Objectives: To examine the ability of the urinary prostate cancer gene 3 (PCA3) assay to predict biopsy-detected cancers in men receiving dutasteride in the Reduction by Dutasteride of Prostate Cancer Events (REDUCE) study cohort. Methods: Urine and serum samples from 930 men in the active arm were acquired at years 2 and 4 of the biopsy visits. In addition to univariate logistic regression and receiver operating characteristic analysis, multivariate analysis for association with biopsy outcome was performed for PCA3 score in the presence of serum prostate-specific antigen (PSA), age, prostate volume, and family history of prostate cancer. Results: At year 2, the univariate PCA3 score area under the receiver operating characteristic curve (AUC) was 0.668 versus 0.603 for PSA. At year 4, the PCA3 assay significantly predicted the biopsy outcome (AUC 0.628, 95% confidence interval 0.556-0.700), and the PSA level was not predictive (AUC 0.556, 95% confidence interval 0.469-0.642). The year 2 multivariate model yielded an AUC of 0.712. Removing the PCA3 score decreased the AUC to 0.660 (P =.0166 vs the full model). The median PCA3 scores in the dutasteride arm were not different from those in the 1072 men in the placebo arm (16.2 and 17.2 at year 2, P =.1755; and 18.8 and 18.1 at year 4, P =.2340, respectively). However, the PSA values were reduced >50% in the dutasteride arm at both visits (both P <.0001 vs placebo). At a PCA3 score cutoff of 35, the sensitivity and specificity were equivalent between the 2 arms. Conclusions: In the present study, the PCA3 assay outperformed PSA for cancer detection in men undergoing dutasteride treatment and improved the diagnostic accuracy when combined with the PSA level and other clinical variables. In addition, no adjustment in PCA3 score was needed to yield equivalent clinical performance between the dutasteride and placebo arms. These findings are particularly important in light of the potential role of dutasteride for prostate cancer chemoprevention.