Evaluation of combined determinations of Epstein-Barr virus antibodies for nasopharyngeal carcinoma assessed with receiver operating characteristic curve based on logistic regression.
- Author:
Yong-lin CAI
1
;
Yu-ming ZHENG
;
Ji-ru CHENG
;
Jun LI
;
Yong-kun MO
;
Qing-yan ZHONG
Author Information
- Publication Type:Journal Article
- MeSH: Adult; Aged; Aged, 80 and over; Antibodies, Viral; blood; immunology; Carcinoma; diagnosis; immunology; virology; Enzyme-Linked Immunosorbent Assay; methods; standards; Female; Herpesvirus 4, Human; immunology; Humans; Logistic Models; Male; Middle Aged; Nasopharyngeal Neoplasms; diagnosis; immunology; virology; Viral Proteins; immunology; Young Adult
- From: Chinese Journal of Experimental and Clinical Virology 2009;23(5):384-387
- CountryChina
- Language:Chinese
-
Abstract:
OBJECTIVEThis 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.
METHODSSerum 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.
RESULTSThe 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 + EBNA1/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.
CONCLUSIONThe 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.