[A model for evaluating screening tests. The case of colorectal tumors]

The results of numerous studies on screening are restricted to crude estimates of the predictive value of different tests. This is notely the case for early diagnosis of colorectal adenoma and cancer. The reference examination is in principle a total colonoscopy. Usually, it is not performed on the subjects for which the screening test gives a negative result. Moreover, the distribution according to age and sex is not taken into account. Both variables are strongly associated with the prevalence of the disease. After a presentation of the problems arising from screening evaluation performed in such situations, a method is proposed for comparing different tests in several populations. An estimate of the expected number of cases within a fictitious control group matched for age and sex with the people whose the test result is positive is computed according to an hypothesis on the duration of asymptomatic cancer and using available descriptive data on symptomatic cancer incidence and on the prevalence of adenoma in the population. The selection power of the test, i.e. the number of observed cases divided by the number of expected cases, is proposed as a standardized parameter for measuring and comparing the validity of screening tests. The method is illustrated by an example of its application procedure. The discussion stresses the limitations of such an approach.