Timo Soeterik

105 External validation of the Martini Nomogram BACKGROUND In patients with prostate cancer who undergo robot-assisted radical prostatectomy (RARP), the trade-off between optimal oncological results versus an attempt to retain erectile function using nerve sparing techniques is an important feature in medical decision-making. 1 The decision to use a nerve sparing approach is particularly based on the probability of present extraprostatic extension (EPE), as this increases the risk of a positive surgical margin during nerve sparing on this ipsilateral side. To exclude the presence of EPE and to ensure the oncological safety of a nerve sparing approach, multiparametric magnetic resonance imaging (mpMRI) of the prostate for local staging is recommended by the European Association of Urology Guidelines, especially in high-risk prostate cancer. 1 However, the sensitivity of mpMRI for detection of EPE is low, which potentially makes a nerve sparing procedure, planned using solely mpMRI information, oncologically unsafe. 2 To improve preoperative estimation of the presence or absence of side-specific EPE and thus the safety of an ipsilateral nerve sparing procedure, several nomograms have been developed. 3–5 The predictors included per model vary, but generally serum prostate- specific antigen (PSA), clinical tumour stage assessed by digital rectal examination, highest International Society of Urological Pathology (ISUP) grade and other biopsy characteristics (e.g., highest percent tumour involvement, number of positive cores) are included. In addition, these models include specific biopsy features of transrectal ultrasonography-guided systematic biopsies. However, due to current developments and recommendations regarding the use of prebiopsy MRI with subsequent target biopsies, these nomograms are becoming less applicable in modern urological practice. 6–9 In a recent study of Martini et al, a nomogram with a potentially higher applicability in daily practice was developed. 10 Besides mpMRI information, the model includes oncological parameters such as serum PSA level, highest ISUP, and highest percent tumour involvement of the corresponding biopsy core. Due to these features, the model can be applied in urological practices that use either transrectal ultrasonography- guided systematic, solely MRI-guided target biopsies or the combination of both techniques. The authors observed that inclusion of these parameters into a prediction model, combined with mpMRI results, improves EPE prediction. 10 Reported discriminative ability after internal validation was good, with an area under the receiver operating characteristic curve (AUC) of respectively 82.11% (95% confidence interval [CI] 78.49 − 85.73). Also, calibration was reported to be good given the excellent concordance of predicted probabilities and observed prevalence of EPE. To determine the performance of the Martini model in another population cohort, and thus its generalisability, we performed an external validation study using data of a large cohort of Dutch prostate cancer patients that underwent RARP. 6