Optimal number of harvested lymph nodes for curatively resected gallbladder adenocarcinoma based on a Bayesian network model

Background and objectives To identify the optimal range and the minimum number of lymph nodes (LNs) to be examined to maximize survival time of patients with curatively resected gallbladder adenocarcinoma (GBAC). Methods Data were collected from the surveillance, epidemiology, and end results database on patients with GBAC who underwent curative resection between 2004 and 2015. A Bayesian network (BN) model was constructed to identify the optimal range of harvested LNs. Model accuracy was evaluated using the confusion matrix and receiver operating characteristic (ROC) curve. Results A total of 1268 patients were enrolled in this study. Accuracy of the BN model was 72.82%, and the area under the curve of the ROC for the testing dataset was 78.49%. We found that at least seven LNs should be harvested to maximize survival time, and that the optimal count of harvested LNs was in the range of 7 to 10 overall, with an optimal range of 10 to 11 for N+ patients, 7 to 10 for stage T1-T2 patients, and 7 to 11 for stage T3-T4 patients. Conclusions According to a BN model, at least seven LNs should be retrieved for GBAC with curative resection, with an overall optimal range of 7 to 10 harvested LNs.