Question

In a randomized controlled trial of screening for two forms of cancer (breast and colon), more than 150,000 men and women were recruited to take part. Follow-up of these individuals took place over an average of 5 years for cancer incidence and cause-specific mortality. In an analysis unrelated to those bearing on the efficacy of screening, incidence and mortality rates in all trial participants combined were compared to those of demographically comparable individuals in the population as a whole. The incidence of cancer (excluding cancer of the breast and colon) in the trial participants was 89% that of the general population, whereas the corresponding figure for cancer mortality was 56%. Because of the large number of events, chance is a highly unlikely explanation for the difference between the relative risk for cancer incidence (0.89) and that for cancer mortality (0.56). What do you believe to be the most likely explanation for the difference? (hint: how and what type of selection bias could be at play here?)

Answer 1

This scenario represents a type of "Healthy screening bias".

The participants in this program of cancer screening are unlikely to be patients dying due to cancer. They are excluded from the study population but not from the comparison population and expected to have a more impact on relative cancer mortality but less or none on the relative cancer incidence.