Prostate MRI: evaluating tumor volume and apparent diffusion coefficient as surrogate biomarkers for predicting tumor Gleason score.

PURPOSE To investigate whether tumor volume derived from apparent diffusion coefficient (ADC) maps (VolumeADC) and tumor mean ADC value (ADCmean) are independent predictors of prostate tumor Gleason score (GS). EXPERIMENTAL DESIGN Tumor volume and GS were recorded from whole-mount histopathology for 131 men (median age, 60 years) who underwent endorectal diffusion-weighted MRI for local staging of prostate cancer before prostatectomy. VolumeADC and ADCmean were derived from ADC maps and correlated with histopathologic tumor volume and GS. Univariate and multivariate analyses were performed to evaluate prediction of tumor aggressiveness. Areas under receiver-operating characteristics curves (AUC) were calculated to evaluate the performance of VolumeADC and ADCmean in discriminating tumors of GS 6 and GS ≥7. RESULTS Histopathology identified 116 tumor foci >0.5 mL. VolumeADC correlated significantly with histopathologic tumor volume (ρ = 0.683). The correlation increased with increasing GS (ρ = 0.453 for GS 6 tumors; ρ = 0.643 for GS 7 tumors; ρ = 0.980 for GS ≥8 tumors). Both VolumeADC (ρ = 0.286) and ADCmean (ρ = -0.309) correlated with GS. At univariate analysis, both VolumeADC (P = 0.0325) and ADCmean (P = 0.0033) could differentiate GS = 6 from GS ≥7 tumor foci. However, at multivariate analysis, only ADCmean (P = 0.0156) was a significant predictor of tumor aggressiveness (i.e., GS 6 vs. GS ≥7). For differentiating GS 6 from GS ≥7 tumors, AUCs were 0.644 and 0.704 for VolumeADC and ADCmean, respectively, and 0.749 for both parameters combined. CONCLUSION In patients with prostate cancer, ADCmean is an independent predictor of tumor aggressiveness, but VolumeADC is not. The latter parameter adds little to the ADCmean in predicting tumor GS.