Head-to-head comparison of lymph node density and number of positive lymph nodes in stratifying the outcome of patients with lymph node-positive prostate cancer submitted to radical prostatectomy and extended lymph node dissection

Abstract Objective The aim of this study was to compare the predictive ability of lymph node density (LND) and number of positive lymph nodes in patients with prostate cancer and lymph node invasion. Materials and methods We included 568 patients with lymph node invasion treated with radical prostatectomy and extended pelvic lymph node dissection between January 1990 and July 2011 at a single center. The Kaplan-Meier method and multivariable Cox regression models tested the association between the number of positive lymph nodes or LND and cancer-specific survival (CSS). The predictive accuracy of a baseline model was assessed using Harrell's concordance index and then compared with that of a model including either the number of positive nodes or LND. Results The median number of positive lymph nodes was 2, whereas the median LND was 11.1%. At 5, 8, and 10 years, CSS rates were 92.5%, 83.9%, and 82.8%, respectively. At multivariable analyses, number of positive lymph nodes and LND, considered as continuous variables, were independent predictors of CSS (all P ≤0.01). A 30% LND cutoff was found to be highly predictive of CSS ( P = 0.004), and a cutoff of 2 positive nodes was confirmed to be a strong predictor of CSS ( P = 0.02). The number of positive nodes and LND similarly, continuous or dichotomized, increased the accuracy for CSS predictions (0.68–0.69 vs. 0.61 of baseline model). LND cutoff of 30% increased the discrimination the most (0.69; +0.083). Conclusions The number of positive lymph nodes and LND showed comparable discriminative power for long-term CSS predictions. A cutoff of 30% LND might be suggested for the selection of patients candidate for adjuvant systemic therapy, because it increased the model's discrimination the most.