Invasive candidiasis infections remain a major complication in I.C.U. patients. Numerous attempts have been made to evaluate potential prophylactic methods. Various agents have been tested. Gastrointestinal tract constitutes one of the major reservoirs for Candida species. One major problem is the difficulty in establishing an accurate, early, diagnosis of invasive fungal infection. A prospective randomized, controlled blind study was performed to assess the ability of oral Amphotericin B to prevent candidiasis in selected I.C.U. patients. Fifty one patients with serious infection and antibiotherapy were randomized to receive either oral Amphotericin B (2 g/day) or placebo, and observed until discharge. All patients were screened weekly for sites culture positive, sero-conversion and oesophagitis. Invasive candidiasis developed in 45% of patients receiving Amphotericin B compared with 41% receiving placebo. C. Albicans persists in the surveillance cultures. However a significant reduction of the colonization by the yeasts and a significant reduction of oesophagitis was demonstrated among the Amphotericin B group. No benefit was found in the total number of hospital days. Digestive decontamination can be successfully managed by Amphotericin B in I.C.U. patients but failed to prevent invasive candidiasis.