Permissive underfeeding, cytokine profiles and outcomes in critically ill patients

Abstract

Background: During critical illness in humans, the effects of caloric restriction on the inflammatory response are not well understood. The aim of this study is to examine the associations of caloric restriction, inflammatory response profiles and outcomes in critically ill patients.

Methods: This is a sub-study of the PermiT trial (Permissive Underfeeding or Standard Enteral Feeding in Critically Ill Adults Trial- ISRCTN68144998). Serum samples were collected on study days 1, 3, 5, 7 and 14 and analyzed for a panel of 29 cytokines. We used principal component analysis to convert possibly correlated variables (cytokine levels) into a limited number of linearly uncorrelated variables (principal components). We constructed repeated measures mixed linear models to assess whether permissive underfeeding compared to standard feeding was associated with difference cytokine levels over time.

Results: A total of 72 critically ill patients were enrolled in this study (permissive underfeeding n = 36 and standard feeding n = 36). Principal component analysis identified 6 components that were responsible for 78% of the total variance. When adjusted to principal components, permissive underfeeding was not associated with 90-day mortality (adjusted odds ratio 1.75, 95% confidence interval 0.44, 6.95, p = 0.43) or with incident renal replacement therapy. The cytokines did not differ with time between permissive underfeeding and standard feeding groups.

Conclusions: The association of permissive underfeeding compared to standard feeding with mortality was not influenced by the inflammatory profile. Permissive underfeeding compared to standard feeding was not associated with differences in the serum levels of cytokines in critically ill patients.

Fig 1. Serial measurements for selected acute pro-inflammatory cytokines in patients randomized to permissive underfeeding and standard feeding groups.

The differences between groups, with time and between groups with time (group*time) were tested by repeated measures mixed linear models. Box plots are displayed with medians and quartiles 1 and 3. The error bars refer to 10^th and 90^th percentiles.