Timo Soeterik

148 CHAPTER 8 FIGURE 3. Decision curves for the four performed validation scenarios compared to the default strategies DISCUSSION Use of mpMRI T-stage for nomogram-based LNI risk assessment resulted in higher AUC, comparable agreement between predicted and observed probabilities, and higher net benefit compared with DRE T-stage, in both the MSKCC 2018 and Briganti 2012 nomograms. In our study population, use of DRE T-stage would lead to overall LNI risk underestimation in the clinically relevant range of risk thresholds (0-30%). In the head-to-head comparison, combined use of the mpMRI T-stage with the Briganti 2012 nomogram resulted in the most accurate LNI risk prediction. Our study acknowledges the robustness of both the MSKCC 2018 and Briganti 2012 nomograms, since model performance was still fair to good, even when the model was applied in a patient population with a considerably higher prevalence of the predicted outcome compared with the development populations. In our cohort, LNI prevalence (28%) was substantially higher compared to both MSKCC (7% [internal communication MSKCC research team]) and Briganti (8%) populations. 10 Therefore, our results show both models are applicable in a contemporary patient cohort. In addition, our analysis confirmed that mpMRI T-stage can be safely used as impute parameter for these nomograms, even leading to improved accuracy of the predicted LNI risk compared with DRE T-stage. The present study’s main findings add up to the available body of literature supporting the additional value of mpMRI information for predicting presence of LNI in prostate cancer. For example, Porpiglia et al. showed MRI has an important role in LNI risk prediction in patients with a nomogram-predicted risk <5%. 22 Huang et al. demonstrated