Personalized Preoperative Nomograms Predicting Postoperative Risks after Resection of Perihilar Cholangiocarcinoma.

INTRODUCTION

Curative treatment of perihilar tumors requires major hepatectomy responsible for high morbidity and mortality. Current nomograms are based on definitive pathological analysis, not usable for patient selection. Our aim was to propose preoperative predictors for severe morbidity (Dindo-Clavien ≥3) and mortality at sixth month after resection of perihilar tumors.

PATIENTS AND METHODS

We reviewed perioperative data of 186 patients operated with major hepatectomy for perihilar tumors between 2012 and 2018 in two high-volume centers. Univariate and multivariate analysis were performed to determine the preoperative predictors of morbidity and mortality. A stepwise regression in forward direction was developed to select variables for definitive models. Hosmer-Lemeshow test, Akaike information criteria and area under the ROC curves were calculated to validate both nomograms.

RESULTS

Resections were indicated for perihilar and intrahepatic cholangiocarcinoma in 125 and 61 cases, respectively. Severe complications occurred in 76 patients (40.8%). Nineteen patients (10.2%) deceased before the sixth postoperative month. The predictors of severe morbidity were: male gender, portal vein embolization, planned biliary resection, low psoas muscle area/height and low hemoglobinemia. The predictors of early mortality were: age, high bilirubinemia, hypoalbuminemia, biliary drainage and long drainage-to-surgery interval. For both models, the p values of Hosmer-Lemeshow tests were of 0.9 and 0.99, respectively, the Akaike information criteria were of 35.5 and 37.7, respectively, and area under the curves was of 0.73 and 0.86, respectively.

CONCLUSION

We developed two accurate and practical nomograms based on exclusively preoperative data to predict early outcomes following the resection of perihilar tumors. If validated in larger series, these tools could be integrated in the decision-making process for patient selection.