Updated nomogram predicting lymph node invasion in patients with prostate cancer undergoing extended pelvic lymph node dissection: the essential importance of percentage of positive cores

Background: Few predictive models aimed at predicting the presence of lymph node invasion (LNI) in patients with prostate cancer (PCa) treated with extended pelvic lymph node dissection (ePLND) are available to date.

Objective: Update a nomogram predicting the presence of LNI in patients treated with ePLND at the time of radical prostatectomy (RP).

Design, setting, and participants: The study included 588 patients with clinically localised PCa treated between September 2006 and October 2010 at a single tertiary referral centre.

Intervention: All patients underwent RP and ePLND invariably including removal of obturator, external iliac, and hypogastric nodes.

Measurements: Prostate-specific antigen, clinical stage, and primary and secondary biopsy Gleason grade as well as percentage of positive cores were included in univariable (UVA) and multivariable (MVA) logistic regression models predicting LNI and formed the basis for the regression coefficient-based nomogram. The area under the curve (AUC) method was used to quantify the predictive accuracy (PA) of the model.

Results and limitations: The mean number of lymph nodes removed and examined was 20.8 (median: 19; range: 10-52). LNI was found in 49 of 588 patients (8.3%). All preoperative PCa characteristics differed significantly between LNI-positive and LNI-negative patients (all p<0.001). In UVA predictive accuracy analyses, percentage of positive cores was the most accurate predictor of LNI (AUC: 79.5%). At MVA, clinical stage, primary biopsy Gleason grade, and percentage of positive cores were independent predictors of LNI (all p≤0.006). The updated nomogram demonstrated a bootstrap-corrected PA of 87.6%. Using a 5% nomogram cut-off, 385 of 588 patients (65.5%) would be spared ePLND. and LNI would be missed in only 6 patients (1.5%). The sensitivity, specificity, and negative predictive value associated with the 5% cut-off were 87.8%, 70.3%, and 98.4%, respectively. The relatively low number of patients included as well as the lack of an external validation represent the main limitations of our study.

Conclusions: We report the first update of a nomogram predicting the presence of LNI in patients treated with ePLND. The nomogram maintained high accuracy, even in more contemporary patients (87.6%). Because percentage of positive cores represents the foremost predictor of LNI, its inclusion should be mandatory in any LNI prediction model. Based on our model, those patients with a LNI risk<5% might be safely spared ePLND.