Assessing the risk of lymph node invasion in patients with intermediate risk prostate cancer treated with extended pelvic lymph node dissection. A novel prediction tool

BACKGROUND Controversy exists regarding the need for extended pelvic lymph node dissection (ePLND) in patients with intermediate risk prostate cancer (PCa). MATERIALS AND METHODS. The study included 982 consecutive men with intermediate risk PCa (PSA 10-20 ng/ml or cT2b-c or biopsy Gleason 3 + 4/4 + 3) treated with ePLND and radical prostatectomy (RP) at a single center. All patients underwent an anatomically defined ePLND. A novel risk stratification tool was developed by applying the nonparametric tree modeling technique of classification and regression tree analysis (CART) which relied on pre-operative PSA, clinical stage, biopsy Gleason score, and percentage of positive cores. The area under the receiver characteristic curve (AUC) method was used to quantify the accuracy of the model. RESULTS. Lymph node invasion (LNI) was found in 81 (8.2%) patients. The CART analyses identified three risk groups of having LNI: a) Low risk: Gleason 3 + 3, cT1c/cT2, PSA 1020 ng/ml, or Gleason 3 + 4/4 + 3, <= 63% of positive cores and PSA < 5 ng/ml (risk of LNI: 3.7 and 5.2%, respectively; 64.8% of patients included); b) Moderate risk: Gleason 3 + 4/4 + 3, >63% of positive cores and PSA >= 5 ng/ml (risk of LNI: 14.4%; 23% of patients included); c) High risk: Gleason 3 + 4/4 + 3, % positive cores >63% (risk of LNI: 20.1%; 12.% of patients included; P < 0.001). The accuracy of the model was 71%. CONCLUSIONS. The risk of having LNI varies significantly (3.7-20.1%) in patients with intermediate risk PCa. Our predictive tool might help selecting those patients suitable for ePLND, allowing to spare this approach in about 60% of intermediate risk patients. Prostate 72: 499-506, 2012. (C) 2011 Wiley Periodicals, Inc.