Statistical considerations in cancer screening programs.

Abstract

The goal of cancer screening is the early detection and treatment of disease, with a consequent reduction in the mortality rate. Evaluation of whether a particular screening program can achieve this goal is a difficult task. Two components of the screening process must be assessed. The first is the ability of the screening test to detect cancer early while minimizing the number of false-positive results. In this regard, the specificity of the test ordinarily must be very high, approaching 99%. No screening test for prostate cancer has yet been reported to have a specificity this high, indicating that any prostate cancer screening program using currently available tests will have to deal with the problem of a large number of false-positive findings. To evaluate the overall impact of a screening program, the best procedure is the randomized controlled trial with cancer-specific mortality as the endpoint. This endpoint is used because it avoids the lead time and length biases inherent in other outcome variables such as stage shift and case survival. The screening randomized controlled trial must be carefully planned and implemented, because it is lengthier and more costly than the usual therapy trial because of differences in study populations, trial design relative to the planned population intervention, and the extent of knowledge of disease natural history. A further important component of screening evaluation is cost. The decision to implement or continue a screening program can be aided by using cost-effectiveness analysis, which bases a decision on the ranking of cost-to-benefit ratios for the various programs contending for limited funds. Screening cost includes the cost of the test, the cost of side effects of the test, and the costs of biopsy and treatment, while screening benefit can be measured in terms of lives saved, life years saved, or quality-adjusted life years.