Abstract Forty-eight neutropenic patients with acute leukaemia were randomly allocated to receive, as antifungal prophylaxis, either ketoconazole, 400 mg once daily (K), or amphotericin B tablets and lozenges (A), or both ketoconazole and amphotericin B together (K + A). Antifungal prophylaxis was considered to have failed if (1) there was evidence of increasing colonization of the oropharynx or faeces with Candida spp. or other yeasts, or (2) if systemic antifungal therapy was begun empirically. Prophylaxis failed in nine of 17 patients given K, in four of 19 given A, and in four of 12 given K + A. The differences between the three regimens were not statistically significant, neither was there any significant difference in the mean duration of neutropenia before prophylaxis failed. The absorption of ketoconazole was impaired when patients were neutropenic. We conclude that ketoconazole was neither more nor less effective than amphotericin B in the prevention of yeast colonization in neutropenic patients.