应用基于Logistic回归的ROC曲线评价鼻咽癌EB病毒抗体联合检测  

Evaluation of combined determinations of Epstein-Barr virus antibodies for nasopharyngeal carcinoma assessed with receiver operating characteristic curve based on Logistic regression

作　　者：蔡永林[1] 郑裕明[1] 成积儒 李军[1] 莫永坤 钟青燕[1] 

机构地区：[1]广西梧州市红十字会医院、梧州市肿瘤防治研究所,543002

出　　处：《中华实验和临床病毒学杂志》2009年第5期384-387,共4页Chinese Journal of Experimental and Clinical Virology

基　　金：广西壮族自治区卫生厅计划课题（Z2008472）;梧州市科技计划项目（200801063） 衷心感谢中山大学公共卫生学院柳青教授在统计学处理方面给予的帮助.

摘　　要：目的以基于Logistic回归的受试者工作特征（ROC）曲线分析方法，评价VCA／IgA、EA／IgA、Rta／IgG和EBNAl／IgA等EB病毒抗体不同组合在鼻咽癌诊断中的价值。方法收集211例初治鼻咽癌和203例相似症状的非鼻咽癌病例的血清，采用免疫酶法检测VCA／IgA及EA／IgA，酶联免疫吸附法检测Rta／IgG和EBNAl／IgA。对各种的抗体组合建立Logistic回归模型，以预测概率为分析指标，应用ROC曲线分析，评价不同组合对鼻咽癌的诊断价值。结果单一指标评价，VCA／IgA敏感度最高（98．1％），EA／IgA特异度最高（98．5％）。以基于Logistic回归的ROC曲线分析各项组合，敏感度和特异度均有所提高。双指标组合中，VCA／IgA＋Rta／IgG组合的ROC曲线下面积（AUC）为0．991，诊断效能最高，敏感度、特异度及约登指数分别为94．8％、98．0％及0．928。VCA／IgA＋Rta／IgG＋EBNA1／IgA组合和4项指标组合的敏感度、特异度及约登指数分别为94．8％、98．5％、0．933和96．7％、97．0％、0．937；这两种多指标组合的AUC与VCA／IgA＋Rta／IgG组合比较差异均无统计学意义（P〉0．05）。结论基于Logistic回归的ROC曲线分析方法可以为多指标联合诊断试验提供更客观的综合分析，VCA／IgA和Rta／IgG联合检测具有互补作用．是鼻咽癌血清学诊断的合适组合。Objective This study was aimed to investigate the diagnostic value of combined determination of Epstein-Barr virus （EBV） antibodies for nasopharyngeal carcinoma （ NPC）, including immunoglobulin （Ig） A against EBV capsid antigens （VCA）, IgA against early antigens （EA）, IgG against BRLF1 transcription activator （Rta） and IgA against EBV nuclear antigen-1 （EBNA1）, assessed with receiver operating characteristic （ROC） curve based on logistic regression. Methods Serum samples derived from 211 untreated patients with NPC and 203 non-NPC ENT patients were examined for the presence of VCA/IgA and EA/IgA by immunoenzymatic assay, Rta/ IgG and EBNA1/IgA by enzyme-linked immnunosorbent assay （ELISA）. The different Logistic regression models were established for various combined determinations of antibodies, respectively. Using the predicted probability as the analyzed variable, ROC curve was applied to evaluate the diagnostic accuracy of different combined determinations. Results The sensitivity of VCA/IgA （98.1% ） and the specificity of EA/IgA （98.5 % ） were the highest while detecting solely. The results which were analyzed by ROC curve based on Logistic regression showed that the sensitivity and specificity were improved. In two-marker combinations, VCA/IgA ＋ Rta/IgG whose area under ROC curve （AUC） was 0.991 had the highest diagnostic accuracy, and its sensitivity, specificity and Youden index were 94.8%, 98.0% and 0.928 respectively. ＂No significant difference of AUC were found comparing VCA/IgA ＋ Rta/IgG with VCA/IgA ＋ Rta/IgG ＋ EBNAi~/IgA and four-marker combination（ P 〉 0.05）, of which sensitivity, specificity and Youden index were 94.8%, 98.5%, 0.933 and 96.7%, 97.0%, 0.937,respectively. Conclusion The approach of ROC curve based on Logistic regression can improve synthetic efficiency for combined determination of multiple markers. The combined determination of VCA/IgA and Rta/IgG with a complementary effect is optimal for NPC serodiagnosis.

关 键 词：鼻咽肿瘤 疱疹病毒4型 人 抗体 肿瘤 LOGISTIC模型 

分 类 号：R739.63[医药卫生—肿瘤]