Abstract

Objectives: To determine whether a minimum number of lymph nodes (LNs) exist to detect lymph node invasion (LNI) in patients undergoing radical nephroureterectomy (RNU) for upper tract urothelial carcinoma. Methods: The study included 551 consecutive patients, from 13 centers worldwide, who underwent RNU and lymphadenectomy (LND) between 1992 and 2006. LND was performed at the discretion of the surgeon. All pathological slides were re-reviewed by uropathologists according to strict criteria. Receiver-operating characteristic curve coordinates were used to determine the probability of diagnosing LNI according to the total number of nodes removed. Additionally, the relationship between the number of nodes removed and the rate of positive LNs was tested in univariate and multivariate logistic regression models. Results: Median patient age was 68 years (range: 27-97). Of 551 patients, 140 (25.4%) had positive lymph nodes. Median number of lymph nodes removed was 5 (mean 6.7, range 1-41). The Receiver-operating characteristic coordinates plot indicated that the removal of 13 nodes yielded a 90% probability to detect =1 positive LNs. The removal of 8 nodes resulted in a 75% probability of finding =1 positive nodes. Removal of >8 LNs (P = .03; odds ratio 1.49) was independently associated with LNI after adjusting for pathological stage and grade. Conclusions: Our data indicate that 8 LNs need to be removed at radical nephroureterectomy to achieve a 75% probability of finding =1 positive nodes. Further improvement of the specificity of LND will require the removal of more lymph nodes. © 2009 Elsevier Inc. All rights reserved.