Abnormal Flagging of Prostate Specific Antigen Screening Tests: A Regression Discontinuity Design.

Abstract

There is no explicitly normal or abnormal level of prostate specific antigen (PSA) in the blood, yet the lab reported "upper limit of normal" (ULN) represents the most accessible guidance available to physicians interpreting these tests and is agnostic to risk factors and life expectancy. We hypothesize that "abnormal flagging" on "elevated" results is a potent driver of downstream care irrespective of a patient's underlying prostate cancer risk. Regression discontinuity design was applied to a cohort of patients undergoing PSA screening at a single center to isolate and estimate the impact of "abnormal flagging" of results above the PSA ULN on subsequent early repeat ("confirmatory") PSA testing, urology referral, and prostate MRI or biopsy (or both). Amongst 38,042 screening PSA tests, 1041 (2.7%) were flagged as abnormal. Adjusting for patient age and PSA gap, "abnormal" flagging of PSA (versus "normal") was associated with an increased odds of repeat PSA testing (OR: 14.17; 95% CI 11.53-17.43; p < 0.001), referral (odds ratio (OR): 10.03; 95% confidence interval (CI) 7.53-13.47; p < 0.001), and either prostate MRI or biopsy (or both; OR 13.61; 95% CI 8.81-21.64; p < 0.001). While the use of PSA ULN has recognized shortcomings, abnormal flagging of results above this threshold strongly influences additional testing, referral, and biopsy, potentially leading to risk-discordant over- and under-evaluation. An improved reporting standard may include risk estimates of meaningful endpoints such as the risk of prostate cancer metastasis or death.