We agree with Dr Caironi that hypoalbuminemia, epiphenomenon or not, is associated with increased morbidity and mortality in critically ill patients. However, no study has shown that correcting hypoalbuminemia or using albumin in resuscitation alters these outcomes.
In the spirit of accepting our opponent’s challenge to debate the theory rather than the details of trial results, we keep our statistical concerns brief, as they have been outlined elsewhere. We agree that shock defines a well-characterized patient subset; however, subgroup analysis, even in a well-defined subpopulation, has the potential for selection bias and sampling error. It is plausible that data exploration in ALBIOS to identify a subgroup with a positive treatment effect resulted in an overestimation of albumin’s potential effect. Moreover, subgroup analysis (prespecified or post hoc) can lead to conclusions of false benefit arising from multiple testing. It would certainly not be the first time that biologically plausible treatments fail to show benefit when tested in critically ill patients. A study on albumin replacement in patients with septic shock found no significant mortality difference.