Objective To explore explanations for the numerical imbalance of biopsy-detected Gleason 8-10 prostate cancers (PCa) diagnosed in years 3-4 in the dutasteride and placebo groups of the Reduction by Dutasteride of Prostate Cancer Events (REDUCE) study. Methods REDUCE was a 4-year, randomized, double-blind, placebo-controlled trial of dutasteride (0.5 mg/d) vs placebo for PCa risk reduction. We modeled the incidence of Gleason 8-10 cancer and used logistic regression analysis to evaluate the effects of baseline predictors of PCa, as well as post-baseline prostate volume at the time of biopsy, on PCa diagnosis. We compared needle biopsy Gleason scores with corresponding surgery Gleason scores. All statistical tests conducted were 2-sided. Results Had there been a scheduled biopsy occurring only at year 4, we estimated a similar incidence of Gleason 8-10 PCa in the dutasteride (n = 45) and placebo (n = 46) groups. Two biopsy Gleason 7 cancers in the placebo group (n = 150) were upgraded to Gleason 8-10 cancer on prostatectomy, and no patients in the dutasteride group (n = 111) were upgraded. Logistic regression analysis demonstrated the effect of prostate volume on Gleason 8-10 cancer diagnosis. Conclusion Although modeling of REDUCE data showed a similar incidence of Gleason 8-10 cancer in the dutasteride and placebo groups at year 4, an association between dutasteride and Gleason 8-10 cancer cannot be definitely excluded. It is likely that several biases, notably study design and prostate size at the time of biopsy, contributed to the numerical imbalance in Gleason 8-10 cancers observed between the treatment groups in years 3-4.