Untargeted Antifungal Treatment in Nonneutropenic Critically Ill Patients: Should Further Studies Be Performed Based on Trial Sequential Analysis Results?

LETTER

We read with great interest the recently published systematic review by Wang et al. (1) about untargeted antifungal treatment in nonneutropenic critically ill patients. This review nicely updated previous meta-analyses reporting similar results in terms of all-cause mortality and incidence of proven invasive fungal infections (IFI) (2, 3). Moreover, the authors performed further important analyses, such as trial sequential analysis (TSA), network meta-analysis, and a pharmacoeconomic analysis. In particular, the results of TSA caught our attention, and we believe that, due to the potential importance, these data should be further commented upon.

TSA is a methodology that combines information size calculation for a meta-analysis with the threshold of statistical significance (4). This analysis is able to determine if an estimate effect is large or small enough that the conclusions are unlikely to change with more evidence (5). In brief, it might theoretically answer the question “Should we go on with trials on this topic or not?” Due to the high mortality rates of Candida spp. infections in critically ill patients and the difficult interpretation of data from recent trials and meta-analyses, this aspect seems to be crucial for further research.

Wang et al. reported that since Z-curves crossed the futility boundary (for all-cause mortality) and the trial sequential monitoring boundary for benefit (for incidence of IFI), evidence should be considered sufficient and conclusive and hence indicated that further trials would not be required. However, we were not able to assess, from authors' data, which was the minimum selected desired effect size for their analysis for both outcomes. This is a very important decision when performing TSA and has a substantial impact on results (5). It should be considered crucial as the selection of the minimum detectable difference in a primary outcome when planning a randomized controlled trial (5). This aspect may merit particular attention because the main outcome of the analysis is all-cause mortality for which little change may justify a therapeutic intervention.

Wang et al. nicely reported that one of the main limitations of trials and meta-analyses about untargeted antifungal treatment is the heterogeneity of included populations and the relatively low incidence of IFI. The evidence now seems to indicate that we should try to find and test new ways for better patient selection and reconsider the impact of IFI in nonneutropenic critically ill patients (6). A complete and reliable TSA on available data may be a strong indication to abandon old research questions on this topic.