Abstract
BACKGROUND: The benefit-to-harm ratio of PSA-based prostate cancer screening may be improved through implementation of PSA reference ranges that consider innate individual characteristics. We evaluated variation in PSA levels by factors that may influence PSA levels among men eligible to undergo prostate cancer screening. METHODS: We identified men ages 40 to 79 years in the All of Us Research Program who had no history of prostate cancer or elevated PSA at cohort enrollment. PSA distributions were compared across age groups, self-identified race or ethnicity (SIRE), body mass index (BMI), polygenic risk score (PRS) for prostate cancer risk, and PRS for PSA level. Multivariable associations between these factors and PSA percentiles were evaluated using quantile regression. RESULTS: Among 13,749 eligible men, 95th percentiles (p95) of PSA values increased with age (40-49 years: 1.81 ng/mL, 50-59 years: 3.23 ng/mL, 60-69 years: 4.15 ng/mL, and 70-79 years: 5.53 ng/mL). p95 of PSA was 0.83 ng/mL lower [95% confidence interval (CI), 0.44-0.1.22] among participants with BMI 35 to 39 kg/m2 versus <25 kg/m2. p95 of PSA was 2.32 ng/mL higher (95% CI, 1.20-3.44) among those with PRS for prostate cancer >90th percentile versus ≤50th percentile and 1.21 ng/mL higher (95% CI, 0.50-1.92) among males with PRS for PSA >90th percentile versus ≤50th percentile. SIRE was not consistently associated with PSA levels. CONCLUSIONS: PSA levels vary by age, BMI, and PRS but not SIRE. Further work is needed to understand how tailoring PSA reference ranges based on these characteristics would affect screening outcomes. IMPACT: Consideration of factors that endogenously influence PSA levels may lead to improved benefit-to-harm ratios of prostate cancer screening.