比较进展期肝癌预后系统（ALCPS）、香港中文大学预后指数（CUPI）、奥田邦雄分期系统（OKUDA）、法国肝癌协会（GETCH）分期、巴塞罗那临床肝癌（BCLC）分期5个分期系统对晚期肝细胞肝癌（肝癌）患者3个月生存率和累积生存率的预测价值。

回顾性分析2008年9月1日至2010年6月1日在中山大学附属第三医院就诊，未进行标准抗肿瘤治疗的255例晚期肝癌患者临床资料。其中男220例，女35例；中位年龄56岁。所有患者均签署知情同意书，符合医学伦理学规定。患者初次诊断时按照ALCPS、CUPI、OKUDA、GETCH、BCLC分期系统分别进行评分，并接受随访。应用受试者工作特征（ROC）曲线，获得每条曲线最佳特异度及灵敏度的预测切点；ROC曲线下面积（A）两两比较采用Delong非参数法检验，评价5个分期系统对患者3个月生存率的预测价值。应用Kaplan-Meier生存曲线和Log-rank检验法比较5个分期系统对晚期肝癌患者累积生存率预测情况，应用似然比检验（LRT）法评价其预测价值。

ALCPS的A值为0.823，在5个分期系统中最高，以14分为预测切点，对应灵敏度为0.76，特异度为0.79；其次是CUPI的A值为0.779，以5分为预测切点，对应灵敏度为0.67，特异度为0.78；OKUDA的A值为0.756，1分为预测切点，对应灵敏度为0.79，特异度为0.62；GETCH的A值为0.687，4分为预测切点，对应灵敏度为0.59，特异度为0.67；BCLC的A值为0.615，以C期为预测切点，对应灵敏度为0.49，特异度为0.74。ALCPS的A值与CUPI、OKUDA、GETCH和BCLC比较差异有统计学意义（Z=2.251，2.577，4.600，5.906；P<0.05），CUPI明显优于GETCH和BCLC，OKUDA则优于BCLC（Z=3.059，4.715；P<0.05）。ALCPS、OKUDA及GETCH的各分组或分期生存曲线比较差异有统计学意义（P<0.05），各分组或分期生存曲线无交叉，有较好的区分价值，而CUPI、BCLC各分组或分期生存曲线比较差异有统计学意义（P<0.05），但其生存曲线存在一定的交叉，区分价值欠佳。LRT检验显示ALCPS、CUPI、OKUDA、GETCH和BCLC的似然比（χ²值）分别为75.70、52.82、45.22、16.87、14.36。ALCPS的χ²值最高，其预测晚期肝癌患者累积生存率的价值最高。

在5个分期系统中，ALCPS对晚期肝癌患者3个月生存率预测能力最好，可作为晚期肝癌患者入组临床试验的选择指标。ALCPS对晚期肝癌患者累积生存率预测价值亦最高，亦可作为晚期肝癌患者长期预后的预测指标。

To evaluate the five current staging systems, including Advanced Liver Cancer Prognostic System (ALCPS), Chinese University Prognostic Index (CUPI) , OKUDA staging system, Groupe d'Etude et de Traitement du Carcinome Hepatocellulaire (GETCH) and Barcelona Clinic Liver Cancer (BCLC) in predicting the 3-month survival rate and cumulative survival rate of patients with advanced hepatocellular carcinoma (HCC).

The subjects of this retrospective study were 255 patients (220 males and 35 females with the median age of 56 years old who did not receive any standard anti-cancer treatment in the Third Affiliated Hospital of Sun Yat-sen University from September 1, 2008 to June 1, 2010. The informed consents of all patients were obtained and the ethical committee approval was received. All the patients were scored according to ALCPS, CUPI, OKUDA, GETCH and BCLC staging system at their first diagnosis and were followed up afterwards. Receiver operating characteristic (ROC) curve analysis was carried out to find out the cut-off point of each scoring curve with best sensitivity and specificity. Nonparametric method Delong test was used to compare Area under curve (A) of ROC in order to evaluate different staging systems in predicting the patients' 3-month survival rate. Kaplan-Meier survival curve and Log-rank test were used to compare the cumulative survival rate of advanced HCC patients’ of different staging systems and the likelihood ratio test (LRT) was used to assess the prediction value of 5 systems.

The A value of ALCPS was 0.823, which was the highest among the 5 staging systems. The cut-off point is 14 with the corresponding sensitivity of 0.76 and specificity of 0.79. The A value of CUPI was 0.779. The cut-off point was 5 with the corresponding sensitivity of 0.67 and specificity of 0.78. The A value of OKUDA was 0.756. The cut-off point was 1 with the corresponding sensitivity of 0.79 and specificity of 0.62. The A value of GETCH was 0.687. The cut-off point was 4 with the corresponding sensitivity of 0.59 and specificity of 0.67. The A value of BCLC was 0.615. The cut-off point was stage C with the corresponding sensitivity of 0.49 and specificity of 0.74. There were significant differences between the A value of ALCPS and that of CUPI, OKUDA, GETCH and BCLC (Z=2.251, 2.577, 4.600, 5.906; P<0.05). The A value of CUPI was significantly higher than GETCH and BCLC and that of OKUDA was significantly higher than BCLC (Z=3.059, 4.715; P<0.05). There were significant differences for the survival curve among the subgroups of ALCPS, OKUDA and GETCH (P<0.05) . There was no cross among the subgroups survival curves with distinguishable value. The differences among the subgroups survival curve of CUPI and BCLC are statistically significant(P<0.05) . Some crosses were observed among the subgroups survival curve of CUPI and BCLC and the distinguishable value was poor. The LRT result showed that the χ² value of ALCPS, CUPI, OKUDA, GETCH and BCLC were 75.70, 52.82, 45.22, 16.87, 14.36 respectively. The χ² value of ALCPS was the highest, which indicated the best predicting value of cumulative survival.

The ALCPS staging system is the best in predicting 3-month survival rate of advanced HCC and can serve as an inclusion criteria for clinical trials of advanced HCC. The ALCPS staging system is also the best in predicting cumulative survival rate of advanced HCC and can be used as a prognostic index for patient's long-term outcome.