[Fall in ICU mortality due to selective decontamination not yet proven].

Selective decontamination reduces the mortality rate in ICU patients by about 3%. This was the conclusion of a recently published trial by De Smet et al. Because of possible ecological changes in the ICU by selective decontamination, a cluster-randomized design was used. As a consequence of this cluster randomization, the study lacked concealed allocation. Treatment allocation may therefore have been determined by prognostic factors. The discrepancies in baseline characteristics between the treatment groups are likely to be due to non-concealed allocation. For this reason the analysis was performed after making statistical adjustments for these discrepancies. The conclusion that selective bacterial decontamination reduces mortality is based on this adjusted analysis. This assumes that all differences between treatment groups were measured. This assumption cannot be proven and the possibility remains that the reduction in mortality found in this study is not a proper reflection of treatment effect.