Impact of surgical volume on the rate of lymph node metastases in patients undergoing radical prostatectomy and extended pelvic lymph node dissection for clinically localized prostate cancer.

BACKGROUND: Previous trials have shown that the number of procedures done by a single surgeon, that is, surgical volume (SV), is associated with several outcomes after radical prostatectomy (RP).
OBJECTIVE: To test the association between SV and the detection of lymph node metastases during extended pelvic lymph node dissection (ePLND). DESIGN, SETTING, AND PARTICIPANTS: The study cohort consisted of 1020 men surgically treated for clinically localized prostate cancer. INTERVENTION: All patients underwent RP and ePLND by a group of six surgeons who were trained by the surgeon with the highest SV. All surgeons performed an anatomically extended PLND, including removal of obturator, external iliac, and hypogastric nodes. MEASUREMENTS: Univariable and multivariable logistic regression models tested the association between SV (either continuously coded or dichotomized according to the most informative cut-off, namely >144 vs < or =144 ePLNDs) and the rate of lymph node invasion (LNI) after accounting for preoperative (baseline prostate-specific antigen [PSA], clinical stage, biopsy Gleason sum) and postoperative (pathologic stages and Gleason score, surgical margin status) patient characteristics. RESULTS AND LIMITATIONS: Mean number of nodes removed was 19.1 (median, 16; range, 7-63). Mean overall SV was 227 (range, 87-379). Overall, LNI rate was 11.8% (120/1020). No significant differences were found among patients treated by the different surgeons in terms of clinical and pathologic characteristics (all p> or =0.06). Conversely, the surgeon with the highest SV removed more nodes and found more nodal metastases compared with the other surgeons (21.1 vs 17.9 mean number of nodes removed; p<0.001, and 15 vs 9.8% of LNI; p=0.01, respectively). At univariable logistic regression analysis, either continuously coded or dichotomized SV was a significant predictor of LNI (p=0.007 and p<0.001, respectively). In multivariable models, continuously coded as well as dichotomized SV maintained a significant association with the rate of LNI, after accounting for preoperative (p=0.04 and p=0.009, respectively) as well as for postoperative variables (p=0.03 and p=0.002, respectively).
CONCLUSIONS: After adjusting for clinical and pathologic case-mix differences, patients treated by the highest-volume surgeons (>144 ePLNDs) were more likely to have LNI than those treated by low-volume surgeons, even though all surgeons used a similar extended template for node removal.