Critical care nutrition: a Bayesian re-analysis of trial data

Nutritional support and optimal glucose control are key components of care during critical illness, yet evidence surrounding their use remains conflicting, making translation into clinical practice challenging. This review explores Bayesian methods to enhance the interpretation of frequentist critical care trials, particularly for interventions with inconclusive outcomes.

Bayesian re-analysis frameworks may clarify conflicting evidence in critical care, thus enhancing interpretability and supporting clinical decision-making. This review focuses on the Bayesian re-analysis of three recent trials with indeterminate results for mortality - NUTRIREA-3, EFFORT Protein, and TGC-Fast - that examined the effects of enteral nutrition and glucose control strategies.

We re-analyzed the mortality outcomes of these trials within a Bayesian framework, contrasting our findings with the original trial results to illustrate how Bayesian methods can enhance the clinical applicability of trial outcomes. Although Bayesian and frequentist analyses generally agree on the direction and magnitude of effect, Bayesian methods offer the advantage of providing posterior probabilities of benefit and harm, thus identifying promising and potentially harmful interventions. This review underscores the value of Bayesian analysis in re-evaluating clinical trial data and guiding clinical practice.