Probiotics for prevention of necrotizing enterocolitis in preterm infants: systematic review and meta-analysis

Necrotizing enterocolitis (NEC) affects predominantly preterm infants, who have specific risk factors leading to intestinal dysbiosis. Manipulations of gut microbiota through probiotics have the potential to prevent NEC. The aim of this systematic review and meta-analysis was to evaluate the effect of probiotics for NEC prevention in preterm infants, with a focus on specific strains, microbiological strength of currently available studies, and high-risk populations. PubMed and the Cochrane Library were searched for trials published within 4th February 2015. Randomized-controlled trials reporting on NEC and involving preterm infants who were given probiotics in the first month of life were included in the systematic review. Twenty-six studies were suitable for inclusion in the meta-analysis. Data about study design, population, intervention and outcome were extracted and summarized independently by two observers. Study quality and quality of evidence were also evaluated. Fixed-effects models were used and random-effects models where significant heterogeneity was present. Subgroup analyses were performed to explore sources of heterogeneity among studies. Results were expresses as risk ratio (RR) with 95 % confidence interval (CI). The main outcome was incidence of NEC stage ≥2 according to Bell’s criteria. Probiotics prevented NEC in preterm infants (RR 0.47 [95 % CI 0.36–0.60], p < 0.00001). Strain-specific sub-meta-analyses showed a significant effect for Bifidobacteria (RR 0.24 [95 % CI 0.10–0.54], p = 0.0006) and for probiotic mixtures (RR 0.39 [95 % CI 0.27–0.56], p < 0.00001). Probiotics prevented NEC in very-low-birth-weight infants (RR 0.48 [95 % CI 0.37–0.62], p < 0.00001); there were insufficient data for extremely-low-birth-weight infants. The majority of studies presented severe or moderate microbiological flaws. Probiotics had an overall preventive effect on NEC in preterm infants. However, there are still insufficient data on the specific probiotic strain to be used and on the effect of probiotics in high-risk populations such as extremely-low-birth-weight infants, before a widespread use of these products can be recommended.


Background
Necrotizing enterocolitis (NEC), which is one of the most devastating neonatal diseases, has become a priority for research [1]. Despite great advances in neonatal care, the morbidity, mortality and health-care costs directly related to the disease are substantial: during hospital stay, the economic burden of NEC in the United States has been estimated as high as several billions USD per year, which is approximately 20 % of the costs for Neonatal Intensive Care Units in the country; furthermore, this estimate is likely to be much higher when the costs of long-term care of survivors are taken into account [2].
NEC is a multifactorial disease: prematurity is a wellrecognized risk factor, and approximately 90 % of the infants who develop NEC are born preterm [3]. This is probably due to specific comorbidities of prematurity, such as immunodeficiency, use of broad-spectrum antimicrobials, delayed enteral feeding and low availability of human milk.
Recently, research has focused on the role of gut microbiota and its manipulations, such as the use of probiotics, on disease and health status. Probiotics are live-microorganisms which, when ingested in adequate amounts, confer a health-benefit to the host through an interaction with gut microbiota [4]. The intestinal microbiota undergoes dynamic changes during childhood. Gut colonization in preterm infants occurs differently than in healthy term newborns [5], and preterm infants frequently have delayed and aberrant acquisition of the "normal" digestive flora. Recent studies performed in preterm foetuses and infants demonstrated that amniotic fluid and meconium are not sterile, suggesting an intrauterine origin of gut microbiota [6,7]; after birth, the preterm infant's immature intestine is exposed to an unique environment and to several iatrogenic manipulations, including the use of broad-spectrum antibiotics. The subsequent intestinal dysbiosis is recognized as a risk factor for NEC: actually, it has been shown that preterm infants with NEC have reduced bacterial gut diversity and different bacterial strains compared to healthy controls [8]. In this perspective, provision of probiotics to preterm infants has the potential to "normalize" the abnormal colonization pattern, thus preventing the occurrence of the disease [9].
The use of probiotics for the prevention of NEC in preterm infants has been extensively investigated in many randomized-controlled trials, whose results have been summarized in several systematic-reviews and metaanalyses [10,11]. The authors of these meta-analyses, which show that probiotics reduce NEC and mortality in preterm infants, strongly encourage a change in practice, promoting a widespread use of probiotics in this population [11], and also claim that withholding probiotics from high-risk neonates would be almost unethical [10].
However, the position of the American Academy of Paediatrics is more cautious, highlighting the need for more studies to address unanswered questions on the amount and specificity of which probiotic or mixture of probiotics should be used [12]. In addition, a recent systematic review, which analyzed the level of evidence of randomized-controlled trials on probiotics in preterm infants, concluded that there is still insufficient evidence to recommend routine probiotics use, but also that present data are encouraging and justify further research on specific probiotic products [13].
Actually, the beneficial effects of probiotics appear to be strain-specific, and pooling data from studies using different strains can result in misleading conclusions [14]. Furthermore, currently available studies often lack specificity in reporting correct identification of probiotic strain [15], dosage regimen and duration, and gut colonization, which are all fundamental to assess the ability of a probiotic to confer a health benefit to the host [16].
The aim of this meta-analysis is thus to evaluate in detail the effect of probiotics for the prevention of NEC in preterm infants, with a focus on specific strains, on microbiological strength of currently available studies, and on high-risk populations.

Literature search
The study protocol was designed jointly by the members of the Task Force on Probiotics of the Italian Society of Neonatology.
A systematic review of published studies reporting the use of probiotics for the prevention of NEC in preterm infants was performed, following PRISMA guidelines [17].
Criteria for inclusion in the meta-analysis were the following: randomized and quasi-randomized controlled trials involving preterm infants (gestational age <37 weeks) and reporting on NEC (any stage, according to modified Bell staging criteria [18,19]); enteral administration of any probiotic starting within one month of age, compared to placebo or no treatment. Being the search strategy focused specifically on NEC, data on different outcomes, such as sepsis or mortality, which were reported in the studies retrieved by the literature search, were not evaluated by meta-analysis.
A search was conducted in PubMed (http://www. ncbi.nlm.nih.gov/pubmed/) for studies published before 4th February 2015, using the search string reported in Fig. 1. This string was built up combining all the terms related to NEC and probiotics, using PubMed MeSH terms and free-text words and their combinations through the most proper Boolean operators, in order to be as comprehensive as possible. Similar criteria were used for searching the Cochrane Library. The review was limited to studies written in English and involving human subjects.
The search was conducted by AA and LC: relevant studies were identified from the abstract, and reference lists of papers retrieved were searched for additional studies. "Snowballing" technique was also used [20].

Data extraction and meta-analysis
Study details, including study population, characteristics of the intervention, use of placebo, and outcome, were assessed independently by AA and LC, and checked by DG. Study quality was evaluated independently using the risk of bias tool as proposed by the Cochrane collaboration (Chapter 8 of the Cochrane Handbook of Systematic Reviews) [21]. In addition, an assessment of the body of evidence using the GRADE working group approach was used in order to grade the quality of evidence. The evaluation was carried out following the Chapter 12 of the Cochrane Handbook [21] and classifying the evidence as high,      P prospective, DB double-blinded, R randomized, C controlled, Multic multicentric, B blinded, M masked, Bic bicentric, BW birth weight, GA gestational age, NEC necrotizing enterocolitis, HM human milk, CFU colony forming unit, OD once daily, BD twice daily, OMM own mother's milk, PFM preterm formula, DM donor milk, FM formula, MDX maltodextrin moderate, low and very low (as suggested by the GRADE Working Group) [22]. The association between probiotic use and NEC was evaluated by meta-analyses, conducted by AA and DG, using the RevMan software (version 5.3.5; downloaded from the Cochrane website: http://tech.cochrane.org/ revman/download). Risk ratio (RR) was calculated using the Mantel-Haenszel method, and reported with 95 % confidence interval (CI).
The following sub-meta-analyses were also performed, in order to evaluate the effect of probiotics: in specific subgroups of patients (very-low-birth-weight [VLBW] infants); in surgical NEC; according to NEC incidence in different populations: the incidence of NEC stage ≥2 in the control population was used as a reference, because only a minority of studies reported NEC incidence in the general population. Studies were arbitrarily divided into three groups defined as "low-risk" (NEC incidence <5 %), "medium-risk" (incidence 5-10 %), and "high-risk" (incidence >10 %); according to probiotic strain: studies were divided according to the specific probiotic strain used, and were considered as suitable for inclusion in the sub-meta-analyses when the same probiotic strain was used in at least two studies. Studies which used a probiotic mixture were considered together.
Microbiological quality of all the studies was evaluated by MLC and LM. Studies were defined as having severe, moderate or minor microbiological flaws according to the evaluation of proper strain identification and microbiological assessment. Specifically, the lack of proper strain identification was considered as a severe flaw; the lack of microbiological assessment regarding the probiotic persistence in stools was considered as a moderate flaw, whereas a low flaw was defined when the presence of the probiotic in stools was evaluated by indirect approaches such as the quantification of its species belonging.
A fixed-effect model was used for the analyses. Heterogeneity was measured using the I 2 test. If significant heterogeneity was present (p < 0.05 from the χ 2 test), a random-effects model was used [23]. The randomeffects model was also used when heterogeneity was not significant but the number of studies was ≤ 5, because the test for heterogeneity is known to have low power when the number of studies is small [24].
Forest plots were used to illustrate results from metaanalyses, and funnel plots to investigate bias.
The online version of GraphPad Quickcalcs software was used to calculate number needed to treat (NNT).

Literature search
Three-hundred-sixty-eight papers were identified through the literature search (310 through PubMed and 58 through the Cochrane Library). Thirty-eight studies met the inclusion criteria: 22 were identified through the PubMed search  and 16 through the Cochrane Library search [25, 26, 28-32, 34-37, 40-43, 45]. Ten additional papers were identified from the reference lists of included studies [49][50][51][52][53][54][55][56][57][58]. Of these 47 studies, 16 were excluded, as they were duplicates retrieved both by PubMed and Cochrane Library search. Six additional studies were excluded after examining the full-texts: one study reported maternal probiotic supplementation during pregnancy [29], one cohort was reported twice [31], one study included both term and preterm infants [36], two studies did not report NEC data [55,56], and in one study randomization was not declared [54].
All the studies reported NEC data in a form suitable for meta-analysis, except one [53], for which data included in a previous Cochrane review were used [59].  2), which means that 33 infants needed to be treated with probiotics in order to prevent one more case of NEC stage ≥2. Heterogeneity among trials was absent (I 2 = 0 %, p = 0.63; Fig. 2a). The funnel plot did not show any clear asymmetry (Fig. 2b).

VLBW infants
Twenty-two studies [

Probiotic strain
Lactobacillus GG was used in 2 studies [42,51] and Lactobacillus reuteri in 2 other studies [25,30]: the effect of these probiotics in reducing NEC was not significant, either for Lactobacillus GG and Lactobacillus  Fig. 4b]). One study used Lactobacillus sporogenes [34]. The results of all the studies including Lactobacilli were pooled, except for the study by Sari et al. [34]: Lactobacillus sporogenes is a species which has not an international recognition, shows characteristics of both genera Lactobacillus and Bacillus, and its strains should be better classified as Bacillus coagulans [60]. Thus, when the results of studies using Lactobacillus GG and reuteri were pooled, no significant reduction in the RR for NEC in the probiotic group was observed (0.62 [95 % CI 0.37-1.05], p = 0.07, Fig. 4c).

Study quality
Evaluation of the quality of the studies included in the meta-analysis according to the risk of bias tool as proposed by the Cochrane Collaboration is showed in Table 4, which also shows the level of evidence evaluated following the recommendations of the GRADE Working Group.

Discussion
The results of this systematic review and meta-analysis show an overall benefit of probiotic supplementation for the prevention of NEC in preterm infants. These results are strengthened by the absence of significant statistical heterogeneity among studies and by the low-risk of publication bias documented by the funnel plot.
However, despite the overall benefit, it is remarkable that the 26 studies included in the meta-analysis were extremely heterogeneous in terms of probiotic strain, dosage, duration of intervention, and target population. Furthermore, only few studies documented an effective colonization of the infants' gut with the probiotic strain. Thus, the proposal made by the authors of the recent Cochrane review of a "change in practice" in the use of probiotics in preterm infants [11] might require further investigation.
Currently available literature does not provide any definite conclusion on which probiotic strain should be used, and which group of preterm infants would benefit most from a probiotic intervention. It is important to note that the effect of a live-microorganism used as a probiotic is strictly strain-specific [61]. In this paper we aimed to perform strain-specific sub-meta-analyses but our efforts were weakened by the fact that in very few studies the same probiotic strain was used. For this reason, we were unable to draw definite conclusions on which single-strain of probiotics would be more effective in reducing NEC. When studies using single strains were pooled according to the probiotic genus, no significant effect was documented for Lactobacilli and Saccharomyces. This is partially in contrast with the recent Cochrane review on probiotics and NEC [11], which showed a beneficial effect of Lactobacilli: this discrepancy appears to be due mainly to differences in the studies included in the two sub-meta-analyses. Actually, the present meta-analysis included the study by Oncel et al. [25], which was on-going when the Cochrane review was published, but excluded the study by Manzoni et al. [57], where probiotics were used in addition to lactoferrin, and the study by Sari et al. [34], which used a probiotic product which is not properly a Lactobacillus [60].
The analysis of studies using Bifidobacteria showed a significant effect of Bifidobacterium breve in reducing NEC. This is also in contrast with the results of the Cochrane review; however, the discrepancy is explained by the inclusion in the present meta-analysis of the recent study by Dilli et al. [44], which appears to drive the beneficial effect documented for Bifidobacteria. Similarly to the Cochrane review [11], the analysis of studies in which more than one strain was used documented a strong and significant effect of these products in the prevention of NEC. No definite conclusion can be drawn from these results, even if it could be suggested that further research should be focused on mixed rather than on single-strain products; a potential rationale for this approach could be that a mix of strains might be more effective in providing an ecological barrier than a single strain.
The evidence that probiotics are effective in reducing NEC in VLBW infants does not necessarily apply also to extremely LBW infants (ELBWs), who are the highestrisk population. Only three studies [26,33,58] reported the rate of NEC in ELBWs: in two of these studies [33,58], the same number of ELBWs in the probiotic and control group developed NEC [33], while in the Pro-Prems trial NEC incidence was slightly lower in the probiotic group [26]. Given the relatively small number of ELBWs and the inconclusive results, no specific recommendation can be drawn from the analysis of these two studies. Similarly, no study reported separate data for intrauterine-growth-restricted (IUGR) infants, and thus  In the analysis of trials evaluating a specific intervention, it is pivotal to understand whether the results of these trials are generalizable or applicable only in specific clinical settings. According to our data, the common belief that probiotics are more effective in populations with a high rate of NEC [62] can be called into question: actually, when studies were divided according to NEC incidence in the control population, NEC reduction was striking and significant also when NEC rate in controls was extremely low. NEC rate in controls can be considered as a proxy for the quality of neonatal care: in this perspective, it is interesting to note that, in contrast with previous data, probiotics appear to confer a preventive benefit also in high quality-of-care settings. NEC rate in control populations was used for the analysis, rather than the baseline NEC rate stated by the authors and used in several studies for sample size calculation: this approach was considered more appropriate, because baseline NEC rate was not provided in many studies and, when provided, there was often a discrepancy with NEC rate detected in controls.
The analysis of included studies according to their microbiological quality points out that clinical studies aiming at evaluate the preventive effect of probiotics on NEC often lack an adequate microbiological assessment and this represents a major limitation of these studies. Actually, it is well known that the correct identification of a probiotic at species level corresponds to evaluate its safety, whereas the identification at strain level is extremely relevant as probiotic beneficial properties are strain-specific. Furthermore, the evaluation of probiotic colonisation, even if temporary,  The development of gut microbiota in preterm infants is known to be influenced by several factors, including gestational age, mode of delivery, diet, and antibiotic exposure [63]. All these factors are likely to be significant confounders in the relationship between probiotics and NEC: actually, it is well documented that infants fed maternal or donor breast milk have a lower risk of NEC compared to formula-fed infants [64], and that caesarean delivery is associated with a disruption in gut microbiota [65]. Quite surprisingly, however, in published studies data are not analyzed taking these confounders into account [66]. Given the definite protective role of human milk feeding and the symbiotic properties of human milk, it would be fundamental to understand whether the use of probiotics should be encouraged also in human-milk fed infants, or if this intervention should be directed towards exclusively formula-fed infants.
The studies included in the meta-analysis did not report any short-term adverse effect of probiotic supplementation (i.e., bloodstream infection with the probiotic strain). Growing evidence suggests the influence of gut microbiota on long-term health and disease, including both type 1 and type 2 diabetes mellitus, atherosclerosis, asthma, colon cancer, and inflammatory bowel disease [67]. However, at present little is known on the longterm outcome possibly related to the alteration of gut flora in preterm infants, which is the result of the supplementation with exogenous strains.
The choice to investigate a single outcome might be viewed as a limitation of the study: however, this choice was deliberate, as the literature search strategy was focused exclusively on NEC. Any speculation on different outcomes such as sepsis or mortality would have been inevitably misleading, because it would have been impossible to be sure to have identified all the studies reporting on those outcomes.

Conclusions
Meta-analyses give a valuable contribution in guiding researchers to focus future clinical studies on specific unanswered questions. The results of the present metaanalysis confirm that research on probiotics and NEC is on the right track, but also suggest that there are several unanswered questions which should be addressed before radically changing clinical practice. Our data highlight the need for further, well-designed studies aimed at clarifying the specific effect of probiotics in high-risk populations (i.e., ELBWs, IUGRs) and at addressing the choice of the most effective probiotic product, at the proper dose and duration of supplementation. For this reason, we encourage, for future studies, the publication of study protocols detailing study population and characteristics of the intervention, in order to narrow probiotic research to the most promising strains or combination of strains and to the most vulnerable populations, thus allowing a confirmative individual patient data analysis.

Competing interests
None of the authors has any conflict of interest to declare in connection with this paper.

Authors' contributions
All the authors, as part of the Task Force on Probiotics of the Italian Society of Neonatology, conceived and designed the study protocol. AA and LC performed the literature search and assessed study details, which were checked by DG. AA and DG evaluated study quality and performed the meta-analyses. MLC and LM evaluated microbiological quality of the included studies. AA and LC wrote the first draft of the paper, which was critically revised by all the other authors. All the authors gave final approval of the version to be submitted and agreed to be accountable for the whole paper.