The clinical and microbiologic efficacy of trimethoprim alone and trimethoprim/sulfamethoxazole for infection prevention was evaluated in 75 patients during 92 episodes of granulocytopenia. Ultimately, 60 patients were evaluable during 77 episodes of granulocytopenia, 36 episodes in the trimethoprim group and 41 episodes in the trimethoprim/sulfamethoxazole group. The incidence of infection was higher in the trimethoprim group (50 percent) than in the trimethoprim/sulfamethoxazole group (39 percent), but this did not reach statistical significance. Trimethoprim did not appear to be as protective as trimethoprim/sulfamethoxazole when the granulocyte count was less than 100/mm3. In patients receiving trimethoprim/sulfamethoxazole, aerobic gram-negative bacilli cleared from fecal surveillance cultures more often and new aerobic gram-negative bacilli were acquired less often than in those receiving trimethoprim alone (p less than 0.05). More myelosuppression was observed among patients receiving trimethoprim/sulfamethoxazole (p less than 0.001). These observations suggest that trimethoprim alone may not be optimal for preventing colonization and infection in granulocytopenic patients and that combination with other agents may be necessary to increase the spectrum of activity. Trimethoprim/sulfamethoxazole itself may predispose toward an increased risk of infection by prolonging myelosuppression.