Background: The American College of Rheumatology (ACR) established criteria to discriminate among patients with seven types of vasculitis. Although designated as "classification criteria" for research, these criteria are often used for diagnosis.

Objective: To examine the operating characteristics of the 1990 ACR classification criteria in the diagnosis of Wegener granulomatosis, giant-cell arteritis, polyarteritis nodosa, and hypersensitivity vasculitis.

Design: Prospective cohort study.

Setting: University medical center and Veterans Affairs medical center.

Patients: 198 consecutive patients referred to rheumatologists for evaluation of possible vasculitis.

Measurements: Blinded chart audits were done to classify patients according to the 1990 ACR classification criteria for Wegener granulomatosis, polyarteritis nodosa, giant-cell arteritis, and hypersensitivity vasculitis on the basis of the patients' initial presentation. Chart audits done 2 to 8 months after baseline provided the patients' final diagnoses, which were considered the gold standard, as in the development of the ACR criteria. Test operating characteristics of the ACR classification criteria were calculated according to 2 x 2 tables for the entire cohort and for only the patients with a final diagnosis of vasculitis.

Results: Vasculitis was diagnosed in 51 (26%) patients. Thirty-eight (75%) of 51 patients with vasculitis and 31 (21%) of 147 patients without vasculitis met ACR criteria for one or more types of vasculitis. The positive predictive values for the four vasculitides according to ACR criteria were 17% to 29% for the entire cohort and 29% to 75% for only the patients with a final diagnosis of vasculitis.

Conclusion: The 1990 ACR classification criteria function poorly in the diagnosis of specific vasculitides.