The paradox of the evidence about invasive fungal infection prevention.

Invasive fungal infections (IFIs) are characterized by high morbidity and mortality in non-neutropenic critically ill patients. Attributable mortality due to Candida spp. infections ranges from about 42 to 63 % [1, 2]. Data from large observational and retrospective studies show an association between early antifungal treatment and improved survival [3, 4]. Updated clinical practice guidelines for the management of candidiasis have been recently published [5]. In 2006, Playford et al. published a Cochrane systematic review investigating the use of antifungal agents for prevention of IFIs in non-neutropenic critically ill patients [6]. In that review, the outcome of proven IFI was defined as a clinical illness consistent with the diagnosis and either histopathological evidence of IFI or a positive fungal culture from one or more sterile site specimens (including blood). Notably, funguria (as indicated by a positive urine fungal culture), in the absence of a complicated urinary tract infection, and fungal esophagitis were classified not as IFIs but as superficial fungal infections. The review included 12 studies and 1606 patients, and the use of antifungal agents was associated with a mortality reduction of about 25 % and with an IFI reduction of about 50 %. Recently, we updated the original review by Playford et al., including 22 randomized controlled trials (RCTs) and 2761 patients [7]. We modified the definition of the outcome-proven IFI excluding positive culture of Candida spp. from the respiratory tract, even in the presence of systemic or respiratory signs of infection, and classifying it as colonization instead of IFI. Untargeted antifungal treatment, encompassing prophylactic, pre-emptive, and empiric regimens [8], was not associated with a significant mortality reduction (moderate-quality evidence). However, antifungal agents reduced IFIs by about 45 % with low-grade-quality evidence. From these data, three clinical questions may arise.