See also the Guide to the Plot of the CPDB and Guide to the Screen Version of the CPDB.
During the past forty-five years, thousands of chronic, long-term animal cancer tests have been conducted to identify chemicals that might be hazardous to humans. The results provide important information about carcinogenesis of individual chemicals, and are a source of data that are used in efforts to protect public health. The published results of these experiments constitute a diverse literature that varies widely with respect to experimental and histological protocols as well as to how and which information is reported in published articles.
The Carcinogenic Potency Database (CPDB) is a systematic and unifying analysis of chronic, long-term animal cancer tests. It standardizes the published literature and creates an easily accessible research resource that can be and has been used to address a wide variety of research and regulatory issues in carcinogenesis. The CPDB includes results reported in 1426 papers in the general literature through 1997 and 429 Technical Reports of the National Cancer Institute/National Toxicology Program (NCI/NTP) through 1998.
An experiment is defined in the CPDB as the control and dose groups under a set of bioassay conditions that includes a single species, sex, strain of mammal administered a test agent by a given route for a given exposure and experiment length. Results are reported for 6153 experiments on 1485 chemical agents; these are displayed in a plot format organized by chemical name. Of 1485 chemicals in the CPDB 415 (28%) have results in the CPDB from NCI/NTP. Detailed information that is important in the interpretation of bioassays, is reported on each experiment (whether positive or negative for carcinogenicity), including: qualitative information on strain, sex, route of compound administration, target organ, histopathology, author’s opinion about carcinogenicity, and reference to the published paper, as well as qualitative information on daily dose-rate, duration of dosing, length of experiment, tumor incidence, dose-response curve shape. A measure of carcinogenic potency, TD50 (tumorigenic dose-rate for 50% of experimental animals), its confidence limits, and statistical significance are estimated for each site reported in the database. Each set of experimental results references the original published paper. A supplementary database details dosing and survival information for each dose group in each experiment.
A word of caution is necessary about the limitations of the CPDB. No attempt has been made by CPDB staff to evaluate whether or not a compound induced tumors in any given experiment; rather, the opinion of the published author is presented for each target site as well as the statistical significance of the TD50 calculated from the experimental results. Additionally, the database contains only long-term tests which fit a set of inclusion criteria, and therefore does not cover all cancer tests.
This document describes the methods used to develop the CPDB. Also available on the Website are a guide to each of the variables in the plot of cancer test results, the plot itself, a screen version of the plot, appendices of codes and definitions, and a bibliography of the original papers and NCI/NTP Technical Reports that serve as the source of data. The supplementary database of dosing and survival provides experimental details and can be linked to CPDB.
Two additional tables systematically summarize the CPDB results 1) by target organ, listing all chemicals in each species that induce tumors at a specific site such as kidney; and 2) by chemical name summarizing for each chemical in each species the carcinogenic potency, positivity, and target organs from all experiments on the chemical. The summary table by chemical also reports mutagenicity in Salmonella for each chemical. These two summary tables are also provided for results of NCI/NTP bioassays separately.
Together, the plot, bibliography, supplementary database and the two summary tables provide a guide to the literature of chronic, long-term animal cancer tests as well as an accessible research resource.
A numerical description of carcinogenic potency, the TD50, is estimated for each set of tumor incidence data reported in the CPDB, thus providing a standardized quantitative measure of the tumorigenic dose-rate that can be used for comparisons and analyses of many issues in carcinogenesis. In a simplified way, TD50 may be defined as follows: for a given target site(s), if there are no tumors in control animals, then TD50 is that chronic dose-rate in mg/kg body wt/day which would induce tumors in half the test animals at the end of a standard lifespan for the species. Since the tumor(s) of interest often does occur in control animals, TD50 is more precisely defined as: that dose-rate in mg/kg body wt/day which, if administered chronically for the standard lifespan of the species, will halve the probability of remaining tumorless throughout that period. TD50 is analogous to LD50, and a low value of TD50 indicates a potent carcinogen, whereas a high value indicates a weak one. TD50 can be computed for any particular type of neoplasm, for any particular tissue, or for any combination of these. Our numerical index of carcinogenic potency, the TD50, and the statistical procedures adopted for estimating it from experimental data, have been described in detail elsewhere (Sawyer et al., 1984; Peto et al., 1984). See Statistical Methods for Estimating TD50.
One goal of the CPDB has been to obtain data which would give the best estimates of carcinogenic potency. Information on the time of death and tumor pathology for each animal were available from bioassays of the NCI/NTP, from a set of NCI bioassays on aromatic amines (Russfield et al., 1973, Russfield et al., 1975, Weisburger et al., 1978), and from tests in nonhuman primates by the NCI Laboratory of Chemical Pathology ( Thorgeirsson et al., 1994), and potency values were calculated using lifetable analysis. For estimates of potency from the general literature, we have calculated TD50 using the final proportions of animals with tumors, since only this summary information is consistently published. When reported in the published paper, the number of animals alive at the time of the first tumor is used to calculate TD50, and if that is not given in the paper then the number of animals examined histologically.
The inclusion criteria for the CPDB are designed to identify reasonably thorough, chronic, long-term tests of single chemical agents (whether positive or negative). The two sources of data are the bioassays of the NCI/NTP and the general published literature.
The CPDB provides a concise summary of results of the NCI/NTP bioassays that meet the inclusion rules of the CPDB. All Technical Reports are included except those for which no comparable dose in mg/kg/day could be calculated e.g., because the test agent was a particulate, the route of administration was dermal or subcutaneous injection, or exposure occurred in utero. The NCI/NTP program was designed to use a similar experimental protocol for a large number of compounds, and to report results in a consistent format including full histopathology. Since the 1970s, the NCI/NTP has provided us with computer files that contain information about the time of death and full histopathology for each animal in each experiment. Currently, the files are on the NTP Web site. By using these files in combination with the series of Technical Reports published by NCI/NTP we have been able to include in the CPDB each site in each Technical Report that was evaluated as evidence of carcinogenicity, as well as each site reported in the Technical Report Statistical Analysis Tables as statistically significant but not considered by NCI/NTP as “carcinogenic”, “clear,” “some” or “equivocal” evidence of carcinogenicity. Tissue-tumor combinations reported in the CPDB thus depend on which statistical tests are reported in Technical Reports, and these have changed over time.
The standard NCI/NTP bioassay protocol in the 1970s recommended that tests be conducted in two species of rodents (rats and mice) with both sexes tested individually at the maximally tolerated dose (MTD) and half that dose, using a control group and a vehicle control where appropriate (Sontag et al., 1976). In NCI/NTP Technical Reports since the early 1990s the standard number of dose groups has been increased to 3, and the standard range of doses tested is 4-10 fold. The MTD is generally accepted to be defined as the maximum dose level which is not expected to shorten the normal longevity from non-neoplastic causes, and which is expected to result in no more than a 10% weight decrement in animals receiving this dose when compared to controls (Sontag et al., 1976). The actual conduct of the bioassays published prior to July 1980 varied from one experiment to another, particularly with respect to the number of animals per group and the experiment length. Comparisons of results using NCI/NTP bioassays are particularly important because of the quality of the experiments and histopathology, peer review, the detailed reporting of results and the standardization of protocols. Complete histopathology and time of death for each animal are reported; terminal sacrifice was usually performed at 2 years; dosing was continued for the majority of life; the same mouse hybrid, B6C3F1 was used throughout; rat bioassays utilize the Fischer 344 rat (for early studies by gavage Osborne-Mendel was used, and for early tests of chemotherapy agents by intraperitoneal injection Sprague Dawley was used).
In the general literature, experimental designs as well as the authors’ choice of which information to report, are quite diverse. For the CPDB we have developed a set of standard inclusion criteria, and experiments from the literature have been included only if they meet all of the following conditions:
A search of the published literature has been conducted through 1997 for all bioassays which met the standard criteria, whether or not the authors considered the test agent related to tumor induction. The literature search covers: the Survey of Compounds Which Have Been Tested for Carcinogenic Activity (formerly PHS 149), Medline, Current Contents, the Monographs on chemical carcinogens prepared by the International Agency for Research on Cancer, the FDA database of food additives, and the Japanese Science and Technology Database.
Separate journal searches have also been done: British Journal of Cancer, Cancer Letters, Cancer Research, Carcinogenesis, Chemosphere, Environmental Health Perspectives, European Journal of Cancer and Clinical Oncology, Food and Chemical Toxicology, Fundamental and Applied Toxicology, International Journal of Cancer, Japanese Journal of Cancer Research, Journal of Cancer Research and Clinical Oncology, Journal of Environmental Pathology and Toxicology, Journal of Toxicology and Environmental Health, Journal of the National Cancer Institute, Regulatory Toxicology and Pharmacology, Toxicology, and Toxicology and Applied Pharmacology. From the literature search, 4471 experiments on 1207 chemicals (excluding NCI/NTP) met the inclusion criteria and are in the database.
Since we have adhered quite strictly to the standard inclusion criteria, bioassays of particulate or fibrous matters are not included, e.g., asbestos, cigarette smoke and dusts. There are no single injection or skin painting tests, and no co-carcinogenesis studies in which more than one chemical is administered to the same animal.
Although we have excluded co-carcinogenesis experiments from the database, some mixtures to which humans are exposed are included, i.e., commercial preparations and technical grade materials including pesticides, industrial compounds, and drugs.
Including experiments in both NCI/NTP Technical Reports and the general literature, results of 6153 experiments are summarized in the CPDB. About 3% of the experiments were conducted in hamsters, less than 1% in nonhuman primates or dogs, and the rest are about evenly divided between rats and mice. Experiments with 120 different mouse strains and 91 rat strains are included. For any single chemical, the number of experiments in the database may vary. Some chemicals have only one test in one sex of one species, while others have multiple tests including both sexes of a few strains of rats and mice, possibly using quite different protocols. For example, among the 1165 chemicals tested in rats, 26% have only 1 test and 51% have 2 tests; however, 26 chemicals have more than 10 tests. For the 957 chemicals tested in mice, the parallel numbers are 11% with 1 test, 59% with 2 tests, and 18 chemicals with more than 10 tests. Additionally, the number of tissues and tumors reported for an experiment varies, depending upon the results of the experiment and the extent of detail in the published paper.
A wide variety of experimental protocols meet the inclusion rules of the CPDB, e.g. with respect to the number of doses, the range of doses tested, and group size. The standard NCI/NTP protocol in Technical Reports in the 1970s and 1980s was 2 dose groups and a control (only 5% of experiments had 3 groups). In contrast, in Reports in the 1990s, 59% of experiments had 3 dose groups. Among experiments in rats or mice from the general literature, 57% have only one dose group and a control, 17% have 2 dose groups, and 26% have 3 or more.
There is also variation in the CPDB with respect to the range of doses tested within an individual experiment: Among NCI/NTP bioassays, 74% are tested within a 2-fold range of dose, 22% within a >2 to 10-fold range, and 4% more than 10-fold. The parallel percentages for literature experiments in rats or mice are: 57% have only one dose group, 9% are within a 2-fold range, 21% >2 to 10-fold, and 13% greater than 10-fold. The standard NCI/NTP protocol is 50 animals per group of dosed animals, but in the literature the group size varies widely: 42% of experiments have 20 or fewer animals per dose group, and 35% have more than 40 per group.
In order to standardize the diverse bioassay literature, a great many decisions have had to be made in the construction of the database. These decisions were based on two major considerations: 1) to provide a large resource with comparable standardized data to facilitate the use by researchers, government agencies and the public of results of chronic, long-term animal cancer tests conducted in different laboratories around the world, and 2) to provide estimates of carcinogenic potency that are based on the most complete information available.
Our choice of which sites and pathology to report from the published literature is limited by what individual authors have reported, and this varies considerably from paper to paper. We have been able to augment the published results through personal communication with authors of published papers, particularly for experiments conducted recently, e.g., to clarify opinions as to carcinogenicity at particular sites, to obtain additional tumor data or separate data for interim sacrifice experiments, or the number of animals alive at the time of the first tumor. Thus, the CPDB improves the published literature and includes some results or more accurate results that are not reported in the original papers. All such cases are indicated on the plot of the CPDB as having had personal communication with a published author, “pers.comm.”
For reasons of both accuracy and consistency throughout the database, our general approach has been to include each category of neoplasm, benign or malignant, which an author evaluated as treatment-related, regardless of the statistical or biological basis for the evaluation. (Hyperplasia and non-neoplastic lesions are not included in the database.) In addition, for comparative purposes, the category “all tumor-bearing animals” (TBA) is included wherever this was reported. We have not included results that indicate a reduction of tumor incidence in dosed animals compared to controls.
In order to provide information which would permit comparisons of target sites across experiments, an additional category of tissue/tumor results is included in the database, “mandatory sites.” Whenever there is adequate documentation in the published paper or Report, results are reported for hepatocellular tumors in rats, mice and hamsters; also for tumors of the lung in mice and hamsters. Negative experiments (without evidence of carcinogenicity) are included in the CPDB with results for these mandatory sites. These tissues were selected as mandatory sites because they occurred most often in a frequency count of positive sites in the species in the NCI/NTP bioassays before 1980. These sites continue to be the most frequent target sites in rats and mice in the CPDB to date.
The selection of tissue-tumor combinations to report in the CPDB for each experiment is determined by a set of rules used throughout the database. Whenever the published paper has the following information, it is included:
Some special considerations about selection of results to include in CPDB from each source of data are as follows:
All sites are reported in the CPDB for which the NCI evaluation was “carcinogenic,” or “associated with carcinogenicity”; or the NTP evaluation was “clear evidence”, “some evidence” or “equivocal evidence” of carcinogenic activity. In the absence of such an evaluation, sites which are reported in the statistical tables but which are not evaluated as treatment-related, are considered for inclusion in the CPDB based on the statistical significance of the tests reported in the Technical Report. The statistical tests in Technical Reports have changed over time. Statistical sites are included in the CPDB if the slope of the dose response is different from zero (two-tailed p-value < 0.05). We refer to these cases as “statistical sites” to indicate that the results were not evaluated in the Technical Report as evidence of carcinogenicity but are included in the Statistical Analysis Tables.
Because the computer files from the individual animal pathology tables of the NCI/NTP contain pathology on each animal, it has been possible to report a composite category for the incidence of all tumor types which the NCI/NTP evaluated as evidence of carcinogenicity. It has generally not been possible to formulate such a composite category for other experiments in the database, with the exception of a group of studies on aromatic amines, bioassays in monkeys, and a few other experiments.
In the NCI/NTP data, the liver mandatory site for Technical Reports prior to 1993 included any of the following: neoplastic nodule, hepatocellular adenoma, and hepatocellular carcinoma. Subsequently, the liver mandatory site for NTP includes hepatocellular adenoma, hepatoblastoma, and hepatocellular carcinoma. This change reflects diagnoses and nomenclature given in the Technical Reports. For the lung, alveolar/bronchiolar adenoma and alveolar/bronchiolar carcinoma are combined. In all cases, the incidence represents the proportion of animals with any of the tumors; animals with multiple tumors of the given tissue are counted only once. For the category TBA, we have excluded interstitial-cell tumors of the testis for Fischer 344 rats, since these tumors occur spontaneously in nearly all male animals by the end of their lifespan.
For several recent NTP bioassays, multiple additional sections were taken of the kidneys. We include incidence data from the standard protocol and also incidence data that includes the step section results. We indicate the step section results on the CPDB plot by the words “with step”, and we report these results as a separate line number on the plot.
The great diversity over time in the published bioassay literature necessitates many decisions about inclusion of experiments or particular tissue-tumor combinations. We have been able to clarify and augment the results in published papers through personal communication with hundreds of authors of published papers, e.g., to clarify opinions as to carcinogenicity at particular sites, to obtain additional tumor data or separate data for interim sacrifice experiments, and to correct inconsistencies in the results reported in the paper. We correspond with about 50% of researchers and indicate this correspondence in the reference field of the plot with the words "pers.comm."
Authors of papers in the general literature rarely indicate which animals were diagnosed as having tumors at more than one site. Therefore, it has been generally impossible to combine data on various tumors within a single tissue. To attempt to combine incidence data would risk multiple counting of animals.
In the general literature, when we report results separately for benign and malignant tumors at a given site in a single experiment, we often do not know whether the pathologist scored animals for the most malignant tumor. For example, we may not know whether the proportion of animals with adenomas represents only those animals that did not have carcinomas or rather all animals with adenomas regardless of whether they had a carcinoma. Therefore, in such cases no mix could be created of benign and malignant tumors at a given site. We have frequent correspondence with researchers to obtain tumor incidence rates for benign and malignant tumors combined at a given site, and it has been possible to report such combinations in those cases. Tumor incidence data are based on the number of animals alive at the first tumor in the experiment whenever that information is provided in the published paper. Additional information on survival for each experiment is given in the Supplementary dataset on dosing and survival.
Because a variety of routes of administration, dosing schedules, species, strains, and sexes are used in carcinogenesis bioassays, some standardization of dose is required. Our convention is to determine for each dose group in an experiment the daily dose-rate in mg/kg of body weight averaged over the duration of the experiment, using standard intake values for each sex-species group.
To convert ppm or percent administered in food, water, or air, into mg/kg body weight during the dosing period, we assume 100% absorption and then use a set of standard values for each sex of each species, including body weight and average intake per day (Table 1). Using standard values, the daily dose-rate is calculated as follows:
Dose rate = (administered dose × intake/day × number of doses/week) / (animal weight × 7 days/week)
Dose rate = dose × intake/day as a proportion of body weight × proportion of week dose is administered
|Experimental Animal||Sex||Standard lifespan (year)b||Weight (kg)c||Food (g)c||Food as % body weight/day||Water (mL/day)d||Inhalation volume (L/min)e|
|Cynomolgus (Macaca fascicularis)||Both||20|
|Rhesus (Macaca mulatta)||Both||20|
|Bush babies (Galago crassicaudatus)||Both||10|
|Tree shrews (Tupaia glis)||Both||4.5|
sometimes vary depending on the source, those given here are within
reasonable limits of those usually found in the published
literature. No value is given when this information was not
necessary for our dose calculation.
bRat and mouse: based on NCI trichloroethylene bioassay (NCI, 1976); hamster: data of Williams (1976); nonhuman primates: data of S. M. Sieber (Laboratory of Chemical Pharmacology, NCI, National Institute of Health, Bethesda, MD), personal communication; bush babies: ages adapted from Dittmer (1973); tree shrews: data of D. J. Reddy (Northwestern University, Chicago, IL), personal communication; dog: data of M. S. Redfearn (Division of Animal Resources, University of California, Berkeley), personal communication.
cRat and mouse: based on NCI trichloroethylene bioassay (NCI, 1976); hamster and dog: data of D. Brooks (University of California, Davis), personal communication. dMouse, rat and dog: data from NIOSH (Sweet, 1993); hamster: data from Hoeltge, Inc.
eMouse: data of Sanockij (1971); rat: data of Baker et al. (1979); hamster: data of Guyton (1947).
In an experiment where the animals were dosed the entire time on test this value would equal the average daily dose level. For example, in a bioassay of male mice fed 50 ppm of some test agent in the diet for the entire time on test, the calculation would be:
Dose rate = 50 ppm × 0.12 × 1
= 6 mg/kg body weight/day
In this example, the exposure time is equal to the experiment time.
In using standard values we recognize that there is no single factor which precisely reflects the entire experimental literature. For example, strains within a species will vary in weight; younger animals have lower body weight and food intake than adults; animals within a group differ in food consumption and body weight; some test agents will reduce appetite due to taste; or illness may result in loss of appetite. However, the values used here for proportion of body weight consumed daily, are within reasonable limits of those usually found in the published literature and are unlikely to produce substantial error. We note that bioassays in the general literature and NCI/NTP provide food ad libitum, which can itself affect tumor rates compared to calorie restricted diets.
In many experiments the administration of the test compound is stopped before the terminal sacrifice or before the death of the last animal. In such cases, averaging the dose over the course of the experiment will lower the daily dose rate. By convention we take the total dose administered during the exposure period and spread this over the entire experimental period; thus, for male mice dosed at 50 ppm in the diet for 70 weeks and then continued on test to 100 weeks, the dose rate is equivalent to 35 ppm for the entire 100 weeks, (i.e., 50 ppm × 70/100), resulting in a 4.2 mg/kg body wt/day average dose. In calculating the average daily dose rate, we utilize the concept of exposure time (the period of active treatment) and experiment time (the actual time on test). When terminal sacrifice is performed, the experiment time is the length of time from the start of the experiment to sacrifice. In some bioassays, animals are permitted to survive to natural death, and then experiment time is defined as the time of death of the last dosed animal. If all animals in a given dose group die early, the dose-rate in that group is calculated based on the exposure time and the time to death of the last dosed animal.
Carcinogenic potency is defined in the CPDB in terms of the average daily dose-rate that will halve the probability of remaining tumor-free at the end of a standard lifespan (TD50). For each species the assumed value for the standard lifespan is given in Table 1; these values are within reasonable limits usually found in the published literature. The use of two years as standard lifespan for rats and mice reflects both the standard NCI/NTP protocol and values that appear frequently in the literature.
When an experiment is terminated before the standard lifespan, animals are not at risk of developing tumors later in life. Thus, the number of tumors found will be reduced, and the TD50 will be greater than the true lifetime TD50, i.e., the compound will appear to be less potent. Because tumor incidence increases markedly with age, our convention has been to adopt as a correction factor f2, where f=experiment time/standard lifespan (Peto et al., 1984).
Note that the correction factor f2 is based on the time the animals are on test, rather than upon age. In an experiment which began when the animals were 6 weeks of age, and which terminated when the animals were 100 weeks of age, the experiment time is 94 weeks. Thus, TD50 is defined in terms of the dose-rate which would be administered throughout the entire standard lifespan.
Taking the example above of male mice fed some test agent at 50 ppm for 70 weeks and then continued on test for 30 more weeks, the experiment time would be 100 weeks. The standard lifespan for mice is 104 weeks, so the extrapolation factor would be (100/104)2, or 0.92.
By omitting from the database any experiments lasting less than half the standard lifespan for that species, the necessity for great extrapolation has been reduced. Bioassays are usually not continued for much longer than the period we have adopted for a standard lifespan, so the reverse correction is minimal. For nearly all NCI/NTP bioassays and 80% of literature tests, the value of the extrapolated TD50 is within a factor of 2 of the unextrapolated value, and for the remaining 20% of literature tests it is within a factor of 4.
The results we have estimated according to the rules and conventions described, are presented in the Plot of the CPDB, the Screen Version of the plot and the Tab-Separated Version. The plot format provides a systematic means of distinguishing bioassay results for a variety of chemicals. The range of TD50s across chemicals is at least 10-million-fold for carcinogens in each sex of rat or mouse.
For female rats, the range of carcinogenic potency is shown in Figure 1, where we present the most potent TD50 for a selected group of compounds which were evaluated as tumorigens in either an NCI/NTP Technical Report or the general literature. In each case, we have indicated the value for the most potent TD50 for a target site(s) which was considered evidence of carcinogenic activity by the published author, and for which the statistical significance of TD50 is less than 0.01. The range is more than 100-million-fold in female rats.
Return to the Carcinogenic Potency Project Home Page:
Last updated: May 17, 2005