Baseline characteristics of the study population separated by critical illness and appropriateness for gestational age are presented in . Critically ill AGA infants (MV days 1–7) weighed less at birth, had a lower gestational age, were more likely to be male and non-white, and were less likely to have received antenatal steroids than less critically ill AGA infants (MV < 7 d). There were no differences between the groups in the proportion of intrapartum antibiotics given. Results were similar for SGA infants, except that there were no differences in gender, or in the proportion who received antenatal steroids.
Outcome data by severity of critical illness (unadjusted analyses for Step 1 in the mediation framework) are presented in . More critically ill AGA infants started enteral feedings later, and reached full feeds later, than less critically ill infants. A higher proportion of more critically ill infants had feeding interruptions for at least 24 hours, but there was no difference in the incidence of NEC. More critically ill infants also experienced a higher incidence of moderate and severe BPD, intraventricular hemorrhage, and late onset sepsis, were more likely to be treated with postnatal steroids for pulmonary disease, and were hospitalized longer than less critically ill infants. Furthermore, they grew more slowly, and were smaller at 36 weeks PMA when compared to less critically ill infants. Among survivors to follow-up, more critically ill AGA infants were more likely to have Bayley MDI and PDI scores less than 70, moderate to severe CP, and to be classified as NDI. Results were similar for SGA infants, although follow-up outcomes were not significantly different.
Outcome Variables by Degree of Critical Illness (Step 1: Unadjusted Analyses)
Results of adjusted analyses for Step 1 in the mediation framework from the GLM and logistic regression model series for various outcomes are summarized in . Among AGA infants critical illness was independently and significantly associated with slower growth velocity and with significantly increased odds ratios (ORs) for all the tested adverse outcomes, except for NEC; late-onset sepsis was of borderline significance (p=0.075). Thus, growth velocity in the more critically ill cohort was, on average, about 2 g/kg/day slower compared to those in the less critically ill cohort. More critically ill infants also had almost 2.5 times higher odds for NDI or death at follow up. Results were similar for SGA infants, except for growth velocity for survivors at 36 weeks’ PMA, late-onset sepsis, late-onset sepsis or death, and NDI.
Results of adjusted analyses in AGA and SGA infants from general linear model (GLM) and logistic regression models (Step 1: Adjusted Analyses)
Step 2 in the mediation framework is established in , in which weekly energy and total fluid intakes for the first 3 weeks of life are presented by severity of critical illness. Medians are included for enteral energy measures due to skewness of these data. For AGA infants, all energy measures, with the exception of parenteral protein energy over days 1–7, were significantly different by critical illness status. Specifically, compared to more critically ill AGA infants, less critically ill AGA infants received significantly more nutritional support during the first 3 weeks of life. Furthermore, the transition from parenteral to enteral nutritional support was clearly evident among the less critically ill infants. Results were similar for SGA infants. Results from covariate-adjusted analyses using GLMs indicate that critical illness status remained an independent and statistically significant predictor of total energy intake even after accounting for birth weight, gender, race, antenatal steroids, intrapartum antibiotics, and center. Measures for total fluid intake variables were significantly different by severity of critical illness for DOL 1–7 for both AGA and SGA infants and for DOL 15–21 for SGA infants. Adjusted results for total fluid intake were statistically significant only for SGA infants for DOL 15–21.
Energy and Fluid Intake by Degree of Critical Illness (Step 2: Unadjusted and Adjusted Analyses)
Results from testing Step 3 in the mediation framework, for models containing both the critical illness and energy intake variables, are summarized for AGA infants in . The effect of critical illness on the magnitude of the ORs for the tested adverse outcomes was significantly decreased once the total daily energy intake (kcal/kg/d) variable was included in the model (). These analyses indicate that critical illness during the first weeks of life and early nutritional practices are both independently associated with most of the outcomes examined, even after adjustment for several relevant covariates. For example, while the odds in favor of BPD or death increase more than threefold for more critically ill babies even after adjusting for total energy intake, an increase in total daily energy intake of only 1 kcal/kg/d was associated with a 2% decrease in the odds for BPD or death upon adjustment for critical illness and other baseline factors. Early nutrition practice as characterized by total daily energy intake was thus found to be a significant mediator of the association between critical illness during the first weeks of life and later outcomes.
Results of adjusted analyses in AGA infants from general linear model (GLM) and logistic regression models (Step 3: Adjusted Analyses)
In addition, although the interaction term for critical illness and total daily energy during days 1–7 of life was not significant for growth velocity, there was significant interaction between critical illness and total daily energy intake during days 1–7 of life for NDI and NDI or death. Thus, the effect of severity of critical illness on neurodevelopmental outcomes depends on the level of total daily energy during days 1–7 of life, with the effect of total daily energy on these outcomes significant for more critically ill infants only (). In order to illustrate the significant interaction between critical illness and total energy intake, ORs for critical illness (more vs less) are shown for varying levels of total daily energy intake, and ORs for total energy are shown separately for more and less critically ill infants.
Results from testing Step 3 in the mediation framework, for models containing both the critically ill and energy intake variables, are shown for SGA infants in . Unfortunately, the sample size was not large enough to include center in the models and no interaction terms were significant.
Results of adjusted analyses in SGA infants from general linear model (GLM) and logistic regression models (Step 3: Adjusted Analyses)