BACKGROUND: HIV-infected patients with sustained immunologic improvement from antiretroviral therapy may be able to discontinue chemoprophylaxis against Pneumocystis carinii pneumonia (PCP). OBJECTIVE: To compare PCP incidence in HIV-infected patients who had sustained CD4+ lymphocyte counts greater than 200 cells/mm3 and who either discontinued or continued PCP prophylaxis. DESIGN: Nonrandomized prospective cohort study. SETTING: 10 HIV clinics in eight U.S. cities. PATIENTS: 146 patients had follow-up visits for a mean of 18.2 months after discontinuation of PCP prophylaxis, and 345 patients who continued PCP prophylaxis had follow-up visits for a mean of 14.0 months. MEASUREMENTS: Incidence of PCP. RESULTS: Patients who discontinued PCP prophylaxis had higher maximum and minimum CD4+ cell counts and lower vira loads than patients who continued PCP prophylaxis. Pneumocystis carinii pneumonia did not develop in either group (upper 95% exact binomial confidence limit of incidence for those who discontinued PCP prophylaxis, 2.3/100 person-years). CONCLUSIONS: Discontinuation of PCP chemoprophylaxis may be appropriate for some HIV-infected ambulatory patients.