To investigate the effects of 3-week versus 4-week administration of benzathine penicillin G (BPG) on the incidence of Group A streptococcal infections and the recurrences of rheumatic fever (RF).We started, in 1979, randomly allocating all patients with RF to a 3-week or 4-week BPG prophylaxis program. They were examined at the RF clinic, every 3 to 6 months, and at any time they did not feel well. During 1979 to 1989, throat cultures and sera for antistreptolysin O and streptozyme titers were obtained at each clinic visit. Chest radiographs, electrocardiogram, color Doppler echocardiograms, and acute phase reactants were obtained.Two hundred forty-nine patients fulfilled the revised Jones criteria and were followed until December 1991: 124 in the 3-week and 125 in the 4-week program. Their age, sex, weight, percentage with history of RF, severity of cardiac involvement, follow-up duration, and compliance to program were comparable. Eight hundred eighty throat cultures were collected in the 3-week program and 770 were collected in the 4-week program. Six hundred sixteen and 627 sera were determined in each program for antistreptolysin O, and 582 and 592 sera for streptozyme titers.True streptococcal infections occurred in both programs: 39 infections in the 3-week program, and 59 infections in the 4-week program (7.5 vs 12.7 per 100 patient-years). Four infections with no antibody response occurred in the 3-week program, and three such infections in the 4-week program. Nine RF recurrences occurred in 8 patients in the 3-week program, and 16 recurrences in 16 patients in the 4-week program. Prophylaxis failure occurred in 2 of 124 patients in the 3-week program, and in 10 of 125 patients in the 4-week program (0.25 vs 1.29 per 100 patient-years). The overall recurrences/infections rate in each program was comparable, 13.6% vs 15.5%, but the recurrences/ infections rate due to prophylaxis failure was higher in the 4-week program than in the 3-week program, 3.0% versus 9.7%.This 12-year prospective and controlled study documented that streptococcal infections and RF recurrences occurred more often in the 4-week program than in the 3-week program. The risk of prophylaxis failure was fivefold greater in the 4-week program than in the 3-week program.