The impact of high versus standard enteral protein provision on functional recovery following intensive care admission: A pre-planned Bayesian analysis of the PRECISe trial

Abstract

Background and aims: High protein nutrition may improve outcomes after critical illness. We recently published the primary frequentist analysis of the PRECISe trial, showing that high (2.0 g/kg/day) compared with standard (1.3 g/kg/day) protein provision led to statistically significant worse health-related quality of life. The study, however, was not powered to draw definitive conclusions about clinical and other functional outcomes under a frequentist framework. We present a pre-planned and pre-specified Bayesian analysis to facilitate the clinical interpretation of these paramount endpoints. Methods: The trial enrolled 935 patients and used the EQ-5D-5L health utility score as the primary endpoint. We performed Bayesian analyses of the primary and selected secondary endpoints, and relevant subgroups, under weakly informative priors. Sensitivity analyses were performed using skeptical and enthusiastic priors, and informed priors (when available) based on existing literature. Thresholds for clinically relevant differences were predefined. Results: The posterior probability of benefit from high (2.0 g/kg/day) protein targets with respect to the EQ-5D-5L health utility score was 0 %. Concerning 60-day mortality, the posterior probability of any benefit from high protein provision was 8 %, with a posterior probability of clinically important harm (>5 % absolute risk difference) of 47 %, which varied between 1 and 21 % across various sensitivity analyses under reference or literature-based priors. Conclusions: This pre-planned Bayesian re-analysis of the PRECISe trial shows that high (2.0 g/kg/day) compared to standard (1.3 g/kg/day) protein provision in critically ill patients has a low probability to yield any benefit and results in a high probability of an increase of 60-day mortality. Registration number of clinical trial: NCT04633421. (c) 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).