Decision of surgical approach for advanced gallbladder adenocarcinoma based on a Bayesian network.

BACKGROUND AND OBJECTIVES: To determine whether radical resection can benefit patients with advanced gallbladder adenocarcinoma using a Bayesian network (BN) with clinical data.

METHODS: In total, 362 patients who had undergone surgical treatment of gallbladder adenocarcinoma at a tertiary institute were evaluated to establish two BN models using a tree-augmented naïve Bayes algorithm. We then chose 250 patients with T3-4N0-2M0 stage gallbladder adenocarcinoma to test the posterior probability after the surgical type was taken into account.

RESULTS: In total, 170 patients (≤7 months) and 137 patients (>7 months) were correctly classified in the median survival time model (accuracy, 84.81%), and 204 patients (≤12 months), 15 patients (12-36 months), 17 patients (36-60 months), and 34 patients (>60 months) were correctly classified in the 1-, 3-, and 5-year survival model (accuracy, 74.59%), respectively. Every posterior probability in the two models upregulated the ratio of the longer survival time and suggested a better prognosis for gallbladder adenocarcinoma that can be improved by R0 resection.

CONCLUSIONS: These BN models indicate that stages T4 and N2 gallbladder adenocarcinoma are not contraindications for surgery and that R0 resection can improve survival in patients with advanced gallbladder adenocarcinoma.