An epidemic of isoniazid (INH)- and streptomycin (SM)-resistant tuberculosis began among Boston's homeless population in 1984. Individuals with skin test conversions who agreed to preventive therapy received either INH, rifampin, or a combination of INH and rifampin. A total of 204 individuals with documented tuberculin skin test conversions who did not have active tuberculosis at the time of the clinical evaluation for their positive skin test were eligible for preventive therapy. Data on type and length of preventive therapy were obtained from the Tuberculosis Clinic and the Boston Tuberculosis Registry records at Boston City Hospital. The individuals were followed for development of active tuberculosis. Six of 71 (8.6%) individuals who received no therapy, 3 of 38 (7.9%) in the INH group, and none in the rifampin or rifampin plus INH groups (49 and 37 persons, respectively) developed active tuberculosis. Patients in the rifampin group were significantly less likely to develop tuberculosis than patients in the no therapy group (p = 0.04; odds ratio [OR] = 0.00, 95% confidence interval [CI] = 0.00-0.91). Treatment with any rifampin-containing preventive therapy (rifampin or rifampin plus INH) was effective (p < 0.01 ) in preventing development of active disease. The three INH failures were with organisms that were resistant to INH.