Commonly used "nutrition" indicators do not predict outcome in the critically ill: a systematic review

In everyday practice, clinicians use a variety of anthropometric, biochemical, and clinical indicators to monitor nutrition therapy, but these have limitations in the critically ill. This systematic review of randomized controlled trials aimed to assess whether commonly used anthropometric, biochemical, and clinical nutrition indicators are predictive of patient outcomes in the critically ill.

A computerized bibliographic search was performed using MEDLINE, EMBASE, and CINAHL from 1950 to December 2012, as well as a citation review of relevant articles. Randomized clinical trials of any nutrition interventions in critically ill patients were included if they reported any nutrition indicator after baseline and any clinically meaningful outcome variables. Information about study quality, setting, and findings was extracted using standardized protocols. Because of the heterogeneity of study characteristics, only a narrative synthesis was undertaken.

Of 223 studies obtained with the search strategy, 2 independent reviewers identified selected 51 studies meeting the eligibility criteria. These reported indicators such as serum albumin, serum prealbumin (transthyretin), retinol-binding protein, transferrin, and lymphocytes. Thirty studies did not report a significant difference in clinical outcomes. Of the remainder, the number of studies supporting a statistical relationship between outcome and particular nutrition indicators was equal to, or outnumbered by, the studies not supporting such a relationship.

None of the commonly used nutrition monitoring parameters demonstrated consistent associations with outcome in randomized controlled trials. Development of nutrition indicators other than laboratory tests that are more closely linked to the patient's clinical progress should be a priority.