O007 - MODEL-BASED PERSONALISED VS ESPEN-GUIDED NUTRITION IN THE ICU: AN EMULATED TRIAL ANALYSIS

O007

MODEL-BASED PERSONALISED VS ESPEN-GUIDED NUTRITION IN THE ICU: AN EMULATED TRIAL ANALYSIS

F. Bustin¹, M. Seret^1,*, V. Uyttendaele¹, J. G. Chase², T. Desaive¹

¹Model-based therapeutics, GIGA Institute, University of Liège, Liège, Belgium, ²Department of Mechanical Engineering, Centre for Bio-Engineering, University of Canterbury, Christchurch, New Zealand

Rationale: Interaction between nutrition, glycaemia, and insulin is complex in ICU patients. STAR is a model-based glycaemic control protocol personalising both insulin and nutrition delivery via insulin sensitivity identification. While its benefits on glycaemic control are well established, the effect of its personalised nutrition strategy on patient outcomes remains unexamined.

Methods: This study uses emulated trials, a statistical framework reproducing RCTs from observational data collected for other purpose. It uses the g-formula, a causal inference method suited for time-varying exposure and confounders, to compare the causal effect on ICU mortality of two nutrition strategies: ESPEN-guided versus STAR personalised nutrition delivery, both expressed as %goal feed and observed in the cohort. The retrospective cohort consisted of ICU patients treated with STAR in New-Zealand (2019-2024). Primary outcome was ICU mortality over ICU stay duration.

Results: Of the initial 1972 patients, 1227 ICU patients met inclusion criteria for episode close to ICU admission and complete demographic and nutrition data. ESPEN-guided nutrition was associated with a 0.3% absolute reduction in 10-day ICU mortality compared to STAR. Observed 10-day mortality was low in both arms: 14.8% [13.6–16.5] under ESPEN and 15.1% [13.9–17.1] under STAR. Mortality curves showed minimal difference in the early days of ICU stay, showing limited impact of nutrition strategy in short-stay patients.

Image:

Conclusion: In this trial emulation, ESPEN-guided nutrition was associated with a not clinically significant reduction in 10-day ICU mortality compared to STAR. Low, comparable mortality suggests STAR model-based, patient-specific nutrition delivery is safe regarding mortality, supporting its clinical use, while its impact on other outcomes should be further studied. These findings support the growing evidence for personalised nutrition approaches in the ICU.

Disclosure of Interest: None declared