Achieving enteral nutrition during the acute phase in critically ill children: Associations with patient characteristics and clinical outcome.

BACKGROUND & AIMS In the absence of methodologically sound randomized controlled trials (RCTs), current recommendations for timing and amount of enteral nutrition (EN) in critically ill children are based on observational studies. These studies have associated achievement of a higher EN intake in critically ill children with improved outcome. Inherent to the observational design of these underlying studies, thorough insight in possible confounding factors to correct for is essential. We evaluated the associations between EN intake and 1) patient and daily clinical characteristics and 2) clinical outcomes adjusted for these patient and clinical characteristics during the first week of critical illness with a multivariable mixed model. METHODS This secondary analysis of the multicentre PEPaNIC RCT investigated a subgroup of critically ill children with daily prospectively recorded gastrointestinal symptoms and EN intake during the first week with multivariable analyses using two-part mixed effect models, including multiple testing corrections using Holm's method. These models combined a mixed-effects logistic regression for the dichotomous outcome EN versus no EN, and a linear mixed-effects model for the patients who received any EN intake. EN intake per patient was expressed as mean daily EN as % of predicted resting energy expenditure (% of EN/REE). Model 1 included 40 fixed effect baseline patient characteristics, and daily parameters of illness severity, feeding, medication and gastrointestinal symptoms. Model 2 included these patient and daily variables as well as clinical outcomes. RESULTS Complete data were available for 690 children. EN was provided in 503 (73%) patients with a start after a median of 2 (IQR 2-3) days and a median % of EN/REE of 38.8 (IQR 14.1-79.5) over the first week. Multivariable mixed model analyses including all patients showed that admission after gastrointestinal surgery (-49%EN/REE; p = 0.002), gastric feeding (-31% EN/REE; p < 0.001), treatment with inotropic agents (-22%EN/REE; p = 0.026) and large gastric residual volume (-64%EN/REE; p < 0.001) were independently associated with a low mean EN intake. In univariable analysis, low mean EN intake was associated with new acquired infections, hypoglycaemia, duration of PICU and hospital stay and duration of mechanical ventilation. However, after adjustment for confounders, these associations were no longer present, except for low EN and hypoglycaemia (-39%EN/REE; p = 0.018). CONCLUSIONS Several patient and clinical characteristics during the first week of critical illness were associated with EN intake. No independent associations were found between EN intake and clinical outcomes such as mortality, new acquired infection and duration of stay. These data emphasize the necessity of adequate multivariable adjustment in nutritional support research and the need for future RCTs investigating optimal EN intake.