In a multicenter trial of 2 regimens for treatment of pulmonary tuberculosis, all patients received 300 mg of isoniazid (INH) and 600 mg of rifampin (RIF) daily for 6 months (the Initial Phase). During the next 9 months (the Maintenance Phase) patients received either daily INH (300 mg) and ethambutol (EMB) (15 mg per kg body weight) or matching placebos. Of the 672 patients who met the admission criteria, only 309 (46%) completed the Initial and Maintenance Phases. Approximately 20% of the patients failed to keep their appointments. Adverse drug reaction, most commonly hepatotoxicity, accounted for the withdrawal of 37 patients (5.5%). No visual toxicity caused by EMB was observed. During the Maintenance Phase, 3 patients who were taking INH and EMB, and 16 who were taking placebos, developed relapses, i.e., 2 or more positive cultures. The significant difference in relapse rate between regimens (Fisher's exact test, p less than 0.001) demonstrates the inadequacy of INH-RIF given alone for only 6 months.