Introduction

Multiparametric MRI (mpMRI) gains widespread acceptance in PC diagnosis and improves detection of significant PC (sPC). We added pre-biopsy mpMRI data to a European Randomised study of Screening for PC (ERSPC) risk calculator 4 (RC4) and developed a nomogram to predict individual sPC risk.

Methods

First, clinical parameters of 755 men who underwent mpMRI prior to transperineal MRI/TRUS-fusion-biopsy between 2012 and 2014 were retrospectively analyzed as training sample. SPC was defined according to NCCN criteria (GS=3+3 and PSA>=10 ng/ml or GS>=3+4). A multivariate regression model was used to determine significant predictors of sPC in the training set and to develop a nomogram. The accuracy was compared to ERSPC RC4 and mpMRI alone (PI-RADS Likert score) by Receiver operating characteristics (ROC) curve. Based on the difference in accuracy, a sample size calculation was performed for a validation set. Subsequently, accuracy, discrimination and calibration of the nomogram were prospectively analyzed in 404 men.

Results

Overall, PC occurred in 732 (63%) and sPC in 560 (48%) men. In the trainings set, 50% of men harbored PC and 78% of them sPC. In multivariate analysis, PSA, PSA-density, Likert score and ERSPC RC4 (each p<0.001) were significant predictors of sPC and used for the prediction model. In ROC analysis, Area under the curve (AUC) was highest for the novel nomogram (0.82), compared to 0.74 for ERSPC RC4 and 0.76 for Likert scoring. Based on the 0.08 benefit of the nomogram, 404 men were enrolled as prospective validation sample. In that subgroup, accuracy of the nomogram was best (0.79), compared to Likert scoring (0.78) and ERSPC RC4 (0.60). Calibration was analyzed using a calibration plot, demonstrating a good slope (0.94). However, the plot demonstrates slight overestimation of the prediction model.

Conclusions

We provide a new prostate cancer prediction model. This model incorporates both ERSPC RC4 and mpMRI data. Compared to clinical and MRI parameters alone, the model provides significantly more reliable non-invasive risk prediction of sPC. Thus, it can help to avoid unnecessary prostate biopsies.