Early Enteral Feeding for Preterm or Low Birth Weight Infants: a Systematic Review and Meta-analysis

We registered the protocol with the international prospective register of systematic reviews PROSPERO CRD42021241600. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses-Protocol (PRISMA-P) guidance. The search strategy was developed in consultation with librarians, and references were managed using Covidence software (Appendix 1).

Randomized controlled trials (RCTs) were the only study design included in this review. We excluded nonhuman studies. There were no language restrictions. Participants were all LBW (<2500g) or preterm (<37 weeks gestational age) infants. The intervention was early initiation of enteral feeding within the first 72 hours of life, and the comparator was delayed initiation of enteral feeding. Enteral feeding was defined as milk feeding with any type of milk (mother’s own milk, donor human milk, or preterm or term formula). Infants could be fed via any method including naso or oro gastric tube, bottle, cup, paladai, or direct breastfeeding. It also included types of feeding such as ‘gastrointestinal priming,’ ‘minimal enteral feeding,’ or ‘trophic feeding.’

Electronic databases were searched to April 2021. Databases included Medline (Ovid), Embase (Ovid), Web of Science, and CINAHL. We also conducted manual checks of existing systematic reviews.^6,7  Selection of studies and data extraction was done by 2 authors, and conflicts were resolved in consultation with a third person. We extracted data using a standardized form and included: setting, design, participants, sample size, intervention, comparison, subgroups, and outcomes of interest.

We collected data for all outcomes at 28 days chronological age and at latest follow-up. Primary outcomes were all-cause and cause-specific mortality, morbidity (NEC, duration of hospitalization, sepsis, feeding intolerance, intraventricular hemorrhage, chronic lung disease, diarrhea), growth (weight, length, head circumference), and adverse events (hyperglycemia).

Two review authors judged the risk of bias using the Cochrane risk of bias tool (ROB-2).⁹  Funnel plots and Egger tests were done for outcomes with >10 studies to assess publication bias and small sample effects.

We used intention-to-treat estimates where possible. For binary outcomes, we summarized data using risk ratios with 95% confidence intervals (95% CI). For continuous outcomes, we summarized data using the pooled mean difference with 95% CIs.

For studies that reported median and interquartile ranges only, we used formulas derived Lou (2018) and Wan (2014) to impute the missing mean and variance information.^10–12  For studies with no events in 1 arm, we added 0.5 to each cell for continuity correction. Studies with no events in either arm were not pooled, because they would not contribute information to the pooled estimate.

We used Dersimonian and Laird random-effects models to calculate pooled estimates for all outcomes.¹³  We assessed forest plots visually for heterogeneity and considered I² values >50% to represent substantial heterogeneity. All analyses were conducted by using Stata 16C and RevMan.

Our a priori subgroup analyses were: (i) gestational age and weight at birth (studies enrolling only infants ≤ 32 weeks’ gestation or ≤ 1.5 kg at birth compared to studies that did not restrict enrollment based on gestational age or birth weight), (ii) fortified and nonfortified breast milk, (iii) parenteral nutrition, (iv) site of care, (v) health care provider, (vi) geographic regions, and (vii) income status of country. Additionally, we added subgroups for (i) timing of feed initiation (day 1, 2, 3); (ii) milk volume (higher [≥15 ml/kg/day] versus lower volume [<15 ml/kg/day]); and (iii) milk type (human milk, formula, and mixed human milk with formula).

We prepared a summary of findings table for each outcome using GRADE (Grading of Recommendations Assessment, Development, and Evaluation) and GRADEPro GDT software to assess risk of bias, consistency of effect, imprecision, and indirectness for each outcome.^14,15 

We made several changes from our initial protocol based on the data we identified and to provide policy-relevant data. These changes included: excluding observational studies because there was sufficient RCT data available; including abstract-only data from previous Cochrane systematic reviews (Becerra et al 1996)¹⁶ ; refining the primary analysis to compare timing of feeding initiation <72 hours vs ≥72 hours; adding subgroups including by time of feeding initiation, type of milk, and by volume of feeding.

We identified 5712 studies (Appendix 2). Of these studies, 22 were RCTs that were eligible for inclusion in the meta-analysis. We conducted the primary analysis using data from 14 studies which included 1505 preterm or LBW infants that compared the effect of enteral feeding initiation <72 hours of age to enteral feeding initiation ≥72 hours.^6,7,16–18  This analysis allowed us to include the maximum number of studies where the time of initiation was distinct in the early and delayed initiation groups.

For the primary analysis, studies took place in North America (n = 6), South America (n = 2), Asia (n = 2), and Europe (n = 4). None of the studies were from Africa. Eight trials restricted enrollment to infants ≤32 weeks or ≤1.5 kg.^16,18–24  The remaining 6 studies did not restrict to ≤32 weeks or ≤1.5 kg.^17,25–29  Three studies enrolled preterm or LBW infants were also born small-for-gestational age.^24,25,27 

Early feeding initiation time ranged from day 1 to 3 after birth. Delayed initiation time ranged from 4 to 15 days after birth. Infants in 1 studies initiated feeding by day 1 (<24 hours).^17,20  Infants in 8 studies initiated feeding by day 2 (<48 hours).^{16,18,22–24,26,27,29}  Infants in 4 studies initiated feeding by day 3(<72 hours).^19,21,25,28  12 studies provided ‘low’ volume feeds (<15 ml/kg/day),^{18–21,23–29}  and 2 provided feeds with volumes ≥15 mL/kg/day (volumes were 15–25 mL/kg/day).^16,22  Three studies gave the infants formula milk,^19,20,22  two gave mother’s own milk,^24,25  and the remaining nine gave a mixture of milks (ie, mother’s own, donor, or formula).^{16–18,21,23,26–29}  All studies provided supplemental parenteral nutrition to the delayed initiation group, except for Pérez et al 2011, which did not specify.¹⁸ 

Risk of bias was assessed for all outcomes for the 14 trials included in the primary analysis and we present the overall judgement by outcome (Appendix 4). No studies had high risk of bias. 75% of the assessments had a low risk of bias for the randomization process, and 25% had some concerns. Fifty-six percent showed some concerns due to deviations from intended interventions; the remaining 43.9% had low risk of bias for that domain. We ranked 95% of studies as low risk of bias because of missing data. For measurement of the outcome, 70% of assessments had a low risk of bias, and 30% had some concerns. Finally, for the selection of reported results 77% of assessments had some concerns of risk of bias.

Outcomes that were reported on by 10 trials or more included all-cause mortality, necrotizing enterocolitis, and duration of hospitalization. There was no obvious publication bias or small study effects in any analyses (Funnel plots and Egger’s tests, Appendix 5).

Meta-analysis results are summarized in Table 1, and forest plots are shown in Appendix 6. At latest follow-up (discharge or 28 days), the relative risk of mortality comparing the intervention (early initiation of enteral feeding <72 hours) to the comparison group (delayed feeding initiation ≥72 hours) was 0.69, (95% CI 0.48 to 0.99, I²= 0%, moderate certainty evidence, 12 trials, 1292 participants).

The relative risk of NEC comparing early (<72 hours) to delayed (≥72 hours) feeding was 1.05 (95% CI 0.75–1.46, I²= 0%, low certainty evidence, 13 trials, 1484 participants). The relative risk for sepsis was 0.90 (95% CI (0.54–1.52), I²= 51%, low certainty evidence, 5 trials, 626 participants), and for intraventricular hemorrhage the relative risk was 0.48 (95% CI 0.18–1.25, I²= 0; very low certainty evidence; 1 trial; 84 participants).

At discharge, the mean difference in weight was −49.02 g (95% CI −149.65 to 51.61, I²= 0%, low certainty evidence, 3 trials, 142 participants). The mean difference, measured at discharge or 32 weeks chronological age, in head circumference was −0.56 (95% CI −1.18 to 0.06, I² = 0%, low certainty evidence, 2 trials, 82 participants). The mean difference in length measured at discharge was -0.62cm (95% CI −1.51 to 0.27), I²= 0%, very low certainty evidence, 2 trials, 82 participants. The mean difference in time days to regain birth weight was 0.26 days (95% CI -0.63 to 1.15, I² = 18%, low certainty evidence, 7 trials, 569 participants).

There were no data on other primary outcomes. No trials assessed neurodevelopment or serious adverse events such as hypoglycemia.

At discharge there was little or no effect of the intervention on feed intolerance (risk ratio 1.03, 95% CI 0.66–1.60; I² = 0%, low certainty evidence, 2 trials, 187 participants). The intervention (early initiation of enteral feeding) decreased duration of hospitalization days (mean difference −3.20 days, 95% CI −5.74 to −0.66, I²= 19%, moderate certainty evidence, 10 trials, 1100 participants, Appendix 6).

There were no apparent differences in the effect of the intervention on primary outcomes in studies that enrolled infants ≤32 weeks’ gestation or ≤1.5 kg at birth compared to studies which did not restrict enrollment based on gestational age or birth weight (Appendix 6). There were also no obvious differences in the effect of the intervention on critical outcomes by timing of feed initiation (day 1, 2, 3); milk volume (volume (higher [≥15 ml/kg/day] versus lower volume [< 15 ml/kg/day]) or milk type (human, formula or mixed) (Appendix 6).

Results were generally similar for the outcomes of mortality, necrotizing enterocolitis, sepsis, intraventricular hemorrhage, head circumference gain, length gain, and weight gain, when we included all studies comparing early versus delayed feeding (Appendix 7).

Our systematic review of 14 trials of 1505 preterm or LBW infants found that early initiation of enteral feeding within the first 72 hours of life likely decreased the risk of neonatal mortality and may reduce the risk of sepsis when compared with delayed feeding initiation after 72 hours for hospitalized preterm and low birth weight infants. We also found that early intervention had little impact on necrotizing enterocolitis, intraventricular hemorrhage, feeding intolerance, time to regain birth weight, head circumference, and length at discharge, although the evidence was generally very uncertain for these outcomes. Early initiation of enteral feeding also decreased the number of days an infant stayed in the hospital by just over 3 days. Our results are directionally consistent for all primary outcomes with the most recent (2014) Cochrane review which examined the effect of delayed (>4 days) introduction of progressive enteral feeds compared to early introduction in very low birth weight infants (<1.5 kg).⁷  We included 9 additional studies, and the certainty of evidence on all primary outcomes (including duration of hospitalization) was higher than the Cochrane review.

We found no apparent sources of heterogeneity in our subgroup analyses. There was no difference in effect by volume of initial enteral feeding (<15 ml/kg/day compared to ≥15 mL/kg/day), although there were only 2 studies of 229 infants which used volumes above 15 mL/kg/day, and our results are inconclusive. However, a 2015 Cochrane review by Walsh and colleagues compared the effect of early full enteral feeding (60–80 ml/kg/day) to lower volume feeds (20–30 mL/kg/day) and found that use of higher volumes was associated with 3 days shorter hospital duration than lower volumes (mean difference −3.09, 95% CI −4.13, −2.02; 4 trials; 290 infants), and there was little to no difference in mortality, morbidities such as necrotizing enterocolitis, and growth.⁸  We also found no difference in effect between trials that enrolled infants who were ≤ 32 weeks or ≤ 1.5 kg and trials that did not, but additional data are needed. We also found no difference in the effect of early initiation between the trials that used human or formula milk in our review. This is in contrast to the evidence that infant formula likely increases the risk of necrotizing enterocolitis in preterm infants.³⁰  However, only 2 of our included studies used human milk, and only 3 used infant formula as the sole diet; sample sizes were limited, and results are inconclusive.

We also found no difference by day of initiation and no evidence of a dose response by day of initiation of feeding, although there were only 2 studies including 74 infants that initiated feeding with the first 24 hours of life. This is in contrast to community-based studies including both preterm and term infants which found a dose-response relationship between increasing delay in initiation (<1 hour, 2–23 hours, ≥24 hours after birth) and increasing risk of neonatal mortality.^3,5  These observational studies are subject to selection bias and confounding, although the quality of evidence was graded at high certainty because of the consistent, large effect size, and evidence of a dose-response relationship.^3,5 

Our findings that early feeding initiation is linked to reduced risk of mortality and sepsis are clinically plausible. When infants are not fed enterally, they must receive parenteral nutrition or water and dextrose through intravenous or enteral routes. Use of intravenous fluids and parenteral nutrition increases risks of blood stream infection and metabolic derangements. This may also explain why the trials showed a reduced length of hospital stay.

Health and survival benefits, as well as reduced time in the hospital, have important economic benefits for the health system and caregivers. A 2021 Canadian study including nearly 30 000 preterm infants in the NICU estimated the costs related to longer length of hospital stays; an infant who has a gestational age of 27 weeks had a mean cost per day of CAD $1759.³¹  Costs associated with infant health complications can be astronomical. In the United States, preterm and low birth weight infants with NEC could cost the hospital USD $100 000 compared to USD $20 300 for those without NEC.³²  Additionally, out-of-pocket payments in low- and middle-income countries can be very high. In Ghana, Enweronu-Laryea et al found that costs associated with preterm and low birth weight hospital stays accounted for 8.1% of a family’s average annual income.³³  Further research is needed to estimate the potential impact, including reduced economic burden (direct and indirect costs), of implementing improved infant feeding protocols for infants in low-and middle-income countries.

There are several important limitations to our study. In our review, all included studies took place in hospital settings and the average length of hospital stay for study participants was >50 days. Most study participants were very preterm or very low birth weight infants. The majority of studies were conducted in middle- or high-income countries, except the study conducted by Tewari and colleagues in India.²⁴  Thus, this evidence should be applied to other contexts with caution. However, a systematic review of observational, community-based studies from Tanzania, Ghana, India, and Nepal found that early breastfeeding initiation within the first 24 hours of life was associated with reduced neonatal mortality.⁵  Our results were heterogeneous, and evidence was of low or very low certainty for all outcomes except mortality and duration of hospitalization. There were no studies on neurodevelopment, and evidence was insufficient to understand effects on short- and long-term growth outcomes.

In conclusion, we found that early enteral feeding initiation likely decreases the risk of mortality and the duration of hospitalization. This evidence comes from studies of hospitalized newborns and adds to evidence supporting early breastfeeding initiation for preterm and LBW infants in all settings, including babies born and managed at home.

Thank you to the George Washington University librarians Stacy Brody, MI, and Susan Koening, MA, for their contribution to the design and execution of search for this meta-analysis.

GRADE

Grading of Recommendations Assessment, Development, and Evaluation

LBW

low birth weight

NEC

necrotizing enterocolitis

RCT

randomized controlled trial