Logistic regression and discriminant analyses of hepatic failure after liver resection for carcinoma of the biliary tract

Abstract

Stepwise logistic and discriminant analyses were used to define the risk factors for hepatic failure after liver resection for carcinoma of the biliary tract and to predict this complication. The logistic analysis identified the linear pattern of oral glucose tolerance test, cholangitis, pancreatoduodenectomy, and indocyanine green disappearance rate as the factors most related to hepatic failure among 18 pre- and perioperative parameters of the past 84 hepatectomized cases (55 hilar bile duct carcinomas and 29 gallbladder carcinomas). Our discriminant formula using five variables, including the above-mentioned four plus the liver resection rate, could distinguish the patients with posthepatectomy liver failure (n = 25) from those without (n = 59) with 86.9% accuracy, 96.0% sensitivity, and 83.1% specificity. The cross-validation test has confirmed that this model was robust for discriminant analysis. The results of this study show that statistical multifactorial analysis makes possible the preoperative prediction of hepatic failure after liver resection for carcinoma of the biliary tract.