Objective: To evaluate the effect of the type of Pneumocystis carinii pneumonia (PCP) prophylaxis on the development of community-acquired bacteremia. Design: Case-control study using all cases of community-acquired bacteremia identified prospectively during a longitudinal study of all infections in a cohort of HIV-infected persons. Setting: University-affiliated Department of Veterans Affairs Medical Center HIV program. Patients: All patients with community-acquired bacteremia seen at the facility between January 1990 and December 1995 were included. Controls, seen at the same facility and matched by date and CD4 count, were used to assess risk factors. A total of 57 cases and 114 controls were analysed. Main outcome measures: Risk of development of bacteremia, distribution of organisms, and effect of specific prophylactic regimens for PCP. Results: Bacteremia was caused by Staphylococcus aureus (23%), Pseudomonas aeruginosa (18%), Escherichia coli (16%), Streptococcus pneumoniae (14%) and others (31%). Groups were similar by age, race, HIV risk factors and CD4 count. The presence of an intravenous catheter was mildly predictive of the development of bacteremia [odds ratio (OR), 2.67; P= 0.024]. Type of PCP prophylaxis in cases and controls with CD4 < 200× 106/l included co-trimoxazole (trimethoprimsulfamethoxazole, TMP-SMX; 31 and 60%, respectively), dapsone (33 and 24%, respectively) and aerosolized pentamidine (27 and 13%, respectively). Use of TMP-SMX (but not dapsone or aerosolized pentamidine) was associated with the absence of bacteremia (OR, 0.28; P = 0.001). A similar protective effect was found when controlling for the presence of an intravenous catheter. Conclusion: PCP prophylaxis with TMP-SMX apparently protects against communityacquired bacteremia in HIV-infected persons.