Abstract
Different approaches to screening for cancer of the cervix by cervical cytology have been evaluated using a computer simulation model developed by Knox and data on the natural history of carcinoma-in-situ (or worse) from a cohort study of women screened in British Columbia, 1949-69. The natural history input parameters and the output parameters without screening were modified to reflect the earlier onset of carcinoma-in-situ in younger cohorts now being experienced in British Columbia, resulting in simulated mortality from carcinoma of the cervix approximately 50% greater than that experienced in Canada in 1955. The simulations showed that the sensitivity of the test and the proportion of women in the population who accept invitations to attend for screening materially influence the extent to which programmes reduce mortality. Missed screens also have an important impact. With a 75% test sensitivity, and an 80% population acceptance, a programme designed to reduce mortality by 90% would commence at age 25, involve triennial screens to age 52, or triennial screens to age 40 and quinquennial screens to age 60, a total of 10 tests in a lifetime. A repeat test at age 26 contributes nothing to the mortality benefit. Nevertheless, additional modifications of the natural history specifications to accommodate high-risk younger women would require a more frequent schedule of examinations under the age of 35, though at a substantial 'cost' in terms of the total number of examinations required in a population.