Introduction

The ERSPC risk calculator estimates the risk of positive prostate biopsy. We aimed at evaluating whether the inclusion of mpMRI data might improve the predictive accuracy of this calculator in detecting positive biopsy in patients undergoing mpMRI targeted biopsy

Methods

214 consecutive patients underwent mpMRI of the prostate with subsequent targeted and concomitant systematic biopsy at a single centre between 2013 and 2016. A 1.5 T mpMRI study using an endorectal coil was performed in all patients who had also complete data required by ERSPC risk calculator 4. Four multivariable logistic regression analyses (MVA) were performed to assess the predictors of positive biopsy. Predictors included in Model 1 were those of the ERSPC risk calculator, namely transrectal ultra-sonography (normal vs. abnormal), digital rectal examination (normal vs. abnormal), prostate volume (ml), the logarithmic transformation of PSA (ng/ml), previous negative biopsy (no vs. yes). Model 2 included as predictor only by PIRADS v.2 (<4 vs. ≥4). Predictors of Model 3 and Model 4 were the same of Model 1 plus PIRADS (<4 vs. ≥4) for Model 3 and PIRADS (<4 vs. ≥4) + patient age for Model 4. Leave-one-out cross validation (LOOCV) and calibration plots were used to internally validate each model. Decision curve analyses (DCA) were performed to evaluate and compare the net benefit associated with the use of each model

Results

Overall, 63% had a positive biopsy. At MVA of Model 1, prostate volume (OR: 0.97) and PSA (OR: 2.8), were independent predictors of positive biopsy (all p≤0.001). In all models including mpMRI results (Model 2-4), the presence of PIRADS ≥4 was significantly associated with positive biopsy (OR range: 3-4.6; all p≤0.001). At MVA of Model 3 and 4, prostate volume and PSA also reached the independent predictor status for positive biopsy prediction (all p≤0.009). Moreover, at MVA of Model 4, older age was significantly associated with positive biopsy (OR: 1.08, p<0.001). In LOOCV, inclusion of age and mpMRI results into the ERSPC risk 4 calculator was associated with the highest accuracy (C-index 77.3%) as compared to all other models tested (C-index of model 1, 2, 3: 65.8 vs. 53.7 vs. 73.7%, respectively). At DCA, the combined use of age, mpMRI and ERSPC calculator resulted into higher net-benefit relative to all the other 3 Models

Conclusions

Our data demonstrated the importance of inclusion of both PIRADS v.2 and patient age into the ERSPC risk calculator to better identify patients with higher risk of positive biopsy

Funding

none