Estimating the cumulative risk of a false-positive under a regimen involving various types of cancer screening tests

Abstract

When evaluating screening for the early detection of cancer, it is important to estimate both harms and benefits. One common harm is a false-positive (FP), which is a positive screening result, perhaps followed by an invasive test, with no cancer detected on the diagnostic work-up or within a specified time period. An important goal is to estimate the risk of at least one FP, which we call the cumulative risk of an FP, if persons took a regimen of various screening tests, as is commonly recommended. The estimation is complicated because the data come from a study in which subjects are offered various screening tests in rounds with some missing tests in most subjects. Previous methods for estimating cumulative risk of FPs with a single type of test are not directly applicable, so a new approach was developed. The tests were ordered by appearance, where the last test was either the first FP (analogous to a failure time) or the last test taken with no FPs having occurred on that test or previously (analogous to a censoring time). We applied a Kaplan-Meier approach for survival analysis with the innovation that the hazard for a first FP for a given test depends on the type of test and number of previous tests of that type which were taken. The method is illustrated with data from the screening arm of the randomized Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. With an FP defined as a diagnostic work-up in the absence of cancer (or advanced adenoma) within three years, the probability of at least one FP among 14 tests in men was 60.5% with 95% confidence interval of (59.3%, 61.6%). A simple estimate is proposed for the probability of at least one FP if persons took a regimen of multiple screening tests of different types. The methodology is useful for summarizing the burden of multiphasic screening programmes.