We examined the effect of different fluconazole treatment regimens on the emergence of azole drug resistance among Candida species recovered from the mouths of 54 HIV-infected individuals. Patients were assigned to one of three treatment groups depending on their history of oral candidosis and fluconazole use. Mouthwashes obtained at regular intervals were cultured and isolates identified using standard methods. Antifungal broth micro-dilution tests were performed to determine IC30s of fluconazole and ketoconazole. Sixty-four Candida albicans isolates from 20 patients with no evidence of oral candidosis who had not received fluconazole all had IC30s of < or = 4 mg/L. Thirty-four (83%) of 41 C. albicans isolates from ten patients receiving intermittent, short-term fluconazole treatment for oral candidosis had IC30s of < or = 4 mg/L, but only two isolates (5%) had IC30s > or = 64 mg/L. In contrast, 26 (40%) of 65 C. albicans isolates from 15 patients given long-term fluconazole (50-200 mg/day or 150 mg/week) were classified as resistant having IC30s of fluconazole of > or = 64 mg/L. Ten of these 26 fluconazole-resistant isolates were susceptible to ketoconazole with IC30s of < or = 4 mg/L suggesting azole drug cross-resistance is not inevitable. Tests on multiple colonies from individual isolation plates showed that it was not unusual to obtain differing IC30 values, indicating that a sweep inoculum is essential if resistance is to be detected. Nine (60%) of the 15 patients given long-term fluconazole harboured isolates of C. albicans that were resistant to fluconazole at some time during the study period. All had low CD4 counts and were approaching the final stage of their illness. Three patients on long-term treatment had resistant organisms at the outset of the study; in the remainder, resistant strains emerged during the study period. In six of the nine cases, emergence of resistance in vitro correlated with persistent clinical signs of oral infection. Thirty-six isolates of Candida species other than C. albicans were also recovered from patients receiving long-term fluconazole and 29 (81%) of these had IC30s of > or = 64 mg/L. Our experience with C. albicans in patients with HIV infection, suggests that the long-term azole drug use may be an important factor in the development of fluconazole resistance as such resistance was rare and transient in patients on intermittent short-term treatment.