Statistical issues in the evaluation of screening and early detection modalities

Thorough evaluation of a screening test requires conducting a series of studies to ascertain its ability to detect accurately disease, as well as its benefits and costs. In this article, I review the steps involved in evaluating a screening test, using the case of prostate-specific antigen (PSA) screening for prostate cancer as a case study. I discuss designs for quantifying the diagnostic properties of a screening test and compare several different studies that have produced quite different estimates of the diagnostic accuracy of PSA screening. I also review methods that may be used to combine other markers or tests with PSA to improve test accuracy. Determining the benefits of a screening test is complex, particularly when information from randomized trials is lacking. I review several observational studies of PSA benefit and discuss the use of computer models for inferring the impact of screening from trends in population mortality. (C) 2008 Elsevier Inc. All rights reserved.