An Assessment of the Agreement Between Calculated and Predicted Cancer Potencies. Sara M. Hoover, Triton Environmental Consultants Ltd., #120-13511 Commerce Parkway, Richmond, B.C. V6V 2L1, Canada
Testing the carcinogenicity of chemicals through animal bioassays is time-consuming and costly. As a result, cancer bioassays are not available for the majority of chemicals in use. Without these toxicological studies potential carcinogenic hazards may go undetected and unregulated, exposing individuals to serious health risks. To address this problem, a number of researchers have developed alternative methods for identifying carcinogens. In 1990, researchers at the National Institute for Environmental Health Sciences initiated a prospective prediction exercise to challenge the various methods as to their accuracy by inviting predictions on chemicals being tested by the National Toxicology Program at that time. The bioassays are now complete and several papers have been published comparing the experimental evidence of carcinogenicity with the predictions. Some prediction methods can also provide estimates of cancer potencies, which are required for most regulatory actions on carcinogens. This paper presents the cancer potencies calculated from the animal bioassays for a subset of chemicals in the prediction exercise, and compares them to forecasted potencies. Kendall’s rank-order correlation coefficient, a nonparametric measure of agreement was used to formally assess the agreement between the calculated and predicted potencies.