Effects of neonatal nutrition interventions on neonatal mortality and child health and development outcomes: A systematic review

What is this review about?

This review assesses the efficacy of synthetic vitamin A, dextrose, and probiotic supplementation during the neonatal period. These interventions were assessed separately and not in combination with each other.

What is the aim of this review?

This Campbell systematic review assesses the efficacy of three neonatal nutritional interventions that have the potential of reducing morbidity and mortality during infancy in LMICs: synthetic vitamin A, dextrose, and probiotic supplementation.

What studies are included?

Sixteen studies that assessed the effect of vitamin A supplementation during the neonatal period were included. Thirteen of these studies were conducted in the community setting and three studies were conducted in the hospital setting. All the included studies on neonatal vitamin A supplementation were conducted in LMICs. Most of the studies had a low ROB.

No experimental studies were found that evaluated the use of dextrose for the prevention or treatment of low blood sugar during the neonatal period.

Thirty-three studies assessed the use of probiotics during the 1st month of life. All included studies on probiotic supplementation were randomized and conducted in LMICs. Most of the included studies had a low ROB. The probiotics studies mainly included babies born early and/or with low birth weight, and these studies were mostly conducted in hospital settings.

Key results

Combined results from thirteen vitamin A studies conducted in the community settings showed that there was no significant effect of vitamin A supplementation for reduction of death in young infants at 1, 6, or 12 months of age. Neonatal vitamin A supplementation increases the risk of bulging fontanelle by 53%. The pooled data from probiotics studies showed that this intervention reduced the risk of death by 20% compared to controls. Further analysis showed that compared to controls, probiotic supplementation reduced the risk of a severe form of gastrointestinal illness in neonates called NEC by 54%. Probiotic supplementation also reduced the risk of blood infection called sepsis by 22% compared to controls. The quality grade ratings for these outcomes were “high.”

What are the main findings of the review?

Combined results from thirteen vitamin A studies conducted in the community settings showed that there was no significant effect of vitamin A supplementation for reduction of death in young infants at 1, 6, or 12 months of age. However, neonatal vitamin A supplementation increased the incidence of bulging fontanelle by 53%.

The pooled data from probiotics studies showed that this intervention reduced the risk of death by 20% compared to controls. Further analysis showed that compared to controls, probiotic supplementation reduced the risk of a severe form of gastrointestinal illness in neonates called NEC by 54%. Probiotic supplementation also reduced the risk of blood infection called sepsis by 22% compared to controls. The quality grade ratings for these outcomes were “high.”

What do the findings of this review mean?

Vitamin A supplementation during the 1st month of life does not reduce the risk of death during the 1st year of life in LMICs. However, neonatal vitamin A supplementation increases the risk of bulging fontanelle, which may cause damage to the brain.

We did not find any experimental studies from LMICs that assessed the use of dextrose gel supplementation during the 1st month of life for the prevention or treatment of low blood sugar.

Probiotic supplementation during the 1st month of life to babies born preterm and/or low birthweight can reduce the risk of death, blood infection and bowel sickness (NEC).

There was clinical heterogeneity in the use of probiotics and we could not recommend any single strain or combination of probiotics for wider use based of these results.

There is a lack of studies on probiotic supplementation in the 1st month of life in community settings. More research is needed to assess the effect of probiotics administered to neonates in home/community settings in LMICs.

How up-to-date is this review?

The review authors searched for studies published up to November 2019.

Neonatal vitamin A deficiency (VAD)

Globally, about 190 million children and 19.1 million pregnant women are vitamin A deficient based on serum retinol levels (i.e., serum retinol <0.70 μmol/L) (WHO, 2009a). VAD is most prevalent in South Asia and Africa (Stevens et al., 2015). VAD is associated with increased risk of blindness, infections, and mortality (Imdad et al., 2017). Most of the newborns are vitamin A deficient and rely on supplementation from maternal breast milk (Haider et al., 2017). High prevalence of maternal VAD in LMICs increases the risk of neonatal VAD. There has been interest in vitamin A supplementation during neonatal period to assess if it reduces risk of illness and death (Haider et al., 2017; WHO, 2009b), as it has been shown to reduce morbidity and mortality in children 6–59 months of age (Imdad et al., 2017).

Hypoglycemia during the neonatal period

Hypoglycemia (low blood sugar) is common during the immediate neonatal period. The definition of neonatal hypoglycaemia varies. The American Academy of Pediatrics defines neonatal hypoglycaemia as blood glucose below 47 mg/dl (2.61 mmol/L); however, other societies such as the Pediatric Endocrine Society define neonatal hypoglycaemia as blood glucose <50 mg/dl (2.77 mmol/L; Thompson-Branch & Havranek, 2017; Thornton et al., 2015). Recurrent, severe, and/or persistent hypoglycaemia can lead to complications such as death; there is limited evidence to show that blood sugars below a certain level leads to long-term brain damage (Kaiser et al., 2015; McKinlay et al., 2017; Thornton et al., 2015). About 10–15% of otherwise healthy newborns have low blood sugar, and the rate is much higher among infants with additional risk factors such as large for gestational age, small for gestational age, low birth weight, preterm birth, infant of diabetic mother, and newborns with perinatal asphyxia (Thompson-Branch & Havranek, 2017). Additional risk factors for neonatal hypoglycaemia include neonatal sepsis, prolonged labor, and maternal medication use such as use of β-agonists and β-blockers (Thompson-Branch & Havranek, 2017). The recommended initial intervention to treat neonatal hypoglycaemia is to offer feeding in the form of breastfeeding followed by formula feeding if breastfeeding is unsuccessful. Persistent hypoglycaemia may require IV dextrose supplementation and admission to a neonatal intensive care unit (Thompson-Branch & Havranek, 2017; Thornton et al., 2015). In LMIC, where a significant proportion of births happen at home and incidence of low birth weight and preterm birth is high, prevention and treatment of hypoglycaemia encounters additional challenges (Singhal et al., 1991, 1992; WHO, 2017b; Williams, 1997). The instruments to test blood sugar might not be available in low-resource settings; In addition, formula, IV dextrose, and intensive care units might not be available to treat persistent and/or severe hypoglycaemia. Recent studies have tested simple interventions such as oral dextrose gel to treat neonatal hypoglycaemia and to prevent hypoglycaemia in high-risk newborns (Hegarty et al., 2016; Weston et al., 2016).

Neonatal sepsis and NEC

Neonatal sepsis and NEC are neonatal morbidities that can be fatal (Oza et al., 2015; WHO, 2017b). Neonatal sepsis is the presence of an infection leading to systemic illness. Bacterial sepsis is common in LMIC and is a significant risk factor of morbidity and mortality in these countries (WHO, 2017a). NEC is a condition that occurs in newborns and can lead to intestinal injury and death. The extent of injury may vary from mucosal injury to full thickness intestinal wall injury. NEC happens most commonly in preterm babies and especially in extremely preterm babies (<28 weeks gestational age; AlFaleh & Anabrees, 2014; Patel & Denning, 2015). Multiple factors lead to the development of NEC in preterm infants including altered bacterial gut flora affecting the protective intestinal barrier, decreased intestinal motility, and the increased susceptibility of preterm infants to inflammation and infections (Patel & Denning, 2015). Recent studies have shown that an imbalance between commensal bacteria and pathogenic bacteria (intestinal dysbiosis) makes babies vulnerable to pathogenic bacterial growth in the intestine which then causes inflammation that may contribute to neonatal sepsis and/or NEC (Arrieta et al., 2014; Deshmukh et al., 2014; Gewolb et al., 1999; Panigrahi et al., 2017). There is an increasing interest in correction of intestinal dysbiosis by probiotics to prevent NEC and neonatal sepsis. Data from early studies on probiotic use in neonates from LMIC is encouraging (AlFaleh & Anabrees, 2014; Rao et al., 2016).

Neonatal vitamin A supplementation

Vitamin A is a term used for a subclass of the family of fat soluble compounds named retinoic acids. Vitamin A is found in nature in two forms, provitamin A carotenoids and preformed vitamin A, which is essential to human bodily function. Plant-based foods are the main source of provitamin A carotenoids, of which β-carotene is the most commonly known. Animal-based foods are the main source of preformed vitamin A (Bates, 1995; Haider & Bhutta, 2011). Vitamin A from animal sources (retinol, retinal, retinoic acid, and retinyl esters) is the most active form, and synthetic vitamin A retinol has been used in most intervention trials in the past (Haider & Bhutta, 2011; Imdad et al., 2017). Plant-based foods may not be an adequate source of vitamin A, as the gastrointestinal conversion ratio from carotenoid-to-retinol varies from 6:1 to 26:1 (US Institute of Medicine, Food and Nutrition Board). VAD may, therefore, exist in areas even when there is high consumption of plant-based foods such as in South Asia and Africa (Imdad et al., 2017; Stevens et al., 2015).

Oral dextrose gel supplementation during neonatal period

Dextrose gel is a thickened aqueous solution that contains the concentrated simple carbohydrate dextrose. It can be administered by direct application to oral, buccal, or sublingual mucosa and can increase blood sugars rapidly by absorption through the highly vascularized and thin mucus membranes of the oral mucosa (Hegarty et al., 2016). Detxrose gel is a low cost nonproprietary intervention, and the gel can be prepared in hospital pharmacies. The typical ingredients include water, glucose, a gelling agent, and preservatives (Hegarty et al., 2016). The decision to use dextrose gel in a neonate should be taken on individual basis and should be avoided in neonates with compromised neurological or respiratory status (Hegarty et al., 2016; Weston et al., 2016).

Probiotic supplementation during neonatal period

Prebiotics are supplements that promote the growth of commensal bacteria (AlFaleh & Anabrees, 2014; Panigrahi et al., 2017). Probiotics contain live bacteria that enrich pools of commensal bacteria (AlFaleh & Anabrees, 2014; Millar et al., 2003; Panigrahi et al., 2017). Synbiotics are a combination of prebiotics and probiotics and might have synergistic effect (Johnson-Henry et al., 2016, Nandhini et al., 2016; Panigrahi et al., 2017). These supplements are meant to optimise gut health and their hypothesized mechanisms of actions include enhanced gut barrier function, inhibition of gut colonization with pathogenic bacteria, improvement in colonization with healthy commensal bacteria that protect the infant from enteropathogenic infection through production of acetate, enhanced innate immunity, and increased maturation of the enteric nervous system (Rao et al., 2016). Recent data have shown that probiotic supplements can prevent the incidence of NEC in preterm babies (AlFaleh & Anabrees, 2014; Millar et al., 2003; Patel & Denning, 2015; van den Akker et al., 2018). There is also promising data on use of probiotocs/synbiotics for prevention of neonatal sepsis (Rao et al., 2016; Panigrahi et al., 2017). The most commonly used strains of probiotics include Lactobacillus and Bifidobacterium (Rao et al., 2016).

Neonatal vitamin A supplementation

Vitamin A has an effect on cell differentiation and helps maintain normal functioning of epithelial cells (Bates, 1995; Bhutta et al., 2013; Haider & Bhutta, 2011). It is considered anti-infective because it helps to maintain the protective epithelial barrier of the skin and mucosa, which protects the body from infections. Vitamin A helps in the regeneration of the epithelium and therefore maintains the integrity of the body's first line of defence. These mechanisms may help prevent infections in newborns (McCullough et al., 1999; Wolbach, 1933). Synthetic vitamin A supplementation has been shown to reduce morbidity and mortality in children 6–59 months of age (Imdad et al., 2017). The potential side effects of synthetic vitamin A supplementation include vomiting and bulging fontanelle (Imdad et al., 2016, 2017; Haider & Bhutta, 2011; Haider et al., 2017). Excess vitamin A supplementation can cause toxicity that presents in the form of a bulging fontanelle in children under 1 year, headaches, vomiting, diarrhea, loss of appetite, and irritability (Haider et al., 2017; Imdad et al., 2017).

Oral dextrose gel supplementation during neonatal period

The absorption of dextrose gel through the oral mucosa leads to entry of glucose into lingual veins and into the internal jugular vein. This pathway provides almost immediate delivery of glucose to the systemic circulation and bypasses first pass liver metabolism through the portal circulation. If proven effective in preventing and treating hypoglycaemia, dextrose gel can avoid the need of intravenous glucose and reduce separation of baby from mother (Hegarty et al., 2016; Weston et al., 2016). The intervention is simple enough that it does not require special skills (such as IV placement) and can be administered by community, lay health workers, or the caregiver herself. Potential adverse effects include vomiting, choking, gagging, respiratory distress, and delay of treatment for severe hypoglycaemia (Hegarty et al., 2016; Weston et al., 2016).

Neonatal vitamin A supplementation

Randomized trials on neonatal vitamin A supplementation have produced conflicting results with some studies (mostly from South Asia) showing a mortality benefit while no major benefit in other studies (mostly from Africa) (Haider et al., 2017) and some studies showing even an increased risk of infant mortality in certain populations (Smith et al., 2016).The exact reason for this difference in results is not clear. Previous reviews (Haider et al., 2017; Gogia & Sachdev, 2009) and a WHO technical consultation (WHO, 2009b) have hypothesized on what factors may explain these varied results. Our group has previously published a Cochrane review on the evidence on neonatal vitamin A supplementation (Haider & Bhutta, 2011), and we wanted to update the previous review. The previous review included studies conducted in the community setting. In this review, we considered studies conducted in both the community and the hospital setting in LMIC. We also included neurodevelopment outcomes for this review that were not covered in the previous Cochrane review.

Oral dextrose gel supplementation during neonatal period

Oral dextrose gel has been studied in the prevention and treatment of neonatal hypoglycaemia in high-income countries (Hegarty et al., 2017; Weston et al., 2016); however, it was not clear if similar studies were available from LMICs. Our objective was to consider both randomized and nonrandomized observational studies with a control arm. We hypothesized that the use of dextrose may be more beneficial in LMIC than in high-income countries, as the incidence of neonatal hypoglycaemia might be higher in these countries due to an increased rate of preterm and low birth weight birth.

Probiotics supplementation during neonatal period

The effect of probiotic supplementation for the prevention of NEC and neonatal sepsis has been assessed in previous reviews (AlFaleh & Anabrees, 2014; Rao et al., 2016; van den Akker et al., 2018). Most of these reviews included studies from both high- and LMICs. Deshpande et al. (2017) reviewed studies from LMIC where neonates were supplemented with probiotics. More studies (Amini et al., 2017; Chowdhury et al., 2016; Guney-Varal et al., 2017; Hernández-Enríquez et al., 2016; Hussain et al., 2016) have been published since the publication of Deshpande et al.'s (2017) review. Overall, our objective was to assess the current evidence for the effect of probiotic supplementation during the neonatal period in the hospital and community setting in LMIC.

Types of studies

We included the following study designs:

Randomized controlled trials (RCTs), where participants were randomly assigned either individually or in clusters to intervention and comparison groups. Cross-over designs were also eligible for inclusion.

Types of participants

Participants for this review included neonates (aged 0–28 days) from LMICs. We included neonates regardless of their health status. This includes low birth weight and preterm babies. However, studies that focused on neonates with congenital anomalies were excluded. We considered studies that included older age population groups in addition to neonates only if we could disaggregate relevant data for the neonatal population. For example, a study might include infants up to 6 months of age. We included such a study if the disaggregated data were available for neonates (0–28 days). Even though we planned to assess later childhood outcomes, we did not plan to include studies that recruited participants after the neonatal period.

Types of interventions

The following interventions were included in the review: 1.

Neonatal vitamin A supplementation compared to no supplementation or placebo: we considered only oral synthetic vitamin A supplementation. There was no restriction on the dosage and frequency of the medicine. The comparison group could include a placebo or standard of care.

Each of the above interventions (i.e., vitamin A, dextrose, or probiotics) was summarized separately, and the interventions were not compared to each other directly or indirectly.

Types of outcome measures

Electronic searches

The electronic search for relevant studies was done in the following databases: PubMed, CINAHL, LILACS, SCOPUS, and CENTRAL (Cochrane Central Register for Controlled Trials).

Appendix 1 gives the search strategy for PubMed, CINAHL, LILACS, SCOPUS, and CENTRAL. It includes keywords and MeSH terms as appropriate. This approach includes a search strategy for the population (neonates) and interventions of interest. We planned to run searches for each intervention separately. We first ran the search for the population, which is the same for each intervention. Then we ran the search for each intervention. We then combined both searches by using “AND” and kept the searches in a separate EndNote file.

An example of a search strategy for vitamin A for PubMed was as follows:

(((((“Vitamin A”[Mesh]) OR (Vitamin A[tiab] OR Aquasol A[tiab] OR Retinol[tiab] OR All Trans Retinol[tiab] OR All-Trans-Retinol[tiab] OR Vitamin A1[tiab] OR Vitamin A 1[tiab] OR 11-cis-Retinol[tiab] OR 11 cis Retinol[tiab] OR Tretinoin[tiab])AND Supplement*[tiab]))AND ((“Infant”[Mesh] OR “Premature Birth”[Mesh]) OR (Neonat*[tiab] OR neo nat*[tiab]) OR (newborn* OR new Born*[tiab] OR newly born*[tiab]) OR (preterm[tiab] OR preterms[tiab] OR pre term[tiab] OR pre terms[tiab]) OR (premature*[tiab] AND (birth*[tiab] OR born[tiab] OR deliver*[tiab])) OR (low[tiab] AND (birthweight*[tiab] OR birth weight*[tiab])) OR (lbw[tiab] OR vlbw[tiab] OR elbw[tiab]) OR infant*[tiab] OR (baby[tiab] OR babies[tiab])))) NOT (“Animals”[Mesh] NOT (“Animals”[Mesh] AND “Humans”[Mesh]))

We applied restriction of “humans” to searches. We did not apply any restrictions on searches based on outcomes, study design, or language. There was no restriction on date of publication.

The searches were conducted for vitamin A on December 10, 2018 (updated on November 13, 2019); probiotics on February 8, 2019 (updated on November 27, 2019); and dextrose on April 25, 2019 (updated on November 26, 2019).

Searching other resources

Other resources included the search for ongoing trials at www.clinicaltrials.gov and WHO's ICTRP trials database. We also searched websites of international agencies such as WHO (including WHO's Reproductive Health Library), UNICEF, Global Alliance for Improved Nutrition (GAIN), International Food Policy Research Institute, International Initiative for Impact Evaluation (3ie), Nutrition International (NI), World Bank, USAID and USAID affiliates (e.g., FANTA, SPRING), and the World Food Programme.

Grey literature search sources included NI, GAIN, International Food Policy and Research Institute (IFPRI), and the WHO library database (WHOLIS).

We searched the reference lists of all included studies. We did citation searches of included studies in Google Scholar and Web of Science. We also searched the reference sections of previously published systematic reviews and the latest published studies. We contacted the experts and authors of the newest published studies to ask about any additional studies. Duplicates were removed.

Selection of studies

Two authors independently screened titles/abstracts using prespecified inclusion/exclusion criteria. A full text was reviewed for the studies selected in the initial screening, and the same inclusion/exclusion criteria were applied. If there was a conflict about the inclusion of a study between the two reviewers, a third reviewer (ZAB) was consulted. We used a web-based software “Covidence” (Covidence, 2019) to do both title/abstract and full-text screening. This software allows simultaneous independent screening of studies, and inter-reviewer reliability can be assessed by checking the number of conflicts in the resolved conflict page following each stage of screening.

Data extraction and management

Assessment of ROB in included studies

We used the Cochrane ROB tool (Higgins & Green, 2011) for randomized studies. The Cochrane ROB tool includes the following items:

Selection bias: random sequence generation and allocation concealment

Two authors independently performed the ROB assessments for each study. A third reviewer was involved to resolve any disagreements (ZAB). An overall score was not provided.

Measures of treatment effect

We performed a meta-analysis for the synthesis of quantitative data when the included studies had comparable participants, interventions, and outcomes. We did not assess the effect on outcome across the interventions, such as is done in network meta-analysis. Each intervention was analysed separately. We analysed continuous and dichotomous data separately. For dichotomous outcomes, results were presented as summary risk ratios with 95% CI. We combined risk ratios (events per child) and rate ratios (events per child year) for incidence data because of their similar interpretation and scale. For continuous outcomes, we presented the summary results as the mean difference with 95% CI when data were available on the same scale across the studies. We used the standardized mean difference with 95% CI when data were presented in different scales across the studies.

To avoid reviewer bias, we planned to predetermine the preference for specific data for certain outcomes. For example, for mortality outcomes, we gave preference to denominators in the following order: number with the definite outcome known, number randomized, and child-years. For morbidity data such as neonatal sepsis where both survivors and nonsurvivors might have contributed data, we gave preference to child years, number with the definite outcome known, and number randomized. For randomized trials, we gave preference to data that required the least manipulation by authors or inference by reviewers. We extracted the raw values (e.g., means and SDs) and built the intention-to-treat (ITT) analysis where applicable.

We anticipated that cause-specific morbidity or mortality data might not be readily available, as febrile illness due to respiratory, urinary, or central nervous system infection during the neonatal period are often categorized under a broader term of neonatal sepsis (WHO, 2017b).

Unit of analysis issues

As we planned to include multiple interventions, all interventions and outcomes within those interventions, were meta-analyzed separately.

For randomized trials, we meta-analyzed individual and cluster-randomized trials in the same analysis. We assessed analyses in the cluster-randomized trials to ensure that clustering was appropriately accounted for within the analysis of the primary study, such that study precision was not over or under-estimated within our analysis. If the authors adjusted for cluster randomization, no further adjustment was made. In case a cluster-randomized study was not adjusted by primary authors, we adjusted effect estimates by using the mean cluster size (M) and the intra-cluster correlation coefficient (ICC) to calculate the design effect as follows: design effect = 1 + (M − 1) ICC. We then used the design effect to adjust the study data such that a trial was reduced to its effective sample size or standard error of the summary estimate was inflated. We used the ICC given in the published studies. If the ICC was not available from the published study, we contacted the authors for the same. If the ICC was not available from the authors, we used ICC from the similar studies done in the similar region and on a similar population or took it from the previously published reviews (Haider et al., 2017).

Dealing with missing data

Any missing data were noted including loss to follow-up and dropouts. The reasons for the missing data were taken from the studies, and if it was not mentioned in the studies, the authors were contacted for the same. If the authors reported the adjusted values for missing data, we used the adjusted values.

Assessment of heterogeneity

Statistical heterogeneity was assessed using τ ², I ², and significance of the χ ² test. We also assessed statistical heterogeneity by visually inspecting the forest plots.

Assessment of reporting biases

A funnel plot and its symmetry were used to assess publication bias if the number of included studies for intervention was more than 10. If the funnel plot was suggestive of publication bias, we further investigated the publication bias with the use of Egger's test (Higgins & Green, 2011).

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

1.

High quality studies versus low quality studies. The quality of study was subjectively based on the ROB assessment. Even though we considered all the domains included in the Cochrane ROB tool, we gave higher importance to sequence generation and allocation concealment, as most of the outcomes for this review were objective, and it was less likely that the results of the included studies would have been biased by a lack of blinding.

Summary of findings and assessment of the certainty of the evidence

Results of the search

Figure 1 shows the PRISMA flow diagram for our literature search.

OPEN IN VIEWER

Figure 1

PRISMA flow diagram

Included studies

The characteristics of included studies are available in the table Characteristics of included studies.

Excluded studies

Overall, 138 studies were excluded. See the Characteristics of excluded studies for reasons for exclusion of studies.

Among excluded studies, 86 studies were excluded because of the wrong settings, and most of these studies were conducted in high-income countries. Fifteen studies were excluded because no relevant clinical outcomes were available from the abstract or full text of the studies. Fourteen studies were excluded for the wrong population, and 10 studies were excluded because of the wrong study design. Nine studies had a wrong intervention, and four studies had a wrong comparator.

Vitamin A supplementation during the neonatal period

Figure 2 shows the ROB in the 16 included studies that addressed vitamin A supplementation during the neonatal period.

OPEN IN VIEWER

Figure 2

Risk of bias: neonatal vitamin A supplementation

Probiotic supplementation during the neonatal period

Figure 3 show the ROB in the 33 included studies that addressed probiotic supplementation during the neonatal period.

OPEN IN VIEWER

Figure 3

Risk of bias: probiotic supplementation during neonatal period

Allocation (selection bias)

Blinding (performance bias and detection bias)

Incomplete outcome data (attrition bias)

Selective reporting (reporting bias)

Other potential sources of bias

VItamin A supplementation during neonatal period

Probiotic supplementation during the neonatal period

Data were available for the effect of probiotic supplementation for all-cause mortality, NEC, sepsis, and sepsis-specific mortality.

All-cause mortality at longest follow-up

Twenty-five studies that included 10,998 subjects (probiotics 5548, control 5450) reported data for all-cause mortality for the effect of probiotic supplementation on all-cause mortality. Our meta-analysis found a reduction of 20% in all-cause mortality in the probiotic group compared to control (RR, 0.80; 95% CI, 0.66–0.96; heterogeneity: τ ² = 0.00; (p = 0.55); I ² = 0%) (Figure 4). The number needed to treat was 100. The GRADE rating for this outcome was “high.” Summary of findings Table 3.

OPEN IN VIEWER

Figure 4

(Analysis 2.1) Forest plot of comparison: 2 probiotics versus control, outcome: 2.1 all-cause mortality

OPEN IN VIEWER

Table 3

Probiotics supplementation during neonatal period

Probiotics supplementation compared to control during neonatal period
Patient or population: neonates (Mmost of the included studies had preterm and low birth weight neonates)
Setting: low and middle income countries
Intervention: probiotics/synbiotics
Comparison: control
Outcomes	No. of participants (studies) Follow up	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Anticipated absolute effects* (95% CI)
				Risk with control	Risk difference with probiotics (intervention)
All-cause mortality	10904 (25 RCTs)	⊕⊕⊕⊕ HIGHa,b,c,d	RR, 0.80 (0.66–0.96)	Study population
				47 per 1000	9 fewer per 1000 (15 fewer to 1 fewer)
Neonatal sepsis	8918 (21 RCTs)	⊕⊕⊕⊕ HIGHa,d,e	RR, 0.78 (0.70–0.86)	Study population
				205 per 1000	45 fewer per 1000 (62 fewer to 29 fewer)
Necrotizing enterocolitis	55574 (29 RCTs)	⊕⊕⊕⊕ HIGHa,d,f	RR, 0.46 (0.35–0.61)	Study population
				101 per 1000	55 fewer per 1000 (66 fewer to 41 fewer)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;
GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Even though three (Dongol Singh et al., 2017; Fernández-Carrocera et al., 2013; Kaban et al., 2019) of the included studies in the analysis had high ROB related to randomizations, the exclusion of these studies did not have much effect on the magnitude of the summary estimate or its statistical significance.

I ² was 0%.

All-cause mortality is an objective outcome and there were no concerns about the indirect measurement of the outcome.

The confidence interval of the summary estimate did not include 1.

I ² was 23% and the p. value for heterogeneity was 0.16.

The I ² was 24%.

Publication bias

A funnel plot for publication bias looked symmetrical (Figure 5).

OPEN IN VIEWER

Figure 5

(Analysis 2.1) Funnel plot of comparison: 2 probiotics versus Control, outcome: 2.1 All-cause mortality

Subgroup analyses

Table 1 gives the summary of data for subgroup analyses. For the outcome of all-cause mortality, data were available to perform subgroup analyses based on settings (hospital vs. community-based studies), type of probiotics (single strains vs. multiple strain vs. synbiotic), type of participants (term vs. preterm/low birth weight), type of feeding (breastfeeding vs. formula feeding vs. mixed feeding), and probiotic preparation (preparation containing Lactobacillus vs. Bifidobacterium vs. both Lactobacillus and Bifidobacterium vs. S. boulardii). No significant difference was noted among the subgroups; however, the number of studies varied for each group within the subgroup analysis.

Sensitivity analysis

Random versus fixed effect models

Use of a fixed-effect model did not change the summary estimate for the effect of probiotics on all-cause mortality (RR, 0.80; 95% CI, 0.66–0.96; heterogeneity: χ ² = 22.55, (p = 0.55); I ² = 0%) Analysis 2.7.

Risk of bias

Exclusion of three studies (Dongol Singh et al., 2017; Fernández-Carrocera et al., 2013; Kaban et al., 2019) that were at high ROB for randomizations/allocation concealment did not change the summary estimate to a great extent, and results remained statistically significant Analysis 2.8.

Incidence of NEC

This outcome was reported by twenty-nine studies that included a total of 5574 (probiotics, 2843; control, 2731) participants. The combined results showed that the probiotics group had a relative reduction in NEC prevalence of 54% compared to control group (RR, 0.46; 95% CI, 0.35–0.59; heterogeneity: τ ² = 0.11; (p = 0.12); I ² = 24%) (Figure 6). The number needed to treat was 17. We have high certainty in this evidence. Summary of findings Table 3.

OPEN IN VIEWER

Figure 6

(Analysis 2.9) Forest plot of comparison: 2 probiotics versus control, outcome: 2.9 necrotizing enterocolitis (any type)

Publication bias

A funnel plot for publication bias looked symmetrical.

Subgroup analyses

Table 1 shows the results of subgroup analyses. For NEC, data were available to perform the following 3 subgroup analyses: type of probiotics (single strain vs. multiple strain vs. synbiotic), type of feeding (breastfeeding vs. formula feeding vs. mixed feeding), and probiotic preparation (preparation containing Lactobacillus vs. Bifidobacterium vs. both Lactobacillus and Bifidobacterium vs. S. boulardii). No significant difference was noted among the subgroup analyses except the one based on probiotics preparation (p value for subgroup difference 0.05). The probiotics preparation that has Lactobacillus in it, either as a single strain (RR, 0.39; 95% CI, 0.25–0.61; 13 studies) or in combination with Bifidobacterium (RR, 0.44; 95% CI, 0.36–0.68; 13 studies), had a significant effect compared to Bifidobacterim alone (RR, 0.20; 95% CI, 0.09, −0.47) or S. boulardii alone (RR, 0.94; 95% CI, 0.45–1.95; two studies). This subgroup analysis should be interpreted carefully, as the number of studies was not uniformly distributed among all the subgroups, and the statistical difference might be due to the small number of studies in two of the subgroups.

Sensitivity analyses

Random versus fixed effect models

The use of a fixed effect model did not change the summary estimate for the effect of probiotics on NEC (RR, 0.45; 95% CI, 0.37–0.56; heterogeneity: χ ² = 37.92, df = 29 (p = 0.12); I ² = 24%).

Risk of bias

The exclusion of three studies (Dongol Singh et al., 2017; Fernández-Carrocera et al., 2013; Kaban et al., 2019) with a high ROB due to randomization/allocation concealment did not change the summary estimate significantly.