Presentation Authors: Ahmed Eissa, Tanta, Egypt, Maria Chiara Sighinolfi*, Modena, Italy, Marco Sandri, Brescia, Italy, Pietro Torricelli, Federica Fiocchi, Guido Ligabue, Luca Reggiani Bonetti, Stefano Puliatti, Cosimo De Carne, Salvatore Micali, Modena, Italy, Vipul Patel, Orlando, FL, Giampaolo Bianchi, Bernardo Maria Rocco, Modena, Italy
Introduction: Prediction of the presence of extracapsular extension (ECE) of prostate cancer (PCa) before surgery is of paramount importance to tailor the amount of nerve-sparing during radical prostatectomy (RP). A novel nomogram to predict ECE has been recently developed with the integration of a multiparametric magnetic resonance imaging (mpMRI) derived variable (Martini et al, BJUI 2018, 1-9). Authors defined the &[Prime]presence of ECE&[Prime] as the loss or irregularity of the capsule, whereas contact, bulge or abutment are considered as negative for ECE. We aimed to externally validate this nomogram on 137 prostatic lobes from 106 patients undergoing mpMRI-targeted biopsy plus saturation sampling.
Methods: We applied the model from Martini to the most recent cases of PCa patients (n=106) with a positive mpMRI submitted to RP. PCa was diagnosed in all cases by mpMRI-targeted plus systematic saturation biopsy. According to Martini&[prime]s model, we considered only lobes with a positive biopsy (137). The primary endpoint was to perform an EV; the secondary endpoint was to explore the incremental role of the mpMRI-variable added to conventional clinical-pathological ones.AUC was used to assess the nomogram&[prime]s discriminative performance. The comparison between AUCs of two-nested models was performed using the test of Heller.
Results: The AUC at EV was 67.6% (95%CI:57.4%-77.8%). Sensitivity and specificity at the 20% cutoff suggested by Authors were 53.6% (95%CI:33.9%-72.5%) and 77.1% (95%CI:68%-84.6%), respectively. The model showed a poor calibration with tendency towards underestimation. As far as the secondary endpoint, the tool without mpMRI-variable showed a discrimination of 66.5% (95% CI:56.5%-76.7%) and the difference between the two AUCs was not statistically significant (p=0.113).
Conclusions: On External validation, the predictive performance of Martini&[prime]s model seems to be suboptimal. A possible explanation could be the subjective approach of ECE depiction at mpMRI used by Authors; actually, the ideal variable predicting ECE from imaging is far to be defined.Further EV studies on larger sample size are required to definitely assess the generalizability of this nomogram.