Test characteristics of immunofluorescence and ELISA tests in 856 consecutive patients with possible ANCA-associated conditions.

Abstract

To examine the test characteristics of immunofluorescence (IF) and enzyme-linked immunosorbent assays (ELISA) in a consecutive series of patients under evaluation for anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV). Using stored sera, we performed a cross-sectional study on 856 consecutive patients tested prospectively for ANCA by IF, Based on guidelines from the 1994 Chapel Hill Consensus Conference (CHCC), we determined each patient's underlying diagnosis by a medical records review without regard to their ANCA status (the CHCC guidelines do not require ANCA as a prerequisite for diagnosis). We grouped patients with forms of vasculitis commonly associated with ANCA into one of 4 types of AAV: Wegener's granulomatosis (n = 45), microscopic polyangiitis (n = 12), Churg-Strauss syndrome (n = 4), and pauci-immune glomerulonephritis (n = 8). We also classified patients without clinical evidence of AAV (92% of all patients tested) into 5 predefined categories of disease (including "other") and an additional category for no identifiable disease. In a blinded fashion, we then performed ELISAs on the stored serum for antibodies to proteinase-3 (PR3) and myeloperoxidase (MPO) and calculated the test characteristics for both ANCA assay techniques. Sixty-nine of the 856 patients (8.1%) had clinical diagnoses of AAV based on CHCC guidelines. The positive predictive value (PPV) of ELISA for AAV was superior to that of IF, 83% versus 45%. For patients with both positive IF tests and positive ELISA tests, the PPV increased to 88%. Both IF and ELISA had high negative predictive values (97% and 96%, respectively). Positive ELISA tests were associated with higher likelihood ratios (LR) than IF (54.2 [95% CI = 26.3, 111.5] versus 9.4 [95% CI = 6.9, 12.7]). The LR of both a positive IF and a positive ELISA was 82.1 (95% CI = 33.3, 202.5). Compared with IF, an ELISA test fo ANCA was associated with a substantially higher PPV and LR for AAV. This fact, combined with the greater sensitivity of IF, suggests that an effective testing strategy is to perform ELISA tests only on samples that are positive for ANCA by IF.