No method currently exists to predict which patients with acute AIDS-associated cryptococcal meningitis can be effectively treated with fluconazole. The objective of this study was to determine the relationship of cryptococcal susceptibility to fluconazole, along with clinical variables, to the risk of treatment failure for patients with acute AIDS-associated cryptococcal meningitis. Results of in vitro fluconazole susceptibility testing of cryptococcal isolates and data from two clinical trials were analyzed. Susceptibility to fluconazole was determined by means of both microtiter and macrobroth (M27-P) dilution methods. Treatment was defined as successful if the patient was alive at 10 weeks and if a cerebrospinal fluid culture was sterile at that time. Seventy-six patients receiving fluconazole +/- flucytosine were included; therapy failed for 19. Patients whose therapy failed were more likely to have a positive blood and urine culture and a higher titer in serum and cerebrospinal fluid of cryptococcal antigen, and the MIC of fluconazole against their isolates (as determined by the microtiter method) was more likely to be higher; they were less likely to have received flucytosine. Logistic regression modeling revealed that a negative blood culture, a low MIC of fluconazole (per the microtiter method), and treatment with flucytosine were factors independently associated with successful treatment.