Determining optimal prostate-specific antigen thresholds to identify an increased 4-year risk of prostate cancer development: an analysis within the Veterans Affairs Health Care System.


Journal Article

PURPOSE: To assess the prostate-specific antigen (PSA) threshold value that optimally predicts future risk of prostate cancer (overall and by race) for a dispersed US population. METHODS: This was a retrospective analysis of men in the Veterans Affairs (VA) Health Care System database. Men ≥ 40 years with a baseline PSA ≤ 4.0 ng/mL, not receiving 5-alpha reductase inhibitors, and without a prostate cancer diagnosis prior to baseline PSA date were included and followed for 4 years. Patients diagnosed with prostate cancer within 6 months of baseline were excluded. The optimal PSA threshold value for future 4-year prostate cancer risk was determined by maximizing Youden's index. RESULTS: The eligible population for the final analysis included 41,250 Caucasian (n = 24,518; 59.4 %) and African American (n = 16,732; 40.6 %) patients. The 4-year prostate cancer rate was 3.08 % overall, and race-specific rates were 3.02 and 3.17 % for Caucasian and African American men, respectively. Mean time to prostate cancer diagnosis was 2.01 years across all patients. Race-specific PSA thresholds that optimally predicted future prostate cancer were 2.5 ng/mL [area under the curve (AUC) = 80.3 %] in Caucasians and a 1.9 ng/mL (AUC = 85.4 %) in African Americans; across all patients, a 2.4 ng/mL threshold was optimal (AUC = 82.5 %). CONCLUSIONS: In the VA population, a relatively low PSA threshold of ~2.5 ng/mL was optimal in predicting prostate cancer within 4 years overall and for Caucasian men, but an even lower threshold of 1.9 ng/mL was applicable for African American men.

Full Text

Duke Authors

Cited Authors

  • Sutton, SS; Crawford, ED; Moul, JW; Hardin, JW; Kruep, E

Published Date

  • August 2016

Published In

Volume / Issue

  • 34 / 8

Start / End Page

  • 1107 - 1113

PubMed ID

  • 26753559

Pubmed Central ID

  • 26753559

Electronic International Standard Serial Number (EISSN)

  • 1433-8726

Digital Object Identifier (DOI)

  • 10.1007/s00345-015-1754-6


  • eng

Conference Location

  • Germany