目的：比较Child-Pugh分级标准（CP）、终末期肝病模型（MELD）和白蛋白-胆红素评分模型（ALBI）预测肝癌患者R0切除术后发生肝衰竭（PHLF）的能力。方法：收集2010年9月—2015年11月符合研究标准的275例肝癌行肝癌R0切除术的患者资料，应用受试者工作特征曲线（ROC）分析方法，比较3种评分预测发生PHLF的准确性。结果：依据50-50标准，275例患者中43例被诊断为PHLF，ALBI、CP、MELD诊断PHLF的ROC曲线下面积（AUROC）（95% CI）分别为0.930（0.893~0.957）、0.795（0.743~0.841）、0.736（0.680~0.787）（ALBI vs. CP：P=0.0003；ALBI vs. MELD：P=0.0001；CP vs. MELD：P=0.2420）。依据国际肝脏外科学组（ISGLS）标准，275例患者中48例被诊断为PHLF，ALBI、CP、MELD诊断PHLF的AUROC（95% CI）分别为0.884（0.840~0.920）、0.828（0.778~0.871）、0.762（0.707~0.811）（ALBI vs. CP：P=0.1542；ALBI vs. MELD：P=0.0064；CP vs. MELD：P=0.2010）。按肝切除范围分亚组分析的结果与全组分析的结果基本一致。结论：3种评分系统中，ALBI评分预测肝癌患者R0切除术后PHLF优于CP和MELD，MELD评分（由于分数构成中肌酐权重较大）相对不适用于轻症早期肝癌患者。
Comparison of Child-Pugh, MELD and ALBI scoring systems in predicting postoperative liver failure after liver cancer R0 resection
Objective: To compare the accuracy of the scoring systems of Child-Pugh (CP), model for end-stage liver disease (MELD) and albumin-bilirubin scoring model (ALBI) in predicting posthepatectomy liver failure (PHLF) in liver cancer patients after R0 resection. Methods: The data of 275 liver cancer patients undergoing R0 resection and meeting the study criteria from September 2010 to November 2015 were collected. The abilities in predicting the occurrence of PHLF among the three scoring systems were compared by using Receiver operating characteristic (ROC) curves. Results: According to 50-50 criteria, 43 cases in the 275 patients were diagnosed with PHLF; the area under the ROC curve (AUROC) (95% CI) for ALBI, CP and MELD in diagnosis of PHLF was 0.930 (0.893–0.957), 0.795 (0.743–0.841) and 0.736 (0.680–0.787), respectively (ALBI vs. CP: P=0.0003; ALBI vs. MELD: P=0.0001; CP vs. MELD: P=0.2420). According to criteria of International Study Group of Liver Surgery (ISGLS), 48 cases in the 275 patients were diagnosed with PHLF; the AUROC (95% CI) for ALBI, CP and MELD in diagnosis of PHLF was 0.884 (0.840–0.920), 0.828 (0.778–0.871) and 0.762 (0.707–0.811), respectively (ALBI vs. CP: P=0.1542; ALBI vs. MELD: P=0.0064; CP vs. MELD: P=0.2010). The results of subgroup analysis stratified by liver resection scope were generally consistent with those of whole group analysis. Conclusion: Among the three scoring system, ALBI scoring is superior to CP and MELD scoring in predicting the occurrence of PHLF in liver cancer patients after R0 resection, and MELD scoring (due to greater value of weighting coefficient of the creatinine in the scoring system) might not be applicable for early-stage patients with mild symptoms.