Value of general surgical risk models for predicting postoperative liver failure and mortality following liver surgery

Abstract

This study evaluated the ability of general surgical models to predict postoperative morbidity and mortality in liver surgery. The postoperative course and mortality rates predicted by general surgical models were investigated in 960 patients who underwent hepatectomy or ablation therapy for primary or metastatic liver carcinoma. The area under the receiver operative characteristic curve (95% confidence intervals) for detecting postoperative liver failure was 0.89 (0.84-0.94), 0.85 (0.78-0.92), and 0.78 (0.72-0.85) for the estimation of physiologic ability and surgical stress (E-PASS) model, the modified E-PASS (mE-PASS) model, and the Portsmouth Physiologic and Operative Severity Score for the enUmeration of Mortality and morbidity (P-POSSUM) model, respectively, and those for detecting in-hospital mortality were 0.85 (0.76-0.93), 0.85 (0.78-0.92), and 0.79 (0.71-0.87), respectively. Nevertheless, all of the models overpredicted the overall mortality rate (by 2.3-fold for E-PASS, 2.3-fold for mE-PASS, and 2.9-fold for P-POSSUM). The general surgical risk models demonstrated high discriminatory power for predicting postoperative outcomes in liver surgery, but overpredicted the overall mortality rate by more than twofold. Therefore, these models should be refined to make them more suitable for predicting liver surgery outcomes.