乙型肝炎病毒相关性肝细胞癌预测模型的优化

Optimization of risk-predicting models for hepatitis B virus-related hepatocellular carcinoma

目的优化已建立的HBV相关性肝细胞癌（HCC）预测模型（REACH—B评分模型）。方法收集2004年10月1日至2014年5月1日于福建医科大学附属第一医院肝病中心初次入院、HBsAg阳性超过半年的患者，分为HCC组和对照组（非HCC组），回顾性收集相关指标，进行评分评估。用受试者工作特征曲线判断各模型预测价值。结果预测3年HBV相关性HCC发生，共纳入627例患者，其中HCC组151例，对照组476例。REACHB评分模型预测3年HCC发生的曲线下面积（Auc）为0．78（95％CI：0．74～0．82），敏感度为73．00％，特异度为78．70％。联合甲胎蛋白（AFP）建立R—AFP评分模型，对3年HCC发生的AUC升至0．80（95％CI：0．76～0．83，Z=2．50，P=0．01），敏感度为71．03％，特异度为79．13％。联合甲胎蛋白异质体3与AFP之比（AFP—L3％）建立R—AFP—L3％评分模型，对3年HCC发生的AUC进一步升至0．83（95％CI：0．80～0．87，Z=2．45，P=0．01），敏感度为75．01％，特异度为79．32％。预测5年HBV相关性HCC发生，共纳人159例患者，其中HCC组65例，对照组94例。REACH—B评分模型预测5年HCC发生的AUC为0．79（95％CI：0．72～0．87），敏感度为73．60％，特异度为75．43％。R-AFP评分模型对5年HCC发生的预测价值AUC升至0．84（95％CI：0．77～0．90，Z=2．70，P=0．006），敏感度为83．12％，特异度为77．89％。结论联合AFP及AFP—L3％能优化REACHB评分模型对3年和5年HBV相关性HCC发生的预测价值。

Objective The present study aimed to optimize the established predictive models （REACH-B scoring model） for hepatitis B virus （HBV）-related hepatocellular carcinoma （HCC）. Methods The hepatitis B surface antigen （HBsAg） positive （〉6 months） patients who were firstly admitted in the Liver Center of First Affiliated Hospital, Fujian Medical University between Oct 1st 2004 and May 1st 2014 were selected as the subjects and divided into two groups, namely, the case group （HCC group） and the control group （non-HCC group）. Clinical data of all the subjects were retrospectively collected and analyzed. Receiver operating characteristic curves were used to evaluate the predictive values of the various models. Results To predict the development of HBV-related HCC within 3 years, a total of 627 patients （151 HCC cases and 476 non-HCC controls） were enrolled. Area under curve （AUC） of HBV-related HCC （REACH-B） scoring model was 0.78 （95%CI： 0.74-0.82）, with the sensitivity of 73.00% and specificity of 78. 70% in predicting 3-year risk of HCC occurrence. By combining alphafetoprotein （AFP） and REACH-B, the R-AFP scoring model was constructed. The AUC increased to 0.80 （95%CI： 0. 76-0. 83, Z= 2. 50, P〈0. 01）, with the sensitivity of 71. 03% and specificity of 79.13% in predicting 3-year HCC development. By combining AFP isoform 3 （AFP-L3%） and REACHB, the R-AFP-L3% scoring model was constructed. The AUC further increased to 0.83 （95%CI： 0.80- 0. 87, Z=2.45, P=0.01）, with the sensitivity of 75.01% and specificity of 79.32%in predicting 3-year HCC development. To predict the development of HBV-related HCC within 5 years, a total of 159 （65 HCC cases and 94 non-HCC controls） were enrolled. The AUC of REACH-B scoring model was 0. 79 （95%CI： 0. 72-0. 87）, with the sensitivity of 73.60%and specificity of 75.43G. The R-AFP scoring model had an AUC of 0.84 （95%CI： 0. 77-0. 90, Z=2.70, P=0. 006）, with the sensitivity of 83.12% and specificity of 77. 89%. Conclusion Combination of AFP or AFP-L3% may optimize the predictive values of REACH B scoring model in predicting 3-year and 5-years risks of developing HBV-related HCC.