The optimal design of batteries of short-term tests for detecting carcinogens

The optimal design of a battery of short-term tests for detecting carcinogens has been explored by constructing tables describing the theoretical accuracy of various batteries of tests. The analysis has revealed that the expected accuracy of the testing battery is greatly influenced by the design of the battery, particularly by the number of tests in the battery and by the number of tests which must give a positive result before the compound is considered positive in the battery. The recommended optimal design is not affected by the expected accuracy of the tests comprising the battery, at least not for reasonably accurate short-term tests. Two types of analyses of published data have been presented in support of these theoretical considerations. The first is a statistical analysis (chi-square test) showing that most short-term tests independently detect carcinogens and noncarcinogens, thus fulfilling a key assumption for the use of batteries of short-term tests to improve the detection of carcinogens and noncarcinogens. The second comes from the results of a recent international collaborative study showing that the recommended design of testing batteries would optimize the identification of carcinogens and noncarcinogens by five of the most accurate tests in the study.