Application of linear discriminant analysis in performance evaluation of extractable nuclear antigen immunoassay systems in the screening and diagnosis of systemic autoimmune rheumatic diseases.

Abstract

This study applied a linear discriminant analysis model to evaluate the performance of 2 types of commercially available extractable nuclear antigen (ENA) immunoassays for the screening and diagnosis of systemic autoimmune rheumatic diseases (SARDs) in a large tertiary hospital reference laboratory: (1) an enzyme-linked immunosorbent assay (ELISA) and (2) a multiplex bead-based immunoassay (MPBI). The results of the study showed both ENA immunoassays had comparable sensitivity for the detection of SARDs compared with the antinuclear antigen immunofluorescence (ANA-IF) method (ANA-IF: 85.6%, ENA-ELISA: 91.5%, ENA-MPBI: 83.1%, pairwise comparisons with ANA-IF: P > .05). However, both ENA immunoassays offered improved specificity compared with the ANA-IF (ANA-IF: 24.2%; ENA-ELISA: 39.8%; ENA-MPBI: 53.1%; pairwise comparison with ANA-IF: P < .001). The use of a more specific screening immunoassay with comparable sensitivity to ANA-IF is important in a tertiary hospital with high prevalence of non-SARD immune diseases. Diagnostic performance of the ENA/dsDNA components by the MPBI and ELISA methods did not differ significantly (area under the curve [AUC], 81.0% vs 83.0%, respectively, P > .05), but the key ENA/dsDNA variables contributing to the discriminating power of the assays for the diagnosis of specific SARDs were reagent/method dependent.