Describing the validity of carcinogen screening tests.

SEVERAL issues pertinent to the validation of screening tests and to the evaluation of screening programmes have evolved, largely in the epidemiological and general medical literature (Vecchio, 1966; Holland, 1974; Henderson, 1976). For this reason, laboratory scientists interested in the development and validation of carcinogen screening tests may not be fully aware of some useful concepts. The paper reviews several such concepts and their relationships. These relationships explain, for example, why the "predictive value", although frequently used as an index of the utility of a screening test, is not well suited for that purpose. A carcinogen screening test is usually evaluated by applying it to a group of substances, each of which is considered to be, or not to be, a carcinogen according to some selected criteria (i.e. animal or human carcinogenesis). The specific criteria chosen are beyond the scope of this discussion; however, the validity measures of a screening test are meaningful only in the context of these criteria. After some number, N, of substances has been tested, the results may be described using the format of the Table. (Terms used to describe screening tests and screening programmes are defined below the Table.) A screening test is expected to designate carcinogens as "positive" and to designate non-carcinogens as "negative". Each of these 2 distinct functions has its own measure of validity. The proportion of carcinogens which give a positive result in Accepted 16 October 1978