Breastfeeding and Risk of Multiple Sclerosis: A Systematic Review and Meta-Analysis of Observational Studies

Multiple sclerosis (MS) is an inflammatory and degenerative disease of the central nervous system [1, 2]. The degenerative component of the disease is presumed to be an immune-mediated process, potentially caused by a viral infection [3]. MS is characterized by relapses, disseminated lesions in the central nervous system, and progression of neurological disability resulting in multiple neurological symptoms and signs (walking and cognitive impairment, sensory dysfunction, visual impairment but also fatigue and depression), which have negative effects on quality of life and participation [4].

In 2016, approximately 2.2 million people worldwide suffered from MS which is equivalent to 30.1 cases per 100,000 population. This corresponds to an increase of approximately 10.4% or 22.47 cases per 100,000 population in the age-standardized prevalence since 1990 [5]. A more recent study by Walton et al. [6] estimated MS prevalence to amount to 2.8 million people (35.9 cases per 100,000 population) worldwide in 2020 with an increase of 30% since 2013. MS incidence was estimated to be 2.1 per 100,000 persons/year [6].

The causes of MS are not fully understood, yet [7]. Genetic predisposition, environmental and lifestyle factors as well as an interplay thereof constitute relevant factors in the development of MS [8]. Especially early-life risk factors such as having been breastfed may also be of relevance. Breast milk provides the newborn not only with essential nutrients and vitamins but also with numerous immune-active molecules, metabolites, oligosaccharides, and microbial components that are important for the development of the immune system [9]. This, in turn, is involved in the demyelination of the brain and spinal cord typical of MS. Moreover, having been breastfed has already been shown to be associated with a reduced incidence of other immune disorders such as type 1 diabetes (DT1) and asthma [9].

Some studies have been published on the influence of having been breastfed on subsequent risk of developing MS, but to our knowledge, no systematic review including a meta-analysis has yet been conducted to summarize the evidence. Given the reported increasing prevalence [5] and incidence [6] of MS, a better understanding of the underlying causal factors is needed to possibly reduce the risk of disease and influence the course of MS in those already ill. Being aware of the multitude of factors triggering MS, evidence synthesis on single factors is helpful in weighing the impact of that factor. Hence, the objective of this work was to systematically search for studies and summarize their findings on the impact of having been breastfed on subsequent MS risk to evaluate whether having been breastfed during infancy is a protective factor for the development of MS.

A systematic review of observational studies including a subsequent meta-analysis was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement [10].

A systematic search strategy was developed based on the PICO (population, intervention, comparison, outcome) framework applying the following search terms ([multiple sclerosis] AND [breastfeeding OR breast milk OR maternal OR perinatal OR neonatal] AND [case-control studies OR cohort studies OR cross-sectional studies OR prospective studies OR systematic review OR meta-analysis]) in the form of a free-text search and controlled vocabulary, i.e., MeSH Terms with automatic explosion to search for more specific terms. The search strategy was adapted to the syntax and subject headings of the corresponding database. MEDLINE, Web of Science, and Cochrane Library were searched on October 12, 2021, for publications published by that date. In addition, a search in Google Scholar (https://scholar.google.com) was conducted using the same search terms and the 100 most relevant studies were considered. We updated the database search on May 18, 2022, using the same search strategy. The search strategy for the respective databases can be found in online supplementary Table 1 (see www.karger.com/doi/10.1159/000526895 for all online suppl. material). Reference lists of included studies were searched for additional studies not identified through the initial search. Prior to the abstract and title screening, duplicates were removed. Selected studies were then subject to full-text screening for eligibility. If there was no full-text access, the authors of the respective study were contacted. The search and study selection were performed by two independent researchers (A.H. and M.R.). In case of discordance, a third researcher (H.B.) was consulted.

Observational studies, i.e., case-control, cohort, and cross-sectional studies, reporting on the influence of having been breastfed on MS risk in children or adults were included if subjects were breastfed in a period from birth to 2 years of age as this is the period recommended by the World Health Organization [11] for breastfeeding. Additionally, MS had to be diagnosed according to established diagnostic criteria by Schuhmacher, Poser, or McDonald criteria [2]. Studies based on self-reported MS diagnoses were also included but were downgraded in the quality assessment accordingly. Studies on disease activity of MS, i.e., relapse rates or progression of neurological symptoms and signs, were excluded. Only peer-reviewed studies published in English or German were included. No restrictions were made regarding the study setting or publication year.

For the data extraction, an adaption of the data collection form of the Cochrane Developmental, Psychological and Learning Problems Review Group [12] was used. Data were extracted by two independent reviewers (A.H., M.R.) who were blinded to each other’s decision. Disagreement was resolved by consultation of a third reviewer (H.B.). Extracted data included author, year, country, diagnostic criteria, number of people per exposure group (breastfed vs. not breastfed), cases (persons with MS [pwMS]) and controls from case-control studies, and number of events (diagnosis of MS) at the end of the study period in cohort studies as well as confounding variables. Crude and adjusted risk estimates, reported below as OR [13], were extracted from the original publications if available. If crude risk estimates were missing, these were calculated by the authors from published data.

Both pooled OR_crude and OR_adjusted with 95% confidence intervals (CIs) were calculated. Overall and between-study heterogeneity was evaluated by calculation of τ² applying the DerSimonian-Liard estimator [14] and of I² statistics [15], respectively. As considerable between-study heterogeneity was anticipated, a random-effects model applying the Mantel-Haenszel method [16] for dichotomous outcomes was chosen to calculate pooled effect sizes, for both crude and adjusted risk estimates. Subgroup and sensitivity analyses were performed to identify the influence of each respective study on the robustness of the overall pooled effect estimates. Individual study quality was assessed using the Newcastle-Ottawa Scale (NOS) for non-randomized studies [17], and publication bias was examined visually with funnel plots. All analyses were run in R (version 4.1.1) using the packages meta (version 5.1–1), metafor (version 3.0–2), and dmetar (version 0.0.9000).

The search in the databases MEDLINE, Web of Science, and Cochrane Library supplemented by a search in Google Scholar yielded 425 results. After duplicate eradication, 305 studies were subject to the title and abstract screening of which 287 were excluded since they did not fulfil the predefined eligibility criteria. For two [18, 19] of the remaining 18 studies, full-text screening revealed that a shortened (<4 months) versus prolonged duration of having been breastfed (≥4 months) was compared. These studies were also included, with the short duration of having been breastfed considered comparable to “not having been breastfed” and the longer duration comparable to “having been breastfed” (see discussion on this point). Of the 18 studies which underwent full-text screening, three studies were excluded as one was a conference abstract and two studies compared having been exclusively breastfed with having been breastfed supplemented with supplementation but not with non-breastfed individuals. The updated search on May 18, 2022 did not reveal any additional studies that met the eligibility criteria. Therefore, 15 studies were ultimately included in the syntheses of results and subsequent meta-analysis. Figure 1 depicts the study selection process.

Relevant characteristics of included studies are displayed in Table 1. A total of two cohort studies and 13 case-control studies, of which two studies [18, 19] reported results for two distinct samples, were included. Two of the included studies investigated paediatric-onset MS [20, 21], whereas the remaining studies investigated adult-onset MS [18, 19, 22‒32]. Studies were conducted in Europe (n = 8), the Americas (n = 5), the Middle East (n = 3), and Australia (n = 1) and published between 1982 and 2021, with most studies (n = 11) having been published in the last 10 years. Most of the included studies applied one of the following diagnostic criteria to establish a diagnosis of MS: McDonald criteria [18, 19, 21, 22, 24, 26, 29], Poser criteria [27, 30, 32], or International Pediatric Multiple Sclerosis Study Group (IPMSSG) criteria [20, 21]. In each of the cohort studies by Baldin et al. [25] and Goldacre et al. [28], ICD codes were used to confirm a diagnosis of MS. Al Wutayd et al. [23] and Spenceley and Dick [31] stated that only individuals with a clinically confirmed MS diagnosis were included as cases, without further specifying which diagnostic criteria were used. In case-control studies controls (non-pwMS) were enrolled through registries [18, 19, 29], were acquaintances/spouses of the case [23, 24, 27, 31], health-care workers [32], or were non-pwMS from the same or another hospital ward [20‒22, 30, 32] or general practitioner’s practice [26]. Sex, age, and region were frequently used as matching variables for pwMS and controls or included as potential confounders in the statistical models. Otherwise, maternal age, education, and smoking habits were most often considered as confounding variables.

Overall, 15 studies comprising 17 distinct study samples investigated the association of having been breastfed compared to not having been breastfed on MS risk and were included in the pooled analysis. For the assessment of risk of bias, the NOS was categorized as follows: ≥6 points equals low risk of bias and ≤5 points equals high risk of bias. The assessment of the respective NOS domains for each study is depicted in online supplementary Table 2. Risk of bias was low in twelve [18‒22, 24‒30] and high in three studies [23, 31, 32]. Table 2 provides an overview of the results of the meta-analysis of the crude and adjusted ORs including analyses stratified by study design. Unless otherwise specified, all data are from the primary reference of each included study.

The combined crude OR estimates based on a random-effects model showed a protective effect of having been breastfed on MS risk (OR_crude = 0.82 [95% CI: 0.70–0.96]) with substantial heterogeneity (I² = 68.2%). Subgroup analysis stratified by study design yielded an OR_crude of 0.83 (95% CI: 0.70–0.98) and an OR_crude of 0.73 (95% CI: 0.49–1.09) for case-control studies and cohort studies, respectively. There is little heterogeneity when comparing the results based on case-control studies with those based on cohort studies (shown in Table 2). Figure 2 displays for each study included in the meta-analysis the summary statistics (number of events and the sample size) for the breastfed and not breastfed groups, the crude OR and corresponding 95% CI for the dichotomous outcome, MS.

The analysis using adjusted OR, when available otherwise using crude OR, also showed a protective effect of having been breastfed compared to not having been breastfed on MS risk with moderate heterogeneity (OR_adjusted = 0.86 [95% CI: 0.75–0.99], I² = 48.9%). Subgroup analysis stratified by study design did not yield significant risk estimates (OR_adjusted for case-control studies = 0.87 [95% CI: 0.75–1.02]; OR_adjusted for cohort studies = 0.73 [95% CI: 0.49–1.09]). As in the analysis of the crude ORs, there is little heterogeneity between case-control studies and cohort studies (shown in Table 2). Figure 3 displays for each study the adjusted OR and corresponding 95% CI for the risk of MS. Restricting analyses to studies with low risk of bias [18‒22, 24‒30], i.e., including only studies with ≥6 points on the NOS, yielded an OR_crude of 0.79 (95% CI: 0.66–0.94) and OR_adjusted of 0.83 (95% CI: 0.71–0.98). Overall, sensitivity analyses showed consistent results with the primary meta-analysis. Subgroup analyses by time of MS onset, i.e., paediatric versus adult-onset MS, also showed a protective effect of having been breastfed on MS risk both in paediatric- and adult-onset MS, although heterogeneity was substantial in the analysis of paediatric-onset MS (shown in online suppl. Table 3).

Visual inspection of the funnel plot revealed no indication of a relevant publication bias (shown in Fig. 4).

A systematic review and meta-analysis were conducted including 15 studies comprising 17 distinct study samples on the influence of having been breastfed on the risk of developing MS later in life. The meta-analysis included two cohort studies with 246 cases and a total cohort size of 259,934 individuals and 15 case-control studies comprising 6,530 cases and 9,896 controls. To our knowledge, this is the first meta-analysis summarizing the current study evidence on the influence of having been breastfed during infancy on the development of MS later in life.

We found a small protective effect of having been breastfed on subsequent MS risk compared to not having been breastfed in analyses of combined OR_crude as well as combined OR_adjusted. When comparing the results of case-control studies with those of cohort studies, there is little heterogeneity, which adds to the judgement of validity of the overall results. Most included studies were of high (≥6/9 points) quality as assessed with the NOS. Although results need to be interpreted with caution due to moderate to substantial between-study heterogeneity in the case-control studies, findings go in line with the qualitative review by Vieira Borba et al. [9] which concludes that having been breastfed might play a protective role in the aetiology of MS due to the immune-modulatory and toxin-protective effects of breast milk. The protective effect of breast milk has already been shown for allergic and autoimmune diseases other than MS, for instance DT1 [33, 34], and asthma [9, 35, 36]. In their meta-analysis including 43 studies (9,874 patients with DT1), Cardwell et al. [33] found a protective effect of having been breastfed on DT1 risk comparing ever versus never breastfed individuals, irrespective of the mode of breastfeeding, i.e., exclusive or non-exclusive breastfeeding. Restricting the analysis to exclusively breastfed individuals (n = 33 studies, 7,621 individuals with DT1), the protective effect remained with a 26% reduction in DT1 risk. However, it should be taken into account that the between-study heterogeneity was high in these analyses [33]. For asthma, the meta-analysis of three cohort and ten cross-sectional studies by Lodge et al. [35] found a protective effect of having been breastfed on subsequent development of asthma (OR_pooled = 0.88 [95% CI: 0.82–0.95]). These results are supported by the meta-analysis of 15 observational studies comprising 7,406 individuals by Harvey et al. [36] who reported a 32% reduction in the odds of wheezing in the first year of life for ever versus never breastfed children.

Not only the influence of breast milk alone, but also the type and duration of breastfeeding, in particular, could play a role in the development of autoimmune diseases such as MS. The included studies by Ragnedda et al. [18] and Hedström et al. [19] used a different dichotomization for breastfeeding and found a protective effect for longer duration (≥4 months vs. <4 months or no breastfeeding). Note that for other autoimmune diseases such as ulcerative colitis and celiac disease, breastfeeding for up to 6 months is recommended [37, 38]. Moreover, for celiac disease [38] and allergy prevention [39], there is indication that continuing breastfeeding while introducing solid foods is beneficial [38, 39], because this might support the tolerance development of the immune system which is crucial for the prevention of autoimmune disorders as well as allergies.

Since our meta-analysis is based on studies of the general population and MS-specific evidence regarding the influence of duration, mode of breastfeeding and optimal time point for introducing solid foods is insufficient, recommendations should be based on general [40] and national recommendations [41] for breastfeeding. Consistent with recommendations for other autoimmune diseases, the newborn should be exclusively breastfed for 4–6 months and solid foods should be introduced while continuing breastfeeding. By implementing these recommendations, women who wish to have children could additionally accrue the well-known benefits of breastfeeding for both mother and child in relation to various diseases (e.g., reduced risk of obesity and diabetes in breastfed children; reduced risk of breast and ovarian cancer in breastfeeding women [40, 42, 43]). Our study has some limitations concerning both the included studies and the conduct of the meta-analysis itself. The included studies used different diagnostic criteria to assess the outcome of interest, which may have led to differences in the detection of MS; however, it was not possible to perform subgroup analyses by diagnostic criteria used.

The majority of studies considered are case-control studies [18‒24, 26, 27, 29, 29‒32] which are prone to recall bias [44, 45]. However, we do not expect a differential recall in cases and controls, and therefore a possible bias is rather towards a null effect. Breastfeeding was defined differently in terms of duration of breastfeeding, exclusive breastfeeding, and complementary feeding making it difficult to determine the sole effect of having been breastfed on MS risk.

Furthermore, the included studies differed, first, in the set of variables for which they performed adjustment including generally accepted risk factors for MS, e.g., EBV-infection [3], obesity in childhood and adolescence [46‒48], smoking [49], or low blood vitamin D concentrations [50]; and second, other potential confounders were not considered. This is partially explained by the different settings and locations of the studies, however, makes it more challenging to combine the estimates. Following the recommendations made above on breastfeeding for MS prevention, future studies should assess breastfeeding in more detail. Given the known female-male ratio in MS prevalence, possible sex-specific aspects should be investigated, as was only the case with the study by Hedström et al. [19], who reported that the protective effect for MS appeared to be higher in males (OR = 0.5 [95% CI: 0.4–0.7]) than in females (OR = 0.9 [95% CI: 0.8–1.1]).

Moreover, only previously published studies in German or English were included, which may have led to studies being overlooked. However, we assume that the number of unidentified studies is relatively small and would not have changed the results of the meta-analysis, as publishing in English-language journals is common practice.

In summary, we found a small protective effect of breastfeeding on MS risk. This adds to the knowledge that breastfeeding is beneficial for the immunological health of a child. Future studies on the influence of having been breastfed on MS risk should apply a uniform definition of breastfeeding and investigate possible sex-specific aspects.

An ethics statement is not applicable because this study is based exclusively on published literature.

CH has received speaker honoraria and research funding from Novartis, Merck, Roche. The other authors have no known conflicts of interest to declare.

The study was supported by the German Federal Ministry of Education and Research (Grant No. 01ER1901A PERGOLA2). The funders had no role in the design of the review, the data collection and analysis, decision to publish, or preparation of the manuscript.

Concept and design: A.H., H.B.; acquisition, analysis, or interpretation of data: A.H., M.R., N.O., H.B.; drafting of the manuscript: A.H.; critical revision of the manuscript: all authors; statistical analysis: A.H., H.B.; supervision: C.H., N.O., H.B.; all authors read and approved the final manuscript and gave their approval of it to be published.

All data generated or analysed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author.