External validation of the Martini nomogram for prediction of side-specific extraprostatic extension of prostate cancer in patients undergoing robot-assisted radical prostatectomy

Abstract

IntroductionTo establish oncological safe nerve-sparing robot-assisted radical prostatectomy, accurate assessment of extraprostatic extension (EPE) is critical. A recently developed nomogram including magnetic resonance imaging parameters accurately predicted side-specific EPE in the development cohort. The aim of this study is to assess this model's performance in an external patient population. Patients and methodsModel fit was assessed in a cohort of 550 patients who underwent robot-assisted radical prostatectomy in 2014 to 2017 for prostate cancer. Model calibration was evaluated using calibration slopes. Discriminative ability was quantified using the area under the receiver operating characteristic curve. Model updating was done by adjusting the linear predictor to minimize differences in expected and observed risk for EPE. ResultsA total of 792 prostate lobes were included for model validation. Discriminative ability expressed in terms of receiver operating characteristic curve was 0.78, 95%CI 0.75–0.82. Graphical evaluation of the calibration showed poor fit with a high disagreement between predicted probabilities and observed probabilities of EPE in the population. Model updating resulted in excellent agreement between mean predicted and observed probabilities. However, calibration plots showed substantial miscalibration; including both under- and overestimation. ConclusionExternal validation of the novel nomogram for the prediction of side specific EPE developed by Martini and co-workers showed good discriminative ability but poor calibration. After updating, substantial miscalibration was still present. Use of this nomogram for individualized risk predictions is therefore not recommended.