decile values of 0 to 14 to accommodate chemicals like 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) on one extreme and chemicals like C.I. Vat Yellow 4 on the other. Therefore, a scaling factor of 107 was chosen in the conversion equation. Although T C D D is not in the tables in this study because it had been tested in the Osborne-Mendel rat (not the F344 rat), the decile value for the B6C3F1 female mouse is 9.9, the smallest of the four test groups. Since it was active in both sexes of both species and since it produced hepatocellular carcinomas in the female mouse, a tumor with a low historical control tumor incidence, the final Carcinogen Activity-Combined value for T C D D would be 12.9 [see further descriptions of these factors and their values]. To allow for a chemical that might be more potent than T C D D by a factor of 10, the upper range of the positive decile scale was then set at 14. C.I. Vat Yellow 4 had the highest TDs0 value, 10400, which when converted into decile units was 2.96. The scaling factor of 107 allows for a new chemical that might produce a higher TDs0 (by a factor of 3), and also allows for potential down-weighting of that value without creating a negative value. A similar concept was used to convert the H A D D values for the inactive chemicals (noncarcinogens) with the formula: decile value (inactivechemical) = -log[abs(HADD)
X
103]
Decile values for inactive chemicals are negative. It was decided to scale the decile values of inactive chemicals from - 1 to - 10. This was based mainly on the two extreme H A D D values found in Table
88
~k © Z Q
I
©
I
I
|
I
I
I
I
I
I
I
~ ,I
I
I
z
8z
83
MUTREV 07281
INTERNATIONAL
COMMISSION
ENVIRONMENTAL
FOR PROTECTION
AGAINST
MUTAGENS AND CARCINOGENS
ICPEMC Working Paper 1/2
A multi-factor ranking scheme for comparing the carcinogenic activity of chemicals Stephen Nesnow Carcinogenesis and Metabolism Branch, Health Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711 (U.S.A.) (Received 31 January 1990) (Accepted 2 February 1990)
Keywords: Cancer; Carcinogens; Potency; Chemicals; B6C3F1 mice; F344 rats; Carcinogen-ranking scheme; TDs0; Highest average daily dose (HADD); Historical control tumour incidence
Summary A scheme for ranking the quantitative activity of chemical carcinogens is described. This activity scheme uses as its base, dose potency measured as TDs0, the chronic dose rate that actuarially halves the adjusted percentage of tumor-free animals at the end of the study (Gold et al., Environ. Health Perspect., 67, 161-200, 1986). The TDs0 is converted into an inverse log scale, a decile scale, and then adjusted by weighting factors that describe other parameters of carcinogenic activity. These factors include positive or negative weightings for: the induction of tumors at tissues or organs associated with high historical control tumor incidences; the induction of tumors at multiple sites; the induction of tumors in both sexes of the species; and the induction of tumors in more than one species. These factors were chosen as they represented qualitative descriptions of the general specificity or non-specificity of chemicals with regard to the activity in r o d e n t s a n d h a v e s o m e b e a r i n g o n the p o t e n t i a l activity o f c h e m i c a l s in h u m a n s . I n o r d e r to
The research described in this article has been reviewed by the Health Effects Research Laboratory, U.S. Environmental Protection Agency and approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of the Agency and no official endorsement should be inferred. 0165-1110/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
84 construct a measure to express the inactivity of chemicals towards the induction of cancer, a measure analogous to the TDs0 has been developed: the highest average daily dose or H A D D . The H A D D is the highest average daily dose in mg chemical/kg body weight administered in a chronic cancer study and that did not induce a statistical increase in tumors. H A D D values were similarly converted to log decile units and adjusted by weighting factors according to lack of activity in both sexes of a species, and the lack of activity in more than one species. In order to explore the use of this multi-factor activity scheme for both carcinogens and non-carcinogens, a group of 142 chemicals was selected that had been tested according to an oral dosing protocol in two sexes of two species of rodents and whose data was peer-reviewed and available for this analysis. This data came from the National Toxicology Program/National Cancer Institute Bioassay Technical Reports. Three activity ranking schemes were developed: the Carcinogen Activity-F344 Rat, an activity scheme based on cancer data obtained with the F344 rat; the Carcinogen Activity-B6C3F1 Mouse, an activity scheme based on cancer data obtained with the B6C3F1 mouse, and the Carcinogen Activity-Combined, an activity scheme based on selecting data from both the F344 rat and the B6C3F1 mouse. This selection was based on the most sensitive rodent responding to the carcinogenic activity of active chemicals and the least sensitive rodent responding to the toxic effects of inactive chemicals. This paper discusses the construction and development of these ranking schemes and analyzes the results in terms of distributions of values within each ranking scheme, and the relative contributions of TDs0 (or H A D D ) and the weighting factors in the activity schemes.
The concept of carcinogenic activity as it relates to the carcinogenic effects of chemicals is one that has had much discussion and debate in the literature. Several models for estimating the carcinogenic activity of chemicals have been proposed in which activity or potency is described in combinations of terms of administered dose, tumor incidence or tumor latency. In 1939, Iball described a potency scheme for chemical carcinogens, the Iball Index: [tumor incidence/tumor latency] × [100] (Iball, 1939). Druckrey (1967) reported a potency scheme, that was termed "indices of carcinogenic dosage" and that related tumor latency (t) and the daily dose ( d ) by the equation: index = [d] × [t2"3]. Meselson and Russell (1977) suggested that carcinogenic potency could be defined as In 2/D1/2 where D1/2 is the daily dose that gives 50% cumulative single-risk incidence of induced cancer after 2 yr of exposure. Peto et al. (1984) and Gold et al. (1984, 1986a, 1987) have produced a cancer potency data base of 975 chemicals based on the TDs0 (tumorigenic dose rate 50), a value calculated from the tumor incidence data that estimates the chronic dose rate needed to halve the actuarially adjusted percentage of tumor-free animals at the end of the standard experiment time. Two additional models
of carcinogenic potency have recently appeared in the literature: Kaldor et al. (1988) have proposed an index of carcinogenic potency for human data, the 10 year cumulative tumor incidence per gram of total dose administered; and Pitot and Campbell (1987) have proposed indices of tumor initiation and tumor promotion. While each of these potency models is useful, there is a need to develop a cancer activity scheme that incorporates other factors in addition to dose in its construct. These factors should account for the specificity or non-specificity of chemicals with respect to the ability to induce cancer in more than one sex of an experimental test species as well as the ability of the chemical to induce tumors in more than one test species. These characteristics are particularly important if the endpoint of concern is the potential of that chemical to induce cancer in man. It is believed that a chemical which is non-specific with regard to tumor site, sex and species of experimental animal is more likely to be a potential hazard to man. This belief arises, in part, from the observations that the majority of chemicals or processes evaluated by the International Agency for Research on Cancer (IARC) and rated as Sufficient Evidence of Carcinogenicity based on human carcinogenicity data are also carcinogenic
85 in experimental animals (Wilbourn et al., 1986) and that most of these human carcinogens are also highly active as genetic toxins (Garrett et al., 1984; Shelby, 1988). Moreover, IARC in its evaluation criteria describing the strength of evidence for the carcinogenic activity of chemicals places weight on chemicals which induce cancer in more than one species of experimental animal. Therefore, the factors of importance for carcinogenic potency in addition to dose include the ability of the chemical to induce tumors in multiple organs, sexes and species of test animals and also at sites associated with low historical control incidences. The incorporation of a number of these factors into a semi-quantitative ranking scheme for animal carcinogens has been proposed by Squire (1981), while Gold et al. (1986b) have suggested that some of these factors be included as additional descriptors in summarizing the potential hazards of chemical carcinogens. This activity scheme would then allow quantitative comparisons to be made between chemicals possessing a broad range of carcinogenic activities. Therefore, in constructing this scheme to quantitate carcinogenic activity it was of interest to integrate dose-response relationships with factors that describe the generality or specificity of chemicals with regard to target-species, sex, organs and tissue sites. In addition to utilizing this scheme to compare the carcinogenic activities of chemicals, this cancer activity scheme could also be compared with similar activity schemes being developed for short-term tests for genotoxic activity. These short-term tests have the ability to detect many kinds of genotoxic damage including gene mutation, chromosomal aberration, morphological cell transformation, and other related endpoints. The development of this cancer activity scheme is a parallel effort with another ICPEMC Committee 1 effort to derive activity scheme for chemicals using combinations of short-term genetic toxicity tests (Brusick et al., in press). This paper describes the construction of such a scheme and applies it to a data set of 142 chemicals taken from the National Toxicology Program and National Cancer Institute Bioassay Technical Reports.
Methods
General considerations and philosophy In considering a cancer activity ranking scheme for chemicals, it would be advantageous to incorporate the following into its construct: (1) An estimate of dose potency as the major determinant. (2) A weighting for tumorigenic responses in organs or tissues that have a low historical control tumor incidence. (3) A weighting for tumorigenic responses in multiple tissues or organs. (4) A weighting for tumorigenic responses in both sexes of the same species. (5) A weighting for tumorigenic responses in multiple species. Since dose potency would be the major determinant, the most reasonable method for applying the weightings would be to directly adjust the dose potency value after converting it to a log-type numeric scale. In applying these weightings, it was necessary to select a base increment adjustment value and the concept of doubling dose was employed for this purpose (National Research Council, 1972). Therefore, weighting factors increasing or decreasing the dose potency value of chemicals by factors of 2 were established. In a relative weighting scheme, a determination must first be made as to the maximum change that will be acceptable when all weighting factors are operating. In the present case, the comparison would be between two chemicals, that both induced tumors at the same dose level (equivalent dose potency). However, one chemical was active in both sexes of one species and at least one sex of a second species. This chemical also produced tumors at multiple sites. The second chemical was active in only one sex of one species and produced only one tumor type. In this analysis, the maximum difference between these two extremes in activity was arbitrarily set to a factor of 2 l° or approximately 1000-fold. Another consideration is the relative weight placed on the presence or absence of data. If an activity scheme evaluates data from multiple sexes and species, a decision needs to be made to accommodate chemicals that have not been tested in all the components of the matrix test group. It was decided not to down-weight for the absence of data but to adjust the activities of chemicals that
86 had complete data sets by the concordance or discordance of results between sexes and between species. In order to construct an activity scheme that compares the responses of different chemicals, it would be desirable to compare data generated by the administration of chemicals by the same route of administration to the same sexes, strain, and species of test animal. Comparisons made between the activity of chemicals would therefore be made using a common set of variables and should be more meaningful. In addition to using the abovementioned variables as a common set, it would be desirable to utilize data derived from standardized protocols, with acceptable numbers of animals treated, sufficient numbers of tissues examined, and appropriate quality control procedures. Selection of N T P / N C I bioassay data as the standard data set
Data that fulfill the criteria stated above can be found in the reports of the National Toxicology Program ( N T P ) / N a t i o n a l Cancer Institute (NCI) Bioassay Program (Ashby and Tennant, 1988; Griesemer and Cueto, 1980; Haseman et al., 1984a, 1986). Generally, chemicals evaluated in that program are administered to both sexes of mice and rats in a two-dose feeding protocol under lifetime exposure and observation periods. The bioassays are performed under strict quality assurance procedures, the bioassay reports are subjected to peer review, and the results are reported in detail. It is for these reasons that the N T P / N C I bioassay data were selected as the standard data set to develop this concept. Selection of oral routes of administration as the standard routes and the F344 rat and the B6C3F1 mouse as the standard rodent strains
In its evolution the N T P / N C I bioassay testing program has used mainly two strains of rats, Osborne-Mendel and F344, one strain of mouse, B6C3F1, and a number of routes of administration. In the construction of this activity scheme, it was decided to utilize only data from the N T P / N C I bioassay reports that described the oral administration (diet, water or gavage) of chemicals to F344 rats and B6C3F1 mice. This would allow the creation of two independent cancer activity
rankings, one based on data generated in the F344 rat and one based on data generated in the B6C3F1 mouse, and each based on one generalized route of administration. The advantages of using this approach are that chemicals are being compared in the same strain of rodent using the same route of administration and using similar protocols for dosing, length of treatment, time of observation, number of animals, number of doses, tissue selection, and levels of histopathologic observation. The use of two separate activity schemes for two specific species and strain combinations circumvents some of the problems inherent with chemicals that have dissimilar potencies in different strains or species of rodent. The disadvantage of this approach is that chemicals that are only active in strains of rodents other than F344 rats or B6C3F1 mice will not be detected or compared. Selection of the TDso values as a measure of activity
In the search for a method to represent the carcinogenic activity of chemicals it was immediately recognized that the major publications of Peto et al. (1984) and Gold et al. (1984, 1986a, 1987) on the development of a carcinogen activity data base by use of the TDs0 concept would greatly aid in this effort. These reports present the calculated TDs0 potencies of 975 chemicals. Of the 975 chemicals, a significant number of chemicals were bioassayed by the N T P / N C I and could be used in this analysis. Selection of the H A D D value as a measure of inactivity
For those chemicals that were bioassayed and not found to induce significant numbers of tumors in specific sex-species combinations, the highest average daily dose ( H A D D ) administered to the test animals was used as a measure of inactivity. The H A D D values in mg chemical administered/ kg body weight/day were calculated according to the methods described by Gold et al. (1984) from the highest dose administered as described in the N T P / N C I Technical Reports. Selection of chemicals for inclusion into the standard data set
Chemicals (mixtures were not included) evaluated by the N T P / N C I in both F344 rats and
87 B6C3F1 mice in which the route of administration was in the diet, by gavage, or in the drinking water were selected for development of this concept and incorporated into this activity scheme according to the following criteria: (1) The summary evaluation for each sexspecies combination (male rat, female rat, male mouse, female mouse) given in the Technical Report or as found in the Gene-Tox Carcinogen Data Base (Nesnow et al., 1987) was used for the categorization of each study as positive, negative, inadequate or equivocal; and only studies that were rated positive or negative were utilized in this report. (2) The selection of individual TDs0 values found in Gold et al. (1984, 1986a) for each sexspecies combination of each chemical was predicated on three rules: (A) If several TDs0 values were available for a sex-species combination then the one with the lowest value was chosen (the most sensitive endpoint). (B) In the Gold et al. (1984, 1986a) reports, several TDs0 values are calculated for each chemical. Only TDs0 values with a probability of p < 0.05 were chosen. (C) TDs0 values were only used for those studies rated as positive in the N T P / N C I technical report. (3) Chemicals that were tested b y the N T P / N C I in the F344 rat and the B6C3F1 mouse by oral routes of administration and that were found not to induce tumors in all four sex-species combinations were also included with their H A D D values. In total, 142 chemicals were selected. Table 1 lists each chemical, the Chemical Abstract Registry (CAS) No., the N T P / N C I Technical Report No., the TDso values in mg chemical administered/kg body w e i g h t / d a y for those chemicals that induced tumorigenic effects in specific sexspecies combinations, and the H A D D values in mg chemical administered/kg body w e i g h t / d a y for the chemicals that were inactive in specific sex-species combinations.
Conversion of TDso values and HADD values into decile units In the use of the TDs0 data base, a method was
needed that could be applied to the TDs0 values and H A D D values to convert them into a simpler numeric scale for comparisons among chemicals. A log-type conversion was chosen to encompass the wide range (109 ) of values encountered in the data base. These converted TDs0 values are termed decile units and are inversely related to TDs0 values by the formula: decile value (active chemical) = log[107/TDs0] Decile values for active chemicals (carcinogens) are positive. It was decided to scale the active chemicals between decile values of 0 to 14 to accommodate chemicals like 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) on one extreme and chemicals like C.I. Vat Yellow 4 on the other. Therefore, a scaling factor of 107 was chosen in the conversion equation. Although T C D D is not in the tables in this study because it had been tested in the Osborne-Mendel rat (not the F344 rat), the decile value for the B6C3F1 female mouse is 9.9, the smallest of the four test groups. Since it was active in both sexes of both species and since it produced hepatocellular carcinomas in the female mouse, a tumor with a low historical control tumor incidence, the final Carcinogen Activity-Combined value for T C D D would be 12.9 [see further descriptions of these factors and their values]. To allow for a chemical that might be more potent than T C D D by a factor of 10, the upper range of the positive decile scale was then set at 14. C.I. Vat Yellow 4 had the highest TDs0 value, 10400, which when converted into decile units was 2.96. The scaling factor of 107 allows for a new chemical that might produce a higher TDs0 (by a factor of 3), and also allows for potential down-weighting of that value without creating a negative value. A similar concept was used to convert the H A D D values for the inactive chemicals (noncarcinogens) with the formula: decile value (inactivechemical) = -log[abs(HADD)
X
103]
Decile values for inactive chemicals are negative. It was decided to scale the decile values of inactive chemicals from - 1 to - 10. This was based mainly on the two extreme H A D D values found in Table
88
~k © Z Q
I
©
I
I
|
I
I
I
I
I
I
I
~ ,I
I
I
z
8z
