Preoperative CT findings for prediction of resectability in patients with gallbladder cancer

Abstract

Objectives To predict residual tumor (R) classification in patients with a surgery for gallbladder (GB) cancer, using preoperative CT. Methods One hundred seventy-three patients with GB cancer who underwent CT and subsequent surgery were included. Two radiologists assessed CT findings, including tumor morphology, location, T stage, adjacent organ invasion, hepatic artery (HA) invasion, portal vein invasion, lymph node metastasis, metastasis, resectability, gallstone, and combined cholecystitis. The R classification was categorized as no residual tumor (R0) and residual tumor (R1 or R2). We analyzed the correlation between CT findings and R classification. We also followed up the patients as long as five years and analyzed the relationship between the R classification and the overall survival (OS). Results There were 134 patients with R0 and 39 patients with R1/R2. On multivariable analysis, liver invasion (Exp(B)= 3.19, p = 0.010), bile duct invasion (Exp(B) = 3.69, p = 0.031), and HA invasion (Exp(B) = 3.74, p = 0.039) were independent, significant predictors for residual tumor. When two of these three criteria were combined, the accuracy for predicting a positive resection margin was 83.38% with a specificity of 93.28%. The OS and the median patient survival time differed significantly according to the resection margin, i.e., 56.0% and 134.4 months in the R0 resection and 5.1% and 10.8 months in the R1/R2 resection group (p < 0.001). Conclusions Preoperative CT findings could aid in planning surgery and determining the resectability using the high-risk findings of residual tumor, including liver invasion, bile duct invasion, and HA invasion. Key Points Liver invasion, bile duct invasion, and HA invasion were significant preoperative CT predictors for residual tumor in GB cancer. HA invasion showed the highest OR on multivariate analysis and the highest predictor point on a nomogram for predicting a positive resection margin. Association of two factors can predict positive resection margin with an accuracy of 83.38% and a specificity of 93.28%.