甲胎蛋白诊断原发性肝癌准确性的系统评价

目的对血清甲胎蛋白（AFP）诊断原发性肝癌进行系统评价。方法通过计算机检索及手工查阅获得研究甲胎蛋白诊断原发性肝癌的中文文献，按纳入与排除标准筛选文献并提取纳入研究中有关准确度的数据，采用MetaDisc1．4软件检验研究问的异质性，按AFP临界值20或25μg／L、200μg/L及400μg/L分组，对诊断准确度指标进行Meta分析并绘制综合受试者工作特征曲线。结果从1062篇文献中筛选出20篇符合纳入标准的研究，病例组6969例为经病理确诊的原发性肝癌，对照组3446例为经临床确诊的非肝癌患者。按AFP临界值20或25μg／L、200μg/L及400μg／L分组，异质性检验提示无阈值效应，但存在其他原因导致的异质性，按随机效应模型做Meta分析。结果显示：与其他两组相比，AFP400μg／L组特异度最高[0．977，95％CI（0．967，0．985）]，但敏感度最低[0．422，95％CI（0．403，0．441）]；AFP400μg／L组阳性似然比值及诊断优势比值（17．691和32．820）低于200μg／L组（19．669，53．599），SROC曲线下面积（0．6575）及Q^＊指数值（0．6338）也明显低于200μg／L组（0．8323和0．7822）。结论AFP400μg／L作为原发性肝癌诊断标准值得商榷，我们初步认为原发性肝癌AFP诊断值200μg／L优于400μg/L。

A Systematic Review of the Accuracy of Diagnostic Test of AFP for Chinese Primary Liver Cancer

Objective To systematically evaluate AFP as the diagnostic standard for Chinese primary liver cancer （PLC）. Methods A comprehensive electronic search and additional manual tracking were performed to retrieve relevant studies on AFP in diagnosis of Chinese PLC. All studies were divided into three groups according to the cutoff value of AFP： 20 or 25 μg/L, 200 μg/L, 400 μg/L （Groups 1, 2, and 3, respectively）. The data about the accuracy of the included studies were extracted for further heterogeneity studies; statistical pooling and SROC （summary receiver operating characteristics） were analyzed using MetaDisc 1.4 software. Results Twenty studies which were selected from 1,062 references met the inclusion criteria. Heterogeneity （except for threshold effect） was found within the three groups. A Meta-analysis was performed using the random effect model. Compared with the other two groups, the specificity of Group 3 （AFP 400 μg/L） was the highest （0.977, 95%CI 0.967 to 0.985） and sensitivity was the lowest （0.422, 95%CI 0.403 to 0.441）. The values of LR＋ and dOR were lower than those of Group 2 （AFP 200Fg/L） （17.691： 19.669; 32.820： 53.599, respectively）. Area under curve （AUC） of SROC and Q^＊ index of Group 3 were also lower than those of Group 2 （0.6575： 0.832 3; 0.633 8：0.782 2, respectively）. Conclusion Four-hundred μg/L of AFP as the diagnostic standard for PLC is not good enough, and we suggest that 200 μg/L may be better than 400 μg/L for PLC diagnosis.