Precision early detection aims to tailor screening policies to disease risk. This ambitious objective rests on (1) accurate risk prediction and (2) translation of the results into tailored screening policies that increase screening intensity in high-risk population strata and reduce intensity in low-risk strata.

When risk predictions are based on populations undergoing screening, results may be subject to detection bias which occurs when screening/biopsy rates or test sensitivity depend on the risk factors of interest. We present an approach for addressing this problem with the objective of developing tailored prostate cancer screening polices based on family history of the disease. Our de-biasing approach is applied to screening history data from a large cohort of prostate patients that were participants in the Selenium and Vitamin E (SELECT) chemo-prevention trial.

To translate risk predictions into tailored screening policies we adapt a previously developed decision model to project critical outcomes including harms such as overdiagnosis and benefits such as deaths prevented under a range of tailored screening strategies. We discuss how harm-benefit tradeoffs may best be used to identify preferred policies and explore the extent to which such policies may be robust to the detection bias issue.