Introduction

Preterm infants are at risk of mortality, and morbidities such as necrotising enterocolitis (NEC) and late-onset sepsis (LOS). An important risk factor for NEC and sepsis in preterm infants is gut dysbiosis.1 Hence attenuating dysbiosis by the use of probiotics has the potential to improve their clinical outcomes.

Probiotics are live microorganisms that when administered in adequate amounts, could confer beneficial effects on the host.2 Systematic reviews of randomised controlled trials (RCTs) and non-randomised studies have shown that probiotic supplementation reduces the risk of NEC (≥Stage II), LOS, and mortality in preterm infants.3 The benefits of probiotics relate to their ability to improve the gut barrier, modulate the immune system and attenuate gut dysbiosis.1 Probiotics have been shown to reduce the relative abundance of pathogens in the gut through various pathways, including blocking the receptors and competing for nutrients.4

Considering that probiotic effects are considered species and strain specific,5 data on individual probiotic species and strains is important for guiding clinical practice and research.

During the early human development, bacteria belonging to the genus Bifidobacterium play an important role.6 Among the bifidobacteria, Bifidobacterium longum subspecies infantis (B. infantis) is considered as an important gut symbiont, especially in infancy. It is considered as a champion coloniser of the gut due to its properties for the consumption of human milk oligosaccharides (HMOs).7 It may have a competitive advantage against other bacteria, allowing increased colonisation and resulting in fewer luminal pathogens.8 B. infantis promotes maturation of the innate immune response9 and improves the anti-inflammatory properties through the production of tryptophan metabolite, indole-3-lactic acid (ILA).10 Given these properties, we hypothesised that supplementation with probiotics containing B. infantis will be more beneficial in preterm infants than those without this sub-species of bacteria.

To our knowledge, no systematic review, including the latest systematic review and the network meta-analysis3 has addressed this specific question. There are no RCTs in preterm infants that have compared B. infantis versus placebo. In addition, apart from the small RCT by our group,11 there are no RCTs that have compared supplementation with probiotics containing B. infantis versus probiotics not containing B. infantis.

Hence, we conducted a systematic review that had two subgroups: Subgroup A: Probiotics containing B. infantis versus placebo/no probiotics; Subgroup B: Probiotics not containing B. infantis versus placebo/no probiotics.

Methods

Guidelines from the Cochrane Neonatal Review Group,12 and the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement13 were followed for undertaking and reporting this systematic review and meta-analysis.

Eligibility criteria

Types of studies

We selected RCTs evaluating probiotics for the prevention of morbidity or mortality in preterm (gestational age <37 weeks) and/or low birth weight (birth weight <2500 g) infants. We excluded studies that enrolled term infants. Non-randomised studies, narrative reviews, systematic reviews, case reports, letters, editorials, and commentaries were excluded but read to identify potentially eligible studies.

Types of participants

Preterm infants born before gestation <37 weeks, low birth weight (<2500 g), or both.

Interventions and comparisons

We included studies assessing enteral administration of any probiotic commenced within the first week of life and continued for at least one week compared to placebo or no treatment. We excluded studies that used prebiotics or synbiotics (i.e. combination of a prebiotic with probiotics).

Outcomes

These included (1) NEC ≥Stage II (Modified Bell’s criteria)14; (2) LOS defined as isolation of a pathogen from blood, cerebrospinal fluid, or a normally sterile body space after 48 h of birth; (3) All-cause mortality.

Search strategy

The Cochrane Central Register of Controlled clinical trials (www.thecochranelibrary.com, through December 2022), PubMed (https://www.ncbi.nlm.nih.gov, 1966–December 2022), EMBASE (Excerpta Medica dataBASE) via Ovid (http://ovidsp.tx.ovid.com, 1980–December2022), EMCARE via OVID (http://ovidsp.tx.ovid.com, 1980–December 2022) databases were searched. We searched https://clinicaltrials.gov and ANZCTR (Australia New Zealand Clinical Trials Registry (www.anzctr.org.au) for ongoing RCTs. Grey literature was searched using Mednar (www.mednar.com). The reference lists of identified studies and key review articles were searched to identify additional RCTs. No language restriction was applied.

PubMed was searched using the following keywords:

((((((((Probiotic) OR (Probiotics)) OR (Bifidobacteria)) OR (Bifidobacterium)) OR (Lactobacilli)) OR (Lactobacillus)) OR (Saccharomyces)) AND ((((((Preterm infant) OR (Preterm infants)) OR (premature infants)) OR (low birth weight infants)) OR (very low birth weight infants)) OR (extremely low birth weight infants))) AND (Trial). PubMed was also searched using relevant MeSH words. Other databases were searched using similar terminologies.

Study selection

Abstracts of the citations obtained from the initial broad search were read independently by two reviewers to identify potentially eligible studies. Full-text articles of these studies were obtained and assessed independently for eligibility by two reviewers, using the predefined eligibility criteria. Differences in opinion were resolved by a group discussion to reach a consensus. Multiple publications of the same study were excluded to avoid duplication of the data.

Data extraction

Two reviewers independently extracted the data using a standardised data collection form. Discrepancies were resolved by discussion and consensus among all authors.

Assessment of risk of bias (ROB) of RCTs

ROB was assessed using the Cochrane “Risk of Bias Assessment Tool”.12 Two reviewers independently assessed the ROB in all domains including random number generation, allocation concealment, blinding of intervention and outcome assessors, completeness of follow up and selectivity of reporting. For each domain, the ROB was assessed as low, high or unclear.

Data synthesis and statistical analysis

Meta-analysis was performed using statistical software, STATA (Version 17.0). Since heterogeneity was expected we used random-effects model (REM) model for meta-analysis. Fixed effect model (FEM) was also used to assure that the choice of the model did not influence the results. Since all outcomes of interest were binary, we used relative risk (RR) and 95% CI to summarise their results.

Heterogeneity

Clinical heterogeneity was assessed and reported by summarising characteristics such as the study population, dose, and duration of probiotic supplementation. Statistical heterogeneity was estimated using the I2 statistic and interpreted as per Cochrane handbook11 as follows: 0–40%: might not be important; 30–60%: may represent moderate heterogeneity; 50–90%: may represent substantial heterogeneity; 75–100%: considerable heterogeneity.

Publication bias

To assess for any publication bias, we used Egger’s,15 Harbord’s,16 Begg’s17 and trim & fill plots.18

Summary of Findings (SOF) table

The key information about the quality of evidence, the magnitude of the effect of the intervention and the sum of available data on the main outcome was presented in the SOF table according to the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) guidelines.19

Results

Initial broad search identified 2315 records, of which full texts of 84 potentially eligible studies were read in detail. Seventeen of these studies were excluded being non-RCTs (n = 4), RCTs of prebiotics or synbiotics (n = 6), being conducted in full-term infants (n = 1), and not reporting our outcomes of interest (n = 6). Finally, 67 RCTs were included.20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86

The flow diagram of the study selection process is given in Fig. 1.

Fig. 1: PRISMA flow diagram of study selection.
figure 1

Identification of studies via databases and registers.

Full size image

Of the 67 RCTs (n = 14,606), 16 (n = 4962) had used B. infantis as a component of the probiotic supplement.20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35 The remaining 51 (n = 9644) did not use B. infantis.36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86 The mean gestation and birth weight ranged from 25.4 weeks to 33.5 weeks and from 727 g to 2262 g, respectively. The duration of probiotic supplementation varied from a minimum of 2 weeks to until discharge/40 weeks corrected gestational age. The probiotic dose ranged from 18 × 106 (18 million) colony forming units (CFU) to 12 × 109 (12 billion) CFU/day. NEC, LOS and all-cause mortality were included as the outcomes in 63, 57 and 54 studies, respectively. The characteristics of the included studies are given in Table 1a, b.

Table 1 (a) Randomised controlled trials using probiotics containing B. infantis; (b) randomised controlled trials using probiotics without B. infantis.
Full size table

ROB of included studies

Of the 16 studies that used probiotics containing B. infantis, 9 (56.2%) studies were considered to have “low ROB” on the domain of random sequence generation, 5 (31.2%) had “high ROB” for performance & detection bias. Fifteen (93.7%) studies and 12 (75.0%) showed “low ROB” for attrition and reporting bias, respectively. Among the 51 studies that used probiotics without B. infantis, 36 (70.5%) and 28 (54.9%) were considered “low ROB” in the selection bias category. 24 (47%) studies had “unclear risk” or “high risk” ROB for blinding. For the attrition and reporting bias categories, 44 (86.2%) and 35 (68.6%) studies showed “low ROB”. Details of the ROB analysis are given in Table 2a, b.

Table 2 (a) Risk of bias in studies using probiotics containing B. infantis; (b) risk of bias in studies using probiotics without B. infantis.
Full size table

Outcomes

The effects of the intervention were compared between the studies that used probiotics containing B. infantis versus those that used probiotics without B. infantis.

(1) NEC ≥Stage II

Meta-analysis of all 63 RCTs found that probiotics decreased the risk of NEC (211/6394 [3.3%] vs 387/6170 [6.3%]); RR 0.59 (CI 0.50–0.70); I2 = 0% (Fig. 2).

Fig. 2: Forest plot.
figure 2

Probiotic supplementation to reduce NEC (≥Stage II).

Full size image

On subgroup meta-analysis, 15 RCTs (n = 3626) that used probiotics containing B. infantis (Subgroup A) showed significant reduction in the incidence of NEC (44/1862 [2.4%] vs 114/1764 [6.5%]; RR 0.38; 95% CI, 0.27–0.55; I2 = 0% (Fig. 2). Subgroup meta-analysis of 48 RCTs (n = 8938) that used probiotics without B. infantis (subgroup B) also found significant reduction in the incidence of NEC (167/4532 [3.7%] vs 273/4406 [6.2%]; RR 0.67; 95% CI, 0.55–0.81; I2 = 0% (Fig. 2).

The p-value for subgroup differences was 0.01, which suggested that the beneficial effects are more pronounced in studies that had B. infantis as a component of the probiotic product.

To determine the publication bias, various statistical tests were used. Harbord (p = 0.203) and Begg’s (p = 0.577) tests showed no publication bias exists but the Egger’s test (p = 0.010) did. Further, trim & fill analysis imputed 9 potentially missing studies (Fig. 5a); however, the final results after including the imputed studies was still significant (RR = 0.60; 95% CI (0.49–0.73).

(2) LOS

Meta-analysis of all 57 RCTs found that probiotics decreased the risk of LOS (905/6472 [14%] vs 1025/6277 [16.3%]); RR 0.84 (CI 0.76–0.94); I2 = 40% (Fig. 3).

Fig. 3: Forest plot.
figure 3

Probiotic supplementation to reduce LOS.

Full size image

On subgroup meta-analysis, 13 RCTs (n = 4123) that used probiotics containing B. infantis (subgroup A) showed that probiotics reduced the incidence of LOS (293/2094 [14%] vs 356/2029 [17.5%]; RR 0.80; 95% CI, 0.63–1.01; I2 = 60% (Fig. 3). Subgroup meta-analysis of 44 RCTs (n = 8626) that used probiotics without B. infantis (subgroup B) also found reduction of LOS (612/4378 [14%] vs 669/4248 [15.7%]; RR 0.86; 95% CI, 0.77–0.97; I2 = 30% (Fig. 3).

The p value for subgroup differences was 0.58, which suggested that the beneficial effects for the prevention of LOS were similar irrespective of whether B. infantis was a component of the probiotic product or not.

In the analysis for any publication bias. Harbord (p = 0.149) and Begg’s (p = 0.188) tests showed no publication bias exists but the Egger’s test (p = 0.088) did. Further, trim & fill analysis imputed 5 potentially missing studies (Fig. 5b); however, the final results after including the imputed studies was still significant [0.87; 95% CI (0.77–0.97).

(3) All-cause mortality

Meta-analysis of all 54 RCTs found that probiotics decreased the risk of all-cause mortality (270/6043 [4.5%] vs 355/5872 [6%]); RR 0.78 (CI 0.67–0.91); I2 = 0% (Fig. 4).

Fig. 4: Forest plot.
figure 4

Probiotic supplementation to reduce mortality.

Full size image

On subgroup meta-analysis, 14 RCTs (n = 4292) that used probiotics containing B. infantis (subgroup A) showed significant reduction of all-cause mortality (72/2174 [3.3%] vs 109/2118 [5.1%]; RR 0.65; 95% CI, 0.48–0.88; I2 = 0% (Fig. 4). Subgroup meta-analysis of 40 RCTs (n = 7623) that used probiotics without B. infantis (subgroup B) also found reduction of all-cause mortality (198/3869 [5.1%] vs 246/3754 [6.5%]; RR 0.83; 95% CI, 0.69–1.00; I2 = 0% (Fig. 4).

The p value for subgroup differences was 0.17, which suggested that the beneficial effects for the reduction in mortality were similar irrespective of whether B. infantis was a component of the probiotic product or not.

In the analysis for any publication bias, Harbord (p = 0.151), Begg’s (p = 0.560) and Egger’s tests (p = 0.100) showed that there is no publication bias. However, the trim & fill analysis imputed 5 potentially missing studies. Meta-analysis after incorporating the results of imputed studies found results that were similar to the primary analysis [(RR = 0.80; 95% CI (0.69–0.93)] (Fig. 5c).

Fig. 5: Trim and fill funnel plots for NEC, sepsis and mortality.
figure 5

a Trim and fill funnel plot for publication bias for NEC. b Trim and fill funnel plot for publication bias for sepsis. c Trim and fill funnel plot for publication bias for mortality.

Full size image

For studies in which B. infantis was a component of the probiotic supplement, the overall GRADE of evidence was high for the outcomes of mortality and NEC and moderate for LOS (Table 3a). For studies in which B. infantis was not a component of the probiotic supplement, the overall GRADE of evidence was high for all the outcomes of NEC, LOS, and mortality (Table 3b).

Table 3 (a) Summary of findings according to GRADE guidelines for RCTs that used B. infantis; (b) summary of findings according to GRADE guidelines for RCTs without B. infantis.
Full size table

Discussion

Our systematic review that included 67 RCTs (n = 14,606) found that probiotic supplementation significantly reduced the risk of NEC≥ Stage II, LOS and all-cause mortality in preterm infants. Specific to our aim, the sub-group meta-analysis of RCTs that used probiotics containing B. infantis showed even more favourable results, especially for the prevention of NEC (≥Stage II). These results provide indirect evidence that probiotics that include B. infantis may be more beneficial in preterm infants than those not including B. infantis.

Our results are supported by a recent non-randomised study by Tobias et al, involving 483 VLBW infants. Supplementation of B. infantis was associated with a significant reduction in NEC (≥Stage II) and NEC-related mortality.87 The B. infantis cohort had a 73% reduction of NEC (≥ stage II) compared with the no B. infantis cohort (adjusted RR, 0.27; 95% CI, 0.094–0.614; p < 0.01).

A prospective study by Nguyen et al. has evaluated the effect of B. infantis administration on gut microbiota, nosocomial acquired antibiotic resistance and enteric inflammation in preterm infants with gestation <32 weeks and/or birth weight <1500 g.88 Infants supplemented with B. infantis had lower enteric inflammation after adjusting for other clinical variables in multivariate modelling. In contrast, Kochjancic et al. reported that probiotic (B. infantis and Lactobacillus acidophilus) supplementation did not reduce the risk of NEC in neonates with duct-dependent congenital heart disease (CHD).89 The lack of benefits of probiotics may relate to the small sample size (n = 15) and retrospective design of the study, lack of concurrence of NEC and duct-dependent CHD, and the fact that the majority of infants were born at term.

Discussing the physiological characteristics of B. infantis is important as probiotic effects are species and strain- specific. Ward et al. and LoCascio et al. reported that given their complex structure, the HMOs cannot be metabolised by the infant or most of the bacteria in the infant’s gut, as they lack the necessary enzymes for the purpose.90,91 B. infantis has been shown to grow in vitro using HMO as the sole carbon source, reaching a cell density 3-fold higher than B. longum subsp. longum, B. breve, B. bifidum, and B. adolescentis. Low pH is a critical factor in preventing the invasion and overgrowth of harmful bacteria in the infant gut, a process known as colonisation resistance. Henrick et al. reported that B. infantis supplementation significantly lowered faecal pH in breastfed infants compared to controls.92 Underwood et al. reported that HMO metabolism by B. infantis produces short-chain fatty acids (SCFA), such as acetate, which play an important role in nutrition and intestinal and immune development, facilitate direct binding to intestinal cells, and stimulate anti-inflammatory/inhibits pro-inflammatory cytokine release by intestinal cells.93 Meng et al. reported that B. infantis contributes to maintaining of gut barrier integrity through indole 3-lactic acid (ILA), a metabolite of tryptophan, and may protect gut epithelium by activating the aryl hydrogen receptor, which can further promote intestinal immune function.10 In addition to their role in the gut, SCFAs produced by B. infantis can enter circulation and directly affect the adipose tissue, lungs, brain, and liver, inducing overall beneficial metabolic effects.94 Animal studies by Bergmann et al. suggested that B. infantis can potentially protect against excessive intestinal inflammation which is implicated in the pathogenesis of NEC in preterm infants.95 Given that probiotics are live organisms, a major concern is the risk of sepsis due to the administered probiotic organism. Although there are few case reports of bacteraemia caused by the Bifidobacteria,96,97 it is reassuring to note that none of the RCTs included in our review that used probiotics reported probiotic related sepsis. However, current evidence is limited for estimating the risk of probiotic sepsis.98,99 In 2007, the European Food Safety Authority (EFSA) assigned qualified presumption of safety (QPS) status to the bacterial species B. longum, which includes subspecies infantis, indicating that this taxonomic group does not carry safety concerns.100 The QPS status, which applies to all strains of B. infantis indicates that none of these has been associated with human clinical disease. However, it should not lead to complacency, and constant microbiological surveillance is essential to identify and treat sepsis that may occur due to the administered probiotic organism.

To our knowledge, ours is the first systematic review related to B. infantis in preterm infants. Our results will help in guiding research for using B. infantis as a probiotic in preterm infants. This is also one of the largest systematic reviews of probiotics in preterm infants involving 14,606 preterm infants, which is close to the recent network meta-analysis (3) and 3800 infants more than the Cochrane review.101

Our systematic review has some limitations. Since only 16 RCTs used probiotics containing B. infantis and the remaining 51 RCTs used a probiotic that did not contain B. infantis, it resulted in a large discrepancy between the number of participants between these two groups. Furthermore, none of the included RCTs used B. infantis as the sole probiotic. Instead, they used a mixture of probiotic organisms with variable doses of B. infantis. Thus, the observed benefits cannot be attributed definitively to B. infantis.

In conclusion, our systematic review of RCTs provides indirect evidence that the beneficial effects for the prevention of NEC are more pronounced if B. infantis is a component of the probiotic product compared to studies in which B. infantis is not a component. However, given the limitations to the evidence, adequately powered RCTs are necessary to confirm the benefits and safety of B. infantis in preterm infants. Such RCTs could compare (1) B. infantis versus Placebo or (2) B. infantis as a component of a multi-strain probiotic product versus the same multi-strain probiotic but without B. infantis.