We have reviewed the medical records of 301/327 consecutive patients in whom anti-neutrophil cytoplasmic antibodies (ANCA) were detected by the Regional Immunology Laboratory in Northern Ireland between January 1988 and October 1991 (45 months). We have collected data for each patient regarding age, sex, smoking habit, area of residence, and details of any other autoantibody activity. Clinical diagnosis was established, with the number of organ systems involved and the evidence for that involvement (symptomatic, biochemical, radiological, and histological). Diagnoses were divided into four groups according to their recognised vasculitic features and these were related to the pattern of immunofluorescence and maximum ANCA titre detected. The most frequent diagnosis was rheumatoid arthritis (18.2% of patients) and the connective tissue disorders as a whole accounted for 27.9% of patients. ANCA were also detected in a wide range of clinical conditions which are not associated with vasculitis and these patients were an important source of 'false-positives'. The positive predictive value (PPV) of ANCA of all patterns and titres for vasculitic conditions was 27%, however, the detection of a classical ANCA pattern at high titre (> or = 1:640) was associated with an increased PPV of 75%. The coexistence of an antinuclear antibody (ANA) reduces the PPV of both classical and perinuclear ANCA, although perinuclear ANCA with antimyeloperoxidase specificity had an improved PPV. We conclude that ANCA testing should not be used as the only screening investigation for vasculitis but should be included in a rational investigative scheme. The interpretation of a positive ANCA result must take into account the presence of other autoantibodies and the full range of non-vasculitic conditions when the clinical situation is not typical of vasculitis.