Feeding in neonates with antenatal doppler abnormalities- a systematic review and meta-analysis

Preterm neonates with antenatal doppler abnormalities are at increased risk of necrotizing enterocolitis (NEC). In these neonates, we did a meta-analysis to compare the impact of early versus late initiation of feeding, and slow versus rapid feed advancement on the important neonatal outcomes. The databases of PubMed, Embase, Cochrane central, CINAHL and google scholar were searched on 6 th September 2020. We included all randomized controlled trials addressing the study objective(s). The risk of bias was assessed using the Risk of Bias tool, version 2. Certainty of evidence was assessed using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach. Early feeding did not increase the incidence of NEC stage 2 or more (odds ratio/OR 1.27, 95% condence interval/CI 0.83, 1.96; 6 studies, 772 participants) and mortality (OR 0.79, 95% CI 0.4, 1.57; 3 studies, 498 participants). A trend was noted towards an increased incidence of feeding intolerance (OR 1.37, 95% CI 0.98, 1.92). There was a signicant reduction in time to reach full feeds, duration of total parental nutrition, duration of hospital stay, and rates of hospital-acquired infections. The time to regain birth weight was not different. Rapid feed advancement decreased the time to reach full feeds, without affecting other outcomes. The overall certainty of the evidence was rated low. Heterogeneity was not signicant. Conclusion: There is low-certainty evidence to recommend early feed initiation in preterm neonates with antenatal doppler abnormalities. The data is insucient to make a recommendation on the rapidity of feed advancement. (probiotics, Future trials could be undertaken to assess the benets and harms of early feeding strategies while other evidence-based practices to reduce necrotizing enterocolitis were followed, namely use of probiotics, mothers’ own milk and rational antibiotic use. Further studies are necessary in neonates with antenatal doppler abnormalities, but no intrauterine growth retardation. More studies are needed comparing slow versus rapid advancement of feeds.


Introduction
Preterm infants are at increased risk of various adverse outcomes compared to the term population in terms of feed intolerance, necrotizing enterocolitis, infections, respiratory morbidities-bronchopulmonary dysplasia, growth failure in short term and abnormal neurodevelopment in the long term. Neonates with abnormal antenatal Doppler of umbilical artery in the form of absent or reverse end diastolic ow (AREDF), or increased resistance to blood ow in ductus venosus and inferior vena cava, or middle cerebral artery redistribution are a subset of preterms with even further increased risk of morbidities [1]. These neonates are usually delivered preterm and are small for gestation (SGA). In the fetal life, they are exposed to chronic hypoxia leading to preferential shunting of blood to the brain at the expense of circulation to the gut. Postnatally, SGA babies have reduced wall thickness, length, villous weight, and crypt depth [2].
Furthermore, this subset of neonates also has intestinal dysbiosis and an alteration of the proliferation-apoptosis homeostasis of the gut. Hence there are structural, functional, and secretory changes in these neonates and enteral feeding in already compromised gut exacerbates feed intolerance and maybe one of the associated etiologies for the development of NEC and thereby poor outcomes [3].
A systematic review of 14 observational studies has shown an increased risk of NEC in neonates with antenatal doppler abnormalities with odds of 2.13 (95% CI 1.4-3.0) compared to controls without AREDF [4]. In preterms to prevent NEC, several strategies like using mother's own milk (MOM), donor milk, standardized feeding regimen, probiotics avoiding prolonged empirical antibiotics, avoiding proton pump inhibitors are being used [5]. However, in growthrestricted babies with antenatal doppler abnormalities, due to fear of NEC, there are considerable variations in initiation and advancement of feeds. Feeds are often delayed and are advanced at a slower rate, thereby delaying the enteral nutrition, and placing them at increased risk of parenteral nutrition induced liver failure and infections [6]. Hence, a focused systematic review of the available evidence is needed to reduce the variations and improve outcomes in this vulnerable subset of preterms. In neonates with antenatal abnormal dopplers, we compared the impact of early feeding and rapid advancement of feeds with delayed feeding and slow advancement of feeds, respectively result on neonatal outcomes (Primary/Critical: Necrotizing enterocolitis/NEC and mortality; Secondary/Important: feed intolerance, total parenteral nutrition days, duration of hospitalization, healthcare associated infections, cost of care, and neurodevelopment at 18-24 months of corrected age).

Methods
The protocol was registered in the PROSPERO database (registration number CRD42020195845) and can be accessed from the website https://www.crd.york.ac.uk/PROSPERO/. We followed the Cochrane handbook for systematic reviews of interventions [7], the PRISMA (preferred reporting items for systematic reviews and meta-analyses) reporting guidelines [8], and the GRADE handbook [9]. The differences between the protocol and nal review are provided in the online supplemental material.
We included all randomized controlled studies comparing early versus delayed feeding, and slow versus rapid feed advancement in neonates with antenatal doppler abnormalities. The population of interest was preterm neonates (<37 weeks) with antenatal doppler abnormalities de ned as absent or reverse end diastolic ow (AREDF), or increased resistance to blood ow in ductus venosus and inferior vena cava, or middle cerebral artery redistribution. Early feeding was de ned as feed initiation within 48 -72 hours of birth, and delayed feeding was de ned as feed initiation beyond this time point. Slow advancement was de ned as <20-30 ml/kg/day advancement per day, and rapid advancement as >20-30 ml/kg/day. The databases of PubMed, Embase, CINAHL, Cochrane CENTRAL, and Google Scholar were searched on the 6th of September 2020 for studies published between 1990 and 5th of September 2020. The search terminology is provided in the online supplemental material. We also searched for grey literature from various sources like dissertation websites and contacting researchers who have worked on this subject.
We included the studies when they provided data on at least one of the pre-speci ed outcomes. These outcomes included feeding intolerance (1 or more of the following: >50% prefeed aspirates, abdominal girth increase by >2 cm, vomiting, bilious aspirates, blood in stools), necrotizing enterocolitis stage 2 or more (modi ed Bell's staging), time to attain full enteral feeds (at least 150 ml/kg/d), duration of total parenteral nutrition (TPN), mortality, health-care associated infections (culture proven sepsis), duration of hospital stay, cost of care, and neurodevelopmental outcomes at 18-24 months corrected age. We excluded conference abstracts and letters to the editor without original data, and studies published in languages other than English. We excluded studies done in growth-restricted babies without data on antenatal doppler ndings.
The database search was performed by 2 reviewers (AS and RP). The studies obtained by database search were entered into Rayyan software. After excluding duplicates and non-English studies, the titles and abstracts were assessed for eligibility by 2 reviewers independently (AS and RP), and any discrepancy was sorted out by mutual discussion, or involvement of the third reviewer (SM). The following data were extracted by RP and AS in a blinded manner: study identi cation details; study characterisitics including the study setting, study design, feeding strategies, de nitions used for feed intolerance, NEC, and hospital-acquired infections; population characteristics including the gestational age, birth weight, gender, antenatal steroid use, mode of delivery, chorioamnionitis, and resuscitation details; feeding strategy in the intervention and control arms (timing of initiation, type of milk used, volume at initiation, rapidity of advancement, and volume of feed advancement); and the data on aforementioned outcomes. When the data were missing/ unclear, we contacted the corresponding author through email. Any discrepancy was resolved by discussion, or by consulting the third reviewer (SM).
We assessed the risk of bias (ROB) of included studies using the Cochrane Risk of Bias tool, version 2 for randomised controlled trials [10]. We assessed the methods used for random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, selective reporting, and completeness of outcome data. An intention to treat analysis was considered desirable. The assessment was done by 2 reviewers (RP and AS) in a blinded manner from all the eligible studies. Any discrepancy was resolved by discussion, or by consulting the third reviewer (SM).
Studies (or outcomes) deemed comparable were pooled to include in the meta-analysis. A random-effect model meta-analysis was performed using the Cochrane statistical package, RevMan 5.4 software, for each outcome if two or more studies were available. The comparative effect sizes were calculated as risk ratios (RR) and risk difference (RD) for categorical outcomes and mean differences (MD) for continuous outcomes with their 95% con dence intervals (CIs). The pooled estimates were depicted using forest plots. We used Cochrane handbook recommended formulae for converting median to mean; and range, interquartile range and 95% con dence intervals to standard deviations (online supplemental material). Heterogeneity was assessed using the I 2 statistic [11].
We planned subgroup and sensitivity analysis based on the type of doppler abnormalities (AEDF or REDF), intrauterine growth status (small-for-gestational age and appropriate-for gestational age), and gestational age (< 28 weeks, 29 -31 weeks, and >32 weeks). We planned to assess heterogeneity by comparing important factors in trial design (bias), intervention (feeding strategy), and co-interventions (probiotics, mothers' own milk use, and antibiotic usage rates).

Assessment of publication bias was planned if the number of studies exceeded 10. The Cochrane Grading of Recommendations Assessment, Development
and Evaluation (GRADE) approach was used to assess the certainty of evidence [9].

SD-standard deviation, IQR-interquartile range, PIH-Pregnancy Induced Hypertension.
In three studies, the neonates were <37 weeks gestational age [12,13,16], while 3 other studies included neonates up to 32 weeks [17], 34 weeks [14], and 35 weeks [15]. In ve studies, all the neonates included were small for gestational age, while in one study [16], it was not reported. In the ve studies that provided data on antenatal steroid use [12-14, 17, 18], no signi cant differences were noted between the experimental and control arms. None of the studies provided data on the duration of doppler abnormalities, chorioamnionitis and sickness scores at admission to NICU. Jain et al have restricted study population to neonates between 30-36 weeks and weight above 1,000 grams; 60% of neonates in the experimental arm and 63% in the control arm were growth-restricted [18].
While donor human milk was provided as a second option when breast milk was unavailable in the three recent studies [15][16][17], infant formula was provided in the other four studies [12][13][14]18]. The human milk was forti ed at feed volumes of 100 -150 ml/kg in three studies [12,13,15], while it was unclear in the others. None of the studies have reported the use of probiotics. The de nitions used for NEC (modi ed Bell's staging) and full feeds (150-180 ml/kg/day sustained for 3 days) were consistent across the studies. Feeding intolerance was variably de ned in ve studies [12][13][14]17, 18], while it was not de ned in the other two studies [15,16]. In all the studies, >50% pre-feed volumes, blood in stool, bilious aspirates and abdominal signs were considered worrisome.
In ve studies, early feeding was de ned as feeding within 48 hours of birth, while in one study, it was de ned as <5 days of birth [14]. The volume at feed initiation varied from 4 -30 ml/kg/day, duration of trophic feeds from 0 -6 days, and volume of feed advancement from 12 -35 ml/kg/day. In 2 studies, the feed initiation and advancement were strati ed according to the birth weight [15,17]. In the study by Jain et al, rapid advancement was de ned as 30 ml/kg/day in neonates weighing 1-1.25 kg, and 40 ml/kg/day in those weighing >1.25 kg [18]. Slow advancement was de ned as 20 ml/kg/day and 30 ml/kg/day in the above weight ranges, respectively. Further details of feed initiation and advancement in the experimental and control arms are summarized in the online supplemental material.
When compared to delayed feeding, the strategy of early feeding did not increase the incidence of NEC stage 2 or more (odds ratio/OR On the risk of bias assessment using RoB2 tool, 5 studies were adjudged to have some concerns [12-14, 16, 17], while 1 study had a low risk of bias [15]. For outcomes with subjectivity in assessment (i.e., feeding intolerance), 4 studies were adjudged to be at high risk of bias due to lack of blinding of outcome assessment [12,[15][16][17]. The assessment is shown in Table 3 and the criteria used in the online supplemental material. Heterogeneity was not noted for the outcomes studied, except the duration of hospital stay (I 2 = 35%). Subgroup analysis, sensitivity analysis and publication bias assessment were not performed because of the small number of eligible studies. a-Methods to generate random sequence and conceal allocation not mentioned.

b-No information provided on allocation concealment
c-Post randomization exclusions were done d-Outcome assessors were not blinded for the intervention. The prior knowledge can in uence assessment of outcomes of feed intolerance. Hence, the assessment for this outcome is high risk of bias. For other outcomes like necrotizing enterocolitis stage 2 or more, mortality, duration of parenteral nutrition, duration of hospital stay, hospital acquired sepsis, time to reach full feeds and time to regain birth weight, prior knowledge is unlikely to in uence assessment of outcomes. They were adjudged to be at low risk of bias.
e-Trail protocol, registration details, and statistical analysis plan not available f-The published protocol mentions this as an observational study for electrogastrography (EGG). Hence, in effect, this protocol was not published/ or is planned as the study progressed g-high risk for outcomes of feeding intolerance and NEC; some concerns for other outcomes.
On GRADE assessment, the certainty of evidence was rated high for days on parenteral nutrition; moderate for mortality, and time to reach full feeds; low for NEC 2 or more, hospital-acquired sepsis, duration of hospital stay, and time to regain birth weight; and very low for feeding intolerance (Table 4).   When compared to slower feed advancement, rapid feed advancement resulted in decreased time to reach full feeds (MD -0.2 days, 95% CI -0.39, -0.01). There was no difference in other outcomes. The overall risk of bias was adjudged to be low. However, because of the small sample size and wide con dence intervals, the certainty of evidence was rated low.

Discussion
Summary of main results: We evaluated the bene ts and harms of early initiation of feeds in neonates with antenatal doppler abnormalities. Six studies with 772 newborns met the inclusion criteria. Early feeding strategy decreased the time to reach full feeds, duration of parenteral nutrition, duration of hospital stay, and rates of hospitalacquired infection. Although there was a trend towards an increase in rates of feeding intolerance, the incidence of NEC did not increase. There was no effect on mortality and time to regain birth weight.
One study with 83 newborns was eligible for assessment of slow versus rapid advancement of feeds. There was a clinically insigni cant reduction in time to reach full feeds. Other outcomes were not affected.
We could not nd any review which addressed the same question, and hence are unable to make any comparisons with existing literature.
Overall completeness and applicability of evidence: The available evidence is su cient to understand that early feeding in neonates with antenatal doppler abnormalities does not increase rates of NEC or mortality. However, as only one study assessed the rapidity of feed advancement, with a small sample size, more evidence is needed to answer this question.
Quality of evidence: The overall quality of evidence was low. The main factors limiting the quality of evidence were the risk of bias and imprecision in estimates. The risk of bias was predominantly due to problems with the randomization process and selective reporting of the outcomes. The assessment of outcome was not blinded in 4 of the 7 studies, but this is unlikely to impact outcomes other than feeding intolerance.
The only eligible study comparing slow versus rapid feed advancement excluded neonates with extremely low birth weight and gestational age <30 weeks, limiting the generalizability. Also, the sample size was very small.
Potential biases in the review process: We are unlikely to miss any published literature with the extensive literature search performed. The process of data screening, data extraction and risk of bias assessment was performed by 2 reviewers in a blinded manner and the discrepancies were resolved on group meetings involving the third reviewer. This would minimize the introduction of additional bias. For two studies, we requested data from the author and received the required information [17,18].

Conclusions
Implications for practice: There is low-quality evidence that early feeding does not increase rates of necrotizing enterocolitis and moderate-quality evidence that it does not impact mortality. There is very low-quality evidence on a trend towards an increased incidence of feeding intolerance. There is low-to moderate-quality evidence on shorter time to reach full feeds, duration of parenteral nutrition, duration of hospital stay, and lower hospital-acquired infections. As bene ts outweigh the harms, it is recommended to use the early feeding strategy in neonates with antenatal doppler abnormalities.
There is low quality evidence that rapid feed advancement reduces time to reach full feeds in neonates with gestational age >30 weeks, and birth weight >1 kg.
The available data is insu cient to recommend one strategy over the other.

Implication for research:
Future trials could be undertaken to assess the bene ts and harms of early feeding strategies while other evidence-based practices to reduce necrotizing enterocolitis were followed, namely use of probiotics, mothers' own milk and rational antibiotic use. Further studies are necessary in neonates with antenatal doppler abnormalities, but no intrauterine growth retardation. More studies are needed comparing slow versus rapid advancement of feeds.