国产甲型肝炎病毒总抗体检测试剂的评价

Assessment of China-made Enzyme Immunoassay Kits for Total Antibodies to Hepatitis A Virus Detection

目的比较国内常用抗甲型肝炎(甲肝)病毒总抗体(Total Antibody to Hepatitis A Virus,Anti-HAV)酶免疫测定(Enzyme Immunoassay,EIA)试剂的性能,为甲肝血清流行病学调查试剂的选择提供参考。方法以雅培(Abbott)HAVAb2.0试剂的检测结果作为金标准,确定参比系统的总Anti-HAV含量。用A、B两种国产总Anti-HAVEIA试剂,平行检测参比系统8遍,记录CO/S值并作ln(CO/S+0.1)转换,通过广义线性化估计方程(Generalized Estimating Equation,GEE)模型估计各次测试样本的阳性概率,计算比较两种试剂的可靠性指标和真实性指标:组内相关系数(Intraclass Correlation Coefficient,ICC)、变异系数(Coefficient of Variability,CV)和全部受试者工作特征(Receive Operation Characteristic,ROC)曲线下面积和部分ROC曲线下面积,以及固定特异度下的灵敏度[Sensitivityata Particular False-Positive-Rate(1-Specificity),Se(FPR=e)],对比阐明两者之优劣。结果 A试剂的ICC和CV分别为0.9971和5.7840%,B试剂的为0.9952和6.2931%(Bootstrap法,P<0.05);A试剂的ROC曲线下面积(Areaunder Curve,AUC)和部分(Partial)ROC曲线(特异度为0.9～1)下面积(pAUC)分别为0.9557和0.0717,B试制的为0.9404和0.0663(Bootstrap法,P<0.05);在特异度为0.920～0.970时,间隔0.005计算的Se(FPR=e),在特异度为0.945、0.950、0.955、0.960几个观察点时,A试剂的灵敏度>B试剂(Bootstrap法,P<0.05);而在特异度为其他观察点时,对应灵敏度的差异尚无统计学意义(Bootstrap法,P>0.05)。结论 A、B两种试剂均表现出良好的诊断能力,虽然某些指标的差异有统计学意义,但差异有无实际意义,值得进一步探讨。在实际应用时,可根据具体情况进行选择。

Objective Two China-made enzyme immunoassays kits that detect total antibody to hepatitis A virus （ Anti-HAV ） are evaluated,and to provide reference for selecting kits to detect antibody of hepatitis A virus.Methods On the gold standard confirmed by assay-Abbott HAVAb2.0,the two kits for testing total Anti-HAV antibody widely sold in China were selected,and each of them was tested 8 times with the total Anti-HAV antibody panel.The CO/S ratios were recorded and transformed by ln （ CO/S＋0.1 ）,and the generalized estimating equation （ GEE ） model was applied to predict the positive probabilities on their gold standard outcomes.Intraclass correlation coefficient （ ICC ） and coefficient of variability （ CV ） were estimated for evaluating reliability;area under curve （ AUC ）,partial area under curve （ pAUC ）,and sensitivity at a Fixed Specificity [ Se （ FPR=e ）] of the receive operation characteristic curve （ ROC ） indices were calculated for validity.Results The ICC were 0.9971 and 0.9952 for assay A and B （ Beetstrap methed,P〈0.05 ） respectively.The CV were 5.7840 % and 6.2931 % for assay A and B （ Bootstrap method,P〈0.05 ） respectively.Similarly,AUC and pAUC （ specificity ranging from 0.9 to 1 ） were 0.9557 and 0.0717 for assay A,and 0.9404 and 0.0663 for assay B,respectively.The accuracy indices of assay A were statistically different from assay B （ Bootstrap method,P〈0.05 ） .When Se（ FPR=0.020 ） to Se（ FPR=0.070 ） were calculated at 0.005 intervals,higher sensitivities can be observed in assay A （ Bootstrap method,P〈0.05 ） if the specificities were 0.945,0.950,0.955 or 0.960;but at other points of the specificities,there were not （ Bootstrap method,P〈0.05 ） .Conclusion Both assays are excellent diagnosis capability.Although there are statistical difference in some indicators,however,they have no realistic difference.In the practice,we will select them based on the specific situation.