Antineutrophil cytoplasmic antibodies (ANCA) are used as diagnostic markers for systemic vasculitis. However, the specificity and sensitivity of ANCA detection differs from centre to centre due in large part to variations in methodology. We compared 8 commercial ELISA kits and an in-house method (HM) for their specificity and sensitivity in detecting ANCA against proteinase 3 (PR3-ANCA, 7 kits) and meyloperoxidse (MPO-ANCA, 8 kits). Sera from 5 patients with systemic lupus erythematosus (SLE), 28 with Wegener's granulomatosis (WG), 22 with microscopic polyangiitis (MPA), 5 with idiopathic rapidly progressive glomerulonephritis (RPGN), and 5 healthy controls were examined by both the indirect immunofluorescence technique (IFT) and the ELISA kits. Sera from healthy controls and patients with SLE or cANCA-negative WG were shown to be PR3-ANCA negative by all 7 PR3-ANCA kits. In 25 cANCA-positive sera from WG patients, PR3-ANCA positivity ranged from 44% to 84%. An absolute concordance among the 7 kits was noted in 56% of the cANCA-positive samples. The PR3-ANCA levels in 5 of the 7 kits correlated with the cANCA titers in IFT. Sera from the healthy controls and 4 out of the 5 SLE and pANCA-negative patients were found to be MPO-ANCA negative in all 8 MPO-ANCA kits. In 20 pANCA-positive sera, MPO-ANCA positivity ranged from 25% to 75%. Thirty-five percent of MPO-ANCA-positive sera were confirmed by capture ELISA, immunoblot and inhibition assay. The concordance rate was only 30% among pANCA-positive sera in the 8 MPO-ANCA kits. No significant correlation was observed between pANCA titers and MPO-ANCA levels. The HM showed that 65% of cANCA-positive sera were PR3-ANCA positive, and 45% of pANCA-positive sera were MPO-ANCA positive. Our results indicate that the sensitivities and specificities for ANCA detection differ significantly among the commercial kits tested and underline the necessity of establishing uniform international standards for ANCA ELISA procedures in order to permit more reliable interpretation and comparison of data.