Objectives: To assess the utility of the percent free prostate-specific antigen (%fPSA) for the prediction of prostate cancer in men undergoing repeat biopsy. Methods: A retrospective review was performed of 1037 patients in an institutional review board-approved repeat prostate biopsy database. A total of 617 patients who underwent 683 biopsies had all their data available for analysis. The patients were categorized as having undergone 1 repeat biopsy or >1 repeat biopsy. Results: The overall cancer detection rate was 27% and 22% in men who underwent 1 and >1 repeat biopsy, respectively. The area under the receiver operating characteristic curve for the %fPSA was 0.65 for men who underwent 1 repeat biopsy. Multivariate analysis demonstrated that a positive family history, decreasing %fPSA, and presence of high-grade intraepithelial neoplasia and/or atypical small acinar proliferation predicted for cancer. The univariate odds ratio for every 5% decrease in the %fPSA was 1.5 (95% confidence interval 1.2-1.7). The performance of %fPSA was further improved in men who underwent >1 repeat biopsy, with an area under the curve of 0.72. In men who underwent >1 repeat biopsy, multivariate analysis showed that a decreasing %fPSA, >20 cores removed, and high-grade intraepithelial neoplasia predicted for cancer. The univariate odds ratio for every 5% decrease in the %fPSA was 1.8 (95% confidence interval 1.4-2.3). A %fPSA cutoff of 10% achieved 90% and 91% specificity in the 1 repeat biopsy and >1 repeat biopsy groups, respectively. Conclusions: %fPSA is useful in predicting for prostate cancer in the repeat biopsy population, particularly for those who have undergone multiple repeat biopsies. A persistently low %fPSA should prompt additional investigation in these men. © 2011 Elsevier Inc.