Probiotics to prevent necrotising enterocolitis in very preterm or very low birth weight infants

Abstract

Background

Intestinal dysbiosis may contribute to the pathogenesis of necrotising enterocolitis (NEC) in very preterm or very low birth weight infants. Dietary supplementation with probiotics to modulate the intestinal microbiome has been proposed as a strategy to reduce the risk of NEC and associated mortality and morbidity. 

Objectives

To determine the effect of supplemental probiotics on the risk of NEC and mortality and morbidity in very preterm or very low birth weight infants.

Search methods

We searched Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 2) in the Cochrane Library; MEDLINE Ovid (1946 to 17 Feb 2020), Embase Ovid (1974 to 17 Feb 2020), Maternity & Infant Care Database Ovid (1971 to 17 Feb 2020), the Cumulative Index to Nursing and Allied Health Literature (1982 to 18 Feb 2020). We searched clinical trials databases, conference proceedings, and the reference lists of retrieved articles for randomised controlled trials (RCTs) and quasi‐RCTs.

Selection criteria

We included RCTs and quasi‐RCTs comparing probiotic supplementation with placebo or no probiotics in very preterm or very low birth weight infants.

Data collection and analysis

We used the standard methods of Cochrane Neonatal. Two review authors separately evaluated trial quality, extracted data, and synthesised effect estimates using risk ratio (RR), risk difference (RD), and mean difference. We used the GRADE approach to assess the certainty of evidence for effects on NEC, all‐cause mortality, late‐onset infection, and severe neurodevelopmental impairment.

Main results

We included 56 trials in which 10,812 infants participated. Most trials were small (median sample size 149). Lack of clarity on methods to conceal allocation and mask caregivers or investigators were the main potential sources of bias in about half of the trials. Trials varied by the formulation of the probiotics. The most commonly used preparations contained Bifidobacterium spp.,Lactobacillus spp., Saccharomyces spp., andStreptococcus spp. alone or in combinations.

Meta‐analysis showed that probiotics may reduce the risk of NEC: RR 0.54, 95% CI 0.45 to 0.65 (54 trials, 10,604 infants; I² = 17%); RD ‐0.03, 95% CI ‐0.04 to ‐0.02; number needed to treat for an additional beneficial outcome (NNTB) 33, 95% CI 25 to 50. Evidence was assessed as low certainty because of the limitations in trials design, and the presence of funnel plot asymmetry consistent with publication bias. Sensitivity meta‐analysis of trials at low risk of bias showed a reduced risk of NEC: RR 0.70, 95% CI 0.55 to 0.89 (16 trials, 4597 infants; I² = 25%); RD ‐0.02, 95% CI ‐0.03 to ‐0.01; NNTB 50, 95% CI 33 to 100. Meta‐analyses showed that probiotics probably reduce mortality (RR 0.76, 95% CI 0.65 to 0.89; (51 trials, 10,170 infants; I² = 0%); RD ‐0.02, 95% CI ‐0.02 to ‐0.01; NNTB 50, 95% CI 50 to 100), and late‐onset invasive infection (RR 0.89, 95% CI 0.82 to 0.97; (47 trials, 9762 infants; I² = 19%); RD ‐0.02, 95% CI ‐0.03 to ‐0.01; NNTB 50, 95% CI 33 to 100). Evidence was assessed as moderate certainty for both these outcomes because of the limitations in trials design. Sensitivity meta‐analyses of 16 trials (4597 infants) at low risk of bias did not show an effect on mortality or infection. Meta‐analysis showed that probiotics may have little or no effect on severe neurodevelopmental impairment (RR 1.03, 95% CI 0.84 to 1.26 (five trials, 1518 infants; I² = 0%). The certainty on this evidence is low because of limitations in trials design and serious imprecision of effect estimate. Few data (from seven of the trials) were available for extremely preterm or extremely low birth weight infants. Meta‐analyses did not show effects on NEC, death, or infection (low‐certainty evidence).

Authors' conclusions

Given the low to moderate level of certainty about the effects of probiotic supplements on the risk of NEC and associated morbidity and mortality for very preterm or very low birth weight infants, and particularly for extremely preterm or extremely low birth weight infants, further, large, high‐quality trials are needed to provide evidence of sufficient quality and applicability to inform policy and practice.

Plain language summary

Probiotics for prevention of necrotising enterocolitis in very preterm or very low birthweight infants

Review question
Does giving very preterm or very low birth weight infants probiotics prevent necrotising enterocolitis?

Background
Very preterm infants (born more than eight weeks' early) and very low birth weight (less than 1.5 kg) are at risk of developing a severe bowel disorder, where a portion of the bowel becomes inflamed, infected, and dies, called necrotising enterocolitis. This condition is associated with death, serious infection, and long‐term disability and developmental problems. One way to help prevent necrotising enterocolitis and associated conditions may be to add probiotics (dietary supplements containing potentially beneficial bacteria or yeasts) to milk feeds.

Study characteristics
The search is up to date as of 18 February 2020. We found 56 trials, with, in total, more than 10,000 infant participants. Trials were mostly small, and some had design flaws that might bias their findings.

Key results
Combined analyses showed that giving very preterm and very low birth weight infants probiotics may reduce the risk of necrotising enterocolitis, and probably reduces the risk of death and serious infection. There is no evidence of an effect on disability or developmental outcomes. Few trials provided data for extremely preterm infants (born more than 12 weeks' early) and extremely low birth weight (less than 1.0 kg), and these analyses did not show effects on necrotising enterocolitis, death and serious infection.

Certainty of evidence
The evidence for an effect on necrotising enterocolitis is "low‐certainty" because of concerns that the effect could have been biased by small trials with unreliable methods.

Background

The intestinal microbiome may play an important role in the pathogenesis of necrotising enterocolitis (NEC) (Embleton 2017). Probiotics are microorganisms that benefit the host by modulating the intestinal microbiome and promoting mucosal barrier functions and resistance to pathogens. Dietary supplementation with probiotics has been proposed as a strategy to reduce the risk of NEC and associated morbidity and mortality in very preterm or very low birth weight infants (VLBW) infants.

Description of the condition

Necrotising enterocolitis, a syndrome of acute intestinal necrosis of unknown aetiology, affects about 5% of very preterm or VLBW infants (Horbar 2012). The major risk predictors for NEC include being extremely preterm or extremely low birth weight (ELBW), and having evidence of intrauterine growth restriction or absent or reversed end‐diastolic flow velocities in Doppler studies of the foetal aorta or umbilical artery (Samuels 2017). Infants who develop NEC experience more infections, have lower levels of nutrient intake, grow more slowly, and have longer durations of intensive care and hospital stay than gestation‐comparable infants who do not (Battersby 2018; Berrington 2012). The associated mortality rate is about 20%, and infants who develop NEC, especially if associated with bloodstream infections, have a higher risk of neurodevelopmental problems and disability compared with their peers (Hickey 2018; Martin 2010).

The pathogenesis of NEC remains incompletely understood but is thought to involve intestinal dysbiosis, infection and inflammation (Eaton 2017; Mara 2018; Morgan 2011). Emerging evidence supports the theory that the intestinal microbiome affects the risk of developing NEC (Masi 2019; Olm 2019; Stewart 2012; Warner 2016). Most very preterm or VLBW infants who develop NEC have received enteral milk feeds. Feeding with human milk rather than cow's milk formula reduces the risk of NEC (Quigley M 2019). One putative mechanism for this protective effect is that “prebiotic” substances in human milk promote the growth of non‐pathogenic probiotic microorganisms, predominantly lactobacilli and bifidobacteria, that modulate the intestinal microbiome and promote mucosal barrier functions (Embleton 2017; Granger 2020; Walsh 2019). Compared with human milk‐fed term infants, however, very preterm or VLBW infants typically harbour fewer probiotic microorganisms and more potential pathogens such as enterococci and Enterobacteriaceae, which might be due to dysbiotic effects of enteral fasting and antibiotic exposure (Stewart 2017).

Given the putative role of probiotics in maintaining the structure, integrity, and function of the intestinal barrier, the possibility that supplemental probiotics might be effective in preventing NEC is of considerable research interest (Berrington 2019; Patel 2018).

Description of the intervention

The probiotic preparations used most commonly as enteral supplements contain one or more strains of bacteria (typically bifidobacteria or lactobacilli) or the fungus Saccharomyces boulardii (Thomas 2010). Other bacteria with probiotic properties include Bacillus clausii,Enterococcus faecium, and Streptococcus thermophilus. Exogenous probiotics can colonise the mucosal surface of the human gastrointestinal tract (Abdulkadir 2016; Zmora 2018). A range of probiotic supplements, as single‐ or multiple‐strain preparations, are available commercially and have been used to prevent and treat infectious or inflammatory gastrointestinal conditions in adults. Despite biological plausibility and underpinning pre‐clinical studies, however, evidence for benefit remains low certainty for most conditions (Bron 2017; Koretz 2018; Kunk 2019; Lerner 2019; Suez 2019). Furthermore, serious, unexpected adverse events and outcomes have been associated with probiotic supplementation for critically‐ill adults (Besselink 2008; Boyle 2006).

Probiotics for very preterm infants

Policies and practices for the use of probiotic supplements to prevent NEC in very preterm or VLBW infants vary within and between countries (Duffield 2019; Viswanathan 2016). Parents have expressed willingness to consider use of probiotics for their very preterm or VLBW infants if evidence of benefit and safety exists (Sesham 2014). Enteral administration of commercially‐available supplements of lyophilised probiotic microorganisms, usually multi‐species preparations containing lactobacilli or bifidobacteria or both, is established in some settings (Robertson 2020). Routine use outwith trials, however, remains limited because of uncertainty about the optimal constitution of preparations (strains of microorganisms and dosing strategies), quality control and safety issues including contamination of products with potential pathogens, and national licensing processes and regulatory requirements (Berrington 2019; Fleming 2019; Pell 2019; van den Akker 2020; Vermeulen 2020). Although probiotic supplementation in immuno‐competent adults is considered to be safe, exogenous probiotic microorganisms have been reported as causing bacteraemia or fungaemia in very preterm or VLBW infants (Bertelli 2015; Esaiassen 2016; Jenke 2012; Zbinden 2015).

How the intervention might work

Intestinal probiotic microorganisms are thought to exert their beneficial effects via several mechanisms. Probiotics may out‐compete pathogens for nutrients and limit pathogen growth via production of inhibitory organic acids ("post‐biotics") and antimicrobial compounds (Embleton 2017; Patel 2015). Infants supplemented with probiotics harbour fewer potential pathogens in the intestine (Alcon‐Giner 2020). Other putative actions include stimulating differentiation and proliferation of enterocytes, enhancing expression of intestinal digestive enzymes, and improving intestinal mucosal barrier integrity (Bron 2017; Johnson‐Henry 2016; Sanders 2019).

Why it is important to do this review

Necrotising enterocolitis and associated complications, particularly infections, are the commonest causes of mortality and serious morbidity beyond the early neonatal period in very preterm or VLBW infants (Berrington 2012). Since probiotic supplementation might reduce the risk of NEC, appraising and synthesising the trial evidence about the effectiveness and safety of probiotic supplementation could inform practice, policy, and research (Embleton 2016; Quigley E 2019). Currently, international policy statements that exist to guide practice do not make unconditional recommendations for use of any probiotic combination for very preterm or VLBW infants (Marchand 2012; van den Akker 2020).

Objectives

To determine the effect of supplemental probiotics on the risk of necrotising enterocolitis (NEC) and mortality and morbidity in very preterm or very low birth weight (VLBW) infants.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) and quasi‐RCTs.

Types of participants

We included very preterm (< 32 weeks' gestation) or extremely low birth weight (VLBW)(< 1500 g) infants (pre‐specified analyses for extremely preterm (< 28 weeks' gestation) or extremely low birth weight (ELBW) (< 1000 g) infants).

Types of interventions

We included enteral administration of any probiotic or probiotic combination for at least one week compared to placebo or no treatment.

We categorised probiotic preparations at the genus level (Bifidobacterium spp.,Lactobacillus spp.,Sacchromyces spp., Streptococcal spp., others, and combinations thereof).

Types of outcome measures

Primary outcomes

• Necrotising enterocolitis (NEC), confirmed at surgery or autopsy or diagnosed by at least two of the following clinical features (Walsh 1986):

  • abdominal radiograph showing pneumatosis intestinalis or gas in the portal venous system or free air in the abdomen;

  • abdominal distension with abdominal radiograph with gaseous distension or frothy appearance of bowel lumen (or both);

  • blood in stool;

  • lethargy, hypotonia or apnoea (or combination of these).

• All‐cause mortality before discharge from hospital.

Secondary outcomes

• Late‐onset invasive infection, as determined by culture of bacteria or fungus from blood or cerebrospinal fluid or from a normally sterile body space (> 48 hours after birth).

• Late‐onset infection with the supplemented probiotic microorganism.

• Duration of hospitalisation (days).

• Neurodevelopmental impairment assessed by a validated test after 12 months' post‐term: neurological evaluations, developmental scores, and classifications of disability, including cerebral palsy and auditory and visual impairment.

Search methods for identification of studies

We used the criteria and standard methods of Cochrane Neonatal.

Electronic searches

We used the standard search strategy of Cochrane Neonatal to search Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 2) in the Cochrane Library; MEDLINE Ovid (1946 to 17 Feb 2020), Embase Ovid (1974 to 17 Feb 2020), Maternity & Infant Care Database Ovid (1971 to 17 Feb 2020), the Cumulative Index to Nursing and Allied Health Literature (1982 to 18 Feb 2020), and clinical trials databases, and conference proceedings (see Appendix 1 for the full search strategies for each database). We searched clinical trials registries for ongoing or recently completed trials (clinicaltrials.gov; the World Health Organization’s International Trials Registry and Platform, and the ISRCTN Registry).

Searching other resources

We searched the reference lists of any articles selected for inclusion in this review.

Data collection and analysis

We used the standard methods of Cochrane Neonatal.

Selection of studies

One review author (SS) screened titles and abstracts of all records identified by the search and coded records as "order" or "exclude". A second review author (WM) assessed all records coded as "order" and made the final decision about which records were ordered as full‐text articles. SS and SO read the full texts and used a checklist to assess each article's eligibility for inclusion on the basis of pre‐specified inclusion and exclusion criteria. WM checked these decisions.

Data extraction and management

Two review authors (SS and WM or SO) extracted data independently using a data collection form to aid extraction of information on design, methods, participants, interventions, outcomes, and treatment effects from each included study. We discussed disagreements until we reached consensus. If data from the trial reports were insufficient, we contacted trialists for further information.

Assessment of risk of bias in included studies

Two review authors (SS and WM or SO), independently assessed the risk of bias (low, high, or unclear) of all included trials using the Cochrane ‘Risk of bias’ tool (Higgins 2011) for the following domains.

• Sequence generation (selection bias).

• allocation concealment (selection bias).

• blinding of participants and personnel (performance bias).

• blinding of outcome assessment (detection bias).

• incomplete outcome data (attrition bias).

We resolved any disagreements by discussion or by a third assessor. See Appendix 2 for a description of risk of bias for each domain.

Measures of treatment effect

We analysed treatment effects in the individual trials using Review Manager 5 (Review Manager 2020), and reported risk ratios (RRs) and risk differences (RDs) for dichotomous data, and mean differences (MDs) for continuous data, with respective 95% confidence intervals (CIs). We determined the number needed to treat for one additional beneficial outcome (NNTB) for analyses with a statistically significant difference in the RD.

Unit of analysis issues

The unit of analysis was the participating infant in individually‐randomised trials. For cluster‐randomised trials, we undertook analyses at the level of the individual while accounting for inter‐cluster correlations in the data using methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). Cross‐over studies were not eligible for inclusion.

Dealing with missing data

We requested additional data from trial investigators when data on important outcomes were missing or were reported unclearly. If unavailable, we planned to undertake sensitivity analyses to assess the potential impact of missing outcome data.

Assessment of heterogeneity

We examined treatment effects in individual trials and heterogeneity between trial results by inspecting the forest plots if more than one trial was included in a meta‐analysis. We calculated the I² statistic for each analysis to quantify inconsistency across studies and to describe the percentage of variability in effect estimates that may be due to heterogeneity rather than to sampling error. If we detected moderate (I² > 50%) or high (I² > 75%) heterogeneity, we planned to explore possible causes (differences in study design, participants, interventions, or outcome assessments).

Assessment of reporting biases

We assessed funnel plot asymmetry visually and with Harbord's modification of Egger's test in meta‐analyses with data from more than nine trials contributing events (Harbord 2006).

Data synthesis

We used a fixed‐effect model for meta‐analysis (as per Cochrane Neonatal recommendations). When moderate or high heterogeneity existed, we planned to examine the potential causes in subgroup (see below) and sensitivity (by methodological quality) analyses.

Subgroup analysis and investigation of heterogeneity

We planned to undertake subgroup analyses by:

• genus of probiotics or combinations (Bifidobacterium spp.,Lactobacillus spp.,Sacchromyces spp., Streptococcal spp., others, and combinations thereof);

• type of enteral feeding permitted for participating infants (human milk versus formula versus mixed).

Sensitivity analysis

We planned sensitivity analyses to determine how estimates were affected by including only studies at low risk of bias: (i) selection bias (adequate randomisation and allocation concealment), (ii) detection or performance bias (adequate masking of intervention and measurement), (iii) attrition bias (< 20% loss to follow‐up for primary outcome assessment), and (iv) reporting bias (selective reporting).

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach, as outlined in the GRADE Handbook (Schünemann 2013), to assess the certainty of evidence of the following (clinically relevant) outcomes: NEC, all‐cause mortality, late‐onset infection, and severe neurodevelopmental impairment.

Three review authors (WM, MXRR and SO) independently assessed the certainty of the evidence for each of the outcomes above. We considered evidence from RCTs as high certainty but downgraded the evidence one level for serious (or two levels for very serious) limitations based upon the following: design (risk of bias), consistency across studies, directness of the evidence, precision of estimates, and presence of publication bias. We used the GRADEpro GDT Guideline Development Tool to create two ‘Summary of findings’ tables to report the certainty of the evidence.

The GRADE approach results in an assessment of the certainty of a body of evidence as one of four grades.

• High certainty: further research is very unlikely to change our confidence in the estimate of effect.

• Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.

• Low certainty: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.

• Very low certainty: we are very uncertain about the estimate.

Results

Description of studies

See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification; Characteristics of ongoing studies

Results of the search

See Figure 1.

1.

Study flow diagram: review update 2020

Included studies

See: Characteristics of included studies.

We included 56 trials. Most were conducted during the past 20 years (four trials pre‐2000). Geographical spread was wide, though predominantly in Europe (23 trials) and Asia (23 trials). Only one trial took place in sub‐Saharan Africa (Zeber‐Lubecka 2016).

Most trials occurred in single centres. Nine were multicentre (Al‐Hosni 2012; Costeloe 2015; Dani 2002; Dilli 2015; Hays 2015; Jacobs 2013; Lin 2008; Manzoni 2009; Totsu 2014).

In all but one of the trials, individual infants were allocated randomly to intervention or control groups. One trial, based in 19 neonatal units in Japan, used a cluster design, with the unit of randomisation being the neonatal unit (Totsu 2014).

Population

In total, 10,812 infants participated in the 56 included trials. The median number of participants in the trials was 149. Twenty‐one trials enrolled fewer than 100 participants. Twenty trials enrolled between 100 and 199 participants. Twelve trials enrolled between 200 and 499 participants. Three trials enrolled 500 participants or more: Costeloe 2015 (N = 1310); Dani 2002 (N = 585); Jacobs 2013 (N = 1099).

Most trials enrolled only very preterm or VLBW infants, with average birth weight among participants typically 1000 g to 1200 g, and average gestation at birth 28 to 32 weeks'. Eight trials enrolled infants of gestational age up to 34 weeks', or birth weight up to 1800 g (Chandrashekar 2018; Dashti 2014; Fujii 2006; Hernandez‐Enriquez 2016; Mohan 2006; Ren 2010; Strus 2018; Tewari 2015). Because the average gestation at birth was < 32 weeks', or the average birth weight < 1500 g, we included these trials.

Two trials restricted participation to extremely low birth weight (ELBW) infants (Al‐Hosni 2012; Wejryd 2019). Four trials excluded infants who were born with birth weight below the 10th percentile for the reference population ("small‐for‐gestation") (Al‐Hosni 2012; Hays 2015; Indrio 2017; Kitajima 1997). None of the trials specified exclusion of infants who had evidence of absent or reversed end‐diastolic flow velocities detected on antenatal Doppler studies of the foetal aorta or umbilical artery.

In most trials, participating infants were permitted human milk or formula feeding. Seven trials enrolled infants who received human milk only (Roy 2014; Samanta 2009; Shadkam 2015; Shashidhar 2017; Tewari 2015; Van Niekerk 2014; Wejryd 2019), and five trials enrolled only formula‐fed participants (Costalos 2003; Chrzanowska‐Liszewska 2012; Indrio 2017; Reuman 1986; Stratiki 2007).

Interventions and comparisons

The probiotic preparations tested varied. Thirty‐three trials used single‐genus probiotics (most commonly, Bifidobacterium spp. or Lactobacillus spp.), and 23 used multi‐genus combinations (most commonly, Bifidobacterium spp. plus Lactobacillus spp.). These were mostly commercially‐available products supplied by the manufacturer for use in the trial.

• Bifidobacterium spp. (14 trials):

• • B. breve (Costeloe 2015; Fujii 2006; Hikaru 2010; Kitajima 1997; Li 2019; Patole 2014; Wang 2007);

  • B. lactis (Dilli 2015; Mihatsch 2010; Mohan 2006; Stratiki 2007);

  • B. bifidum (Totsu 2014);

  • B. adolescentis (Huang 2009);

  • B. lactis, or B. longum, or both (three intervention groups) (Hays 2015).

• Lactobacillus spp. 13 trials):

• • L. rhamnosus (Agarwal 2003; Chrzanowska‐Liszewska 2012; Dani 2002; Manzoni 2006; Manzoni 2009; Millar 1993);

  • L. reuteri (Oncel 2014);

  • L. acidophilus (Reuman 1986).

• Sacchromyces spp. (four trials):

• • Sacchromyces boulardii (Costalos 2003; Demirel 2013; Serce 2013; Zeber‐Lubecka 2016).

• Bacillus spp. (two trials):

• • Bacillus clausii (Tewari 2015);

  • Bacillus coagulans* (Sari 2011).

(*Lactobacillus sporogenes in report.)

• Bifidobacterium spp. plus Lactobacillus spp. (eight trials):

• • B. breve and L. casei (Yakult®) (Braga 2011);

  • B. bifidum, B. longum, B. infantis, L. rhamnosus, L. paracasei , L. casei, L. acidophilus, and L.latis (Cap TS6®) (Chowdhury 2016);

  • B. bifidum and L. acidophilus (Infloran®) (Lin 2005; Lin 2008; Saengtawesin 2014);

  • B. longum and L. rhamnosus (Rougé 2009);

  • B. longum, B. bifidum, B. lactis and L. acidophilus (Roy 2014);

  • B. longum, B. bifidum, B.infantis and L. acidophilus (Samanta 2009).

• Bifidobacterium spp. plus Streptococcus spp. (two trials):

• • B. infantis, B. lactis and S. thermophilus (Jacobs 2013);

  • B. infantis, B. bifidum** and S. thermophilus (Bin‐Nun 2005).

(** Lactobacillus bifidus in report)

• Bifidobacterium spp. plus Lactobacillus spp. plus Sacchromyces spp. (four trials):

• • B. infantis, L. rhamnosus, L. casei, L. plantarum, L acidophilus, and S. boulardii (Dutta 2015);

  • B. bifidum, L acidophilus, and S. boulardii (Hariharan 2016);

  • B. longum, L.acidophilus,L. rhamnosus, and S. boulardii (Chandrashekar 2018; Shashidhar 2017).

• Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp. (five trials):

• • B longum, B. breve, L. acidophilus, L. rhamnosus, L. bulgaricus, L.casei, and S. thermophilus (Dashti 2014);

  • B. infantis, L. rhamnosus, L. casei, L. plantarum, L acidophilus, and S. thermophilus (Fernández‐Carrocera 2013);

  • B. infantis, L acidophilus, and Enterococcus faecium (Kanic 2015);

  • B. infantis, L. acidophilus, Enterococcus faecium, and Bacillus cereus (Ren 2010);

  • Bifidobacterium spp. (not specified), L. acidophilus, L. delbrueckii. and S. thermophilus (Rehman 2018).

Most trials initiated probiotic (and placebo if used) administration during the first week after birth, typically with the first enteral feed. The lyophilised probiotics were reconstituted in water or milk, and administered to supply 10⁸ to 10⁹ colony forming units per dose, once or twice daily via a gastric feeding tube. In most trials, the intervention period was at least six weeks, typically until 34 to 36 weeks' postmenstrual age, or until discharge from hospital. Eleven of the trials administered the intervention for a shorter period (from seven to 30 days) (Braga 2011; Costalos 2003; Dutta 2015; Huang 2009; Kitajima 1997; Millar 1993; Mohan 2006; Ren 2010; Reuman 1986; Shadkam 2015; Van Niekerk 2014). One trial continued the intervention until the infant reached 2000 g body weight (Totsu 2014).

Outcomes

Fifty‐four trials reported the number of infants who developed NEC, and 51 trials reported mortality prior to hospital discharge. Forty‐seven trials reported (or provided unpublished data) the number of infants with at least one episode of culture‐confirmed infection. Other in‐hospital outcomes reported included time to establish full enteral feeding, rate of weight gain, and duration of hospital stay (22 trials). Six trials reported neurodevelopmental or cognitive outcomes (Jacobs 2013; Lin 2005; Oncel 2014; Sari 2011; Totsu 2014; Patole 2014). Two trials did not report any of the review outcomes (Agarwal 2003; Li 2019).

Excluded studies

We excluded 26 reports of studies (Characteristics of excluded studies). The most common reasons for exclusion were ineligible population (most participants not very preterm, or VLBW), intervention (prebiotics or synbiotics) and design (not randomised). A further four screened articles were secondary reports for included trials.

Risk of bias in included studies

Methodological quality varied between the included trials (Risk of bias in included studies; Figure 2).

2.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Twenty‐five of the 56 trials were assessed as being a low risk of selection bias. These employed adequate methods to generate the random sequence, typically computer‐generated, and methods to conceal allocation, typically central or pharmacy allocation, or storage of allocation codes in sealed envelopes (we did not mandate that reports stated that envelopes were "opaque"). Randomisation and allocation concealment methods were not stated in 26 trial reports (unclear risk of bias), and in five "quasi‐randomised" trials, alternate allocation was used (high risk of bias).

Blinding

Twenty‐five trials were assessed as being a low risk of performance bias and detection bias. These were placebo‐controlled (usually maltodextrin), or the report or investigators indicated that preparation of the intervention (mixing the probiotic in milk) was undertaken by staff who were not directly involved in other caregiving duties or outcome assessments (for example, pharmacy staff). In 13 trials, control infants received milk feeds without probiotic supplements, but it was unclear whether staff were aware of the group allocation (unclear risk of bias). Eighteen trials were at high risk of bias due to absence of any masking measures.

Incomplete outcome data

Attrition bias does not appear to be an issue in most trials (outcome data reported for > 80% of randomised cohorts).

Selective reporting

Most reports did not provide access to the trial protocol. It is unlikely, however, that reporting bias was an issue in most trials (low risk of bias) where the review primary and infant‐important outcomes were reported. In trials where the aim was to assess surrogate outcomes such as stool colonisation or intestinal permeability, clinical outcome data were generally available from the investigators.

Effects of interventions

See: Table 1; Table 2

Summary of findings 1. Probiotics compared to control in very preterm or very low birth weight infants.

Probiotics compared to control in very preterm or very low birth weight infants
Patient or population: very preterm or very low birth weight infants Setting: neonatal care centres globally Intervention: probiotics Comparison: control
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Sensitivity analysis of trials at low risk of bias
	Risk with control	Risk with Probiotics
Necrotising enterocolitis (before hospital discharge)	Study population		RR 0.54 (0.45 to 0.65)	10,604 (54 studies)	⊕⊕⊝⊝ Lowa,b	Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias showed a reduced risk of NEC: RR 0.70, 95% CI 0.55, 0.89 (I² = 25%)
	61 per 1000	33 per 1000 (27 to 40)
Mortality (all‐cause before hospital discharge)	Study population		RR 0.76 (0.65 to 0.89)	10,170 (51 studies)	⊕⊕⊕⊝ Moderatea	Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias did not show an effect: RR 0.86, 95% CI 0.69, 1.07 (I² = 0%)
	65 per 1000	49 per 1000 (42 to 58)
Invasive infection (before hospital discharge)	Study population		RR 0.89 (0.82 to 0.97)	9762 (47 studies)	⊕⊕⊕⊝ Moderatea	Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias did not show an effect: RR 0.90, 95% CI 0.79, 1.02 (I² = 8%)
	173 per 1000	154 per 1000 (142 to 168)
Severe neurodevelopmental impairment (18 months to 3 years)	Study population		RR 1.03 (0.84 to 1.26)	1518 (5 studies)	⊕⊕⊝⊝ Lowa,c	Sensitivity meta‐analysis of two trials (913 infants) at low risk of bias did not show an effect: RR 0.99, 95% CI 0.76, 1.27 (I² = 0%)
	194 per 1000	200 per 1000 (163 to 245)
*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.
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.

^aDowngraded one level for serious study limitations (high risk of bias due to uncertainty about methods used to generate random sequence, conceal allocation, and mask outcome assessment) in 12 trials

^bDowngraded one level for serious publication bias (funnel plot asymmetry and statistical evidence consistent with trial size; trials favouring controls missing)

^cDowngraded one level for serious imprecision of effect estimate (95% CI around estimate consistent with substantial harm or benefit)

Summary of findings 2. Probiotics compared to control in extremely preterm or extremely low birth weight infants.

Probiotics compared to control in extremely preterm or extremely low birth weight infants
Patient or population: extremely preterm or extremely low birth weight infants Setting: neonatal care centres globally Intervention: probiotics Comparison: control
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)
	Risk with control (extremely preterm or ELBW)	Risk with Probiotics
Necrotising enterocolitis (before hospital discharge)	Study population		RR 0.90 (0.68 to 1.21)	1712 (8 studies)	⊕⊕⊝⊝ Low,a,b
	100 per 1000	90 per 1000 (68 to 121)
Mortality (before hospital discharge)	Study population		RR 0.91 (0.71 to 1.16)	1661 (6 studies)	⊕⊕⊝⊝ Low,a,b
	137 per 1000	124 per 1000 (97 to 159)
Invasive infection (before hospital discharge)	Study population		RR 0.90 (0.76 to 1.06)	1471 (6 studies)	⊕⊕⊝⊝ Low,a,b
	282 per 1000	254 per 1000 (214 to 299)
*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.
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.

^aDowngraded one level for serious study limitations due to high risk of bias (uncertainty about methods used to generate random sequence, conceal allocation, and mask assessments) in many trials

^bDowngraded one level for serious imprecision of effect estimate (95% CI around estimate consistent with substantial harm or benefit)

Comparison 1. Probiotics versus control

Primary outcomes

Necrotising enterocolitis

Meta‐analysis of data from 54 trials (10,604 infants) showed a reduced risk of NEC (Analysis 1.1; Figure 3):

1.1. Analysis.

Comparison 1: Probiotics versus control, Outcome 1: Necrotising enterocolitis

3.

Forest plot of comparison: 1 Probiotics versus control, outcome: 1.1 Necrotising enterocolitis.

• Risk ratio (RR) 0.54, 95% confidence interval (CI) 0.45 to 0.65 (I² = 17%);

• Risk difference (RD) ‐0.03, 95% CI ‐0.04 to ‐0.02;

• NNTB 33; 95% CI 25 to 50.

There was statistically significant evidence of funnel plot asymmetry consistent with trials favouring controls missing from the meta‐analysis (Harbord's modified Egger test for bias ‐0.78, 95% CI ‐1.51 to ‐0.06; P = 0.04) (Figure 4).

4.

Funnel plot of comparison: 1 Probiotics versus control, outcome: 1.1 Necrotising enterocolitis.

We assessed the certainty of evidence as "low" using GRADE approach, downgraded for serious study design limitations and serious risk of publication bias (Table 1).

Mortality

Meta‐analysis of data from 51 trials (10,170 infants) showed a reduced risk of mortality (Analysis 1.2; Figure 5):

1.2. Analysis.

Comparison 1: Probiotics versus control, Outcome 2: Mortality

5.

Forest plot of comparison: 1 Probiotics versus control, outcome: 1.2 Mortality.

• RR 0.76, 95% CI 0.65 to 0.89 (I² = 0%);

• RD ‐0.02, 95% CI ‐0.02 to ‐0.01;

• NNTB 50; 95% CI 50 to 100.

There was some evidence of funnel plot asymmetry (Harbord's modified Egger test for bias ‐0.52, 95% CI ‐1.15 to 0.12, P = 0.11) (Figure 6).

6.

Funnel plot of comparison: 1 Probiotics versus control, outcome: 1.2 Mortality.

We assessed the certainty of evidence as "moderate" using GRADE approach, downgraded for serious study design limitations (risk of bias in included trials) (Table 1).

Secondary outcomes

Invasive infection

Meta‐analysis of data from 47 trials (9762 infants) showed a reduced risk of infection (Analysis 1.3):

1.3. Analysis.

Comparison 1: Probiotics versus control, Outcome 3: Invasive infection

• RR 0.89, 95% CI 0.82 to 0.97 (I² = 19%);

• RD ‐0.02, 95% CI ‐0.03 to ‐0.01;

• NNTB 50; 95% CI 33 to 100.

There was no evidence of funnel plot asymmetry (Harbord's modified Egger test for bias ‐0.07, 95% CI ‐0.86 to 0.73, P = 0.86).

We assessed the certainty of evidence as "moderate" using GRADE approach, downgraded for serious study design limitations (risk of bias in included trials).

Late‐onset infection with the supplemented probiotic microorganism

None of the included studies reported invasive infection caused by the supplemented probiotic microorganisms.

Duration of birth hospitalisation

Meta‐analysis of data from 22 trials (5458 infants) showed a shorter duration of hospitalisation (Analysis 1.4):

1.4. Analysis.

Comparison 1: Probiotics versus control, Outcome 4: Duration of birth hospitalisation (days)

• MD ‐1.93 days, 95% CI ‐3.78 to ‐0.08 (I² = 26%).

There was no evidence of funnel plot asymmetry.

Two other trials reported data that could not be meta‐analysed:

• Oncel 2014 reported shorter median duration of hospitalisation (38 versus 46 days);

• Tewari 2015 reported no difference in duration of hospitalisation.

Neurodevelopmental outcomes

Neurodevelopmental impairment

Five trials reported severe neurodevelopmental impairment (either motor, sensory, or cognitive) in surviving children. Three assessed children using Bayley Scales of Infant Development II (BSID‐II) at 18 to 24 months (Oncel 2014; Sari 2011), or three years (Lin 2005) post‐term. One trial assessed Bayley‐III composite scales, Movement Assessment Battery for Children, and Wechsler Preschool and Primary Scale of Intelligence Full Scale Intelligence Quotient at two to five years (Jacobs 2013). One trial, undertaken in Japan, used the Kyoto Scale of Psychological Development 2001 (similar to the Bayley III scales) and physical examination to assess neurodevelopmental status at 18 months' post‐term (Totsu 2014).

Completeness of neurodevelopmental follow‐up assessment varied (balanced between groups in all trials):

• Lin 2005: 90%;

• Sari 2011: 84%;

• Totsu 2014: 73%;

• Oncel 2014: 68%;

• Jacobs 2013: 48%.

None of the individual trials, nor a meta‐analysis of data from five trials (1518 infants) showed an effect (Analysis 1.5);

1.5. Analysis.

Comparison 1: Probiotics versus control, Outcome 5: Severe neurodevelopmental impairment

• RR 1.03, 95% CI 0.84 to 1.26 (I² = 0%).

We assessed the certainty of evidence as "low" using GRADE approach, downgraded for serious study design limitations (including attrition bias) and for serious imprecision of effect estimate.

Cerebral palsy

None of the individual trials, nor a meta‐analysis of data from five trials (1512 infants) showed an effect (Analysis 1.6):

1.6. Analysis.

Comparison 1: Probiotics versus control, Outcome 6: Cerebral palsy

• RR 1.13, 95% CI 0.74 to 1.72 (I² = 18%).

Visual impairment

None of the individual trials, nor a meta‐analysis of data from four trials (1356 infants) showed an effect (Analysis 1.7):

1.7. Analysis.

Comparison 1: Probiotics versus control, Outcome 7: Visual impairment

• RR 0.50, 95% CI 0.14 to 1.80 (I² = 0%).

Hearing impairment

None of the individual trials, nor a meta‐analysis of data from four trials (1356 infants) showed an effect (Analysis 1.8):

1.8. Analysis.

Comparison 1: Probiotics versus control, Outcome 8: Hearing impairment

• RR 0.46, 95% CI 0.18 to 1.17 (I² = 32%).

Cognitive performance

Patole 2014 assessed 42% of eligible participants aged three to five years using the Mullen's Scale of Early Learning tool. Analysis did not show an effect on the "early learning composite score" (Analysis 1.9):

1.9. Analysis.

Comparison 1: Probiotics versus control, Outcome 9: Continuous early learning composite measure

• RR ‐1.00 (95% CI ‐6.38, 4.38).

Probiotics versus control in extremely preterm or ELBW infants

Two trials restricted participation to ELBW infants (Al‐Hosni 2012; Wejryd 2019). Five trials reported subgroup data for extremely preterm or ELBW infants (Costeloe 2015; Jacobs 2013; Oncel 2014; Roy 2014; Tewari 2015; Wang 2007).

Necrotising enterocolitis

Meta‐analysis of data from eight trials (1712 infants) did not show an effect (Analysis 2.1):

2.1. Analysis.

Comparison 2: Probiotics versus control (extremely preterm or ELBW), Outcome 1: Necrotising enterocolitis

• RR 0.90, 95% CI 0.68 to 1.21 (I² = 0%).

We assessed the certainty of evidence as "low" using GRADE approach, downgraded one level for study limitations due to high risk of bias and one level for imprecision of effect estimate (Table 2).

Mortality

Meta‐analysis of data from six trials (1661 infants) did not show an effect (Analysis 2.2):

2.2. Analysis.

Comparison 2: Probiotics versus control (extremely preterm or ELBW), Outcome 2: Mortality

• RR 0.91, 95% CI 0.71 to 1.16 (I² = 0%)

We assessed the certainty of evidence as "low" using GRADE approach, downgraded one level for serious study limitations due to high risk of bias and one level for serious imprecision of effect estimate (Table 2).

Invasive infection

Meta‐analysis of data from six trials (1471 infants) did not show an effect (Analysis 2.3):

2.3. Analysis.

Comparison 2: Probiotics versus control (extremely preterm or ELBW), Outcome 3: Invasive infection

• RR 0.90, 95% CI 0.76 to 1.06 (I² = 0%)

We assessed the certainty of evidence as "low" using GRADE approach, downgraded one level for serious study limitations due to high risk of bias and one level for serious imprecision of effect estimate (Table 2).

Late‐onset infection with the supplemented probiotic microorganism

None of the included studies reported invasive infection caused by the supplemented probiotic microorganisms.

Duration of birth hospitalisation

Analysis of data from one trial (22 infants) did not show an effect:

• MD ‐5.40 days, 95% CI ‐14.20 to 3.40)

Neurodevelopmental outcomes

None of the trials reports provided subgroup data for meta‐analysis. Three reports stated that there was not an effect of probiotics on the rate of severe neurodevelopmental impairment in the extremely preterm or ELBW subgroup (Jacobs 2013; Sari 2011; Totsu 2014).

Subgroup comparison by genus of probiotics

Necrotising enterocolitis

There was some evidence of subgroup differences depending on genus of probiotics (Chi² = 11.23, df = 7 (P = 0.13), I² = 37.7%; Analysis 1.1; Figure 3). The largest effect size estimates favoured trials using combinations of:

• Lactobacillus spp.

• Bifidobacterium spp. plus Lactobacillus spp.

• Bifidobacterium spp. plus Streptococcus spp.

• Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp.

Mortality

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 4.40, df = 7 (P = 0.73), I² = 0%; Analysis 1.2; Figure 5).

Invasive infection

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 2.57, df = 7 (P = 0.92), I² = 0%; Analysis 1.3).

Duration of birth hospitalisation

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 2.56, df = 6 (P = 0.86), I² = 0%; Analysis 1.4).

Neurodevelopmental outcomes

Neurodevelopmental impairment

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 1.48, df = 4 (P = 0.83), I² = 0%; Analysis 1.5).

Cerebral palsy

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 3.66, df = 4 (P = 0.45), I² = 0%; Analysis 1.6).

Visual impairment

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 1.59, df = 2 (P = 0.45), I² = 0%; Analysis 1.7).

Hearing impairment

There was no evidence of subgroup differences depending on genus of probiotics (Chi² = 3.63, df = 3 (P = 0.30), I² = 17.4%; Analysis 1.8).

Subgroup comparison by type of enteral feed (human milk versus formula versus mixed)

Necrotising enterocolitis

There was no evidence of subgroup differences depending on the type of enteral feed (Chi² = 3.81, df = 2 (P = 0.15), I² = 47.6%; Analysis 3.1).

3.1. Analysis.

Comparison 3: Subgroup analysis by type of feeding, Outcome 1: Necrotising enterocolitis

Mortality

There was no evidence of subgroup differences depending on the type of enteral feed (Chi² = 2.80, df = 2 (P = 0.25), I² = 28.7%; Analysis 3.2).

3.2. Analysis.

Comparison 3: Subgroup analysis by type of feeding, Outcome 2: Mortality

Invasive infection

There was no evidence of subgroup differences depending on the type of enteral feed (Chi² = 3.45, df = 2 (P = 0.18), I² = 42.0%; Analysis 3.3).

3.3. Analysis.

Comparison 3: Subgroup analysis by type of feeding, Outcome 3: Invasive infection

Duration of birth hospitalisation

There was no evidence of subgroup differences depending on the type of enteral feed (Chi² = 1.98, df = 2 (P = 0.37), I² = 0%; Analysis 3.4).

3.4. Analysis.

Comparison 3: Subgroup analysis by type of feeding, Outcome 4: Duration of birth hospitalisation (days)

Neurodevelopmental outcomes

In all trials, participants may have received human milk, or formula, or both.

Sensitivity analyses by risk of bias

Necrotising enterocolitis

There was evidence of subgroup differences depending on risk bias (Chi² = 7.82, df = 2 (P = 0.02), I² = 74.4%). Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias showed a reduced risk of NEC (Analysis 4.1):

4.1. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 1: Necrotising enterocolitis

• RR 0.70, 95% CI 0.55, 0.89 (I² = 25%);

• RD ‐0.02, 95% CI ‐0.03 to ‐0.01;

• NNTB 50; 95% CI 33 to 100.

Mortality

There was no evidence of subgroup differences depending on risk of bias (Chi² = 3.41, df = 2 (P = 0.18), I² = 41.3%). Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias did not show an effect (Analysis 4.2):

4.2. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 2: Mortality

• RR 0.86, 95% CI 0.69, 1.07 (I² = 0%);

• RD ‐0.01, 95% CI ‐0.03 to 0.00.

Invasive infection

There was some evidence of subgroup differences depending on risk of bias (Chi² = 4.62, df = 2 (P = 0.10), I² = 56.7%). Sensitivity meta‐analysis of 16 trials (4597 infants) at low risk of bias did not show an effect (Analysis 4.3):

4.3. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 3: Invasive infection

• RR 0.90, 95% CI 0.79, 1.02 (I² = 8%);

• RD ‐0.02, 95% CI ‐0.04 to 0.00.

Duration of birth hospitalisation

There was no evidence of subgroup differences depending on risk of selection bias (Chi² = 1.30, df = 2 (P = 0.52), I² = 0%). Sensitivity meta‐analysis of six trials (2786 infants) at low risk of bias did not show an effect (Analysis 4.4):

4.4. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 4: Duration of birth hospitalisation (days)

• MD ‐2.44 days, 95% CI ‐5.76 to 1.29 (I² = 52%).

Neurodevelopmental outcomes

Neurodevelopmental impairment

There was no evidence of subgroup differences depending on risk of bias (Chi² = 0.30, df = 1 (P = 0.58), I² = 0%). Sensitivity meta‐analysis of two trials (913 infants) at low risk of bias did not show an effect (Analysis 4.5):

4.5. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 5: Severe neurodevelopmental impairment

• RR 0.99, 95% CI 0.76, 1.27 (I² = 0%);

• RD 0.00, 95% CI ‐0.05 to 0.05.

Cerebral palsy

There was no evidence of subgroup differences depending on risk of bias (Chi² = 0.01, df = 1 (P = 0.92), I² = 0%). Sensitivity meta‐analysis of two trials (913 infants) at low risk of bias did not show an effect (Analysis 4.6):

4.6. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 6: Cerebral palsy

• RR 1.14, 95% CI 0.68, 1.92 (I² = 0%);

• RD 0.01, 95% CI ‐0.02 to 0.04.

Visual impairment

There was no evidence of subgroup differences depending on risk of performance and detection bias (Chi² = 1.53, df = 1 (P = 0.22), I² = 34.6%). Sensitivity meta‐analysis of two trials (913 infants) at low risk of bias did not show an effect (Analysis 4.7):

4.7. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 7: Visual impairment

• RR 2.91, 95% CI 0.12, 71.21 (I² = not applicable);

• RD 0.00, 95% CI ‐0.01 to 0.01.

Hearing impairment

There was no evidence of subgroup differences depending on risk of performance and detection bias (Chi² = 1.96, df = 1 (P = 0.16), I² = 48.9%). Sensitivity meta‐analysis of two trials (913 infants) at low risk of bias did not show an effect (Analysis 4.8):

4.8. Analysis.

Comparison 4: Sensitivity analyses: Risk of bias, Outcome 8: Hearing impairment

• RR 0.30, 95% CI 0.09, 0.98 (I² = 60%); 0.30 [0.09, 0.98)

• RD ‐0.02, 95% CI ‐0.03 to ‐0.00.

Discussion

Summary of main results

Meta‐analyses of data from more than 50 trials, with more than 10,000 participants in total, show that enteral supplementation with probiotics may reduce the risk of NEC, and probably reduces mortality and the risk of late‐onset invasive infection in very preterm or VLBW infants. Sensitivity meta‐analyses of trials at low risk of bias did not show effects on mortality or infection. None of our included studies reported instances of invasive infection caused by the probiotic organisms being tested. Meta‐analyses of data available from five trials do not show an effect on severe neurodevelopmental impairment. According to GRADE assessment, the certainty of the evidence in this review is low to moderate.

Overall completeness and applicability of evidence

Most of the trials were undertaken within the past 20 years in healthcare facilities internationally, but predominantly in Europe and Asia. Few data were available from trials conducted in sub‐Saharan Africa. The findings should be applicable to current care practices for very preterm or VLBW infants including infants 'small for gestation' at birth (only four trials excluded such infants, and none defined evidence of abnormal end‐diastolic flow velocities in fetal Doppler studies as an exclusion criterion). The average event rate for NEC in the control group was 6%, consistent with estimates from prevalence studies in very preterm of VLBW infants in high‐income countries (Battersby 2018). We pre‐specified a comparison including only data for extremely preterm or ELBW infants. Only two small trials, however, restricted participation to this population, and a further five trials reported subgroup data. Meta‐analyses included fewer than 1800 infants, and did not show effects on any of the review outcomes. These estimates are imprecise due to few participants being included in meta‐analyses. The wide confidence intervals around the point estimates do not rule out important benefits or harms in this subpopulation, and are consistent with the effects seen in the meta‐analyses including the entire very preterm or VLBW population.

The review findings are likely to be broadly applicable to infants fed enterally with human milk or formula or both. Formula feeding increases risk of NEC and the risk‐benefit balance of probiotic supplementation could differ between human milk‐ and formula‐fed very preterm or VLBW infants (Quigley M 2019). Pre‐specified subgroup analyses did not show differences in effect sizes between trials that permitted only human milk feeding for participants (seven trials), versus trials in which all infants received only formula (five trials), versus those trials in which infants could be fed with human milk or formula or both (42 trials). The reported data in trials that permitted human milk‐ or formula‐feeding or both were insufficient to analyse subgroups effects at an infant level by type of enteral feeds received.

The main challenge in applying the findings of this review is the heterogeneity of the interventions tested. Subgroup analyses showed some evidence of differences in effect sizes depending on the genus of the probiotics used, with larger effects in trials that used combinations of bifidobacteria and lactobacilli (with or without S. thermophilus). Data from the only two large (> 1000 participants), high‐quality trials support this interpretation (Costeloe 2015; Jacobs 2013). The largest trial of probiotic supplementation yet reported (N = 1310) showed that a single‐strain preparation of Bifidobacterium breve is probably ineffective in reducing NEC (Costeloe 2015). Conversely, the combination of Bifidobacterium infantis, Streptococcus thermophilus and B. lactis used in the other large trial (N = 1099) is probably effective in reducing the risk of NEC (but not mortality or infection) (Jacobs 2013). These findings, although consistent with recent network analyses of different probiotic combinations, should be interpreted cautiously (Bi 2019; Morgan 2020; van den Akker 2018). As indirect comparisons are not randomised, any differences in effect between trials or groups of trials could be due to other factors, including methodological quality, types of participants, setting, and other practices and policies such as feeding protocols and antibiotic stewardship. Effect estimates may be confounded by species and strain level differences that affect how probiotic organisms interact with each other and endogenous microorganisms in the intestine of immature infants (Millar 2012). Consequently, the optimal probiotic composition or combination is unlikely to be determined reliably by analyses of the existing trial data.

Quality of the evidence

We assessed, using GRADE approach, the certainty of evidence as low or moderate for the pre‐specified outcomes (Table 1; Table 2). About half of the trials had methodological quality weaknesses, including in measures used to conceal random allocation and to mask clinicians, parents, and caregivers to the intervention (Figure 2), increasing the risk of bias in outcomes assessment. Knowledge of the intervention group could have affected caregivers' or assessors' subjective perceptions of outcomes, for example, it may have influenced decisions on whether investigate or diagnose NEC or invasive infection.

Most of the included trials were small (median N = 149). The asymmetry evident in the funnel plot for the meta‐analysis of the effect on NEC (and mortality to a lesser extent) was consistent with small‐study bias (Figure 4). One explanation is publication bias ‐ the tendency for articles that report "statistically significant" effects to be submitted and accepted for publication (Gale 2020). Publication bias, as well as other sources of small‐study bias, has become increasingly evident as an important contributor to exaggerated effect size estimates in meta‐analyses of interventions to improve outcomes in very preterm or VLBW infants (Ohlsson 2020; Pammi 2020). Another concern is that in many of the trials that aimed to assess the effect of probiotics on clinical outcomes, it is unclear from most reports how the sample size was defined, and whether trial "stopping rules" existed. If trial investigators were able to monitor accumulating outcome data until an effect on an outcome was detected, this may result a tendency to detect spurious effects that inflate the pooled estimate of effect sizes.

Attrition bias, due to loss of outcome data from randomised participants, was not a concern for the in‐hospital outcomes (NEC, death, infection) assessed in this review. Completeness of long‐term neurodevelopmental outcomes data, however, ranged from 48% to 90% between the trials that reported such assessments. The degree of incomplete "follow up" assessment was balanced across the intervention and control groups in each trial. Although this is reassuring with regard to the impact of attrition bias on effect estimates, some concern remains that the assessed population may not be representative of the entire cohort (Tin 1998). The findings in meta‐analyses that probiotics does not affect neurodevelopmental outcomes are consequently of 'low‐certainty'.

Potential biases in the review process

The main concern with meta‐analysis of the effect on NEC is the possibility that the findings are subject to small‐study biases, including publication bias. There may be a greater availability of data for inclusion in meta‐analyses from trials which reported statistically significant or potentially important effects (Hopewell 2009). We attempted to minimise this threat by searching the proceedings of major international perinatal conferences to identify trial reports that were not published in full form in journals. We cannot be sure that other trials have been undertaken but not reported, and the concern remains that such trials are less likely than published trials to have detected statistically significant or clinically important effects.

We contacted trial investigators for unpublished data (Young 2011). In several cases, authors of "proof of concept" or exploratory trials that aimed primarily to assess whether probiotic administration affected intestinal (stool) colonisation patterns or permeability or immune function were able to provide unpublished clinical outcomes data for inclusion in meta‐analyses.

We did not include any potential risk of bias due to the funding source of the included trials (where reported). In related contexts, such as manufacturers of breast milk substitutes funding infant feeding trials, this conflict is important to note (Cleminson 2015). We did not, however, consider this to be a substantial risk of bias here. Manufacturers of probiotic products supported some of the trials by supplying the intervention at no or low cost (noted in Characteristics of included studies), but we considered that they were unlikely to have a conflict of interest in the trial outcome for this relatively niche indication.

Agreements and disagreements with other studies or reviews

Our findings are broadly consistent with other recent systematic reviews of probiotics for preterm infants (summarised in Jarrett 2019). Our review differs from others in some key respects:

• we restricted the population of interest to very preterm and VLBW infants to enhance applicability to those infants at high risk of developing NEC and associated complications;

• we included trials that assessed probiotics only, and excluded trials that assessed prebiotics or synbiotics;

• we conducted genus‐level subgroup analyses to explore for differences in effect sizes depending upon the probiotic or combination of probiotics assessed;

• we included formal statistical evaluation to assess the risk of small‐study bias for the major outcomes;

• we pre‐specified sensitivity analyses to determine how trial methodological quality affected effect sizes; and

• we included a formal GRADE assessment of the 'certainty' of the evidence at outcomes level to help inform policy, practice, and research (Gephart 2020).

Authors' conclusions

Implications for practice.

Despite the quantity of trial evidence, and the effects shown on necrotising enterocolitis, mortality, and infection, uncertainty remains about how to interpret and apply the trial data of probiotic supplementation for very preterm or VLBW infants. As well as concern that effect size estimates are inflated by biases in the existing trials (including publication bias), the major barrier to implementing the findings is that existing analyses are not able to determine reliably the optimal constitution of probiotics (strains, doses, timing of introduction, duration of use) for routine prophylactic use. A variety of commercially‐available probiotic preparations are in use in a minority of neonatal units internationally, but widespread use appears to be limited by availability and regulatory and licensing issues. Although the data from the included trials are reassuring with regard to safety, probiotic bacteraemia or fungaemia and other adverse effects have been reported in preterm infants (Bertelli 2015; Esaiassen 2016; Jenke 2012; Zbinden 2015). It remains unclear whether different strains or combinations have different safety profiles.

Implications for research.

Given the uncertainty about whether (and which) probiotics affect important outcomes in very preterm or VLBW infants, consideration could be given to further assessment in randomised placebo‐controlled trials. It is essential, firstly, for investigators to determine whether families and clinicians would support a trial of this intervention. Any planned trials should attempt to ensure that caregivers and assessors are masked to the intervention as investigation and diagnosis of important outcomes such as NEC, invasive infection and neurodevelopmental impairment can be subjective. While it may be appropriate to be broadly inclusive of very preterm and VLBW infant participants, trials should ensure sufficient power to assess effects in extremely preterm or ELBW infants, and to explore interactions with the type of enteral feed received.

A key concern in planning any trial is choosing the appropriate intervention to assess. Two options appear favourable. Firstly, a 'confirmatory' trial that uses the probiotic combination (Bifidobacterium infantis, Streptococcus thermophilus and B. lactis) already shown to be likely to reduce the risk of NEC in a large, high‐quality trial in Australasia (Jacobs 2013). Alternatively, investigators may consider a pragmatic choice based on multi‐strain products in established use in their regions (which provides some availability and quality control reassurances with regard to product integrity and safety). Furthermore, investigators could consider whether trials using 'synbiotics' (combinations of probiotics with 'prebiotics' such as human milk oligosaccharides and other milk glycans) are merited alongside trials, or as part of an adaptive design, of probiotics (Underwood 2019).

Unit of randomisation and analysis is another consideration. Although individual infant randomisation is preferred for statistical and analytical reasons, concern exists that cross‐contamination of the trial organisms to infants in the control group will limit the power of the trial to detect an effect (as may have happened in Costeloe 2015). Randomising at the neonatal care centre level (cluster‐RCT) obviates this problem, but inflates the sample size requirement considerably because of inter‐cluster correlation of outcomes.

What's new

Date	Event	Description
4 October 2020	New search has been performed	Inclusion criteria modified to include only very preterm (< 32 weeks' gestation) or very low birth weight infants (< 1500 g) with pre‐specified analyses for extremely preterm (< 28 weeks' gestation) or extremely low birth weight (< 1000 g) infants. The literature was searched in February 2020. Thirty‐two new published trials were identified.
4 October 2020	New citation required and conclusions have changed	Probiotics may reduce the risk of necrotising enterocolitis, but the certainty of the evidence is "low".

History

Protocol first published: Issue 4, 2005
Review first published: Issue 1, 2008

Date	Event	Description
1 October 2013	New citation required but conclusions have not changed	Updated search identified eight new trials for inclusion in this review update.
1 October 2013	New search has been performed	This updates Al Faleh 2011
3 November 2010	New search has been performed	This updates the review "Probiotics for prevention of necrotizing enterocolitis in preterm infants" published in the Cochrane Database of Systematic Reviews (Al Faleh 2008). New authorship: Khalid AlFaleh, Jasim Anabrees, Dirk Bassler, Turki Al‐Kharfi. Updated search identified seven new trials for inclusion in this review update.
3 November 2010	New citation required and conclusions have changed	With the addition of seven new trials to this update, it brings the total to sixteen eligible trials randomizing 2842 infants. The previous review included nine eligible trials, randomizing 1425 infants.
12 November 2008	Feedback has been incorporated	Feedback incorporated
22 July 2008	Amended	Converted to new review format.

Appendices

Appendix 1. Electronic search methodology

Cochrane probiotics search strategies February 2020

Bibliographic databases: Cochrane Central register of Controlled Trials (CENTRAL), CINAHL, Embase, Maternity & Infant Care, MEDLINE

Trial registers: WHO ICTRP & ClinicalTrials.gov

Cochrane Register of Controlled Trials (CENTRAL)

Search date = 18^th February 2020; 126 records

#1 MeSH descriptor: [Probiotics] explode all trees

#2 (probiotic*):ti,ab,kw (Word variations have been searched)

#3 MeSH descriptor: [Bifidobacterium] explode all trees

#4 (bifidobacterium*):ti,ab,kw (Word variations have been searched)

#5 MeSH descriptor: [Lactobacillus] explode all trees

#6 (lactobacill*):ti,ab,kw (Word variations have been searched)

#7 MeSH descriptor: [undefined] explode all trees

#8 MeSH descriptor: [Saccharomyces boulardii] this term only

#9 (Saccharomyces):ti,ab,kw (Word variations have been searched)

#10 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9

#11 MeSH descriptor: [Prebiotics] explode all trees

#12 (prebiotic*):ti,ab,kw (Word variations have been searched)

#13 MeSH descriptor: [Oligosaccharides] explode all trees

#14 (oligosaccharide*):ti,ab,kw (Word variations have been searched)

#15 MeSH descriptor: [Inulin] explode all trees

#16 (inulin*):ti,ab,kw (Word variations have been searched)

#17 ((fructooligosaccharide* or fructo‐oligosaccharide* or FOS or FOSs or galacto‐oligosaccharide* or galactooligosaccharide*)):ti,ab,kw (Word variations have been searched)

#18 MeSH descriptor: [Lactoferrin] explode all trees

#19 (lactoferrin*):ti,ab,kw (Word variations have been searched)

#20 MeSH descriptor: [Lactulose] explode all trees

#21 (lactulose*):ti,ab,kw

#22 #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 or #20 or #21

#23 MeSH descriptor: [Synbiotics] explode all trees

#24 (synbiotic*):ti,ab,kw (Word variations have been searched)

#25 (((probiotic* and prebiotic*) NEAR/4 combin*)):ti,ab,kw (Word variations have been searched)

#26 #23 OR #24 OR #25

#27 #10 AND #22 AND #26

#28 MeSH descriptor: [Infant, Newborn] explode all trees

#29 MeSH descriptor: [Premature Birth] explode all trees

#30 neonat*:ti,ab,kw (Word variations have been searched)

#31 neo‐nat*:ti,ab,kw (Word variations have been searched)

#32 newborn or new born* or newly born*:ti,ab,kw (Word variations have been searched)

#33 preterm or preterms or (pre term) or (pre terms):ti,ab,kw (Word variations have been searched)

#34 preemie* or premie or premies:ti,ab,kw (Word variations have been searched)

#35 prematur* near/3 (birth* or born or deliver*):ti,ab,kw (Word variations have been searched)

#36 low near/3 (birthweight* or birth weight*):ti,ab,kw (Word variations have been searched)

#37 lbw or vlbw or elbw:ti,ab,kw (Word variations have been searched)

#38 infan* or baby or babies:ti,ab,kw (Word variations have been searched)

#39 #28 or #29 or #30 or #31 or #32 or #33 or #34 or #35 or #36 or #37 or #38

#40 #27 AND #39

CINAHL Via EBSCO

27 records; 18^th February 2020

S35 S31 AND S34 (27)

S34 S32 OR S33 (616,583)

S33 TX ( (neonat* or neo nat*) ) OR TX ( (newborn* or new born* or newly born*) ) OR TX ( (preterm or preterms or pre term or pre terms) ) OR TX ( (preemie$ or premie or premies) ) OR TX ( (prematur* N3 (birth* or born or deliver*)) ) OR TX ( (low N3 (birthweight* or birth weight*)) ) OR TX ( (lbw or vlbw or elbw) ) OR TX infan* OR TX ( (baby or babies) ) (616,583)

S32 (MH "Infant, Newborn+") (126,178)

S31 S22 AND S30 (107)

S30 S28 not S29 (628,752)

S29 ( MH animals+ OR MH (animal studies) OR TI (animal model*) ) NOT MH (human) (167,644)

S28 S23 OR S24 OR S25 OR S26 OR S27 (657,363)

S27 AB (cluster W3 RCT) (322)

S26 MH placebos OR PT randomized controlled trial OR AB control W5 group OR MH crossover design OR MH comparative studies (401,674)

S25 MH sample size AND AB ( (assigned OR allocated OR control) ) (3,766)

S24 TI ( (randomised OR randomized) ) OR AB random* OR TI trial (337,314)

S23 MH Randomized Controlled Trials OR MH double‐blind studies OR MH single‐blind studies OR MH random assignment OR MH pretest‐posttest design OR MH cluster sample (192,625)

S22 S9 AND S18 AND S21 (240)

S21 S19 OR S20 (366)

S20 TI ( (probiotic* and prebiotic*) N4 combin* ) OR AB ( (probiotic* and prebiotic*) N4 combin* ) (51)

S19 TI Synbiotic* OR AB Synbiotic* (342)

S18 S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 (4,196)

S17 TI Lactoferrin OR AB Lactoferrin (524)

S16 TI fructooligosaccharide* OR AB fructooligosaccharide* OR TI fructo‐oligosaccharide* OR AB fructo‐oligosaccharide* OR TI galactooligosaccharide* OR AB galactooligosaccharide* OR TI galacto‐oligosaccharide* OR AB galacto‐oligosaccharide* (363)

S15 TI Inulin OR AB Inulin (515)

S14 TI lactulose* OR AB lactulose* (481)

S13 TI Oligosaccharides OR AB Oligosaccharides (778)

S12 (MH "Oligosaccharides") (932)

S11 TI Prebiotic* OR AB Prebiotic* (1,270)

S10 (MH "Prebiotics") (1,408)

S9 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 (10,092)

S8 TI Saccharomyces OR AB Saccharomyces (510)

S7 (MH "Saccharomyces") (47)

S6 TI lactobacillus OR AB lactobacillus (2,281)

S5 (MH "Lactobacillus") OR (MH "Lactobacillus Acidophilus") (2,502)

S4 TI bifidobacterium* OR AB bifidobacterium* (875)

S3 (MH "Bifidobacterium") (946)

S2 TI probiotic* OR AB probiotic* (5,016)

S1 MH "Probiotics" (6,611)

Embase Via OVID

Search date 17^th February 2020; 5600 records

Database: Embase <1974 to 2020 February 14>

1 Probiotic Agent/ (34490)

2 probiotic$.ti,ab,kw. (31301)

3 exp bifidobacterium/ (12860)

4 bifidobacterium$.ti,ab,kw. (9740)

5 exp lactobacillus/ (43379)

6 lactobacill$.ti,ab,kw. (38688)

7 Saccharomyces/ or Saccharomyces boulardii/ or Saccharomyces cerevisiae/ (98260)

8 Saccharomyces$.ti,ab,kw. (77090)

9 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 (183648)

10 Prebiotic Agent/ (7387)

11 prebiotic$.ti,ab,kw. (9900)

12 exp Oligosaccharide/ (546080)

13 oligosaccharide$.ti,ab,kw. (37361)

14 Galactose oligosaccharide/ (961)

15 (galacto‐oligosaccharide$ or galactooligosaccharide$).ti,ab,kw. (1364)

16 Fructose Oligosaccharide/ (2182)

17 (fructooligosaccharide$ or fructo‐oligosaccharide$ or FOS or FOSs).ti,ab,kw. (35709)

18 Lactulose/ (8835)

19 lactulose$.ti,ab,kw. (5550)

20 Inulin/ (7321)

21 inulin$.ti,ab,kw. (9557)

22 Lactoferrin/ (10431)

23 lactoferrin$.ti,ab,kw. (9054)

24 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 (617217)

25 Synbiotic Agent/ (1624)

26 synbiotic$.ti,ab,kw. (1737)

27 ((probiotic$ and prebiotic$) adj4 combin$).ti,ab,kw. (411)

28 25 or 26 or 27 (2333)

29 9 or 24 or 28 (778900)

30 Newborn/ (516866)

31 Prematurity/ (99389)

32 (neonat$ or neo nat$).ti,ab. (334397)

33 (newborn$ or new born$ or newly born$).ti,ab. (189575)

34 (preterm or preterms or pre term or pre terms).ti,ab. (102056)

35 (preemie$ or premie or premies).ti,ab. (257)

36 (prematur$ adj3 (birth$ or born or deliver$)).ti,ab. (21105)

37 (low adj3 (birthweight$ or birth weight$)).ti,ab. (42758)

38 (lbw or vlbw or elbw).ti,ab. (11219)

39 infan$.ti,ab. (487240)

40 (baby or babies).ti,ab. (94958)

41 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 (1110575)

42 Randomized controlled trial/ (590055)

43 Controlled clinical study/ (462890)

44 Random$.ti,ab. (1501724)

45 randomization/ (85807)

46 intermethod comparison/ (256520)

47 placebo.ti,ab. (300990)

48 (compare or compared or comparison).ti. (500389)

49 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. (2058845)

50 (open adj label).ti,ab. (76978)

51 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. (228154)

52 double blind procedure/ (169466)

53 parallel group$1.ti,ab. (24938)

54 (crossover or cross over).ti,ab. (103058)

55 ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. (322434)

56 (assigned or allocated).ti,ab. (379281)

57 (controlled adj7 (study or design or trial)).ti,ab. (339741)

58 (volunteer or volunteers).ti,ab. (243065)

59 human experiment/ (484405)

60 trial.ti. (291075)

61 or/42‐60 (4900385)

62 (random$ adj sampl$ adj7 ("cross section$" or questionnaire$1 or survey$ or database$1)).ti,ab. not (comparative study/ or controlled study/ or randomi?ed controlled.ti,ab. or randomly assigned.ti,ab.) (7961)

63 Cross‐sectional study/ not (randomized controlled trial/ or controlled clinical study/ or controlled study/ or randomi?ed controlled.ti,ab. or control group$1.ti,ab.) (228646)

64 (((case adj control$) and random$) not randomi?ed controlled).ti,ab. (16824)

65 (Systematic review not (trial or study)).ti. (135640)

66 (nonrandom$ not random$).ti,ab. (15874)

67 "Random field$".ti,ab. (2243)

68 (random cluster adj3 sampl$).ti,ab. (1253)

69 (review.ab. and review.pt.) not trial.ti. (777162)

70 "we searched".ab. and (review.ti. or review.pt.) (30687)

71 "update review".ab. (103)

72 (databases adj4 searched).ab. (33664)

73 (rat or rats or mouse or mice or swine or porcine or murine or sheep or lambs or pigs or piglets or rabbit or rabbits or cat or cats or dog or dogs or cattle or bovine or monkey or monkeys or trout or marmoset$1).ti. and animal experiment/ (1045069)

74 Animal experiment/ not (human experiment/ or human/) (2213091)

75 or/62‐74 (3395835)

76 61 not 75 (4366247)

77 29 and 41 and 76 (5600)

Maternity & Infant Care Via OVID

Search date 17^th February 2020; Records 94

Database: Maternity & Infant Care Database (MIDIRS) <1971 to December 2019>

1 probiotic$.ti,ab,de. (430)

2 bifidobacterium$.ti,ab,de. (153)

3 lactobacill$.ti,ab,de. (306)

4 Saccharomyces$.ti,ab,de. (12)

5 1 or 2 or 3 or 4 (643)

6 prebiotic$.ti,ab,de. (145)

7 oligosaccharide$.ti,ab,de. (139)

8 inulin$.ti,ab,de. (13)

9 (fructooligosaccharide$ or fructo‐oligosaccharide$ or FOS or FOSs).ti,ab,de. (39)

10 (galactooligosaccharide$ or galacto‐oligosaccharide$).ti,ab,de. (35)

11 lactoferrin$.ti,ab,de. (156)

12 lactulose$.ti,ab,de. (27)

13 6 or 7 or 8 or 9 or 10 or 11 or 12 (413)

14 synbiotic$.ti,ab,de. (27)

15 ((probiotic$ and prebiotic$) adj4 combin$).ti,ab,de. (5)

16 14 or 15 (28)

17 5 or 13 or 16 (932)

18 (neonat$ or neo nat$).ti,ab. (46156)

19 (newborn$ or new born$ or newly born$).ti,ab. (20773)

20 (preterm or preterms or pre term or pre terms).ti,ab. (27396)

21 (preemie$ or premie or premies).ti,ab. (56)

22 (prematur$ adj3 (birth$ or born or deliver$)).ti,ab. (4126)

23 (low adj3 (birthweight$ or birth weight$)).ti,ab. (11086)

24 (lbw or vlbw or elbw).ti,ab. (3170)

25 infan$.ti,ab. (66564)

26 (baby or babies).ti,ab. (29888)

27 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 (123341)

28 17 and 27 (765)

29 limit 28 to randomised controlled trial (94)

MEDLINE Via OVID

Search date 17^th February 2020; Records 2054

Database: Ovid MEDLINE(R) ALL <1946 to February 14, 2020>

1 Probiotics/ (16413)

2 probiotic$.ti,ab,kw. (23385)

3 exp bifidobacterium/ (5805)

4 bifidobacterium$.ti,ab,kw. (7563)

5 exp lactobacillus/ (28003)

6 lactobacill$.ti,ab,kw. (34222)

7 Saccharomyces/ or Saccharomyces boulardii/ or Saccharomyces cerevisiae/ (109184)

8 Saccharomyces$.ti,ab,kw. (72585)

9 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 (183166)

10 Prebiotics/ (2477)

11 prebiotic$.ti,ab,kw. (8040)

12 Oligosaccharides/ (24163)

13 oligosaccharide$.ti,ab,kw. (33210)

14 (galactooligosaccharides or galacto‐oligosaccharides).ti,ab,kw. (859)

15 (fructooligosaccharide$ or fructo‐oligosaccharide$ or FOS or FOSs).ti,ab,kw. (29851)

16 Lactulose/ (2114)

17 lactulose$.ti,ab,kw. (3524)

18 Inulin/ (6862)

19 inulin$.ti,ab,kw. (8603)

20 Lactoferrin/ (5956)

21 lactoferrin$.ti,ab,kw. (7664)

22 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 (100138)

23 Synbiotics/ (525)

24 synbiotic$.ti,ab,kw. (1327)

25 ((probiotic$ and prebiotic$) adj4 combin$).ti,ab,kw. (313)

26 23 or 24 or 25 (1500)

27 9 or 22 or 26 (276802)

28 exp Infant, Newborn/ (599027)

29 Premature Birth/ (13220)

30 (neonat$ or neo nat$).ti,ab. (258480)

31 (newborn$ or new born$ or newly born$).ti,ab. (163361)

32 (preterm or preterms or pre term or pre terms).ti,ab. (72698)

33 (preemie$ or premie or premies).ti,ab. (166)

34 (prematur$ adj3 (birth$ or born or deliver$)).ti,ab. (15366)

35 (low adj3 (birthweight$ or birth weight$)).ti,ab. (33943)

36 (lbw or vlbw or elbw).ti,ab. (8192)

37 infan$.ti,ab. (428676)

38 (baby or babies).ti,ab. (68784)

39 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 (1039559)

40 randomized controlled trial.pt. (500729)

41 controlled clinical trial.pt. (93588)

42 randomized.ab. (470135)

43 placebo.ab. (205251)

44 drug therapy.fs. (2181901)

45 randomly.ab. (327315)

46 trial.ab. (494771)

47 groups.ab. (2009585)

48 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 (4636061)

49 exp animals/ not humans.sh. (4674306)

50 48 not 49 (4016966)

51 27 and 39 and 50 (2054)

Appendix 2. 'Risk of bias' tool

Sequence generation (checking for possible selection bias). Was the allocation sequence adequately generated?

For each included study, we categorised the method used to generate the allocation sequence as:

• low risk (any truly random process e.g. random number table; computer random number generator);

• high risk (any non‐random process e.g. odd or even date of birth; hospital or clinic record number); or

• unclear risk.

Allocation concealment (checking for possible selection bias). Was allocation adequately concealed?

For each included study, we categorised the method used to conceal the allocation sequence as:

• low risk (e.g. telephone or central randomisation; consecutively numbered sealed envelopes);

• high risk (open random allocation; unsealed or non‐opaque envelopes, alternation; date of birth); or

• unclear risk.

Blinding of personnel (checking for possible performance bias). Was knowledge of the allocated intervention adequately prevented during the study?

For each included study, we categorised the methods used to blind study participants and personnel from knowledge of which intervention a participant received. Blinding was assessed separately for different outcomes or class of outcomes. We categorised the methods as:

• low risk, high risk or unclear risk for personnel.

Blinding of outcome assessment (checking for possible detection bias). Was knowledge of the allocated intervention adequately prevented at the time of outcome assessment?

For each included study, we categorised the methods used to blind outcome assessment. Blinding was assessed separately for different outcomes or class of outcomes. We categorised the methods as:

• low risk for outcome assessors;

• high risk for outcome assessors; or

• unclear risk for outcome assessors.

Incomplete outcome data (checking for possible attrition bias through withdrawals, dropouts, protocol deviations). Were incomplete outcome data adequately addressed?

For each included study and for each outcome, we described the completeness of data including attrition and exclusions from the analysis. We noted whether attrition and exclusions were reported, the numbers included in the analysis at each stage (compared with the total randomised participants), reasons for attrition or exclusion where reported, and whether missing data were balanced across groups or were related to outcomes. Where sufficient information was reported or supplied by the trial authors, we re‐included missing data in the analyses. We categorised the methods as:

• low risk (< 20% missing data);

• high risk (≥ 20% missing data); or

• unclear risk.

Selective reporting bias. Are reports of the study free of suggestion of selective outcome reporting?

For each included study, we described how we investigated the possibility of selective outcome reporting bias and what we found. For studies in which study protocols were published in advance, we compared prespecified outcomes versus outcomes eventually reported in the published results. If the study protocol was not published in advance, we contacted study authors to gain access to the study protocol. We assessed the methods as:

• low risk (where it is clear that all of the study's prespecified outcomes and all expected outcomes of interest to the review have been reported);

• high risk (where not all the study's prespecified outcomes have been reported; one or more reported primary outcomes were not prespecified outcomes of interest and are reported incompletely and so cannot be used; study fails to include results of a key outcome that would have been expected to have been reported); or

• unclear risk.

Data and analyses

Comparison 1. Probiotics versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Necrotising enterocolitis	54	10604	Risk Ratio (M‐H, Fixed, 95% CI)	0.54 [0.45, 0.65]
1.1.1 Bifidobacterium spp.	14	2988	Risk Ratio (M‐H, Fixed, 95% CI)	0.72 [0.54, 0.96]
1.1.2 Lactobacillus spp.	12	2000	Risk Ratio (M‐H, Fixed, 95% CI)	0.45 [0.28, 0.71]
1.1.3 Sacchromyces spp.	4	621	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.44, 1.50]
1.1.4 Bacillus spp.	2	465	Risk Ratio (M‐H, Fixed, 95% CI)	0.61 [0.23, 1.61]
1.1.5 Bifidobacterium spp. plus Lactobacillus spp.	11	2041	Risk Ratio (M‐H, Fixed, 95% CI)	0.36 [0.23, 0.59]
1.1.6 Bifidobacterium spp. plus Streptococcus spp.	2	1244	Risk Ratio (M‐H, Fixed, 95% CI)	0.36 [0.19, 0.68]
1.1.7 Bifidobacterium spp. plus Lactobacillus spp. plus Sacchromyces spp.	4	583	Risk Ratio (M‐H, Fixed, 95% CI)	0.67 [0.28, 1.58]
1.1.8 Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp.	5	662	Risk Ratio (M‐H, Fixed, 95% CI)	0.42 [0.22, 0.77]
1.2 Mortality	51	10170	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.65, 0.89]
1.2.1 Bifidobacterium spp.	12	2761	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.58, 1.09]
1.2.2 Lactobacillus spp.	12	2000	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.60, 1.37]
1.2.3 Sacchromyces spp.	3	534	Risk Ratio (M‐H, Fixed, 95% CI)	1.12 [0.46, 2.70]
1.2.4 Bacillus spp.	2	465	Risk Ratio (M‐H, Fixed, 95% CI)	0.87 [0.45, 1.69]
1.2.5 Bifidobacterium spp. plus Lactobacillus spp.	12	2071	Risk Ratio (M‐H, Fixed, 95% CI)	0.60 [0.45, 0.81]
1.2.6 Bifidobacterium spp. plus Streptococcus spp.	2	1244	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.52, 1.35]
1.2.7 Bifidobacterium spp. plus Lactobacillus spp. plus Sacchromyces spp.	4	583	Risk Ratio (M‐H, Fixed, 95% CI)	0.67 [0.30, 1.49]
1.2.8 Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp.	4	512	Risk Ratio (M‐H, Fixed, 95% CI)	0.74 [0.39, 1.42]
1.3 Invasive infection	47	9762	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.82, 0.97]
1.3.1 Bifidobacterium spp.	12	2736	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.70, 1.02]
1.3.2 Lactobacillus spp.	11	1970	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.76, 1.21]
1.3.3 Sacchromyces spp.	4	621	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.58, 1.22]
1.3.4 Bacillus spp.	2	465	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.67, 1.51]
1.3.5 Bifidobacterium spp. plus Lactobacillus spp.	10	1913	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.78, 1.08]
1.3.6 Bifidobacterium spp. plus Streptococcus spp.	2	1244	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.72, 1.17]
1.3.7 Bifidobacterium spp. plus Lactobacillus spp. plus Sacchromyces spp.	4	583	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.53, 1.18]
1.3.8 Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp.	2	230	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.63, 1.00]
1.4 Duration of birth hospitalisation (days)	22	5458	Mean Difference (IV, Random, 95% CI)	‐1.93 [‐3.78, ‐0.08]
1.4.1 Bifidobacterium spp.	4	1945	Mean Difference (IV, Random, 95% CI)	‐1.05 [‐6.55, 4.45]
1.4.2 Lactobacillus spp.	4	217	Mean Difference (IV, Random, 95% CI)	‐1.95 [‐10.81, 6.90]
1.4.3 Sacchromyces spp.	2	470	Mean Difference (IV, Random, 95% CI)	‐2.88 [‐8.06, 2.29]
1.4.4 Bifidobacterium spp. plus Lactobacillus spp.	7	1265	Mean Difference (IV, Random, 95% CI)	‐1.74 [‐5.22, 1.73]
1.4.5 Bifidobacterium spp. plus Streptococcus spp.	1	1044	Mean Difference (IV, Random, 95% CI)	‐3.00 [‐6.28, 0.28]
1.4.6 Bifidobacterium spp. plus Lactobacillus spp. plus Sacchromyces spp.	2	231	Mean Difference (IV, Random, 95% CI)	‐5.65 [‐11.68, 0.38]
1.4.7 Bifidobacterium spp. plus Lactobacillus spp. plus Streptococcus spp.	2	286	Mean Difference (IV, Random, 95% CI)	1.69 [‐6.73, 10.11]
1.5 Severe neurodevelopmental impairment	5	1518	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.84, 1.26]
1.5.1 Bifidobacterium spp.	1	162	Risk Ratio (M‐H, Fixed, 95% CI)	0.77 [0.34, 1.72]
1.5.2 Lactobacillus spp.	1	249	Risk Ratio (M‐H, Fixed, 95% CI)	1.01 [0.69, 1.48]
1.5.3 Bacillus spp.	1	174	Risk Ratio (M‐H, Fixed, 95% CI)	1.09 [0.58, 2.07]
1.5.4 Bifidobacterium spp. plus Streptococcus spp.	1	664	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.69, 1.36]
1.5.5 Bifidobacterium spp. plus Lactobacillus spp.	1	269	Risk Ratio (M‐H, Fixed, 95% CI)	1.27 [0.81, 1.98]
1.6 Cerebral palsy	5	1512	Risk Ratio (M‐H, Fixed, 95% CI)	1.13 [0.74, 1.72]
1.6.1 Bifidobacterium spp.	1	156	Risk Ratio (M‐H, Fixed, 95% CI)	0.38 [0.10, 1.36]
1.6.2 Lactobacillus spp.	1	249	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.40, 2.08]
1.6.3 Bacillus spp.	1	174	Risk Ratio (M‐H, Fixed, 95% CI)	2.05 [0.38, 10.88]
1.6.4 Bifidobacterium spp. plus Streptococcus spp.	1	664	Risk Ratio (M‐H, Fixed, 95% CI)	1.32 [0.67, 2.58]
1.6.5 Bifidobacterium spp. plus Lactobacillus spp.	1	269	Risk Ratio (M‐H, Fixed, 95% CI)	2.28 [0.62, 8.41]
1.7 Visual impairment	4	1356	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.14, 1.80]
1.7.1 Bifidobacterium spp.	1	174	Risk Ratio (M‐H, Fixed, 95% CI)	0.51 [0.05, 5.54]
1.7.2 Lactobacillus spp.	1	249	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
1.7.3 Bifidobacterium spp. plus Streptococcus spp.	1	664	Risk Ratio (M‐H, Fixed, 95% CI)	2.91 [0.12, 71.21]
1.7.4 Bifidobacterium spp. plus Lactobacillus spp.	1	269	Risk Ratio (M‐H, Fixed, 95% CI)	0.21 [0.02, 1.89]
1.8 Hearing impairment	4	1356	Risk Ratio (M‐H, Fixed, 95% CI)	0.46 [0.18, 1.17]
1.8.1 Bifidobacterium spp.	1	174	Risk Ratio (M‐H, Fixed, 95% CI)	1.02 [0.07, 16.10]
1.8.2 Lactobacillus spp.	1	249	Risk Ratio (M‐H, Fixed, 95% CI)	3.02 [0.12, 73.52]
1.8.3 Bifidobacterium spp. plus Streptococcus spp.	1	664	Risk Ratio (M‐H, Fixed, 95% CI)	0.18 [0.04, 0.79]
1.8.4 Bifidobacterium spp. plus Lactobacillus spp.	1	269	Risk Ratio (M‐H, Fixed, 95% CI)	1.71 [0.16, 18.64]
1.9 Continuous early learning composite measure	1	52	Mean Difference (IV, Fixed, 95% CI)	‐1.00 [‐6.38, 4.38]

Comparison 2. Probiotics versus control (extremely preterm or ELBW).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Necrotising enterocolitis	8	1712	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.68, 1.21]
2.1.1 Bifidobacterium spp.	2	665	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.70, 1.43]
2.1.2 Lactobacillus spp.	2	330	Risk Ratio (M‐H, Fixed, 95% CI)	0.73 [0.36, 1.48]
2.1.3 Bacillus spp.	1	120	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
2.1.4 Bifidobacterium spp. plus Lactobacillus spp.	2	123	Risk Ratio (M‐H, Fixed, 95% CI)	1.01 [0.21, 4.79]
2.1.5 Bifidobacterium spp. plus Streptococcus spp.	1	474	Risk Ratio (M‐H, Fixed, 95% CI)	0.73 [0.33, 1.60]
2.2 Mortality	6	1661	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.71, 1.16]
2.2.1 Bifidobacterium spp.	1	474	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.60, 1.61]
2.2.2 Lactobacillus spp.	2	330	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.42, 1.42]
2.2.3 Bacillus clausii	1	120	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.36, 2.08]
2.2.4 Bifidobacterium spp. plus Lactobacillus spp.	1	100	Risk Ratio (M‐H, Fixed, 95% CI)	0.75 [0.18, 3.18]
2.2.5 Bifidobacterium spp. plus Streptococcus spp.	1	637	Risk Ratio (M‐H, Fixed, 95% CI)	0.94 [0.65, 1.35]
2.3 Invasive infection	6	1471	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.76, 1.06]
2.3.1 Bifidobacterium spp.	2	642	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.73, 1.37]
2.3.2 Lactobacillus spp.	1	134	Risk Ratio (M‐H, Fixed, 95% CI)	1.05 [0.67, 1.66]
2.3.3 Bacillus clausii	1	120	Risk Ratio (M‐H, Fixed, 95% CI)	0.80 [0.43, 1.47]
2.3.4 Bifidobacterium spp. plus Lactobacillus spp.	1	101	Risk Ratio (M‐H, Fixed, 95% CI)	0.83 [0.45, 1.54]
2.3.5 Bifidobacterium spp. plus Streptococcus spp.	1	474	Risk Ratio (M‐H, Fixed, 95% CI)	0.82 [0.64, 1.06]
2.4 Duration of birth hospitalisation (days)	1	22	Mean Difference (IV, Random, 95% CI)	‐5.40 [‐14.20, 3.40]
2.4.1 Bifidobacterium spp. plus Lactobacillus spp.	1	22	Mean Difference (IV, Random, 95% CI)	‐5.40 [‐14.20, 3.40]

2.4. Analysis.

Comparison 2: Probiotics versus control (extremely preterm or ELBW), Outcome 4: Duration of birth hospitalisation (days)

Comparison 3. Subgroup analysis by type of feeding.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Necrotising enterocolitis	54	10604	Risk Ratio (M‐H, Fixed, 95% CI)	0.54 [0.45, 0.65]
3.1.1 Human milk only	8	986	Risk Ratio (M‐H, Fixed, 95% CI)	0.30 [0.16, 0.57]
3.1.2 Mixed‐ human milk or formula or both	42	9364	Risk Ratio (M‐H, Fixed, 95% CI)	0.58 [0.48, 0.70]
3.1.3 Formula only	4	254	Risk Ratio (M‐H, Fixed, 95% CI)	0.43 [0.16, 1.18]
3.2 Mortality	51	10271	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.65, 0.89]
3.2.1 Human milk only	8	986	Risk Ratio (M‐H, Fixed, 95% CI)	0.64 [0.41, 1.00]
3.2.2 Mixed‐ human milk or formula or both	40	9118	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.67, 0.94]
3.2.3 Formula only	3	167	Risk Ratio (M‐H, Fixed, 95% CI)	0.22 [0.04, 1.21]
3.3 Invasive infection	47	9762	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.82, 0.97]
3.3.1 Human milk only	8	986	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.59, 0.96]
3.3.2 Mixed‐ human milk or formula or both	36	8552	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.83, 1.00]
3.3.3 Formula only	3	224	Risk Ratio (M‐H, Fixed, 95% CI)	0.41 [0.11, 1.49]
3.4 Duration of birth hospitalisation (days)	22	5458	Mean Difference (IV, Random, 95% CI)	‐1.93 [‐3.78, ‐0.08]
3.4.1 Human milk only	4	366	Mean Difference (IV, Random, 95% CI)	‐3.95 [‐7.70, ‐0.21]
3.4.2 Mixed‐ human milk or formualor both	16	5002	Mean Difference (IV, Random, 95% CI)	‐1.00 [‐2.84, 0.85]
3.4.3 Formula only	2	90	Mean Difference (IV, Random, 95% CI)	1.50 [‐26.33, 29.32]

Comparison 4. Sensitivity analyses: Risk of bias.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Necrotising enterocolitis	54	10800	Risk Ratio (M‐H, Fixed, 95% CI)	0.55 [0.46, 0.65]
4.1.1 LOW	16	4597	Risk Ratio (M‐H, Fixed, 95% CI)	0.70 [0.55, 0.89]
4.1.2 UNCLEAR	20	3905	Risk Ratio (M‐H, Fixed, 95% CI)	0.43 [0.31, 0.59]
4.1.3 HIGH	18	2298	Risk Ratio (M‐H, Fixed, 95% CI)	0.41 [0.26, 0.63]
4.2 Mortality	51	10170	Risk Ratio (M‐H, Fixed, 95% CI)	0.76 [0.65, 0.89]
4.2.1 LOW	16	4597	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.69, 1.07]
4.2.2 UNCLEAR	19	3818	Risk Ratio (M‐H, Fixed, 95% CI)	0.71 [0.54, 0.94]
4.2.3 HIGH	16	1755	Risk Ratio (M‐H, Fixed, 95% CI)	0.57 [0.38, 0.85]
4.3 Invasive infection	47	9762	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.82, 0.97]
4.3.1 LOW	16	4597	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.79, 1.02]
4.3.2 UNCLEAR	18	3700	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.84, 1.10]
4.3.3 HIGH	13	1465	Risk Ratio (M‐H, Fixed, 95% CI)	0.73 [0.59, 0.90]
4.4 Duration of birth hospitalisation (days)	22	5458	Mean Difference (IV, Random, 95% CI)	‐1.93 [‐3.78, ‐0.08]
4.4.1 LOW	6	2786	Mean Difference (IV, Random, 95% CI)	‐2.24 [‐5.76, 1.29]
4.4.2 UNCLEAR	8	1675	Mean Difference (IV, Random, 95% CI)	‐0.99 [‐4.07, 2.10]
4.4.3 HIGH	8	997	Mean Difference (IV, Random, 95% CI)	‐3.92 [‐7.91, 0.07]
4.5 Severe neurodevelopmental impairment	5	1518	Risk Ratio (M‐H, Fixed, 95% CI)	1.03 [0.84, 1.26]
4.5.1 LOW	2	913	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.76, 1.27]
4.5.2 UNCLEAR	3	605	Risk Ratio (M‐H, Fixed, 95% CI)	1.11 [0.79, 1.54]
4.6 Cerebral palsy	5	1512	Risk Ratio (M‐H, Fixed, 95% CI)	1.13 [0.74, 1.72]
4.6.1 LOW	2	913	Risk Ratio (M‐H, Fixed, 95% CI)	1.14 [0.68, 1.92]
4.6.2 UNCLEAR	3	599	Risk Ratio (M‐H, Fixed, 95% CI)	1.09 [0.52, 2.28]
4.7 Visual impairment	4	1356	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.14, 1.80]
4.7.1 LOW	2	913	Risk Ratio (M‐H, Fixed, 95% CI)	2.91 [0.12, 71.21]
4.7.2 UNCLEAR	2	443	Risk Ratio (M‐H, Fixed, 95% CI)	0.31 [0.06, 1.49]
4.8 Hearing impairment	4	1356	Risk Ratio (M‐H, Fixed, 95% CI)	0.46 [0.18, 1.17]
4.8.1 LOW	2	913	Risk Ratio (M‐H, Fixed, 95% CI)	0.30 [0.09, 0.98]
4.8.2 UNCLEAR	2	443	Risk Ratio (M‐H, Fixed, 95% CI)	1.38 [0.23, 8.29]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Agarwal 2003.

Study characteristics
Methods	RCT
Participants	39 VLBW infants
Interventions	Probiotics (N = 24): Lactobacillus rhamnosus GG once daily with human milk or formula for 21 days or discharge from hospital Control (N = 15): unsupplemented milk feeds
Outcomes	Stool colonisation patterns (NEC, death, infection not reported)
Notes	India (1999 to 2000) Funding: UK National Institute for Health (Fogarty Grant TW‐00601) and Conagra Foods Inc., USA (supplied intervention)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Unlikely
Selective reporting (reporting bias)	Unclear risk	No clinical outcomes reported

Al‐Hosni 2012.

Study characteristics
Methods	RCT
Participants	101 ELBW infants (appropriate for gestational age)
Interventions	Probiotic (N = 50): Lactobacillus rhamnosus GG (LGG) and Bifidobacterium infantis added to the 1st milk feed and continued once daily until discharge or until 34 weeks' postmenstrual age Control (N = 51): unsupplemented milk feeds
Outcomes	Weight gain NEC Death Infection
Notes	USA (2009 to 2011) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Bin‐Nun 2005.

Study characteristics
Methods	RCT
Participants	145 VLBW infants
Interventions	Probiotics (N = 72): "Lactobacillus bifidus" (likely Bifidobactrium bifidum), Streptococcus thermophilus, and B. infantis added to expressed breast milk or formula enteral feeds daily until 36 weeks' postmenstrual age Control (N = 73): unsupplemented milk feeds
Outcomes	NEC Death Infection Time to full enteral feeds
Notes	Israel (2001 to 2004) Funding: Solgar, Wyeth (manufacturer of intervention)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Data published in an abstract form on two previous occasions at the Society of Pediatrics Research (SPR 2003, 2005) with different inclusion criteria and clinical outcomes

Braga 2011.

Study characteristics
Methods	RCT
Participants	231 VLBW infants (birth weight 750 g to 1500 g)
Interventions	Probiotics (N = 119): Lactobacillus casei and Bifidobacterium breve (Yakult ‐ LB) in human milk once daily until day 30 or hospital discharge Control (N = 112): unsupplemented milk feeds
Outcomes	NEC Death Infection Days to full enteral feeds Duration hospital stay
Notes	Brazil (2007 to 2008) Funding: public/state. External Study Committee terminated trial early (quote:) "for a clear benefit" after enrolment of 231 infants
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Sealed envelope with group allocation
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Chandrashekar 2018.

Study characteristics
Methods	Quasi‐RCT
Participants	145 preterm infants of gestation < 34 weeks' (most participants were very preterm or VLBW)
Interventions	Probitics (N = 72): Lactobacillus acidophilus, L. rhamnosus, Bifidobacterium longum, and Saccharomyces boulardii with human milk or formula feeds until discharge from hospital Control (N = 73): unsupplemented milk feeds (no placebo)
Outcomes	NEC Death Infection Duration of hospitalisation
Notes	India (2014 to 2015) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "Simple random sampling method"
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (5 participants withdrawn pre‐analysis)
Selective reporting (reporting bias)	Low risk	Unlikely

Chowdhury 2016.

Study characteristics
Methods	RCT
Participants	119 VLBW Infants (28 to 33 weeks' gestation)
Interventions	Probiotics (N = 60): (quote:) "Cap TS6" containing Lactobacillus rhamnosus GG, L. paracasei , L. casei, L. acidophilus, Lactococcus latis, Bifidobacterium bifidum, B. longum, B. infantis) in human milk once daily until discharge Control (N = 59): unsupplemented milk feeds
Outcomes	NEC Death Infection days to achieve full enteral feeding Length of hospital stay
Notes	Bangladesh (2012 to 2015) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	First allocation by lottery, and subsequent by alternate allocation
Allocation concealment (selection bias)	High risk	Unconcealed
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Chrzanowska‐Liszewska 2012.

Study characteristics
Methods	RCT
Participants	47 very preterm infants (birth weight > 1000 g)
Interventions	Probiotics (N = 21): Lactobacillus rhamnosus GG, added to formula, once daily until day 42 Control (N = 26): maltodextrin placebo added to formula
Outcomes	Microflora of stool measured on day 7, 21, and 42 NEC Death Infection (courtesy of investigators)
Notes	Poland (2008 to 2009) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Coded capsules containing probiotics or placebo
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Costalos 2003.

Study characteristics
Methods	RCT
Participants	87 formula‐fed infants of gestational age at birth 28 to 32 weeks.
Interventions	Probiotics (N = 51): Saccharomyces boulardii added to formula every 12 hours during the 1st week of life when enteral feed are tolerated for 30 days Control (N = 36): maltodextrin placebo
Outcomes	NEC Death Infection Weight gain
Notes	Greece (period of study: not specified) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Low risk	Cards with allocation in sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (5 infants with incomplete data were not included in analyses)
Selective reporting (reporting bias)	Low risk	Unlikely

Costeloe 2015.

Study characteristics
Methods	RCT
Participants	1310 infants born before 31 weeks' gestation
Interventions	Probiotics (N = 650): Bifidobacterium breve BBG‐001 once daily until 36 weeks’ postmenstrual age or discharge from hospital Control (N = 660): corn starch placebo
Outcomes	NEC Death Infection
Notes	UK (24 neonatal units; 2010 to 2013) Funding: by UK National Institute for Health Research Health Technology Assessment programme (ISRCTN 05511098)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Web‐based
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	No

Dani 2002.

Study characteristics
Methods	RCT
Participants	585 VLBW infants (or < 33 weeks' gestation at birth)
Interventions	Probiotics (N = 295): Lactobacillus rhamnosus GG added to milk (human or formula) feeds once daily until hospital discharge Control (N = 290): maltodextrin placebo
Outcomes	NEC Death Infection Duration hospitalisation
Notes	Italy (12 centres; study period not specified) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Low risk	Sealed envelope containing allocation
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Dashti 2014.

Study characteristics
Methods	RCT
Participants	136 preterm infants of birth weight 700 g to 1800 g (most participants very preterm or VLBW)
Interventions	Probiotics (N = 69): Lactobacillus acidophilus, L. rhamnosus, L. bulgaricus, L. casei, Streptococcus thermophilus, Bifidobacterium longum, B. breve added to milk feeds once daily until hospital discharge Control (N = 67): placebo powder (not described)
Outcomes	NEC Death (infection data sought from investigators July 2020)
Notes	Iran (2010 to 2011) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Demirel 2013.

Study characteristics
Methods	RCT
Participants	271 VLBW infants (gestational age ≤ 32 weeks at birth)
Interventions	Probiotics (N = 135): Saccharomyces boulardii added to human milk or formula once a day, starting with the 1st feed, until hospital discharge Control (N = 136): unsupplemented milk
Outcomes	NEC Death Infection
Notes	Turkey (2011) Funding: not stated ClinicalTrials.gov Identifier: NCT01315821
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Allocations sealed in opaque, sequentially‐numbered envelopes
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Dilli 2015.

Study characteristics
Methods	RCT
Participants	200 very preterm or VLBW infants
Interventions	Probiotics (N = 100): Bifidobacterium lactis added to human milk or formula once daily for 8 weeks (or hospital discharge) Control (N= 100): maltodextrin powder placebo
Outcomes	NEC Death Infection Length of hospital stay
Notes	Turkey (5 centres: 2011 to 2014) Funding: not stated NB. This was a 4‐arm RCT‐ 2 other groups were prebiotic (N = 100) and synbiotic (n + 100)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Sealed opaque envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Dutta 2015.

Study characteristics
Methods	RCT
Participants	149 infants (27 to 33 weeks' gestation at birth)
Interventions	Probiotics (N = 114): Lactobacillus acidophilus, L. rhamnosus, Bifidobacterium longum, and Saccharomyces boulardii (3 groups: (quote:) "low‐dose" (109) for 21 days or quote:) "high‐dose" (1010) 2 times daily with human milk or formula feeds for 14 or 21 days Control (N = 35): maltodextrin placebo for 21 days
Outcomes	Probiotic stool colonisation NEC Mortality Infection
Notes	India (study period not stated) Funding: Aristo Pharmaceuticals Pvt Ltd, Madhya Pradesh, India provided the sachets of probiotics and placebo free of cost
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Fernández‐Carrocera 2013.

Study characteristics
Methods	RCT
Participants	150 VLBW infants
Interventions	Probiotics (N = 75): Lactobacillus rhamnosus, L. casei, L. plantarum, L acidophilus, Bifidobacteruim infantis, and Streptococcus thermophilus added to human milk or formula (duration intervention not stated) Control (N = 75): unsupplemented milk feeds
Outcomes	NEC Death Infection
Notes	Mexico (2007 to 2010) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random number table
Allocation concealment (selection bias)	Low risk	Staff unable to predict allocation by number
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Fujii 2006.

Study characteristics
Methods	Quasi‐RCT
Participants	19 preterm infants (most very preterm or VLBW)
Interventions	Probiotics group (N = 11): Bifidobacterium breve 2 times daily with human milk or formula feeds until hospital discharge Control (N = 8): unsupplemented milk feeds
Outcomes	Cytokine levels in plasma NEC Death Infection
Notes	Japan (2000 to 2002) Published: 2004 Funding: Morinaja Milk industry and Meiji Dairies (manufactured intervention)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Complete
Selective reporting (reporting bias)	Unclear risk	Unclear

Hariharan 2016.

Study characteristics
Methods	RCT
Participants	196 very preterm infants with birth weight < 1250 g
Interventions	Probiotics (N = 93): Lactobacillus acidophilus, Bifidobacterium bifidum, Saccharomyces boulardii 2 times daily in milk feeds for 6 weeks Control (N = 103): unsupplemented feeds
Outcomes	NEC Death Infection
Notes	India (study period not stated) Funding: Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Unclear

Hays 2015.

Study characteristics
Methods	RCT
Participants	199 very preterm infants (gestation at birth 25 to 31 weeks), and birth weight 700 g to 1600 g that was appropriate for gestational age
Interventions	Probiotics (3 groups: N = 145): Bifidobacterium lactis, or B. longum, or both once daily in sterile water for 4 to 6 weeks (depending on gestation at birth) Control (N = 52): maltodextrin placebo
Outcomes	NEC Death Infection
Notes	France (three centres: 2007 to 2010) Funding: Nestle France (Marne‐la‐Vallee, France) and Nestec (Vevey, Switzerland)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Consecutively numbered, sealed, opaque envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Hernandez‐Enriquez 2016.

Study characteristics
Methods	RCT
Participants	44 preterm infants < 34 weeks' gestation or ≤ 1550 g birth weight (most infants very preterm or VLBW)
Interventions	Intervention (N = 24): Lactobacillus reuteri once daily for 1st 10 days after birth Control (N = 20): placebo (sterile water)
Outcomes	NEC Death Infection (data courtesy of investigators)
Notes	Mexico (2012 to 2013) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Simple randomisation sequence"
Allocation concealment (selection bias)	Low risk	Seaed opaque envelopes
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Complete
Selective reporting (reporting bias)	Unclear risk	Unlikely

Hikaru 2010.

Study characteristics
Methods	RCT
Participants	208 VLBW infants
Interventions	Probiotics (N = 108): Bifidobacterium breve in human milk or formula once daily until discharge from the intensive care unit Control (N = 100): unsupplemented milk feeds
Outcomes	Infection. (NEC not reported)
Notes	Japan (2001 to 2013) Funding: Morinaga Milk Industry Co. Ltd. (supplied Bifidobacterium breve preparation)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Unclear

Huang 2009.

Study characteristics
Methods	RCT
Participants	183 VLBW infants who survived 7 days after birth and began enteral feeding
Interventions	Probiotics (N = 95): Bifidobacterium adolescentis twice daily with milk feeds daily for 7 days Control (N = 88): unsupplemented milk feeds
Outcomes	NEC (unclear whether death or infection assessed)
Notes	China (single centre, study dates not stated) Translation from Chinese courtesy of Yuan Chi
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Unable to assess
Selective reporting (reporting bias)	Unclear risk	Mortality and infection not reported

Indrio 2017.

Study characteristics
Methods	RCT
Participants	60 preterm infants of gestational age 28 to 32 weeks' at birth
Interventions	Probiotics (N = 30): Lactobacillus reuteri DSM 17938 suspended in sunflower and medium‐chain triglyceride oils, given once daily until day 30 Control (N = 30): identical oils without probiotics
Outcomes	NEC Death Infection Duration of hospital stay (data courtesy of personal communication from investigators)
Notes	Italy (2011 to 2012) Funding: University of Bari, Italy ClinicalTrials.gov no. NCT00985816
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Unlikely

Jacobs 2013.

Study characteristics
Methods	RCT
Participants	1099 very preterm VLBW infants
Interventions	Probiotics (N = 548): Bifidobacterium infantis, Streptococcus thermophilus and B. lactis once daily in human milk or formula until discharge from hospital or term corrected age. Control (N = 551): maltodextrin powder placebo
Outcomes	NEC Death Infection Infection with a probiotic species Duration of birth hospitalisation Major neurodevelopmental impairment comprised any of: moderate or severe cerebral palsy, Bayley‐III Motor Composite Scale < –2SD (or Movement Assessment Battery for Children < 15th centile if > 42 months’ post‐term), Bayley‐III Composite Cognitive or Language Scales <–2 SD (or Wechsler Preschool and Primary Scale of Intelligence Full Scale Intelligence Quotient <–2 SD if > 42 months’ post‐term), blindness or deafness
Notes	Australasia (10 centres; 2007 to 2011) Funding: National Health and Research Medical Council, Australia
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Central allocation
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete for in hospital outcomes (Neurodevelopmental assessment = 48%)
Selective reporting (reporting bias)	Low risk	Unlikely

Kanic 2015.

Study characteristics
Methods	RCT
Participants	80 VLBW infants
Interventions	Probiotics (N = 40): Lactobacillus acidophilus, Enterococcus faecium, Bifidobacterium infantis 2 times daily with milk feeds until discharge from hospital Control (N = 40): unsupplemented milk feeds
Outcomes	NEC Death Infection Duration of birth hospitalisation
Notes	Slovenia (2008 to 2011) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Alternate allocation (quote: "quasi‐randomised")
Allocation concealment (selection bias)	High risk	Unconcealed
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Kitajima 1997.

Study characteristics
Methods	RCT
Participants	91 VLBW infants
Interventions	Probiotics (N = 45): Bifidobacterium breve in distilled water once daily for 28 days Control (N = 46): distilled water
Outcomes	Probiotic colonisation of stool (NEC, death, infection‐ data courtesy of investigators)
Notes	Japan (1990 to 1991) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (4 participants not included in analyses)
Selective reporting (reporting bias)	Low risk	Data

Li 2019.

Study characteristics
Methods	RCT
Participants	30 VLBW infants
Interventions	Probiotics (N = 16): Lactobacillus plantarum, Bifidobacterium longum, B. bifidum once daily with milk feeds until 36 weeks' postmenstrual age. Control (N = 14): 5% glucose solution
Outcomes	Change of gut microbiota Correlation of gut microbial composition Levels of cytokines (NEC, death, infection not reported (author contacted in May 2020))
Notes	China (2014 to 2015) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Quote: "Concealed by the principal investigator according to sequential numbers"
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (intervention and control solutions identical)
Incomplete outcome data (attrition bias) All outcomes	High risk	> 50% outcome data unreported
Selective reporting (reporting bias)	Unclear risk	Unable to determine

Lin 2005.

Study characteristics
Methods	RCT
Participants	367 VLBW infants
Interventions	Probiotics (N = 180): Lactobacillus acidophilus and Bifidobacterium. infantis (Infloran®) 2 times daily with human milk until discharge from hospital Control (N = 187): unsupplemented milk feeds (no placebo)
Outcomes	NEC Death Infection Duration of hospitalisation Neurodevelopmental impairment at aged 3 years, defined as 1 or more of: BSID‐II MDI < 70, PDI < 70, bilateral blindness, bilateral hearing impairment requiring amplification, or moderate or severe cerebral palsy (requiring ambulatory assistance)
Notes	Taiwan (1999 to 2003) Funding: Research Department of China Medical University Hospital
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random‐number table
Allocation concealment (selection bias)	Low risk	Opaque, sequentially numbered, sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled (investigators aware of allocation)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (90% for neurodevelopmental assessments)
Selective reporting (reporting bias)	Low risk	Unlikely

Lin 2008.

Study characteristics
Methods	RCT
Participants	434 VLBW infants
Interventions	Probiotics (N = 217): Bifidobacterium bifidum and Lactobacillus acidophilus, added to human milk or formula 2 times daily for 6 weeks Control (N = 217): unsupplemented milk feeds
Outcomes	NEC Death Infection
Notes	Taiwan (7 centres: 2005 to 2007) Funding: National Science Council of Taiwan ClinicalTrials.gov Identifier: NCT00540033
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Allocated centrally
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Manzoni 2006.

Study characteristics
Methods	RCT
Participants	80 VLBW infants
Interventions	Probiotics (N = 39): Lactoacillus casei subspecies rhamnosus with human milk until 6 weeks or hospital discharge Control (N = 41): unsupplemented milk feeds
Outcomes	NEC Death Infection
Notes	Italy (2004 to 2005) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not described.
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Manzoni 2009.

Study characteristics
Methods	RCT
Participants	485 VLBW infants
Interventions	Probiotics (N = 238): Lactoacillus casei subspecies rhamnosus with human milk or formula until 4 (VLBW) or 6 (ELBW) weeks plus bovine lactoferrin (100 mg/day) Control (N = 247): bovine lactoferrin alone (All doses including placebo were diluted in prepared milk so as to maintain masking)
Outcomes	NEC Death Infection
Notes	Italy (11 centres: 2007 to 2008) Funding: Dicofarm SpA (manufacturer of intervention)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Pharmacy allocation (remote)
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Data for invasive infection in complete cohort not reported in primary publication (available to derive from later publications)

Mihatsch 2010.

Study characteristics
Methods	RCT
Participants	180 VLBW infants (< 30 weeks' gestation)
Interventions	Probiotics (N = 91): Bifidobacterium lactis BB12 mixed with powdered fortifier in human milk or formula once daily for 6 weeks Control (N = 89): powdered fortifier placebo
Outcomes	NEC Death Infection
Notes	Germany (2000 to 2003 ) Funding: Nestlé AG, Frankfurt, Germany
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Millar 1993.

Study characteristics
Methods	RCT
Participants	20 infants < 33 weeks' gestation (most participants very preterm or VLBW)
Interventions	Probiotics (N = 10): Lactobacillus rhamnosus GG mixed with human milk or formula 2 times daily for 14 days, starting with 1st feed Control (N = 10): unsupplemented milk feeds
Outcomes	Stool colonisation Invasive infection (NEC, death (courtesy of investigators))
Notes	UK (1991 to 1992) Funding: Wessex Medical Trust
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Mohan 2006.

Study characteristics
Methods	RCT
Participants	69 preterm infants (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 37): Bifidobacterium lactis in milk feeds from 1st day after birth for 21 days Control (N = 32): unsupplemented milk feeds
Outcomes	No clinical outcomes were presented in the published data (NEC, death, infection (courtesy of investigators))
Notes	Germany (2003 to 2005) Funding: Nestlé, Konolfingen, Switzerland
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Central allocation (web‐based)
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Oncel 2014.

Study characteristics
Methods	RCT
Participants	424 VLBW infants (and gestational age ≤ 32 weeks' at birth)
Interventions	Probiotics (N = 213) Lactobacillus reuteri DSM 17938 once daily with milk feeds until discharge from hospital Placebo (N = 211): placebo containing only oil base
Outcomes	NEC Death Infection Culture‐proven infection with L reuteri (duration of hospitalisation‐ presented as median/range) Neurodevelopmental impairment at 18 to 24 months, defined as 1 or more of: BSID‐II MDI < 70, PDI < 70, moderate‐to‐severe cerebral palsy, bilateral hearing impairment, or bilateral blindness
Notes	Turkey (2012 to 2013) Funding: not stated ClinicalTrials.gov Identifier: NCT01531179
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Opaque, sequentially numbered sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (8 participants withdrawn by family) for in hospital outcomes (Neurodevelopmental assessment = 68%)
Selective reporting (reporting bias)	Low risk	Unlikely

Oshiro 2019.

Study characteristics
Methods	RCT
Participants	35 VLBW infants
Interventions	Probiotics (N = 17): Bifidobacterium breve BBG‐01 in human milk feeds once daily during the hospital stay Control (N = 18): placebo
Outcomes	NEC Death Infection Weight gain
Notes	Japan (2015 to 2017) Funding: Yakult Honsha Company, Japan (manufacturer of intervention) Additional data via personal communication: Dr Yuichiro Yamashiro UMIN Registration No. UMIN000005412
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Sealed, opaque envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (probiotic added to milk by dieticians who were not involved in the care of the infant)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Masked
Selective reporting (reporting bias)	Low risk	Unlikely

Patole 2014.

Study characteristics
Methods	RCT
Participants	159 VLBW infants (< 33 weeks' gestation at birth)
Interventions	Probiotics (N = 79): Bifidobacterium breve M‐16V in milk feeds once daily until term equivalent Control (N = 80): maltodextrin placebo
Outcomes	Probiotic colonisation of stool NEC, death, infection, blood culture‐positive sepsis by B. breve M‐16V (neurodevelopmental outcomes‐ Agrawal 2020)
Notes	Australia (2009 to 2012) Funding: Morinaga Milk Industry Company, Japan supplied the product free for the trial
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Opaque, sealed, coded envelopes
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (6 infants withdrawn)
Selective reporting (reporting bias)	Low risk	Unlikely

Rehman 2018.

Study characteristics
Methods	RCT
Participants	146 VLBW preterm infants (gestational age at birth > 26 weeks')
Interventions	Probiotics (N = 70): Bifidobacterium spp (not specified), Lactobacilli acidophilhis, Streptococcus thermophilus, L. delbrueckii with human milk or formula 2 times daily until hospital discharge Control (N = 70): unsupplemented milk feeds
Outcomes	NEC Death (data courtesy of investigators)
Notes	Pakistan (2014) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Infection not reported

Ren 2010.

Study characteristics
Methods	RCT
Participants	150 preterm infants (most participants were very preterm)
Interventions	Probiotics (N = 79): Bifidobacterium infantis, Lactobacillus acidophilus, Bacillus cereus, and Enterococcus faecalis in milk feeds twice daily from day 7 after birth for 7 days (route translated as "oral or nasal"‐ presumed to refer to oro‐gastric or naso‐gastric tube) Control (N = 80): unsupplemented milk feeds
Outcomes	NEC
Notes	China (single centre, 2006‐2008) Translation from Chinese courtesy of Yuan Chi
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Drawing lots"
Allocation concealment (selection bias)	Unclear risk	Safeguards unclear
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Unable to assess
Selective reporting (reporting bias)	Unclear risk	Mortality and infection not reported

Reuman 1986.

Study characteristics
Methods	Quasi‐RCT
Participants	30 very preterm infants (birth weight < 2000 g)
Interventions	Probiotics (N = 15): Lactobacillus acidophilus in formula daily for 28 days Control (N = 15): unsupplemented formula feeds
Outcomes	Stool colonisation NEC Death Duration of hospitalisation Rate of weight gain
Notes	US (early 1980s) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Random number charts and the last digit of patient's chart number, then alternate allocation of next participant
Allocation concealment (selection bias)	High risk	Unconcealed
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Unclear risk	Infection not reported

Rougé 2009.

Study characteristics
Methods	RCT
Participants	94 very preterm or VLBW infants
Interventions	Probiotics (N = 45): Lactobacillus rhamnosus GG and Bifidobacterium longum with human milk or formula once daily until discharge from hospital Control (N = 49): maltodextrin placebo
Outcomes	NEC Death Infection Duration of hospital stay
Notes	France (2005 to 2007) Funding: Programme Hospitalier de Recherche Clinique of the French Ministry of Health
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Centrally allocated
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Roy 2014.

Study characteristics
Methods	RCT
Participants	112 preterm VLBW infants
Interventions	Probiotics (N = 56): Lactobacillus acidophilus, Bifidobacterium longum, B. bifidum, B. lactis 2 times daily with human milk for 6 weeks or until discharged from hospital Control (N = 56): sterile water as (quote:) "placebo"
Outcomes	NEC Death Infection
Notes	India (2012 to 2013) Funding: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Centrally allocated
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Sadowska‐Krawczenko 2012.

Study characteristics
Methods	RCT
Participants	55 very preterm or VLBW infants
Interventions	Probiotics (N = 30): Lactobacillus rhamnosus 2 times daily in 2 mL of 5% dextrose until discharge from hospital Control (N = 25): maltodextrin placebo
Outcomes	NEC Death Infection
Notes	Poland (2008 to 2009) Funding: Biomed Lublin, Poland supplied the intervention
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Central allocation
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Saengtawesin 2014.

Study characteristics
Methods	RCT
Participants	60 VLBW infants with gestational age ≤ 34 weeks' at birth
Interventions	Probiotics (N = 31): Lactobacillus acidophilus and Bifidobacterium bifidum (Infloran®) once daily with human milk or formula until 6 weeks or hospital discharge Control (N = 29): unsupplemented milk feeds
Outcomes	NEC Death Infection Probiotic (quote:) "sepsis" Duration of hospitalisation
Notes	Thailand (2012 to 2013) Funding: Queen Sirikit National Institute of Child Health, Perinatal Society of Thailand and DKSH (Thailand) Limited
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Samanta 2009.

Study characteristics
Methods	RCT
Participants	186 very preterm or VLBW infants
Interventions	Probiotics (N = 91): Bifidobacteria infantis, B. bifidum, B. longum and Lactobacillus acidophilus with human milk 2 times daily until hospital discharge Control (N = 95): unsupplemented human milk feeds
Outcomes	NEC Death Infection Duration of hospital stay
Notes	India (2007 to 2008) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described.
Allocation concealment (selection bias)	Unclear risk	Not described.
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Sari 2011.

Study characteristics
Methods	RCT
Participants	221 VLBW infants (gestational age < 33 weeks' at birth)
Interventions	Probiotics (N = 110): Lactobacillus sporogenes in human milk or formula once daily until discharge from hospital Control (N = 111): unsupplemented milk feeds
Outcomes	NEC Death Infection Rate of weight gain Neurodevelopmental impairment at 18 to 24 months' post‐term, defined as one or more of: BSID‐II MDI < 70, PDI < 70, cerebral palsy, bilateral blindness, or hearing impairment requiring amplification in both ears
Notes	Turkey (2008 to 2009) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Opaque, sequentially numbered, sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Caregivers masked, investigators not masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete for in hospital outcomes (Neurodevelopmental assessment = 84%)
Selective reporting (reporting bias)	Low risk	Unlikely

Serce 2013.

Study characteristics
Methods	RCT
Participants	208 very preterm or VLBW infants
Interventions	Probiotics (N = 104): Saccharomyces boulardii in human milk or formula once daily until discharge from hospital Control (N = 104): unsupplemented milk feeds
Outcomes	NEC Death Infection Rate of weight gain Duration of hospitalisation Culture proven Saccharomyces boulardii (quote:) "sepsis"
Notes	Turkey (2010 to 2011) Funding: Biocodex supplied the intervention
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Opaque, sequentially‐numbered, sealed envelopes.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Shadkam 2015.

Study characteristics
Methods	RCT
Participants	60 preterm infants born between 28 to 34 weeks' gestation and birth weight 1000 g to 1800 g (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 30): Lactobacillus reuteri DSM 17938 2 times daily with human milk until full enteral feeding was reached (about 2 weeks) Control (N =30): unsupplemented milk feeds
Outcomes	NEC Death Infection
Notes	Iran (2012 to 2013) Funding: Shahid Sadughi University, Iran
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	States that random allocation software was used
Allocation concealment (selection bias)	Unclear risk	No information on concealment
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Shashidhar 2017.

Study characteristics
Methods	RCT
Participants	104 VLBW infants
Interventions	Probiotics (N = 52): Lactobacillus acidophilus, L. rhamnosus, Bifidobacterium longum and Saccharomyces boulardii (Darolac) once daily in human milk until discharge from hospital Control (N = 52): unsupplemented milk feeds
Outcomes	NEC Death Duration of hospital stay
Notes	India (2012 to 2013) Funding: not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Sequentially‐numbered, opaque. sealed envelopes
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (3 infants in each group withdrawn)
Selective reporting (reporting bias)	Low risk	Unlikely

Stratiki 2007.

Study characteristics
Methods	RCT
Participants	77 preterm infants with gestation at birth > 26 weeks' (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 41): Bifidobacterium lactis supplemented formula for 30 days Control (N = 36): unsupplemented formula feeds
Outcomes	Stool colonisation Intestinal permeability NEC Death Infection Rate of weight gain
Notes	Greece (2004 to 2005) Funding: Nestlé, Vevey provide the B. lactis supplemented formula
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random numbers generator
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unsupplemented milk feeds‐ not placebo‐controlled
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete (3 infants not included in analyses)
Selective reporting (reporting bias)	Low risk	Unlikely

Strus 2018.

Study characteristics
Methods	RCT
Participants	181 preterm infants ≤ 34 weeks' gestation and birth weight 750 g to 1800 g (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 90): Lactobacillus rhamnosus KL53A and Bifidobacterium breve PB04 in milk feeds for 6 weeks or until hospital discharge Control (N = 91): maltodextrin placebo
Outcomes	Stool colonisation NEC Death Infection
Notes	Poland (2012 to 2013) Funding: IBSS BIOMED S.A., Krakow, Poland ClinicalTrials.gov no. NCT02073214
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated sequence
Allocation concealment (selection bias)	Low risk	Centrally allocated
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Tewari 2015.

Study characteristics
Methods	RCT
Participants	244 preterm infants < 34 weeks' gestation at birth (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 121): Bacillus clausii 3 times daily with human milk for 6 weeks, or until discharge or death or occurrence of late‐onset invasive infection Control (N= 123): sterile water placebo (probiotic and the placebo were identical in appearance)
Outcomes	NEC, death, infection, duration of hospital stay
Notes	India (2012 to 14) Funding: Enterogermina, Sanofi‐Aventis, Italy supplied intervention
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Web‐based
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Totsu 2014.

Study characteristics
Methods	Cluster‐RCT
Participants	283 VLBW infants in 19 neonatal centres
Interventions	Probiotics (N = 10 centres; 153 infants*): Bifidobacterium bifidum with human milk or formula feeds 2 times daily until infant reached 2000 g body weight Control (N = 9 centres; 130 infants*): maltodextrin placebo *Inter‐cluster correlation correction of data for inclusion in meta‐analyses achieved by dividing numerators and denominator by the design effect (1.2779): Probiotics: adjusted N = 120 for in hospital outcomes; N = 80 for neurodevelopmental assessment outcomes Control: adjusted N = 102 for in hospital outcomes; N = 82 for neurodevelopmental assessment outcomes
Outcomes	NEC Death Infection Duration of hospital stay Rate of weight gain Neurodevelopmental impairment at 18 months, defined as Kyoto Scale of Psychological Development 2001 developmental quotient < 70, hearing (bilateral aids) or visual impairment, cerebral palsy (Gross Motor Function Classification System level II or greater)
Notes	Japan (19 centres: 2010 to 2011) Funding: Meiji, Tokyo, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated (stratified by (quote: "patient volume" of centre)
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete for in hospital outcomes (Neurodevelopmental assessment = 73%)
Selective reporting (reporting bias)	Low risk	Unlikely

Van Niekerk 2014.

Study characteristics
Methods	RCT
Participants	184 VLBW infants (< 1250 g)
Interventions	Probiotics (N = 91): Lactobacillus rhamnosus GG and Bifidobacterium infantis daily with human milk feeds for 4 weeks Control (N = 93): MCT oil placebo in milk feeds
Outcomes	NEC Death Infection
Notes	South Africa (2011 to 2012) Funding: National Research Foundation, Nestle Nutrition Institute Africa, Medical Research Council and the Faculty of Medicine and Health Sciences, Stellenbosch University ClinicalTrials.gov no. NCT01868737
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Independent statistician‐generated
Allocation concealment (selection bias)	Low risk	Pharmacy allocation (stratified by maternal HIV status)
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Wang 2007.

Study characteristics
Methods	Quasi‐RCT
Participants	44 VLBW infants
Interventions	Probiotics (N = 22): Bifidobacterium breve in milk feeds 2 times daily until hospital discharge Control (N = 33): unsupplemented milk feeds
Outcomes	Short chain fatty acid and faecal lactic acid concentration Infection (NEC (courtesy of investigators))
Notes	Japan (2001 to 2004) Funding: intervention provided by Morinaga Milk Industry, Kanagawa, Japan
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Alternate allocation
Allocation concealment (selection bias)	High risk	Unconcealed
Blinding (performance bias and detection bias) All outcomes	High risk	Unmasked
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely (did not aim to assess clinical outcomes)

Wejryd 2019.

Study characteristics
Methods	RCT
Participants	141 ELBW infants (of gestation born < 28 weeks')
Interventions	Probiotics (N = 72): Lactobacillus reuteri DSM 17938 once daily with human milk until 36 weeks' postmenstrual age Control (N = 69): maltodextrin placebo
Outcomes	Time to full enteral feeds NEC Death Infection
Notes	Sweden (10 centres: 2012 to 2015) Funding: Swedish Research Council, the Swedish Society for Medical Research, the Swedish Society of Medicine, the Research Council for the South‐East Sweden, ALF Grants, Region Ostergotland, the Ekhaga Foundation, and BioGaia AB ClinicalTrials.gov no. NCT01603368
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Centrally coded by sequential study number
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Complete
Selective reporting (reporting bias)	Low risk	Unlikely

Zeber‐Lubecka 2016.

Study characteristics
Methods	RCT
Participants	55 preterm infant < 33 weeks' gestation (most participants were very preterm or VLBW)
Interventions	Probiotics (N = 28): Saccharomyces boulardii once daily with human milk or formula feeds for six weeks Control (N = 27): maltodextrin placebo
Outcomes	Stool microbiomic structure (NEC, death, infection‐ no events courtesy investigators)
Notes	Poland (study period not stated) Funding: The National Science Centre, Poland
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described (quote: "randomly divided")
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	Low risk	Masked (placebo‐controlled)
Incomplete outcome data (attrition bias) All outcomes	High risk	Missing data from each group (10 from probiotics and 6 from placebo) – not accounted for
Selective reporting (reporting bias)	Low risk	Unlikely (primary aim to study intestinal microbiome)

BBG‐01: Bifidobacterium breve;BSID: the Bayley Scales of Infant Development; ELBW: extremely low birth weight; g: gram(s); HIV: human immunodeficiency virus; LGG:Lactobacillus rhamnosus GG; MCT: medium chain triglycerides; MDI: Mental Developmental Index; NEC: necrotising enterocolitis; PDI: Psychomotor Development Index; RCT: randomised controlled trial; SD: standard deviation; VLBW: very low birth weight.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Arora 2017	Most participants not very preterm or VLBW.
Awad 2010	Most participants not very preterm or VLBW.
Chi 2019	Not an RCT.
Dasopoulou 2015	RCT of prebiotics.
Deng 2010	Most participants not very preterm or VLBW.
Denkel 2016	Not an RCT.
Di 2010	Most participants not very preterm or VLBW.
Dongol‐Singh 2017	Most participants not very preterm or VLBW.
Hua 2014	Most participants not very preterm or VLBW.
Hussain 2016	Most participants not very preterm or VLBW.
Kaban 2019	Most participants not very preterm or VLBW.
Ke 2008	Most participants not very preterm or VLBW.
Koksal 2015	RCT of synbiotics
Moles 2015	A pilot study with including 5 infants.
Partty 2013	Most participants not very preterm or VLBW.
Qiao 2017	Most participants not very preterm or VLBW.
Rojas 2012	Most participants not very preterm or VLBW.
Romeo 2011	Most participants not very preterm or VLBW.
Shujie 2011	Most participants not very preterm or VLBW.
Sinha 2015	Most participants not very preterm or VLBW.
Thanhaeuser 2014	Not an RCT.
Uhlemann 1999	Most participants not very preterm or VLBW.
Underwood 2014	RCT of prebiotics
Xu 2016	Most participants not very preterm or VLBW.
Zhou 2012	Most participants not very preterm or VLBW.
Zhuang 2007	Most participants not very preterm or VLBW.

RCT: randomised controlled trial; VLBW: very low birth weight

Characteristics of studies awaiting classification [ordered by study ID]

Coleta 2013.

Methods	Randomised controlled trial
Participants	60 preterm infants
Interventions	Probiotics (N = 31): human milk with Lactobacillus reuteri Control (N = 21): human milk alone
Outcomes	Efficacy of probiotics on digestive tolerance to enteral feeding
Notes	Romania (study period not stated) Unlikley to have been reported fully (unable to contact investigators)

Punnahitananda 2006.

Methods	RCT
Participants	VLBW infants
Interventions	Lactobacillus acidophilus and Bifidobacterium infantis
Outcomes	Late‐onset infection, NEC, feeding tolerance, time to full enteral feeding
Notes	Data presented at 14th Congress of the Federation of Asia Oceania Perinatal Societies, 2006, Bangkok,Thailand (report not available)

NEC: necrotising enterocolitis; RCT: randomised controlled trial; VLBW: very low birth weight

Characteristics of ongoing studies [ordered by study ID]

Marisen 2019.

Study name	Efficacy of Bifidobacterium longum, B. infantis and Lactobacillus acidophilus probiotics to prevent gut dysbiosis in preterm infants of 28‐ 32 weeks' gestation: a randomised, placebo‐controlled, double‐blind, multicentre trial: the PRIMAL Clinical Study protocol
Methods	RCT
Participants	Preterm infants (28 to 32 weeks')
Interventions	Bifidobacterium longum, B. infantis, and Lactobacillus acidophilus
Outcomes	Stool colonisation
Starting date	2020
Contact information	Christoph Hartel, Department of Paediatrics, University of Lübeck, Germany
Notes	Trial registration number: DRKS00013197

NCT00977912.

Study name	Necrotizing enterocolitis (Nec) and B. Lactis in premature babies
Methods	RCT
Participants	VLBW infants
Interventions	B. lactis for 6 weeks
Outcomes	NEC, antibiotic administration, stool microbiology
Starting date	November 2009
Contact information	Dr Peter Cooper, University of Witwatetersrand & Charlotte Maxek Johannestburg Academic Hospital, Zambia
Notes	(Quote:) "Terminated" in 2013 ‐ unlikely to have been completed (not reported)

NCT01181791.

Study name	Effects of Lactobacillus reuteri in premature infants (reuteri)
Methods	RCT
Participants	VLBW infant
Interventions	Lactobacillus reuteri during hospitalisation
Outcomes	Time to reach full enteral feeds, stool colonisation and Intestinal immunological response
Starting date	2010
Contact information	Teresa del Moral, University of Miami
Notes	Chile (Quote:) "Terminated" because of slow recruitment‐ unlikely to have been reported

NCT01375309.

Study name	Bifidobacterium supplementation for very low birth weight infants (Bifido(RCT))
Methods	RCT
Participants	VLBW infants
Interventions	Bifidobacterium bifidum (duration not clear)
Outcomes	Time to full enteral feeding, weight gain, NEC
Starting date	2011
Contact information	Satoshi Kusuda, Professor of Neonatology, Tokyo Women's Medical University
Notes	(Quote:) "Completed" 2012 ‐ unlikely to have been reported

NCT04541771.

Study name	The role of Lactobacillus reuteri in preventing necrotizing enterocolitis (NEC) in pre‐term infants (NEC)
Methods	RCT
Participants	Preterm infants (28 to 34 weeks')
Interventions	Lactobacillus reuteri until 35 weeks' of gestation or discharged from hospital
Outcomes	NEC, infection
Starting date	2020
Contact information	Dr Summera Tabasum, The Children Complex & The Institute of Child Health, Multan
Notes	ClinicalTrials.gov identifier: NCT04541771

NEC: necrotising enterocolitis; RCT: randomised controlled trial; VLBW: very low birth weight

Differences between protocol and review

In the 2020 update:

• new authors updated this review;

• we restricted the population of interest to very preterm and VLBW infants in order to enhance applicability to those infants at high risk of developing NEC and associated complications;

• we added the methodology and plan for 'Summary of findings' tables and GRADE recommendations, which were not included in the original protocol (AlFaleh 2005), or in previous publications of the review (Al Faleh 2008; Al Faleh 2011; Al Faleh 2014);

• we updated the search strategy; and

• we updated the "Risk of Bias" assessments.

Contributions of authors

SS and SO screened and appraised reports identified in the updated search, and extracted and analysed data.

NM undertook analyses for small‐study bias.

WM and MXRR arbitrated inclusion and data extraction disagreements, assessed the certainty of the evidence (GRADE), and drafted the review.

All authors contributed to the final manuscript.

Sources of support

Internal sources

• Centre for Reviews and Dissemination, University of York, UK

  Host department

• Department of Clinical Epidemiology and Biostatistics. Faculty of Medicine. Pontificia Universidad Javeriana, Colombia

  Host department

External sources

• Vermont Oxford Network, USA

  Cochrane Neonatal Reviews are produced with support from Vermont Oxford Network, a worldwide collaboration of health professionals dedicated to providing evidence‐based care of the highest quality for newborn infants and their families.

• National Institute of Health Research (NIHR), UK

  This report describes independent research funded by a UK NIHR Cochrane Incentive Award. The views expressed in this publication are those of the review authors and are not necessarily those of the NHS, the NIHR, or the UK Department of Health

Declarations of interest

SS is funded by the UK National Institute of Health Research (NIHR) for the review.

NM: the UK NIHR pays a grant to NM's institution.

MXRR has no interest to declare.

SO: the UK NIHR pays a grant to SO's institution. (SR‐PG 13/89/12).

WM: the UK NIHR pays a grant to WM's institution. WM is co‐coordinating editor of Cochrane Neonatal.

New search for studies and content updated (conclusions changed)