Objective: To determine the rate of tuberculosis relapse among HIV-seropositive and -seronegative persons treated for active tuberculosis with short-course (6-month) therapy.

Design: Consecutive cohort study.

Setting: City of Baltimore tuberculosis clinic.

Patients: Tuberculosis patients treated between 1 January 1993 and 31 December 1996.

Intervention: Patients received 2 months of isoniazid, rifampin, pyrazinamide and ethambutol followed by 4 months of isoniazid and rifampin.

Main outcome measure: Passive follow-up for tuberculosis relapse was performed through September 30, 1998.

Results: There were 423 cases of tuberculosis during the study period; 280 patients completed a 6-month course of therapy. Therapy was directly-observed for 94% of patients. Of those who completed therapy, 47 (17%) were HIV-seropositive, 127 (45%) were HIV-seronegative, and 106 (38%) had unknown HIV status. HIV-infected patients required more time to complete therapy (median 225 versus 205 days; P = 0.04) but converted sputum culture to negative within the same time period (median 77 versus 72 days; P = 0.43) as HIV-seronegative or unknown patients. Relapse occurred in three out of 47 (6.4%) HIV-infected patients compared to seven out of 127 (5.5%) HIV-seronegative patients (P = 1.0). Relapse rates also did not differ when HIV-seropositive patients were compared with HIV-seronegative and patients with unknown HIV status (6.4% versus 3.0%; P = 0.38). Of the 10 patients with tuberculosis relapse, restriction fragment length polymorphism data were available for five; all five relapse isolates matched the initial isolate.

Conclusions: These results support current recommendations to treat tuberculosis in HIV-infected patients with short-course (6-month) therapy.