Timo Soeterik

150 CHAPTER 8 In addition, the trade-off between subjecting node-negative patients to the concomitant risks of ePLND vs the potential advantages of ePLND in the specific node-positive subgroup, remains to be explored. Future studies should focus on finding the optimum risk threshold at which the benefits of ePLND, at best, outweigh the harms. Although this study has several strengths, such as the inclusion of a multicentre cohort representing a real-world prostate cancer population and a large study sample with a sufficient number of events for adequate external validation, it also has some limitations. Firstly, the data used in the study were derived from routine clinical practice, and no central review of DRE, mpMRI and histopathological evaluation was performed. Secondly, the majority of the data were collected retrospectively, which could have led to measurement bias. Lastly, the indication to perform an ePLND in this patient cohort was done using nomogram-based LNI risk estimation (risk of LNI >5%). Even though this is according to current EAU guideline recommendations, and reflects contemporary clinical practice, this could have introduced bias due to the selection of patients for ePLND with higher risk of LNI (reflected by the relatively high LNI prevalence). For instance, selecting patients with higher risk of LNI (and prevalence) could explain the counterintuitive finding on decision-curve analysis, showing that a “treat all” approach would lead to higher net benefit compared with nomogram-based selection for risk thresholds between 0% and 15%. CONCLUSIONS The MSKCC 2018 and Briganti 2012 nomograms were found to be adequate models for the prediction of LNI in patients with prostate cancer when using either mpMRI T-stage or DRE T-stage. The use of mpMRI T-stage led to improved model discrimination, equal calibration, and lower rates of missed LNI cases. Using the mpMRI T-stage with the Briganti 2012 nomogram was shown to be the most accurate strategy for LNI risk prediction.