OBJECTIVE: To assess the safety of discontinuing or withholding primary prophylaxis against disseminated Mycobacterium avium infection (MAC) in HIV infected patients on successful antiretroviral combination therapy. SETTING: National prospective multicentre cohort study. DESIGN: HIV-infected patients were eligible for the analysis if: (i) they had a history of at least two CD4 cell counts < 50 x 10(6)/l; (ii) they had never had MAC; (iii) they had discontinued or never begun primary prophylaxis against MAC; (iv) they received antiretroviral therapy and demonstrated an increase in CD4 cell counts to > or = 100 x 10(6)/l that was sustained for at least 12 weeks. From this time point until last follow-up, incidence of disseminated MAC disease was measured, and 99% confidence intervals were calculated assuming a Poisson distribution of events. RESULTS: Two-hundred and fifty-three patients (22.5% female; median age, 37 years, 30% injecting drug users) were eligible for analysis. Sixty-six per cent were in Centers for Disease Control and Prevention (CDC) stage C, and 28% were in CDC stage B. Their median nadir CD4 cell count was 10 x 10(6)/l, the median duration of CD4 cell count < 50 x 10(6)/l was 12 months. During a total follow-up of 364.3 patient-years there was no case of disseminated MAC. The one-sided 99% confidence limit for incidence density of MAC was 1.3 per 100 person-years. CONCLUSION: Discontinuing or withholding primary prophylaxis against MAC is safe in patients who have a sustained increase in their CD4 cell count to > or = 100 x 10(6)/l.