We performed a retrospective evaluation of whether c-ANCA titres (indirect immunofluorescence) and anti-proteinase 3 (PR3)-ANCA levels (ELISA) at diagnosis and following immunosuppressive treatment are predictive of relapse of ANCA-associated vasculitis. Patients diagnosed with PR3-ANCA-associated vasculitis between 1991 and 2002, with at least 2 yr of follow-up, and treated with cyclophosphamide and corticosteroids only (1991-1996) or switched to azathioprine after induction of remission with cyclophosphamide and corticosteroids (1997-2002) were included. ANCA were assessed by immunofluorescence and direct PR3-specific ELISA at diagnosis and 3, 6, 12, 18 and 24 months after diagnosis. Actuarial relapse-free survival was analysed with the log rank test. We studied 87 patients positive for PR3-ANCA: 46 were on cyclophosphamide maintenance therapy and 41 switched to azathioprine. Overall actuarial relapse-free survival was 72% at 2 yr and 34% at 5 yr. Relapse-free survival did not differ between patients on cyclophosphamide maintenance and patients switched to azathioprine maintenance (P = 0.34). Patients who became and stayed negative for c-ANCA (immunofluorescence) or PR3-ANCA (ELISA) until 24 months after diagnosis had a lower risk of relapse (P = 0.01 and P = 0.02, respectively). Positive c-ANCA (immunofluorescence) titres at 3 [relative risk (RR) 2.0; 95% confidence interval (CI) 1.2-3.8], 12 (RR 1.9; 95% CI 1.1-3.3), 18 (RR 2.9; 95% CI 1.3-4.6) and 24 months (RR 2.6; 95% CI 1.2-5.0) were significantly associated with relapse within 5 yr after diagnosis. PR3-ANCA levels >10 U/ml at 18 (RR 2.7, 95% CI 1.1-4.3) and 24 months (RR 4.6; 95% CI 1.2-6.3) were predictive of relapse within 5 yr. In the azathioprine group, a positive c-ANCA titre at the time of switching to azathioprine (RR 2.2; 95% CI 1.0-5.4) was associated with relapse. Positive c-ANCA (immunofluorescence) and PR3-ANCA (ELISA) titres during early follow-up identify patients at increased risk of relapse.