ABSTRACT
BACKGROUND: A significant number of patients with minimal lymph node disease at radical prostatectomy (RP) and pelvic lymph node dissection (PLND) have better than expected long-term outcomes. We explored whether stratification by number of positive nodes enhances our institutional prediction model for biochemical recurrence after RP. METHODS: A total of 7789 patients underwent RP and pelvic lymph node dissection from 1995 to 2012 at a tertiary referral center. We compared two recurrence prediction models: one incorporated lymph node invasion and the other tracked the number of positive nodes. Existing and updated models' discrimination was assessed using Harrell's c-index and calibration. The 10-fold cross-validation was performed to correct for model overfitting. RESULTS: Of the 491 patients (6.3%) harboring nodal disease, 387 (5.0%) had 1-2 positive nodes and 104 (1.3%) had ⩾3 positive nodes. Data on number of positive nodes did not improve the c-index for the cohort as a whole. When we assessed discrimination for node-positive patients only, c-index for the model with number of positive nodes was 0.01 (95% confidence interval 0.001-0.024) higher than the model with lymph node invasion. Illustrative examples were provided by reclassification tables using number of positive lymph nodes. For instance, 40 of 7789 patients would be reclassified with a cutoff point of 50% for biochemical recurrence at 1 year, and 36 of 7789 patients would be reclassified with a cutoff point of 40% for biochemical recurrence at 10 years. CONCLUSIONS: Stratification by number of positive lymph nodes provided additional discriminative ability for evaluating risk in node-positive patients. Pending external validation, this model could be used for patient counseling and clinical trial stratification in this subpopulation.
