To prospectively determine whether multiparametric magnetic resonance imaging (mpMRI)-based staging is a more accurate independent predictor of outcome than traditional clinical variables for patients undergoing brachytherapy and external beam radiation therapy.

The primary endpoints were biochemical (nadir plus 2 ng/mL) and metastatic failure. Descriptive, univariate, and multivariate competing risks analyses were performed. The cumulative incidence rates were estimated to describe the cumulative risk of the events of interest. The magnitude of the increased risk was estimated using univariate and multivariate subdistribution hazard ratios.

A total of 185 patients had undergone prospective treatment (123 with high risk and 62 with intermediate risk). The median age was 71 years (range 56-82). Of the patients, 20.5% had mpMRI-determined (mrT) stage mrT1-mrT2b, 37.3% had mrT2c, 31% had mrT3a, and 11.2% had mrT3b. The Gleason score was 6 in 22.2%, 7 in 49.5%, and 8 to 10 in 28.2%. The median baseline prostate-specific antigen was 11.7 ng/mL (range 2.9-153). After a median follow-up period of 46 months (range 16-70), 15 patients (8.1%) had developed biochemical failure and 9 (4.9%) had developed distant metastases. None of the traditional clinical variables (prostate-specific antigen, Gleason score, clinical stage) predicted for biochemical or metastatic failure. The multivariate competing risk analysis demonstrated that the 2 independent predictors of biochemical failure were the presence of extraprostatic extension (EPE; mrT3a; hazard ratio [HR] 4.80; P = .035) and presence of seminal vesicle invasion (SVI; mrT3b; HR 10.17; P = .003) on mpMRI. The only independent predictor of metastatic failure was the percentage of positive cores on prostate biopsy (HR 13.95; P = .014). After excluding patients with SVI, the only independent predictor of biochemical failure and metastatic failure was the presence of EPE (stage mrT3a) on mpMRI (HR 4.36; P = .042; and HR 5.76; P = .010, respectively).

The pretreatment mpMRI findings might be more accurate independent predictors of the outcome than traditional clinical variables. In particular, the presence of EPE, SVI and a greater percentage of positive cores on biopsy predicted for a worse prognosis.