Objective: To compare performances for predicting surgical difficulty and postoperative complications.
Background: An expert panel recently proposed a complexity classification for liver resection with 3 categories of complexity (low, medium, or high). We compared this new classification with the conventional major/minor classification.
Methods: We retrospectively reviewed 469 hepatocellular carcinoma patients who underwent liver resection between 1 January 1, 2004 and June 30, 2015. We used receiver-operating characteristic curve analysis to compare the performances of both classifications for predicting perioperative outcomes.
Results: Both classifications effectively differentiated subgroups of patients in terms of their intraoperative findings and short-term outcomes, including blood loss, transfusion rate, operation time, and postoperative hospital stay (all P < 0.05). The ability to predict complications was not significantly different between the major/minor classification and the complexity classification [area under the curve (AUC) 0.625 vs 0.617, respectively; P= 0.754). However, the complexity classification showed stronger correlations with blood loss (AUC 0.690 vs 0.617, respectively; P = 0.001) and operation time (AUC 0.727 vs 0.619, respectively; P < 0.001) compared with the major/minor classification. To check heterogeneity, the minor resection group was further divided into low (n = 184), medium (n = 149), and high complexity (n = 13) groups. Operation time and blood loss were significantly different among these 3 subgroups of patients.
Conclusions: The complexity classification outperformed the major/minor classification for predicting the surgical difficulty of liver resection.