Milk Fat Globule Membrane Supplementation in Children: Systematic Review with Meta-Analysis

(1) Background: Milk fat globule membrane (MFGM), composing fat droplets responsible for lipid transport in breast milk, has been shown to possess immunological and antimicrobial effects. Standard formulas (SF) are devoid of MFGMs during the production process. The study’s aim was to evaluate the safety and benefits of MFGMs supplementation in children. (2) Methods: We searched four databases for randomized controlled trials evaluating the supplementation of MFGMs in children. Growth parameters were chosen as the primary outcome. (3) Results: Twenty-four publications of seventeen studies were included. Meta-analyses assessing the primary outcomes at the age of 4 months included four studies (814 children) comparing the MFGM-supplemented formulas and SF, and two trials (549 children) comparing the MFGM-supplemented formulas and breastfeeding. The primary outcomes were non-inferior in all the experimental MFGM formulas compared to SF, or even represented more similar results to breastfed infants. The promising effects, including a lower incidence of acute otitis media and improved cognitive development, cannot be firmly confirmed due to the small amount of existing evidence. No significant adverse effects were reported in any of the assessed products. (4) Conclusions: The available data signaled beneficial effects and a good safety profile, requiring future research with well-designed trials.


Introduction
Breastfeeding is the most effective and natural method of feeding to provide infants with valuable nutrients, immunomodulatory factors, and energy [1]. It has been long reported to have various protective effects against allergies, asthma, infections, and immune-mediated diseases in children [2,3]. The exact composition of human breast milk depends on various maternal factors and matches specific infant demands. The largest energy source in human milk forms lipids, which contribute approximately 40-55% of the total energy [4]. Lipids are secreted into breast milk by an exclusive mammary gland mechanism and are composed of fat droplets enclosed with a tri-layer membrane [5]. The core lipid consists mainly of triglycerides (98%), whereas the membrane, called milk fat globule membrane (MFGM), comprises a monolayer of polar lipids and a lipid bilayer. MFGM consists of phospholipids such as sphingomyelins, phosphatidylcholines, gangliosides, and different proteins, including lactoferrin or mucins [6].
An increasing body of evidence indicates that MFGM can exert a beneficial effect on immune functions, central nervous system development, and metabolism [7]. Standard formulas are devoid of bovine MFGMs due to the production process and are replaced by vegetable oils. This phenomenon may account for the differences observed between formula-fed and breastfed infants. Therefore, in recent years, efforts have been made to create standard formulas closely imitating human milk by adding bovine MFGM to the composition.
The MFGMs-supplemented formulas are attracting an increasing interest due to the reports on their favorable effectiveness on infants' growth, risk of infections, cognitive development, microbiome, and metabolism in comparison to standard formulas and/or breastfeeding [8].
The purpose of this systematic review was to evaluate the safety and clinical benefits of the MFGMs supplementation in children available from randomized controlled trials.

Search Strategy
The study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and was registered in the National Institute for Health Research's PROSPERO (CRD42020138870) [9].
We performed a systematic search and retrieved studies from PubMed, Web of Science, Embase, and the Cochrane's Library databases from inception to 20 October 2019. Additionally, the hand-search was conducted. We have repeated the same search on 18 July 2020 for new data through all previously used databases to keep our systematic review up to date. The search terms with the filter validated in the Cochrane Handbook for randomized controlled trials can be found in the Supplementary Materials (Table S1).

Study Selection
All extracted papers were imported into EndNote®reference manager Version x8 (Clarivate Analytics; 2016). Two reviewers working independently (D.A., M.R.) screened at first the titles with abstracts and next assessed the potentially suitable full texts. Uncertainty about the eligibility of studies for the review was resolved through discussion at each stage with a third reviewer (P.D.). The final group of selected studies met the inclusion and exclusion criteria pointed out herein.

Inclusion Criteria
Original studies were accepted only if they (1) were published in English; (2) were performed on human beings aged less than 18 years old; (3) introduced MFGMs oral supplementation performed in any available form; (4) involved any of the following control groups: placebo or milk formula without supplementation; and (5) have been performed as randomized controlled trials (RCTs).

Data Extraction
Data were independently extracted by three reviewers (D.A., K.D., and M.R.) using a structured data collection form, collating relevant study details such as date of publication, country of origin, study design, population of the study, intervention and comparison, measurement points, results, and confounding factors. Actions were taken to contact corresponding authors when additional clarification was required. Data were extracted from all selected articles, regardless of whether the results were positive or negative, to present a fair and unbiased account of the available published evidence.

Risk of Bias Assessment
Three independent reviewers (D.A., K.D., and M.R.) assessed the risk of bias, without being blinded to the authors or journal, in each of the included studies. A revised Cochrane risk-of-bias tool for randomized trials was used (RoB 2) [10], Figure S1. Encountered discrepancies were resolved through discussion of all the reviewers.
The RoB 2 tool is structured into five bias domains, which enables judging the randomization process, deviations from intended interventions, any missing outcome data, measurement of the outcome, and selection of the reported result.
The judgments for each domain were to choose between "low risk of bias", "some concerns", or "high risk of bias". In conclusion, the overall bias was determined by reflecting the individual marks.

Data Synthesis
Required data were extracted from included studies to a predesigned Excel sheet in Microsoft Excel. Review Manager (RevMan) v5.3 software from the Cochrane Collaboration (London, UK) was used to perform meta-analyses, generate forest plots, and calculate the I 2 statistic. The data were pooled on baseline characteristics and outcomes using the meta-analysis approach. In the studies that reported continuous variables as means, mean difference (MD) was calculated. Heterogeneity was assessed using the I 2 statistic defined by the Cochrane Handbook for Systematic Reviews. Either Chi 2 test p < 0.10 or I 2 value > 50% indicated substantial heterogeneity.
Growth parameters such as weight, length/height, and head circumference were chosen as the primary outcome of our systematic review.
The secondary outcomes assessed in the review consisted of any additional outcomes reported in the included trials such as validated psychomotor development scales, any side effects, oral or gut microbiota shifts if reported, as well as any potential benefit on the infection/fever prevalence in children.

Description of the Studies
Our initial search yielded 676 randomized controlled trials; after removing duplicates (n = 185), we identified 491 unique citations. The PRISMA flow chart is shown in Figure 1. The screening of titles and abstracts excluded 432 articles, rendering 59 trials to full-text assessment. Finally, 17 publications were included in our systematic review studies (Table 1). Data from three RCTs were eligible for the meta-analysis comparing anthropometric measurements in experimental vs. standard formula-fed infants [11][12][13], and two RCTs were eligible for meta-analysis comparing the same measurements in experimental formula vs. breastfed infants [11,13].
In most of the included trials, the overall risk of bias was considered as low [11][12][13] or only with some concerns [22,24,26] (Figure 2). However, two of the studies were judged as with a high risk of bias [20,23]. Their methodological concerns included poor data availability on randomization and blinding process, deviations from the intended intervention, and potentially missing outcome data, as well as some discrepancies in reporting results. Notably, a few papers did not present the results according the intention to treat (ITT) analysis, which may have led to the increased bias [20,23,25,27]. Another potential confounder was the lack of the beforehand published trial protocol in three cases [23][24][25].

Populations
Most of the studies recruited participants below 2 months of age at inclusion. In the three remaining trials, older children were included (Table 1) [23,24,26].

Outcomes
The outcomes presented in the trials were highly heterogenous in terms of the methods used to evaluate specific outcomes and time points of the assessment. In most of the trials' anthropometric measurements (weight, length, head circumference, daily weight gain), cognitive development or infection prevalence were employed as the primary outcomes. Additional outcomes evaluated in the studies included adverse events, hematological and biochemical assessment, and oral and fecal microbiota.
Due to the heterogeneity of outcomes, most of the data were analyzed descriptively. Solely data on weight, length, and head circumference at four months of age were eligible for meta-analysis.

•
Experimental formula compared to a standard formula Data from the studies eligible for meta-analysis showed no differences in mean weight, length, and head circumference at the age of four months (Figure 2a,c,e) [11][12][13]. Furthermore, studies that were not included in the meta-analyses showed no differences in mean weight, length, head circumference, and mean weight gain [12,25,27]. Dissimilarities were also not shown with respect of weight-for-age, length-for-age, and weight-for-length z-scores [20,25].

• Experimental Formula Compared to Breastfeeding
The conducted meta-analyses indicated that at the age of four months, children from the group fed with experimental formula (EF) had slightly lower mean body weight and head circumference compared to the breastfed (BF) infants (Figure 2b,f). Body length did not differ between the groups (Figure 2d) [11,13].
Moreover, the comparison between BF and formula-fed (both EF + standard formula, SF) infants by Timby et al. revealed the significant interactions for z-scores for the weight (p = 0.025) and length (p = 0.003) [11]. On the other hand, Nieto-Ruiz et al. showed no statistically significant differences in mean weight, growth, daily weight gain, and no differences in several z-scores [20].
In the research conducted by Billeaud et al., one control group of standard formulafed infants served as a comparison to infants fed with MFGM-P and MFGM-L formulas (F-females, M-males, MFGM-P-standard formula enriched with a protein-rich MFGM fraction, MFGM-L-standard formula enriched with a lipid-rich MFGM fraction).

Psychomotor Development
According to the results of the evaluation of 6-month-old infants using the Griffiths Mental Development Scale (GMDS) by Gurnida et al., there was a significant increase in scores for Hand and Eye Coordination (p = 0.006), Performance (p < 0.001), and for General IQ (p = 0.041) in the EF group. However, no significant difference was found between the EF and SF group for the Griffith Locomotor, Personal-Social, and Hearing and Speech scores. Remarkably, all scores in this group were similar to those of the BF group [25]. In two trials, infants' psychomotor development was evaluated at 12 months of age using the Bayley Scales of Infant and Toddler Development (Baley-III) scale [11,27]. Timby et al. reported significantly higher scores of EF compared to SF groups in the cognitive domain (mean difference −4.0, 95%CI (−6.9498 to −1.0502), p = 0.008), but there were no differences compared to the BF group (p = 0.35). Moreover, BF infants performed significantly better in the verbal domain than both EF (mean difference −4.1, 95% CI (−7.6134 to −0.5866), p = 0.029) and SF infants (mean difference −4.2, 95% CI (−7.5807 to −0.8193), p = 0.012). No differences were reported in respect of the motor domain between any groups [11]. According to Li et al. EF-fed infants presented higher scores in the cognitive, language, and motor domains (p < 0.01; p < 0.01; p < 0.01) at 12 months of age. Differences did not persist until 18 months of age [27]. Nieto-Ruiz et al. indicated that children who had consumed EF during the first 18 months of life presented higher scores in the use of language (mean difference −5.86, 95% CI (−10.6048 to −1.1152) p = 0.033) and spontaneous oral expression (mean difference −0.150, 95% CI (−0.2998 to −0.0002) p = 0.024) than SF children according to the Oral Language Task of Navarra-Revised (PLON-R) test at the age of four [21]. Notably, Veereman et al. reported fewer behavioral problems assessed by parents using the Achenbach System of Empirically Based Assessment (ASEBA) questionnaire in children in the experimental group [23].

Risk of Infections
The impact of applied formulas or breastfeeding on the prevalence of infections was evaluated in two manners in the included studies-either as a secondary endpoint or as an adverse event. Timby et al. reported a lower incidence of acute otitis media in the EF group (MFGM-P) compared to the SF group (p = 0.034) as well as the incidence of antipyretic use (p = 0.021) [15]. Likewise, Li et al. showed a lower incidence of respiratory tract infections and diarrhea in the EF group compared to the SF group [27]. According to Li et al. and Gurnida et al., MFGM-formula-fed and SF-fed infants did not differ significantly with respect to various infections [13,25]. By contrast, Breji et al. reported a higher occurrence of diarrhea in the EF group compared to the SF group (13.4% vs. 7.8%), though without statistical significance [22]. Furthermore, Billeaud et al. showed a slight difference in the prevalence of infections between infants in experimental and control groups [12]. Children in the MFGM-P group presented a significantly higher rate of respiratory tract infection symptoms reported by parents after 2 months of the beginning of the study. There were no differences in the infection rate between MFGM formula-fed infants and BF infants (except for one MFGM-P group). Bovine MFGM and phospholipids added to chocolate milk in pre-school children showed a reduction of days with fever compared to the control group (p < 0.028); however, there was no significant influence on the incidence of diarrhea, constipation, cough, doctors' visits, and days of school absence [23].

Prevention of Diarrhea in Developing Countries
Zavaleta et al. demonstrated a reduction of the number of bloody diarrhea episodes in the EF group as compared to the SF group (p = 0.025), although no significant effect on the longitudinal prevalence of diarrhea, duration of diarrhea episodes, and incidence of severe diarrhea episodes were reported in Peruvian infants. Furthermore, no differences in the pathogenesis of diarrhea were found (infection with different Escherichia coli serotypes, Campylobacter jejuni, Giardia lamblia, Blastocystis hominis, Rotavirus) [24]. Poppitt et al. reported no differences in the number, duration, and severity of all-cause diarrhea episodes between the EF and SF groups, however, the duration of rotavirus diarrhea was lower in the EF group (p = 0.03) throughout the intervention period [26].

Metabolic Effects
Metabolic effects of several formulas were extensively evaluated in the cohort of children primarily described by Timby et al [11]. The first analysis revealed that EF and SF infants presented higher growth velocity than BF with concomitant higher plasma insulin and blood urea nitrogen level at six months of age [11]. Further analysis of lipidomics showed higher levels of triglycerides and cholesterol in serum, as well as higher lowdensity lipoprotein cholesterol: high-density lipoprotein cholesterol (LDL:HDL) ratios and leptin:fat mass ratios in BF infants compared to SF. The EF group also had a higher total serum cholesterol concentration than the SF group, reaching the level of the BF group [14]. Grip et al. indicated significant differences in the serum, plasma, and erythrocyte membrane concentrations of sphingomyelins, phosphatidylocholines, and ceramides at 4 and 6 months of age between infants from the EF and SF groups [17]. Moreover, He et al. demonstrated that the formula-fed infants presented a preference for protein metabolism, which was expressed by the higher serum levels of amino-acid catabolism products and low efficiency of amino-acid clearance. In contrast, BF infants had higher levels of fatty acid oxidation products, which account for fat metabolism preference. Interestingly, the EF group was characterized by higher levels of fat metabolism products than SF infants, which may bridge the gap between the EF and BF infants [18]. Metabolic differences were visible only during the exclusive feeding period and did not remain at 12 months of age [14,17,18].
Gurnida et al. revealed higher ganglioside serum concentrations in the EF group than SF (p < 0.01) but no significant differences between EF and BF infants at 6 months of age [25].
On the other hand, Billeaud et al. did not show any differences in plasma and red blood cell membrane levels of phospholipids, cardiolipin, cholesterol, insulin-like growth factor 1 (IGF-1), leptin nor IGF BP3 between EF and SF infants. However, the level of C-peptide was lower in both EF groups compared to SF (p = 0.03) [12].

Microbiome
The oral and fecal microbiome was evaluated in infants from the cohort primarily described by Timby et al. Assessment of oral microbiota using the bacterial 16S rRNA sequencing method revealed significantly higher microbial richness in formula-fed (EF + SF) compared to BF infants at 4 months of age. Moreover, infants fed with EF had a significantly lower level of Moraxella catarrhalis compared to the SF-fed group, which may be associated with fewer episodes of acute otitis media in the EF group [16]. Yet, He showed that BF infants had a more heterogenous fecal microbiome but only during the exclusive formula feeding period. After the introduction of weaning food with either EF, SF, or BF, no significant differences in fecal microbiome were observed [19].

Adverse Events
Treatment-emergent adverse events were reported as respiratory tract and gastrointestinal infections, skin diseases, and formula intolerance. Risk of respiratory tract and gastrointestinal infections was described in the Section 3.4.4, since in some of the trials, they were assessed in the character of primary outcomes. None of the studies raised the concerns of MFGM-supplemented formulas. No adverse events in constipation and skin diseases including eczema were reported by Li et al. [27], whereas an even lower rate of dry skin was noted in the EF group compared to SF (p = 0.08) by Breji et al. [22]. Li et al. reported a low number of adverse events without any differences between MFGM, both for the SF and BF groups. On the other hand, Billeaud et al. (MFGM-P) showed a significantly higher rate of eczema (p = 0.001) in infants consuming MFGM-P formula [12]. Other treatment-emergent adverse events did not differ between groups.

Discussion
The deepening understanding of complex communication between distinct organs, such as gut-brain and gut-lung axes, has expanded the significance of the breast milk to support the optimal cognitive and physical development of a child [28,29]. Therefore, unceasing efforts are being taken to create an infant formula mimicking the composition of the breast milk as closely as possible. It has been suspected that MFGM may be responsible for some of the advantageous properties of breastfeeding. Recently, a novel formula milk enriched with MFGM, naturally occurring in the maternal milk, was placed on the market. Numerous questions have been risen about the tangible evidence and whether to recommend it in the clinical setting. This infant milk is advertised as a product stimulating brain development, along with strengthening the immune system.
The beneficial role of MFGM has already been explored in animals and adults. Studies on rodent models have proven the impact of MFGM on infection, inflammation, brain composition, and gut barrier integrity, as well as intestinal development [30][31][32][33][34]. Interestingly, pups fed control formula demonstrated delayed intestinal growth, while the MFGM supplementation normalized intestinal crypt depths, epithelial cell proliferation, make-up of intestinal epithelial cell subsets, and resembled the make-up of intestinal microbes of mother's milk-fed rats at the phylum level [34]. This finding may be specifically important for premature infants born with immature gastrointestinal tract, resulting in their higher risk of infections and a life-threatening Necrotizing Enterocolitis (NEC) [35]. Moreover, this study showed that the addition of MFGM also afforded significant protection to Clostridium difficile toxins, which may be an attractive feature as there is growing evidence that C. difficile is associated with diarrhea and prolonged hospitalization in younger children [34,36,37].
The results of our primary outcome meta-analyses have successfully confirmed that all the MFGM supplemented formulas have shown the non-inferiority when it comes to growth and weight parameters in children, or it even represented the results that were more similar to the breastfed groups. These conclusions had been anticipated beforehand, as formula feeding has been associated with infant's increased weight gain velocity compared to breastfeeding [38,39]. Another key thing to remember is the very good safety profile of these formulas. Unfortunately, there is a small amount of evidence about the reported promising effects, which are fewer episodes of fever and lower need to use antipyretics, lower incidence of acute otitis media, reduction in episodes or duration of diarrhea, and improved cognitive development, which does not permit us to draw any firm conclusions about the effectiveness of MFGM supplementation in infants and children.
Yet, several issues have arisen during the assessment. Importantly, our findings revealed a limited number of pertinent clinical trials and a considerable heterogeneity among them. Firstly, although the composition of the milk formulas used was comparable, several different MFGM supplement products were applied in these trials. Due to the addition of other supplements enriching concept formulas along with MFGM, the results of two trials could not be analyzed equally [20,21,27]. The long-chain polyunsaturated fatty acids (LCPUFA) supplementation during pregnancy and/or lactation might be associated with a beneficial impact on child's neurodevelopment [40,41]. Moreover, current evidence suggests that prophylactic use of lactoferrin, a protein found in cow and human milk, in preterm infants could significantly reduce the incidence of NEC [42]. Secondly, intervention duration and length of follow-up differed significantly between the studies. Thirdly, the experimental outcomes, such as an incidence of infection or makeup of the oral microbiome, were often inspected in individual trials [13,15,16]. Furthermore, cognitive development was evaluated with various measurement scales. Timby et al. used the Bayley-III score, while Li et al. employed the GMDS at different time points, making the outcomes harder to compare [11,13].
Our systematic review has several strengths. A key asset is a brief presentation, with a quantitative and qualitative synthesis of data, of the novel formula enrichment on the market. The thorough search of the most extensive databases complemented with seeking the information from conference abstracts, contacting study authors and experts in the field makes it the most comprehensive summary of all experimental data about the use of MFGM in children up to date. Furthermore, this review provides an overview of currently ongoing trials registered at the ClinicalTrials, the EU Clinical Trials Register, and the Australian New Zealand Clinical Trials Registry (ANZCTR) ( Table 2). Finally, when there is enough evidence, the mentioned supplementation could become an easy-to-use and cost-efficient add-on approach for many parents to support a child's cognitive development and lower the prevalence of acute otitis media or fever episodes. We are also aware of the limitations of our study. Trials presenting positive results are more eagerly published than those with negative results; therefore, this might be a source of positive-results bias. We could fully compare and contrast the findings of all included trials, as some of them assessed several supplements simultaneously.

Conclusions
Our review is the first systematic attempt to provide a valid answer to the question as to whether the present evidence is convincing enough to commend milk formulas enriched with MFGM. The currently available data signaled beneficial effects and a good safety profile, requiring future research with well-designed randomized double-blind placebo-controlled trials measuring the outcomes in a more unified way.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.