The authors reply
First of all, the inclusion of the most recent trial changes the evidence as it allows for the first time a meta-analysis to have enough power to address the issue of publication bias for mortality.
Second, at the opposite of the Cortegiani et al (2) meta-analysis, cancer patients could be included if they were in the ICU. Patients with major surgery not requiring ICU were not included. In the studies of Slotman et al (3) and Ostrosky-Zeichner et al (4), patients really immunosuppressed were not included (5).
Third, even if nonabsorbable antifungal treatment was analyzed separately in the secondary objectives in the meta-analysis of Cortegiani et al (2), it was also included in the primary analysis. We believe it could include a significant bias in the results.
Fourth, we believe that lung colonization is not a part of the debate for antifungal prophylaxis. These studies have nothing to do in such a meta-analysis.
Fifth, we agree that in the study of He et al (6), ICU was only a part of the inclusion criteria (5). Nevertheless, patients with severe pancreatitis are usually hospitalized in ICU. Concerning the study of Sandven et al (7), patients were admitted in the ICU after surgery (5).
Sixth, we disagree with the comment about classification and definition used. Individual risk factors of fungal infection are now very well-known (8). We believe that it is a key point to differentiate prophylaxis and preemptive treatments. Furthermore, prophylaxis was not defined in the ESCMID guidelines (9). The Italian consensus was cited for their classifications, not for the individual or population risk factors. We agree that Cortegiani et al (2) analyzed separately the groups according to their definition, but it was only on the secondary analysis. Furthermore, they did not separate antifungal use for prevention of fungal infections and that for anti-inflammatory effects of antifungal agents. We believe that it is another important difference in the presentation and in the results.