Are predictive models for cancer volume clinically useful in localized prostate cancer?

Objectives:To evaluate the correlation between preoperatively predicted and pathologically measured prostate cancer volumes and to investigate the clinical use of preoperatively predicted cancer volume in predicting pathological stage. Methods:Correlations between pathological findings and various preoperative parameters, including the cancer volumes as predicted by using two methods (Vca and estimated PCvol), were analyzed in 196 patients who underwent radical prostatectomy for clinically localized prostate cancer. Results:Pathologically measured prostate cancer volume was significantly correlated with the Vca and estimated PCvol, but the correlation coefficients were respectively only 0.46 and 0.35. Prostate-specific antigen (PSA), PSA density (PSAD), primary Gleason score, Vca, Vca fraction (Vcafx), and estimated PCvol were significantly higher in 82 patients with extraprostatic cancer than in 114 patients with organ-confined cancer. Magnetic resonance imaging (MRI) findings were significantly correlated with pathological stage. Multivariate logistic regression analysis indicated that the Vcafx and MRI findings were significant predictors of extraprostatic cancer, but receiver operating characteristic analysis revealed that the combination of Vcafx and MRI findings had no advantage over the combination of Gleason score, PSAD, and MRI findings. Conclusions:Vca and estimated PCvol are significantly correlated with the pathologically measured cancer volume but their ability to accurately predict cancer volume is limited. Vcafx and MRI findings were statistically significant predictors of extraprostatic cancer but their combination was not superior to the combination of Gleason score, PSAD, and MRI findings.