Optimal prostate-specific antigen cutoff for prostate cancer detection: A prospective cohort study.

Abstract

5109 Background: Serum prostate-specific antigen (PSA) testing is crucial for detecting early stage and low-risk prostate cancer. However, standard and uniform PSA thresholds may underestimate risk in males with smaller prostate volumes and distinct epidemiologic characteristics. We aimed to evaluate data-driven, age-specific PSA thresholds for short-term prostate cancer detection in a large screening cohort. Methods: We analyzed 512,992 PSA measurements from 123,607 asymptomatic males undergoing annual or biennial health checkups in Kangbuk Samsung Cohort Study (KSCS) from 2011 to 2020. Incident prostate cancer within 1-year of each PSA measurement was identified through linkage to the Korea National Cancer Incidence Database. The dataset was randomly partitioned into training (70%) and test (30%) subsets. To estimate cancer risk, logistic regression models incorporating synthetic minority oversampling technique were used. Optimal PSA thresholds were determined using the Youden index in test datasets. Performance was evaluated by accuracy, precision, recall, and F1-score with bootstrap validation. Associations between risk factors and prostate-cancer risk were evaluated using Cox regression models. Results: During follow-up, 224 men developed prostate cancer. The optimal PSA cutoff in the overall population was 2.96 ng/mL. Age-specific thresholds increased with age, with males in their 40s, 50s, 60s, and age ≥70 years demonstrating cutoff values of 1.62, 2.54, 2.76, and 3.79 ng/mL, respectively. These thresholds demonstrated high discriminative performance (accuracy, 0.97; F1-score, 0.92) (Table). PSA values exceeding age-specific thresholds were associated with a 20.7-fold elevated risk of prostate cancer (HR 20.65; 95% CI, 16.33–26.12; P<0.001). Older age, family history of prostate cancer (HR 2.27; 95% CI, 1.58–3.27), and dyslipidemia medication use (HR 2.04; 95% CI, 1.49–2.80) were additional independent predictors. Conclusions: In this large prospective Asian cohort, age-specific PSA thresholds between 1.6 and 3.8 ng/mL provided robust short-term risk discrimination and identified male at substantially higher 1-year prostate cancer risk than those below the thresholds. Implementing age-adjusted PSA reference values may improve the precision and efficiency of early detection strategies in Asian male and better inform risk-adapted screening and diagnostic pathways. To determine their effect on long-term clinical outcomes, prospective validation is warranted. Age-specific PSA cutoffs and performance metrics. Cutoff value Accuracy Precision Recall F1-score Overall 2.96 0.97 0.89 0.96 0.92 40 ≤ Age < 50 1.62 0.93 0.70 0.99 0.82 50 ≤ Age < 60 2.54 0.95 0.80 0.95 0.87 60 ≤ Age < 70 2.76 0.91 0.66 0.99 0.80 Age ≥ 70 3.79 0.90 0.65 0.91 0.76