What Is Known/What Is New
What Is Known
- Several randomized controlled trials show that probiotics reduce neonatal morbidity and mortality, but data are inconsistent.
- Multiple different probiotic strains or combinations have been used in these randomized controlled trials.
- Most existing meta-analyses group different strains and fail to adequately account for strain-specific effects.
What Is New
- Network meta-analysis shows that only a minority of probiotic strains have a statistically significant effect in reducing mortality and morbidity in preterm infants.
- The absence of significant effects may reflect a lack of adequately powered randomized controlled trials, or a genuine lack of efficacy for those species or strains.
Necrotizing enterocolitis (NEC) is a devastating disease in preterm infants strongly associated with gestational age, that has a variable incidence but typically averages 5% to 10% in infants weighing <1500 g (1). Mortality rates vary but range from 10% to 30%. Its pathogenesis is not completely understood as its occurrence may be the result of a variety of different etiologies (2–4), and early detection is difficult (5–7). Accumulating evidence shows that in addition to the effect of human milk feeding (8–10), probiotics may be important (11) in preventing NEC and reducing mortality (12,13). The role of the gut microbiota in the pathogenesis of NEC is, however, still unclear (14–16). Caution has therefore been advised until appropriately regulated safe products are available for use in this high-risk population (12,16,17). In its commentary published in 2010 on enteral nutrition in preterm infants, the European Society for Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) Committee on Nutrition did not recommend routine probiotics administration in infants <1800 g (18).
Numerous meta-analyses have recently been published summarizing a large number of randomized clinical trials (RCTs) (19–32). In almost all of these meta-analyses, the experimental group was constructed after pooling a wide variety of different probiotic strains as used in the original trials. This approach, however, does not give the clinician a meaningful answer as to which specific probiotic product has evidence-based efficacy. To partly overcome this problem, genus-specific meta-analyses have been performed as well (25,31). Also within genera or species there, however, may be significant differences in effectiveness depending on precise strain. Therefore, efficacy can only be evaluated at strain level and such meta-analyses have previously only been performed twice (21,22).
Classical pair-wise meta-analyses address the comparative effectiveness among similar or competing interventions against a common comparator (usually placebo or standard care). In addition, these omit direct evidence from the few studies that provide head-to-head comparisons on probiotic strains. Network meta-analyses (NMA), however, can address multiple interventions simultaneously. This method allows a system to visualize and statistically combine evidence from direct comparisons with evidence from indirect comparisons across several competing interventions. By using an NMA on data from RCTs of probiotics in preterm infants, our objective was to develop an updated and clinically meaningful understanding of the relative effectiveness of the different probiotic treatments.
Protocol Registration and Reporting Guidelines
This systematic review was registered on PROSPERO (international prospective registry of systematic reviews) http://www.crd.york.ac.uk/prospero/DisplayPDF.php?ID=CRD42017064847. This manuscript is conducted and reported according to the methods and recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Extension Statement for Reporting of Systematic Reviews incorporating NMA (33).
Inclusion criteria were studies that included only preterm infants, or studies that reported subgroups between term and preterm infants so that only results from the latter could be included. Preterm birth was defined as gestational age of <37 weeks. Studies comparing probiotic treatment against placebo, usual care, or head-to-head with a different probiotic regime were considered eligible. If a study intervention consisted of the combination of a probiotic strain together with another form of intervention, such as prebiotics or lactoferrin, the study was only included if a control group was included which received the same nonprobiotic intervention (eg, prebiotics or lactoferrin), but without the probiotic intervention. Studies could be included when infants were fed own mother's milk, donor milk or formula. Single or multiple strain studies were included. Studies were included if they reported well-described outcome reports of NEC (with Bell stages included (34–36)), blood-culture–proven late-onset sepsis (LOS), postnatal age at reaching full enteral feeding (150 mL · kg−1 · day−1), or in-hospital mortality. Other outcomes were not considered in this report. We included only RCTs (including cluster RCTs), which were fully published and in the English language. Complete blinding was not considered as mandatory. The results from nonrandomized studies, conference papers, abstracts, or other nonpublished studies were not included.
A systematic literature search was conducted using the PubMed database from inception to September 19, 2017. Search terms were: probiotic∗ AND (premature OR preterm OR neonat∗ OR infant∗), with a limit on “Clinical Trial.” In addition, reference lists of previously published meta-analyses were screened to identify additional eligible studies. The literature search was conducted independently by 2 reviewers (C.H.v.d.A. and J.B.v.G.). Inconsistency whether to include a study was resolved by discussion.
Two round table meetings of the group (May 2016, Athens and May 2017, Prague) were held to achieve consensus on the approach, outcomes assessed, methods and to resolve all differences in interpretation of the methodological assessment and results of the eligible trials.
From the eligible studies information regarding inclusion criteria, study groups, key characteristics, and outcomes was extracted by the 2 reviewers independently using a standardized data collection form. Missing data were requested by contacting the authors. Probiotic strains were identified at strain levels (eg, Lactobacillus rhamnosus GG ATCC 53103 or Bifidobacterium animalis subspecies lactis Bb-12), although often this was not available in the original reports. If no e-mail reply was received from the original authors, strain numbers were obtained if possible through contacting manufacturers or internet searches on available product names.
For the remainder of this manuscript probiotic strains are truncated at their genus: Bacillus, Bifidobacterium, Enterococcus, Lactobacillus, Saccharomyces, and Streptococcus are denoted by Ba, B, E, L, Sa, and S, respectively. In addition, subspecies (subsp) names are truncated as well: B animalis subsp lactis is denoted as B lactis; B bifidum subsp infantis as B infantis; B bifidum subsp longum as B longum, and S salivarius subsp thermophilus as S thermophilus. During the past decades multiple reclassifications in taxonomy have been proposed and designations in the historical publications may no longer be accurate. We therefore adhered to the latest nomenclature we were aware of. For example, B infantis 35624 is designated as B longum 35624 (37), B bifidum Bb-12 as B lactis Bb-12 (38), and L sporogenes as Ba coagulans(39). Because the L reuteri strain DSM 17938 is a daughter strain of L reuteri ATCC 55730 in which only resistance plasmids were removed but other characteristics are maintained (40), these strains are analyzed together. Similarly, as B lactis B94 shares many characteristics with B lactis Bb-12 (41), these strains are also analyzed together in our NMA. Except for in our study summary Table 1 , all control groups will be further denoted as placebo, whether true placebo was used, or whether usual care was given.
In case continuous data (ie, data on time until full enteral feeding [TUFEF]) were originally presented as medians with corresponding interquartile percentiles or outer ranges, data were converted to estimated means with standard deviations, as previously described (42,43). Obtained data were compared to previously published systematic reviews and discrepancies were rechecked in the original trials.
Assessment of Study Quality
To assess the methodological quality of the included RCTs the Cochrane Collaboration's tool for assessing risk of bias was used. The tool includes the following criteria: adequacy of sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, incomplete outcome data, selective reporting, and other biases (43). Noteworthy, studies were not excluded based on these results. Items were scored as low, high, or unknown risk of bias. Because of possible probiotic cross-colonization from intervention to control groups, the item on other biases was scored as unknown, except in case of cluster RCTs. Publication bias was assessed by constructing funnel plots for each outcome.
Studies were compared between every studied probiotic strain or combination of studied strains via a comprehensive NMA based on the Bayesian theorem (44,45). This approach can be considered to be an extension of the traditional pair-wise meta-analysis, as it incorporates both direct and indirect information through a common comparator (most often placebo or routine care) to obtain estimates of the relative interventional effects on multiple intervention comparisons.
Network graphs were constructed for each outcome variable. They consisted of nodes (points representing the competing interventions) and edges (adjoining lines between the nodes that show which interventions have been compared among the included studies). The size of the nodes represents the number of infants that received the intervention. The thickness of the edges represents the number of comparative studies between the respective nodes. Reviewing the network geometry graphically summarizes how the evidence base is built up and whether various strains were directly compared or only through indirect network evidence.
We made use of the Aggregate Data Drug Information System 2 (ADDIS2, release 13, Groningen, the Netherlands) software (46,47), which is an open-source online application based on R statistical software (48). The NMA was conducted using a Bayesian framework in combination with a Markov chain Monte-Carlo simulation. Given that most treatments included a limited number of RCTs, we assumed between-study heterogeneity, although this could not formally be tested. Therefore, a conservative random effect approach was employed in our NMA models. The NMA was run with an outcome scale of 5 to set default prior distributions accordingly. Run-lengths were based on at least 80,000 inference iterations for each 4 chains with a burn-in period of the first 10,000 iterations combined with a thinning factor of 10. Convergence was assessed using the Brooks-Gelman-Rubin diagnostic to reassure that all potential scale reduction factors (PSRFs) remained below 1.05. In the rare occasion that the PSRF approached 1.05 or was higher than 1.05, run-length was increased accordingly to improve model convergence.
Relative treatment effect plots were constructed for each studied probiotic strain (or combination of multiple strains) versus placebo. Dichotomous outcomes are expressed as risk ratios (RRs) with their 95% credible interval (95% CrI). Continuous outcome measures are expressed as mean differences with their 95% CrI.
Besides the relative effect plots in the NMA, we constructed classic pair-wise forest plots for those strains or combinations which were compared to placebo if they showed either significant efficacy in the NMA or were evaluated in at least 2 RCTs and at least 250 infants were included in the treatment arm. These forest plots were constructed in RevMan (49) and visualize how the NMA evidence base is built up.
Consistency of the NMA model between direct and indirect network evidence was tested with the node-splitting method (50). A known drawback from this approach, however, is that it cannot properly handle multiarm studies (44). To give insight in possible publication biases, we constructed funnel plots for all studies that compared placebo versus any probiotic strain or combination for each outcome separately.
Our PubMed search yielded 515 citations; a flow diagram on the screening and eligibility process is presented in Fig. 1. In total 56 articles (for 49 different RCTs) were identified which fulfilled our inclusion criteria and reported usable extractable outcome data on at least 1 of our predefined outcome domains (51–106). Five of these articles only showed overlapping data and were further discarded (102–106). Two studies were either a substudy (65) from a multicenter study (51), or an extension (75) of an already published trial (74); from the substudies only data not elsewhere reported was entered in our database. One small RCT only described that NEC, sepsis, and mortality incidence were not different between groups (107), but no response was received from the authors after requesting original data. Eight studies were identified with appropriate study design and inclusion criteria, although no data on our outcome domains were described (108–113) or only long-term follow-up data was presented (114,115). In 4 studies, the use of probiotics was not the only difference between intervention and control groups, that is, prebiotics were used only in the intervention group (116,117), or only the control group received nystatin instead of placebo (118,119). One study included both preterm and term neonates and results were not split between subgroups (120). The studies from which no suitable data on our outcome domains could be extracted were therefore excluded from further analysis in this report (107–120). From the 51 included articles (51–101), data from 11,231 preterm infants could potentially be extracted. Table 1 shows study details and main characteristics of the included patients. Based on broadly overlapping basic inclusion criteria and patient characteristics across the various trials we assumed transitivity (33,44). Identified strain numbers are summarized in Appendix Table S1 (Supplemental Digital Content, http://links.lww.com/MPG/B273). Unfortunately, we were not able to retrieve them for all microorganisms. Results for which no strain number is available must therefore be interpreted with caution as different products may have been used. Appendix Figures S1A&B (Supplemental Digital Content, http://links.lww.com/MPG/B273) show the risk of bias assessment.
Figure 2 shows the network graph comparing the probiotic strains or combinations as used in the original trials for the reduction of in-hospital mortality. The network geometry shows the evidence base comparing 25 different treatments (n = 4788 in total in 39 arms; mortality incidence 5.1%) versus the common comparator placebo (n = 4512 in 36 trials; mortality incidence 7.0%). Original mortality incidence data from all included studies is shown in Appendix Table S2 (Supplemental Digital Content, http://links.lww.com/MPG/B273). There were only 4 direct head-to-head comparisons. Three interventions had to be excluded from the quantitative analysis because there were zero events in the placebo or intervention groups which is not compatible with NMA analyses. In 1 study using Sa boulardii CNCM I-3799 (100), there were no mortality cases in both the intervention and control groups. Three studies (2 treatments: B bifidum OLB6378 (95); and B breve M-16 V (67,81)) were excluded because there were zero events in the intervention groups, although mortality rates in the placebo groups from these studies amounted 2/153 and 4/176, respectively (events/N).
Figure 3 shows the relative-effects plot for efficacy in reducing mortality of the various tested probiotic strains or tested combinations versus placebo treatment. It shows that the RRs for mortality are significantly reduced for 3 interventions: for the combination of B bifidum NCDO 1453 and L acidophilus NCDO 1748 (based on 2 studies with 494 infants (72,87)); the combination of B bifidum, B infantis, B longum, and L acidophilus (based on 1 study with 186 infants (88)); and the combination of B infantis, L acidophilus, L casei, L plantarum, L rhamnosus, and S thermophilus altogether (based on 1 study with 150 infants (63)); see also Table 2. Separate pair-wise forest plots are shown in Appendix Figure S2 (Supplemental Digital Content, http://links.lww.com/MPG/B273) for probiotic strains or combinations which were significant or were tested in at least 250 infants versus placebo. Node-splitting detection of inconsistency was not possible in this model due to only head-to-head trials from multiple arm studies, or head-to-head comparisons with 0 events in 1 or more study arms (50). A funnel plot shows no clear evidence of a publication bias (Appendix Figure S3, Supplemental Digital Content, http://links.lww.com/MPG/B273).
Figure 4 shows the network graph comparing the probiotic strains or combinations as they were used in the original trials for the prevention of NEC grades 2 or 3. The network geometry shows the evidence base comparing 25 different treatments (n = 5550 in total in 50 arms; NEC incidence 3.2%) versus the common comparator placebo (n = 5101 in 43 trials; NEC incidence 6.1%). Original NEC incidence data from all included studies is shown in Appendix Table S3 (Supplemental Digital Content, http://links.lww.com/MPG/B273). There were only 6 direct head-to-head comparisons. Three studies (using Ba clausii 4 strains (94); B bifidum OLB6378 (95); and Sa boulardii CNCM I-3799 (100)) had to be excluded from the NMA because there were zero events in both the placebo and intervention groups. Seven studies (4 treatments: B breve M-16 V (64,67,81,98); B breve and L casei(55); B infantis PTA-5843, E faecium PTA-5844, and L gasseri PTA-5845 (70); and L acidophilus Lb (53)) were excluded because there were zero events in the intervention groups, although NEC rates in the placebo groups from these studies amounted 1/127, 4/112, 5/40, and 5/16, respectively (events/N).
Figure 5 shows the relative-effects plot for efficacy in reducing NEC grade 2 or 3 of the various tested probiotic strains or tested combinations versus placebo treatment. It shows that the RRs for NEC are significantly reduced for 7 treatments: B lactis Bb-12 or B94 (based on 5 trials with 828 infants (61,66,76,78,93)); L reuteri ATCC 55730 or DSM 17938 (based on 4 studies with 1459 infants (79,83,84,91)); L rhamnosus GG (based on 6 studies with 1507 infants (56,59,73,75,84,96)); the combination of B bifidum, B infantis, B longum, and L acidophilus (based on 2 studies with 247 infants (88,96)); the combination of B infantis ATCC 15697 and L acidophilus ATCC 4356 (based on 1 study with 367 infants (71)); the combination of B infantis Bb-02, B lactis Bb-12, and S. thermophilus TH-4 (based on 2 studies with 1244 infants (54,69)); and the combination of B longum 35624 and L rhamnosus GG (based on 2 studies with 285 infants (51,97)); see also Table 2. Separate pair-wise forest plots are shown in Figure S4 (Supplemental Digital Content, http://links.lww.com/MPG/B273) for probiotic strains or combinations which were significant or were tested in at least 250 infants versus placebo. Node-splitting models did not show substantial differences between direct and indirect evidence, so that the consistency model holds (Appendix Figure S5, Supplemental Digital Content, http://links.lww.com/MPG/B273) (50). The funnel plot shows no clear evidence of a publication bias (Appendix Figure S6, Supplemental Digital Content, http://links.lww.com/MPG/B273).
Figure 6 shows the network graph comparing the probiotic strains or combinations as they were used in the original trials for the prevention of culture proven LOS. The network geometry shows the evidence base comparing 25 different treatments (n = 5576 in total in 52 arms; sepsis incidence 15.4%) versus the common comparator placebo (n = 5049 in 45 trials; sepsis incidence 24.9%). Original LOS incidence data from all included studies is shown in Appendix Table S4 (Supplemental Digital Content, http://links.lww.com/MPG/B273). There were only 6 direct head-to-head comparisons. One intervention using Sa boulardii CNCM I-3799 had to be excluded because there were zero events in both the placebo and intervention groups (100).
Figure 7 shows the relative-effects plot for efficacy in reducing LOS of the various tested probiotic strains or tested combinations versus placebo. It shows that the RRs for LOS are significantly reduced for 2 probiotic treatments: for the combination of B bifidum, B infantis, B longum, and L acidophilus (based on 2 studies with 247 infants (88,96)); and for the combination of B longum R00175, L helveticus R0052, L rhamnosus R0011, and Sa boulardii CNCM I-1079 (based on 3 studies with 241 infants (52,62,92)); see also Table 2. Separate pair-wise forest plots are shown in Appendix Figure S7 (Supplemental Digital Content, http://links.lww.com/MPG/B273) for probiotic strains or combinations which were significant or tested in at least 250 infants versus placebo. Node-splitting models did not show substantial differences between direct and indirect evidence, so that the consistency model holds (Appendix Figure S8, Supplemental Digital Content, http://links.lww.com/MPG/B273) (50). The funnel plot shows no clear evidence of a publication bias (Appendix Figure S9, Supplemental Digital Content, http://links.lww.com/MPG/B273).
Time Until Full Enteral Feeding
Figure 8 shows the network graph comparing the probiotic strains or combinations as they were used in the original trials to reduce TUFEF. The network geometry shows the evidence base comparing 13 different treatments (n = 3122 in 24 arms) versus the common comparator placebo (n = 2988 in 21 trials). There were only 2 direct head-to-head comparisons. Original data from all included studies is shown in Appendix Table S5 (Supplemental Digital Content, http://links.lww.com/MPG/B273). Figure 9 shows a relative-effects plot for efficacy in reducing TUFEF of the various tested probiotic strains or tested combinations versus placebo treatment. It shows that the mean difference for TUFEF is significantly reduced for 3 interventions: L reuteri ATCC 55730 or DSM 17938 (based on 3 studies with 626 infants (79,84,91)); and for the combination of B bifidum, B infantis, B longum, and L acidophilus (based on 2 studies with 247 infants (88,96)); and for the combination of B longum BB536 and L rhamnosus GG (based on 1 study with 94 infants (85)); see also Table 2. Separate pair-wise forest plots are shown in Figure S10 (Supplemental Digital Content, http://links.lww.com/MPG/B273) for probiotic strains or combinations which were significant or tested in at least 250 infants versus placebo. Node-splitting models did not show substantial differences between direct and indirect evidence, so that the consistency model holds (Appendix Figure S11, Supplemental Digital Content, http://links.lww.com/MPG/B273) (50). From the funnel plot, a publication bias could not be excluded, as no clear triangular shape could be identified from included studies (Appendix Figure S12, Supplemental Digital Content, http://links.lww.com/MPG/B273).
By using the approach of an NMA, we were able to determine, based on the current literature, which tested probiotic strains were most effective, and which were not, in reducing mortality and morbidity in preterm infants. Only 3 of 25 studied probiotic treatments showed significant reduction in mortality rates. Seven treatments reduced NEC incidence, 2 reduced LOS, and 3 reduced TUFEF. There was no clear overlap of certain strains which were significantly effective on multiple outcome domains. Most strains or combination of strains only showed trends towards efficacy, whereas other strains did not demonstrate efficacy (such as Sa boulardii CNCM I-745 and B breve BBG-001). A lack of effect may either be due to understudied species or a true lack of effect of certain strains.
Although the total number of 51 RCTs included with over 11,000 infants is considerable (51–101), many of the different probiotic treatments were only evaluated in 1 or 2 trials. Only 5 strains were studied in at least 4 RCTs (see also tables S2–5): B breve M-16 V, B lactis Bb-12 or B94, L reuteri ATCC 55730 or DSM 17938, L rhamnosus GG ATCC 53103, and Sa boulardii CNCM I-745. In addition, the evidence base was frequently dependant on small, lower quality, or outdated studies as can be seen in the pair-wise forest plots. We chose to include moderately preterm infants in our analyses noting that the number of studies which focused on the smallest infants was limited. In only 4 of the 51 included studies (compromising 3 different RCTs), the mean birth weight was below 1000 g (Table 1 ) (51,65,76,97). In 10 studies, the mean birth weight was at least 1500 g. There were many studies with unclear risks of bias for the various domains, and 8 studies with high risks in at least 1 domain (64,67,70,82,84,87,98,101). Most of the studies with a high risk of bias assessment did not contribute to the significant strains. An unclear risk indicates that the risk item was not clearly described, but does not necessarily indicate a potentially flawed study. We decided not to exclude studies purely on the basis of the quality assessment criteria, but allow readers to make their own interpretation of the evidence base. A visual inspection of the evidence bases by means of the funnel plots did not show a clear publication bias for most outcomes, although for TUFEF no triangular shape could be identified.
In most of the original studies the primary outcome was not on one of our outcome domains (Table 1 ), but on for example stool colonization, growth rates, or not reported even more frequently. If one performs a power calculation (α 5%; 1-β 80%) on reducing mortality rates while taking the observed rates in our manuscript (7.0% vs 5.1%), one would need almost 2500 infants for each studied strain. For NEC, to demonstrate a reduction from 6.1% to 3.2%, one would need >800 infants per group. These high inclusion rates have not been reached for these outcomes. One must therefore realize that the results here presented are based on exploratory data analysis and thus only have hypothesis generating power.
It must be noted that some interesting results were produced. For example, both L rhamnosus GG and B lactis Bb-12/B94 appeared to be effective in reducing NEC (Fig. 5 and Figure S4). In addition, B longum BB536 showed a clear trend towards a similar effect. Both the combination of L rhamnosus GG with B longum BB536 and the combination of B lactis Bb-12 with B longum BB536, however, showed no measurable effect. This may reflect an antagonistic effect of B longum BB536 together with the other 2 strains, or the relatively poor evidence base on which this NMA is built. A similar pattern was seen with these strains in the reduction of LOS, although much less pronounced. Somewhat un-expectedly, only L rhamnosus GG simultaneously administered with B longum BB536 was able to reduce TUFEF significantly (based on 1 study with 94 infants studied (85)), whereas L rhamnosus GG alone was not (Fig. 9 and Figure S10).
The classic strain versus placebo forest plots show that the effect size was more or less similar to evidence from relative effect plots in the NMA. The small differences can be explained due to the Bayesian statistical approach versus the classic frequentist random-effects models, but it is known that Bayesian techniques better account for trial heterogeneity. In addition, the NMA gained extra power from indirect network evidence, from studies in which there was no suitable placebo group but only provided head-to-head comparisons (96,101), and from studies in which data from multiple treatment arms could be included (66,84). Unfortunately, most network evidence was based on indirect comparisons as the vast majority only compared treatment versus placebo or routine care. There were only a few head-to-head trials available that could be tested for inconsistency by means of the node-splitting method. Nevertheless, inconsistency between direct and indirect network evidence was not apparent in tested cases.
As is the case for almost any meta-analysis, there are small differences in study design in terms of inclusion criteria (eg, birth weight, gestational age, degree of growth restriction) or drug administration regimens (initiation, duration, and dosing). Although most studies did not exclude infants depending on their dietary exposures (formula, own mother's milk, or donor milk), some studies only included infants who either received only breast milk, or only formula. The magnitude of how this affected results is unknown. Nevertheless, transitivity was assumed as inclusion criteria were all broadly overlapping. A further bias could be that we included only RCTs published in the English language due to language barriers for reviewing. On the other hand, most RCTs from other settings such as for example China used different combinations of strains that were not tested in the RCTs included here (29). A major strength of our analysis is that we paid meticulous attention to retrieve the correct probiotic strain in the included RCTs. Regrettably, strain numbers were frequently not mentioned in the original manuscripts and could sometimes not be retrieved despite contacting original authors or companies. The results from species without further designation should therefore be interpreted with caution. It is no longer acceptable in current studies to omit a clear description of the used probiotic drug at subspecies level with strain number according to the latest taxonomic nomenclature (17,121). In addition, many commercial products turned out to contain different bacterial strains than were included on the ingredient list (17,122). Future studies should therefore validate their studied probiotic strains and exclude contamination by other strains. Apart from efficacy, a high degree of quality control and assurance is mandatory as probiotic related sepsis has been regularly reported in preterm infants who can be considered immuno-compromised patients (17,31). An additional safety issue could be that some probiotics strains carry antibiotic resistance genes themselves, and could thus have the potential to pass the antibiotic resistance genes to pathogenic bacteria through horizontal gene transfer (123). These elements need to be taken into account when balancing supposed or true beneficial and harmful effects.
To conclude, our efforts in this study were to present an overview of all published evidence on the use of probiotics in preterm infants at a strain level, and to identify the most promising strains. Most strains were unfortunately only studied once or a few times. In addition, the number of reports in the most preterm neonates was limited. Furthermore, it was not possible to determine optimal probiotic dosages, time of initiation, and duration of treatment course. Nevertheless, we believe that our approach of a strain-specific NMA gives a much more meaningful answer than previously performed meta-analyses in which all probiotic strains were analyzed as one group, or were grouped at a genus or species level, as even these latter analyses do not address strain-specific characteristics. The NMA allowed us to identify potential strains that can reduce NEC and mortality incidence in this vulnerable population. Still, our major and rather disappointing conclusion, is that >10 years from the first RCTs showing that probiotics may reduce a disease as serious as NEC, we remain unable to clearly identify the optimal strain, dose or combination, and that clinicians are left using inadequately tested, potentially un-safe and possibly ineffective treatments.
1. Niño DF, Sodhi CP, Hackam DJ. Necrotizing enterocolitis
: new insights into pathogenesis and mechanisms. Nat Rev Gastroenterol Hepatol
2. Denning TL, Bhatia AM, Kane AF, et al. Pathogenesis of NEC: role of the innate and adaptive immune response. Semin Perinatol
3. Eaton S, Rees CM, Hall NJ. Current research on the epidemiology, pathogenesis, and management of necrotizing enterocolitis
4. Frost BL, Modi BP, Jaksic T, et al. New medical and surgical insights into neonatal necrotizing enterocolitis
: a review. JAMA Pediatr
5. Samuels N, van de Graaf RA, de Jonge RCJ, et al. Risk factors for necrotizing enterocolitis
in neonates: a systematic review of prognostic studies. BMC Pediatr
6. Terrin G, Stronati L, Cucchiara S, et al. Serum markers of necrotizing enterocolitis
: a systematic review. J Pediatr Gastroenterol Nutr
7. Rusconi B, Good M, Warner BB. The microbiome and biomarkers for necrotizing enterocolitis
: are we any closer to prediction? J Pediatr
8. Corpeleijn WE, Kouwenhoven SM, Paap MC, et al. Intake of own mother's milk during the first days of life is associated with decreased morbidity and mortality in very low birth weight infants during the first 60 days of life. Neonatology
9. O’Connor DL, Gibbins S, Kiss A, et al. Effect of supplemental donor human milk compared with preterm formula on neurodevelopment of very low-birth-weight infants at 18 months: a randomized clinical trial. JAMA
10. Shulhan J, Dicken B, Hartling L, et al. Current knowledge of necrotizing enterocolitis
in preterm infants and the impact of different types of enteral nutrition products. Adv Nutr
11. Hill C, Guarner F, Reid G, et al. Expert consensus document. The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic. Nat Rev Gastroenterol Hepatol
12. Underwood MA. Impact of probiotics on necrotizing enterocolitis
. Semin Perinatol
13. Yu W, Sui W, Mu L, et al. Preventing necrotizing enterocolitis
by food additives in neonates: A network meta-analysis revealing the efficacy and safety. Medicine (Baltimore)
14. Peterson LW, Artis D. Intestinal epithelial cells: regulators of barrier function and immune homeostasis. Nat Rev Immunol
15. Pammi M, Cope J, Tarr PI, et al. Intestinal dysbiosis in preterm infants preceding necrotizing enterocolitis
: a systematic review and meta-analysis. Microbiome
16. Warner BB, Tarr PI. Necrotizing enterocolitis
and preterm infant gut bacteria. Semin Fetal Neonatal Med
17. Kolaček S, Hojsak I, Canani RB, et al. ESPGHAN Working Group for Probiotics and Prebiotics. Commercial probiotic products: a call for improved quality control. A Position Paper by the ESPGHAN Working Group for Probiotics and Prebiotics. J Pediatr Gastroenterol Nutr
18. Agostoni C, Buonocore G, Carnielli VP, et al. Enteral nutrient supply for preterm infants: commentary from the European Society of Paediatric Gastroenterology, Hepatology and Nutrition Committee on Nutrition. J Pediatr Gastroenterol Nutr
19. Aceti A, Gori D, Barone G, et al. Probiotics and time to achieve full enteral feeding in human milk-fed and formula-fed preterm infants: systematic review and meta-analysis. Nutrients
2016; 8:pii: E471. doi: 10.3390/nu8080471.
20. Aceti A, Maggio L, Beghetti I, et al. Probiotics prevent late-onset sepsis
in human milk-fed, very low birth weight preterm infants: systematic review and meta-analysis. Nutrients
2017; 9:pii: E904. doi: 10.3390/nu9080904.
21. Athalye-Jape G, Rao S, Patole S. Lactobacillus reuteri
DSM 17938 as a probiotic for preterm neonates: a strain-specific systematic review. JPEN J Parenter Enteral Nutr
22. Athalye-Jape G, Rao S, Simmer K, et al. Bifidobacterium breve
M-16 V as a probiotic for preterm infants: a strain-specific systematic review. JPEN J Parenter Enteral Nutr
2017; doi: 10.1177/0148607117722749.
23. Baucells BJ, Mercadal Hally M, Alvarez Sanchez AT, et al. Probiotic associations in the prevention of necrotising enterocolitis and the reduction of late-onset sepsis
and neonatal mortality in preterm infants under 1,500 g: A systematic review. An Pediatr (Barc)
24. Billimoria ZC, Pandya S, Bhatt P, et al. Probiotics—to use, or not to use? An updated meta-analysis. Clin Pediatr (Phila)
25. Chang HY, Chen JH, Chang JH, et al. Multiple strains probiotics appear to be the most effective probiotics in the prevention of necrotizing enterocolitis
and mortality: an updated meta-analysis. PLoS One
26. Dermyshi E, Wang Y, Yan C, et al. The “Golden Age” of probiotics: a systematic review and meta-analysis of randomized and observational studies in preterm infants. Neonatology
27. Hu HJ, Zhang GQ, Zhang Q, et al. Probiotics prevent Candida colonization and invasive fungal sepsis
in preterm neonates: a systematic review and meta-analysis of randomized controlled trials. Pediatr Neonatol
28. Rao SC, Athalye-Jape GK, Deshpande GC, et al. Probiotic supplementation and late-onset sepsis
in preterm infants: a meta-analysis. Pediatrics
29. Sawh SC, Deshpande S, Jansen S, et al. Prevention of necrotizing enterocolitis
with probiotics: a systematic review and meta-analysis. PeerJ
30. Sun J, Marwah G, Westgarth M, et al. Effects of probiotics on necrotizing enterocolitis
, intraventricular hemorrhage, mortality, length of hospital stay, and weight gain in very preterm infants: a meta-analysis. Adv Nutr
31. Thomas JP, Raine T, Reddy S, et al. Probiotics for the prevention of necrotising enterocolitis in very low-birth-weight infants: a meta-analysis and systematic review. Acta Paediatr
32. Zhang GQ, Hu HJ, Liu CY, et al. Probiotics for preventing late-onset sepsis
in preterm neonates: a PRISMA-compliant systematic review and meta-analysis of randomized controlled trials. Medicine (Baltimore)
33. Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: checklist and explanations. Ann Intern Med
34. Bell MJ, Ternberg JL, Feigin RD, et al. Neonatal necrotizing enterocolitis
. Therapeutic decisions based upon clinical staging. Ann Surg
35. Walsh MC, Kliegman RM. Necrotizing enterocolitis
: treatment based on staging criteria. Pediatr Clin North Am
36. Kliegman RM, Walsh MC. Neonatal necrotizing enterocolitis
: pathogenesis, classification, and spectrum of illness. Curr Probl Pediatr
37. Altmann F, Kosma P, O’Callaghan A, et al. Genome analysis and characterisation of the exopolysaccharide produced by Bifidobacterium longum
subsp. longum 35624. PLoS One
38. European Food Safety Authority (EFSA). Scientific opinion on the substantiation of health claims related to Bifidobacterium animalis
ssp. lactis Bb-12. EFSA J
39. Drago L, De Vecchi E. Should Lactobacillus sporogenes
and Bacillus coagulans
have a future? J Chemother
40. Rosander A, Connolly E, Roos S. Removal of antibiotic resistance gene-carrying plasmids from Lactobacillus reuteri
ATCC 55730 and characterization of the resulting daughter strain, L. reuteri
DSM 17938. Appl Environ Microbiol
41. Crittenden RG, Morris LF, Harvey ML, et al. Selection of a Bifidobacterium strain to complement resistant starch in a synbiotic yoghurt. J Appl Microbiol
42. Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol
43. The Cochrane Collaboration, Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions. Version 5.1.0 [updated March 2011]. 2011; Available at: http://handbook.cochrane.orgs
. Accessed September 4, 2017.
44. Efthimiou O, Debray TP, van Valkenhoef G, et al. GetReal in network meta-analysis: a review of the methodology. Res Synth Methods
45. Turner RM, Davey J, Clarke MJ, et al. Predicting the extent of heterogeneity in meta-analysis, using empirical data from the Cochrane Database of Systematic Reviews. Int J Epidemiol
46. Drugis.org. Aggregate Data Drug Information System (ADDIS), version 2. https://drugis.org/software/addis2/index
47. van Valkenhoef G, Lu G, de Brock B, et al. Automating network meta-analysis. Res Synth Methods
48. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria; 2017. Accessed at: https://www.R-project.org
49. Review Manager (RevMan) [Computer program]. Version 5.3.5. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration; 2014.
50. van Valkenhoef G, Dias S, Ades AE, et al. Automated generation of node-splitting models for assessment of inconsistency in network meta-analysis. Res Synth Methods
51. Al-Hosni M, Duenas M, Hawk M, et al. Probiotics-supplemented feeding in extremely low-birth-weight infants. J Perinatol
52. Arora S, Khurana MS, Saini R. To study the role of probiotics in the prevention of necrotizing enterocolitis
in preterm neonates. Int J Contemp Pediatr
53. Awad H, Mokhtar H, Imam SS, et al. Comparison between killed and living probiotic usage versus placebo for the prevention of necrotizing enterocolitis
in neonates. Pak J Biol Sci
54. Bin-Nun A, Bromiker R, Wilschanski M, et al. Oral probiotics prevent necrotizing enterocolitis
in very low birth weight neonates. J Pediatr
55. Braga TD, da Silva GA, de Lira PI, et al. Efficacy of Bifidobacterium breve
and Lactobacillus casei
oral supplementation on necrotizing enterocolitis
in very-low-birth-weight preterm infants: a double-blind, randomized, controlled trial. Am J Clin Nutr
56. Chrzanowska-Liszewska D, Seliga-Siwecka J, Kornacka MK. The effect of Lactobacillus rhamnosus
GG supplemented enteral feeding on the microbiotic flora of preterm infants-double blinded randomized control trial. Early Hum Dev
57. Costalos C, Skouteri V, Gounaris A, et al. Enteral feeding of premature infants with Saccharomyces boulardii
. Early Hum Dev
58. Costeloe K, Hardy P, Juszczak E, et al. Bifidobacterium breve
BBG-001 in very preterm infants: a randomised controlled phase 3 trial. Lancet
59. Dani C, Biadaioli R, Bertini G, et al. Probiotics feeding in prevention of urinary tract infection, bacterial sepsis
and necrotizing enterocolitis
in preterm infants. A prospective double-blind study. Biol Neonate
60. Demirel G, Erdeve O, Celik IH, et al. Saccharomyces boulardii
for prevention of necrotizing enterocolitis
in preterm infants: a randomized, controlled study. Acta Paediatr
61. Dilli D, Aydin B, Fettah ND, et al. The propre-save study: effects of probiotics and prebiotics alone or combined on necrotizing enterocolitis
in very low birth weight infants. J Pediatr
62. Dutta S, Ray P, Narang A. Comparison of stool colonization in premature infants by three dose regimes of a probiotic combination: a randomized controlled trial. Am J Perinatol
63. Fernández-Carrocera LA, Solis-Herrera A, Cabanillas-Ayon M, et al. Double-blind, randomised clinical assay to evaluate the efficacy of probiotics in preterm newborns weighing less than 1500 g in the prevention of necrotising enterocolitis. Arch Dis Child Fetal Neonatal Ed
64. Fujii T, Ohtsuka Y, Lee T, et al. Bifidobacterium breve enhances transforming growth factor beta1 signaling by regulating Smad7 expression in preterm infants. J Pediatr Gastroenterol Nutr
65. Havranek T, Al-Hosni M, Armbrecht E. Probiotics supplementation increases intestinal blood flow velocity in extremely low birth weight preterm infants. J Perinatol
66. Hays S, Jacquot A, Gauthier H, et al. Probiotics and growth in preterm infants: a randomized controlled trial, PREMAPRO study. Clin Nutr
67. Hikaru U, Koichi S, Yayoi S, et al. Bifidobacteria prevents preterm infants from developing infection and sepsis
. Int J Probiotics Prebiotics
68. Indrio F, Riezzo G, Tafuri S, et al. Probiotic supplementation in preterm: feeding intolerance and hospital cost. Nutrients
69. Jacobs SE, Tobin JM, Opie GF, et al. Probiotic effects on late-onset sepsis
in very preterm infants: a randomized controlled trial. Pediatrics
70. Kanic Z, Micetic Turk D, Burja S, et al. Influence of a combination of probiotics on bacterial infections in very low birthweight newborns. Wien Klin Wochenschr
2015; 127 (suppl 5):S210–S215.
71. Lin HC, Su BH, Chen AC, et al. Oral probiotics reduce the incidence and severity of necrotizing enterocolitis
in very low birth weight infants. Pediatrics
72. Lin HC, Hsu CH, Chen HL, et al. Oral probiotics prevent necrotizing enterocolitis
in very low birth weight preterm infants: a multicenter, randomized, controlled trial. Pediatrics
73. Manzoni P, Mostert M, Leonessa ML, et al. Oral supplementation with Lactobacillus casei
prevents enteric colonization by Candida species in preterm neonates: a randomized study. Clin Infect Dis
74. Manzoni P, Rinaldi M, Cattani S, et al. Bovine lactoferrin supplementation for prevention of late-onset sepsis
in very low-birth-weight neonates: a randomized trial. JAMA
75. Manzoni P, Meyer M, Stolfi I, et al. Bovine lactoferrin supplementation for prevention of necrotizing enterocolitis
in very-low-birth-weight neonates: a randomized clinical trial. Early Hum Dev
2014; 90 (suppl 1):S60–S65.
76. Mihatsch WA, Vossbeck S, Eikmanns B, et al. Effect of Bifidobacterium lactis
on the incidence of nosocomial infections in very-low-birth-weight infants: a randomized controlled trial. Neonatology
77. Millar MR, Bacon C, Smith SL, et al. Enteral feeding of premature infants with Lactobacillus GG. Arch Dis Child
1993; 69 (5 Spec No):483–487.
78. Mohan R, Koebnick C, Schildt J, et al. Effects of Bifidobacterium lactis
Bb12 supplementation on intestinal microbiota
of preterm infants: a double-blind, placebo-controlled, randomized study. J Clin Microbiol
79. Oncel MY, Sari FN, Arayici S, et al. Lactobacillus reuteri
for the prevention of necrotising enterocolitis in very low birthweight infants: a randomised controlled trial. Arch Dis Child Fetal Neonatal Ed
80. Pärtty A, Luoto R, Kalliomaki M, et al. Effects of early prebiotic and probiotic supplementation on development of gut microbiota
and fussing and crying in preterm infants: a randomized, double-blind, placebo-controlled trial. J Pediatr
2013; 163:1272–1277. e1–2.
81. Patole S, Keil AD, Chang A, et al. Effect of Bifidobacterium breve
M-16 V supplementation on fecal bifidobacteria in preterm neonates—a randomised double blind placebo controlled trial. PLoS One
82. Reuman PD, Duckworth DH, Smith KL, et al. Lack of effect of Lactobacillus on gastrointestinal bacterial colonization in premature infants. Pediatr Infect Dis
83. Rojas MA, Lozano JM, Rojas MX, et al. Prophylactic probiotics to prevent death and nosocomial infection in preterm infants. Pediatrics
84. Romeo MG, Romeo DM, Trovato L, et al. Role of probiotics in the prevention of the enteric colonization by Candida in preterm newborns: incidence of late-onset sepsis
and neurological outcome. J Perinatol
85. Rougé C, Piloquet H, Butel MJ, et al. Oral supplementation with probiotics in very-low-birth-weight preterm infants: a randomized, double-blind, placebo-controlled trial. Am J Clin Nutr
86. Roy A, Chaudhuri J, Sarkar D, et al. Role of enteric supplementation of probiotics on late-onset sepsis
by Candida species in preterm low birth weight neonates: a randomized, double blind, placebo-controlled trial. N Am J Med Sci
87. Saengtawesin V, Tangpolkaiwalsak R, Kanjanapattankul W. Effect of oral probiotics supplementation in the prevention of necrotizing enterocolitis
among very low birth weight preterm infants. J Med Assoc Thai
2014; 97 (suppl 6):S20–S25.
88. Samanta M, Sarkar M, Ghosh P, et al. Prophylactic probiotics for prevention of necrotizing enterocolitis
in very low birth weight newborns. J Trop Pediatr
89. Sari FN, Dizdar EA, Oguz S, et al. Oral probiotics: Lactobacillus sporogenes
for prevention of necrotizing enterocolitis
in very low-birth weight infants: a randomized, controlled trial. Eur J Clin Nutr
90. Serce O, Benzer D, Gursoy T, et al. Efficacy of Saccharomyces boulardii
on necrotizing enterocolitis
in very low birth weight infants: a randomised controlled trial. Early Hum Dev
91. Shadkam MN, Jalalizadeh F, Nasiriani K. Effects of probiotic Lactobacillus reuteri
(DSM 17938) on the incidence of necrotizing enterocolitis
in very low birth weight premature infants. Iranian J Neonatol
92. Shashidhar A, Suman Rao PN, Nesargi S, et al. Probiotics for promoting feed tolerance in very low birth weight neonates - a randomized controlled trial. Indian Pediatr
93. Stratiki Z, Costalos C, Sevastiadou S, et al. The effect of a bifidobacter supplemented bovine milk on intestinal permeability of preterm infants. Early Hum Dev
94. Tewari VV, Dubey SK, Gupta G. Bacillus clausii
for prevention of late-onset sepsis
in preterm infants: a randomized controlled trial. J Trop Pediatr
95. Totsu S, Yamasaki C, Terahara M, et al. Bifidobacterium and enteral feeding in preterm infants: cluster-randomized trial. Pediatr Int
96. Underwood MA, Salzman NH, Bennett SH, et al. A randomized placebo-controlled comparison of 2 prebiotic/probiotic combinations in preterm infants: impact on weight gain, intestinal microbiota
, and fecal short-chain fatty acids. J Pediatr Gastroenterol Nutr
97. Van Niekerk E, Nel DG, Blaauw R, et al. Probiotics reduce necrotizing enterocolitis
severity in HIV-exposed premature infants. J Trop Pediatr
98. Wang C, Shoji H, Sato H, et al. Effects of oral administration of Bifidobacterium breve
on fecal lactic acid and short-chain fatty acids in low birth weight infants. J Pediatr Gastroenterol Nutr
99. Xu L, Wang Y, Wang Y, et al. A double-blinded randomized trial on growth and feeding tolerance with Saccharomyces boulardii
CNCM I-745 in formula-fed preterm infants. J Pediatr (Rio J)
100. Zeber-Lubecka N, Kulecka M, Ambrozkiewicz F, et al. Effect of Saccharomyces boulardii
and mode of delivery on the early development of the gut microbial community in preterm infants. PLoS One
101. Zhang AM, Sun ZQ, Zhang LM. Mosapride combined with probiotics on gastrointestinal function and growth in premature infants. Exp Ther Med
102. Costeloe K, Bowler U, Brocklehurst P, et al. A randomised controlled trial of the probiotic Bifidobacterium breve
BBG-001 in preterm babies to prevent sepsis
, necrotising enterocolitis and death: the Probiotics in Preterm infantS (PiPS) trial. Health Technol Assess
103. Demirel G, Celik IH, Erdeve O, et al. Impact of probiotics on the course of indirect hyperbilirubinemia and phototherapy duration in very low birth weight infants. J Matern Fetal Neonatal Med
104. Manzoni P, Stolfi I, Messner H, et al. Bovine lactoferrin prevents invasive fungal infections in very low birth weight infants: a randomized controlled trial. Pediatrics
105. Patole SK, Keil AD, Nathan E, et al. Effect of Bifidobacterium breve
M-16 V supplementation on faecal bifidobacteria in growth restricted very preterm infants—analysis from a randomised trial. J Matern Fetal Neonatal Med
106. Van Niekerk E, Kirsten GF, Nel DG, et al. Probiotics, feeding tolerance, and growth: a comparison between HIV-exposed and unexposed very low birth weight infants. Nutrition
107. Li Y, Shimizu T, Hosaka A, et al. Effects of Bifidobacterium breve
supplementation on intestinal flora of low birth weight infants. Pediatr Int
108. Agarwal R, Sharma N, Chaudhry R, et al. Effects of oral Lactobacillus GG on enteric microflora in low-birth-weight neonates. J Pediatr Gastroenterol Nutr
109. Indrio F, Riezzo G, Raimondi F, et al. The effects of probiotics on feeding tolerance, bowel habits, and gastrointestinal motility in preterm newborns. J Pediatr
110. Indrio F, Riezzo G, Raimondi F, et al. Effects of probiotic and prebiotic on gastrointestinal motility in newborns. J Physiol Pharmacol
2009; 60 (suppl 6):S27–S31.
111. Kitajima H, Sumida Y, Tanaka R, et al. Early administration of Bifidobacterium breve
to preterm infants: randomised controlled trial. Arch Dis Child Fetal Neonatal Ed
112. Mohan R, Koebnick C, Schildt J, et al. Effects of Bifidobacterium lactis
Bb12 supplementation on body weight, fecal pH, acetate, lactate, calprotectin, and IgA in preterm infants. Pediatr Res
113. Stansbridge EM, Walker V, Hall MA, et al. Effects of feeding premature infants with Lactobacillus GG on gut fermentation. Arch Dis Child
1993; 69 (5 Spec No):488–492.
114. Akar M, Eras Z, Oncel MY, et al. Impact of oral probiotics on neurodevelopmental outcomes in preterm infants. J Matern Fetal Neonatal Med
115. Chou IC, Kuo HT, Chang JS, et al. Lack of effects of oral probiotics on growth and neurodevelopmental outcomes in preterm very low birth weight infants. J Pediatr
116. Chowdhury T, Ali MM, Hossain MM, et al. Efficacy of probiotics versus placebo in the prevention of necrotizing enterocolitis
in preterm very low birth weight infants: a double-blind randomized controlled trial. J Coll Physicians Surg Pak
117. Nandhini LP, Biswal N, Adhisivam B, et al. Synbiotics for decreasing incidence of necrotizing enterocolitis
among preterm neonates—a randomized controlled trial. J Matern Fetal Neonatal Med
118. Oncel MY, Arayici S, Sari FN, et al. Comparison of Lactobacillus reuteri and nystatin prophylaxis on Candida colonization and infection in very low birth weight infants. J Matern Fetal Neonatal Med
119. Demirel G, Celik IH, Erdeve O, et al. Prophylactic Saccharomyces boulardii
versus nystatin for the prevention of fungal colonization and invasive fungal infection in premature infants. Eur J Pediatr
120. Sinha A, Gupta SS, Chellani H, et al. Role of probiotics VSL#3 in prevention of suspected sepsis
in low birthweight infants in India: a randomised controlled trial. BMJ Open
121. Hill C, Scott K, Klaenhammer TR, et al. Probiotic nomenclature matters. Gut Microbes
122. Lewis ZT, Shani G, Masarweh CF, et al. Validating bifidobacterial species and subspecies identity in commercial probiotic products. Pediatr Res
123. Imperial IC, Ibana JA. Addressing the antibiotic resistance problem with probiotics: reducing the risk of its double-edged sword effect. Front Microbiol