Misclassifying the indications for prostate-specific antigen testing may bias case-control studies of the efficacy of prostate cancer screening.
BACKGROUND AND OBJECTIVE: In the absence of data from randomized controlled trials, prostate cancer (CaP) screening recommendations may be based on observational studies that contrast exposure to screening between cases and controls. We evaluated the potential bias from mis-classifying indications for PSA testing in observational studies of CaP screening. METHODS: We randomly selected men undergoing PSA testing and obtained data on PSA results and prostate biopsies. Data were linked with a tumor registry to identify incident and prevalent cases of CaP. We abstracted medical records for 45 incident cases with CaP and 118 controls without, recording information on lower urinary tract symptoms (LUTS), constitutional symptoms, and digital rectal examination findings. PSA testing was classified as definitely, likely, or possibly screening, or not screening based on clinical history. RESULTS: Changing the definitions for PSA screening to variably exclude men with LUTS and enlarged prostates differentially lowered the frequency of screening. With more restrictive screening definitions, the odds ratio for screening decreased from 0.47 to 0.07. CONCLUSION: Accurately classifying PSA testing status is difficult because LUTS are common among men targeted for CaP screening. Failing to correctly classify PSA tests may bias study results.