Effects of Exclusive Breastfeeding Promotion Interventions on Child Outcomes: A Systematic Review and Meta-Analysis

Breast milk is the optimum source of nutrition for infants. Promoting and supporting breastfeeding (BF) is an important public health intervention with multiple benefits for infants and mothers, including a reduced risk of infant gastrointestinal and respiratory infections and reduced risk of breast cancer in the mother [1]. In 2001, following a systematic review on the effects on child and maternal health of exclusive breastfeeding (EBF) for 6 months versus EBF for 3 to 4 months [2], the World Health Organization (WHO) made a global recommendation that infants should be exclusively breastfed until 6 months of age. After 6 months, safe complementary foods are recommended with maintenance of BF up to 2 years of age or beyond. However, despite numerous initiatives over many years in different settings, globally only 44% of infants 0–6 months old are EBF [3], with particularly low rates in many Western countries.

Much of the available evidence focuses on health effects of BF rather than the impact of EBF or EBF for 6 months. Furthermore, defining causal effects of BF per se, or specific periods of EBF on health outcomes is problematic as most studies are observational, given the ethical and practical difficulties of conducting randomized trials in this field. Therefore, the optimal duration of EBF [4] and the magnitude of benefits in different settings remain uncertain. Observational studies attempt to control for possible confounding factors, but it is difficult to completely account for the complex biological, social, economic, and cultural factors that influence BF and health outcomes.

Although the process of randomizing mothers to EBF for different durations is challenging, it is possible to randomize mothers to interventions aimed at promoting a longer duration of EBF. These trials provide an opportunity to systematically synthesize evidence for health outcomes in infants and children with different EBF exposure. A systematic review and meta-analysis examining the efficacy of interventions aimed at increasing exposure to EBF found that mothers who received interventions were 2.77 times more likely to EBF up to 6 months [5]. However, this analysis did not consider the impact on infant health outcomes. To address this gap, we conducted a systematic review to evaluate the effect of interventions that aimed to increase infant exposure to EBF on their growth, morbidity and mortality up to 7 years of age in both low- and high-income settings.

The full protocol is available on PROSPERO [CRD42020203796]. A systematic search of the literature was performed using medical subject headings (MeSH) and key text words pertaining to EBF interventions and infant/child mortality and health outcomes (see online supplement 1; for all online suppl. material, see https://doi.org/10.1159/000535564). The search was conducted in August 2020, and re-run on April 27, 2022. The Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trails, WHO International Clinical Trials Registry Platform (ICTRP), Cumulative Index of Nursing and Allied Health Literature, EMBASE, and MEDLINE were searched from inception. Reference lists of included articles were reviewed manually to identify additional sources. Studies were limited to those published in English and French.

Randomized controlled trials (RCTs) and cluster-randomized controlled trials (CRCTs) that focused on increasing infant exposure to EBF and reported at least one of the specified outcomes were included.

Healthy mother-infant pairs and/or infants who were followed at any time up to age 7 years were eligible. Studies that exclusively included preterm infants or mothers living with HIV were excluded.

Any intervention(s) aiming to increase exposure to EBF.

Controls could have no intervention or receive standard care for the study setting.

The primary outcome was infant and child growth, including anthropometric measurements such as weight, height/length, body composition, and BMI at the latest reported time point and at 3–4 months, 6 months, and at ≥18 months, and changes in weight and length/height. Secondary outcomes were infant and child morbidity and mortality and EBF rates up to 6 months. Morbidity included prevalence or incidence of disease and/or infections at the latest available time point and at 0–3 and 4–6 months. For all outcomes, we used author definitions but this was evaluated in the quality assessment.

The review included studies from low-middle income countries (LMIC) and high-income (HIC) countries, defined according the World Bank Classification.

Two review authors independently ran the initial and updated search using Covidence (https://www.covidence.org/), scanning titles and abstracts of retrieved records; those meeting the selection criteria were assessed further through full text appraisal. Discrepancies were discussed with a third reviewer to reach consensus. Additionally, three other members of the review team checked a random sample of 10% of both included and excluded studies to ensure agreement with methodology and logic. Figure 1 shows the PRISMA [6] flowchart.

Data extraction was performed by two independent review authors. All data included in the meta-analysis were additionally checked by a third author. Key information was extracted using a pre-defined template on Covidence. Disagreements were resolved through discussion with the review team. For studies with incomplete data, the authors of the original study were contacted.

Two authors independently assessed the risk of bias (ROB) for included studies using the Cochrane Handbook for Systematic Review of Interventions with disagreements resolved by discussion [7]. In addition, for CRCTs we assessed the risk of (i) recruitment bias; (ii) baseline imbalance; (iii) loss of clusters; (iv) incorrect analysis; and (v) comparability with individually randomized trials as outlined in the Cochrane Handbook for Systematic Reviews of Interventions [8]. Each ROB category was classified “high risk,” “low risk,” or “unclear risk.” As specified in the protocol, “high quality” was defined as a trial having adequate sequence generation, allocation concealment and an attrition rate <25%.

For two or more studies reporting similar outcome measures, we undertook statistical analysis using the Review Manager software (RevMan 2014). We analyzed outcomes on an intention-to-treat basis whenever possible. Random-effects meta-analysis was used given variability in baseline characteristics, interventions, and outcome reporting [9]. Results were presented as mean treatment effect with 95% confidence intervals (CIs) using odds ratio for dichotomous data and standardized mean difference for continuous data.

For CRCTs that did not adjust for clustering in their analyses (or did not report adjusted results), we adjusted the sample sizes using the methods described in the Cochrane Handbook [Section 23.1] [10] using an estimate of the intracluster correlation co-efficient (ICC) derived from the trial (if possible), from a similar trial or from a study in a similar population. Meta-analysis using the generic inverse variance method was undertaken in RevMan using the adjusted estimates from CRCT that adequately accounted for the cluster design within the analysis, alongside an approximated effect estimate for the CRCTs that did not.

Heterogeneity was assessed through the χ² and I² statistics. An I² statistic of 50% or greater or χ² significance level p < 0.1 was regarded as indicating substantial heterogeneity between studies [7].

To identify potential sources of substantial heterogeneity, subgroup analysis according to country income classification, type of intervention, and timing of intervention was performed when at least two studies were available in each comparator. We also planned subgroup analysis according to maternal BMI, baby-friendly initiative accreditation and family socioeconomic status but data did not allow for these analyses. Additionally, subgroup analysis for child growth, morbidity, and mortality outcomes was performed, stratified by whether the intervention resulted in a significant improvement in the odds of EBF at latest available timepoint, 6 months, 3–4 months, or 1–3 months.

Sensitivity analysis was performed based on the quality of the included trials determined by ROB to identify the impact on the overall results. For cluster-randomized trials, sensitivity analysis was also used to investigate the effect of variation in the ICC and unit of randomization. We initially adjusted the sample size for unadjusted outcomes using the ICC of 0.1 [11] then undertook sensitivity analysis using varying ICC from 0.01 to 0.3 [12, 13].

For outcomes that were reported in ≥10 trials, we investigated reporting biases by visually examining asymmetry in funnel plots. We also undertook the Egger, Smith [14] regression asymmetry test using Stats Direct software, with p < 0.10 taken as evidence of small study effects.

We initially identified 9,158 studies, of which 7,158 were screened and 106 full-text articles were reviewed (Fig. 1). A total of 32 studies (40 publications) were eligible for inclusion [11‒13, 15‒51].

Of the included studies (Table 1), 17 were RCTs and 15 were CRCTs, 28 were in LMIC and only 4 were in HIC, with sample size between 108 and 140,048 participants. Year of publication was 1994–2021, with 23 studies published in the last 10 years. Most interventions lasted longer than 1 month (25 interventions), four lasted 1 week to 1 month and four lasted 1 week or less (one study contained two interventions, one <1 week and one >1 month). The type of intervention varied, with most (n = 20) involving peer counselling and/or home visits while five studies targeted hospital/clinic practices. Three studies investigated the effects of longer versus shorter durations of EBF by randomizing infants to complementary feeding from 4 months or EBF until 6 months. Two studies provided mobile phone-based support to BF mothers. Two studies involved BF support provided by community representatives or members. One study was undertaken in three separate countries; effect size estimates were given separately for each country, so they were entered into the meta-analyses separately.

The quality of included studies based on pre-defined criteria was poor, with 24/32 studies of low or uncertain quality and only eight of high quality (online suppl. 2). Eighty-seven percent (14/16) of CRCTs had high risk of recruitment bias where individuals were recruited after randomization of clusters. All studies had low ROB when considering imbalance between randomized groups. Two studies had high ROB for not undertaking adjustment for clustering and two had potentially high ROB due to loss of clusters.

EBF promotion interventions did not have an overall effect on infant weight (Fig. 2; SMD = −0.01 [95% CI: −0.11, 0.09], 11 studies, 11,556 participants, I² 67%), infant length/height (Fig. 2; SMD = 0.01 [95% CI: −0.03, 0.05], 12 studies, 28,817 participants, I² 21%) nor BMI/weight-for-length (Fig. 2; SMD = −0.04 [95% CI: −0.10, 0.03], 11 studies, 27,702 participants, I² 54%) at the latest time-point reported. There were also no effects at 3–4 months, 6 months nor at ≥18 months (online suppl. 3). Similarly, there were no effects on gain in length or weight (online suppl. 3).

There was a trend toward reduced odds of infant diarrhea with EBF promotion interventions at the latest available time-point (Fig. 3; OR = 0.84 [95% CI: 0.70, 1.00], 12 studies, 24,060 participants, I² 42%), but not at 0–3 months and 4–6 months (online suppl. 3). There was no effect of EBF promotion interventions on respiratory illness at the latest time-point (Fig. 3; OR = 0.80 [95% CI: 0.60, 1.06], 5 studies, 19,718 participants, I² 44%) nor at 4–6 months; however, there was a significant reduction in the odds of respiratory illness at 0–3 months (online suppl. 3; OR = 0.41 [95% CI: 0.20, 0.84], 2 studies, I² 0%). No differences were found in infant hospitalization (Fig. 3; OR = 0.56 [95% CI: 0.31, 1.02], 5 studies, 3,162 participants, I² 53%) nor infant mortality (Fig. 3; OR = 0.98 [95% CI: 0.75, 1.28], 5 studies, 60,918 participants, I² 52%).

The interventions were successful at improving the odds of EBF (pooled odds ratio 3.15 [95% CI: 2.36, 4.19, 25 studies, 202,644 participants, I² 85%]) at any time point from 0 to 6 months (Fig. 4). Similar effects were observed at 6 months, 3–4 months, and 1–3 months separately (online suppl. 3). Across all settings, the median EBF rate at 3–4 months was 66.7% (IQR 25.4; 10.4–83.1) and 34.6% (IQR 26.7; 6.2–80.1) in the intervention and control groups (17 studies), respectively. This rate was lower at 6 months (34.5% [IQR 45; 2.0–73.0] vs. 13.7% [IQR 27.6; 0.0–54.8]), respectively (18 studies).

Studies conducted in LMIC yielded a significantly higher OR of EBF in the intervention group compared to those in HIC (3.31 [2.44, 4.49] vs. 1.66 [0.98, 2.81]; p = 0.02) with 80% of the heterogeneity in EBF explained by income classification (online suppl. 4). There were significant subgroup differences in the OR of respiratory illness between interventions that started prenatally (0.51 [95% CI: 0.33, 0.79]) versus postnatally (0.97 [95% CI: 0.78, 1.20]) with 85% of heterogeneity explained by timing of the intervention. There were no significant differences in growth outcomes between subgroups stratified by whether the intervention resulted in an improvement in EBF (at latest available time point, 6 months, 3–4 months, or 1–3 months). There were insufficient studies to test these subgroup differences for respiratory illness and mortality outcomes.

Confining analyses to RCTs resulted in significantly lower pooled odds of diarrhea (OR = 0.69 [95% CI: 0.53, 0.89]) and a trend toward higher length/height (SMD = 0.13 [95% CI: 0.00, 0.26]) with interventions. Conversely, limiting inclusion to high-quality studies did not notably affect the magnitude or direction of the intervention effect and CIs. Using an ICC of 0.01 resulted in significantly lower pooled odds of diarrhea (OR = 0.83 [95% CI: 0.70, 0.99]). However, overall there was minimal impact on the magnitude of the intervention effect and CIs, and no change in the direction of effect for analyses using different ICC.

There was no evidence for publication bias for weight, length/height, BMI/weight-for-length, or diarrhea (online suppl. 5). There was evidence of publication bias (Egger bias = 1.81; 95% CI: −0.09, 3.71; p = 0.06 < 0.1) for EBF when using data for the latest available time point within each study, but this was not evident when including data at different timepoints (6 months, 3–4 months, or 1–3 months).

This review examined the effects of exposure to more versus less breast milk as a result of EBF promotion interventions. The interventions significantly improved EBF rates at various time-points, however, EBF rates at 6 months were low overall. Promotion of EBF reduced the odds of respiratory illness at 0–3 months. There was also a trend toward a reduction in the overall odds of diarrhea which was significant in the sensitivity analysis. No significant differences were found in infant/child growth, even when results were stratified by whether the intervention improved EBF. There were also no significant differences in hospitalizations or mortality.

Our results for infant/child growth are consistent with a previous systematic review and meta-analysis published in 2012 that concluded that EBF for a longer duration was not associated with growth deficits [4] and add to the body of evidence that promotion of EBF does not give rise to growth concerns. Interestingly, our analyses confined to RCTs also suggested better linear growth in infants whose mothers received an EBF promotion intervention.

Overall, there was a significant reduction in the odds of respiratory illness at 0–3 months following EBF interventions but not at 4–6 months or at the latest available time point. The inconsistent effects of EBF interventions on respiratory infection in different studies [11, 13, 16, 28, 31] are similar to those reported by Kramer and Kakuma [4] and contrast with previous studies that consistently show strong protective effects of BF per se against respiratory infection [52]. This might suggest that promotion of BF rather than focusing specifically on EBF could yield larger benefits in reducing respiratory infection, although this was not addressed in our meta-analysis.

There was a trend toward lower odds of diarrhea with EBF promotion interventions. Contrary to what might be expected due to clustering “herd effects,” the effect on diarrhea was significant when considering only RCTs (and excluding CRCTs) and also when adjusting for clustering using a smaller ICC. Nevertheless, with the exception of three CRCTs (from Ethiopia [11], DR Congo [13], and South Africa [39, 44]) which showed higher prevalence/incidence of diarrhea in the intervention group, all other studies favored the intervention group. It is biologically plausible that promotion of EBF would protect against gastrointestinal infections as it avoids contamination from unsafe preparation of breast milk substitutes or other foods, and breast milk in general provides several anti-microbial and anti-inflammatory compounds. We were not able to explore this further as data on the infant feeding practices of participants who were not EBF was not consistently reported.

Previous BF meta-analyses of mostly observational studies such as Victora et al. [1] found that BF offers strong protection against infections and hospitalizations due to infections. They also found that BF was associated with some reductions in overweight/obesity. Direct comparisons between these findings and ours are difficult as our focus was the promotion of EBF whereas Victora et al. [1] compared groups with BF defined in several ways (for example, any BF vs. none, predominant vs. partial, EBF vs. partial/predominant). We also only included interventional studies and excluded observational studies. Furthermore, most studies included in our meta-analysis reviewed the effects of exclusive EBF promotion interventions rather than the effects of EBF; only three trials directly randomized mothers to EBF for specific durations. Therefore, the magnitude of effects on infant and child outcomes might be underestimated by non-compliance with the interventions and depends on the success of the interventions in promoting EBF. It is also possible that had we been able to subdivide the control groups according to BF intensity (predominant, vs. partial vs. not BF) or according to definition of not EBF (introduction of solids, liquids, and/or infant formula), we may have seen larger effect sizes.

Strengths of our review are the inclusion of studies from all settings and a range of health outcomes assessed up to 7 years of age, allowing a comprehensive review of the effects of EBF promotion. However, there are several limitations. We examined the effects of exposure to more versus less breast milk as a result of EBF promotion interventions and did not directly address whether EBF for 6 months has greater benefits than EBF for 4 months; indeed, we noted that EBF rates at 6 months were low even in the intervention groups with a median of only 34.5% for studies that reported this outcome. As expected, included studies were heterogenous and country income classification, type of intervention and timing of the intervention were significant sources of heterogeneity for some outcomes. We were also unable to conduct planned subgroup analysis by maternal BMI, Baby-Friendly Initiative accreditation status, or family socioeconomic status. There were insufficient studies investigating other potential benefits of EBF promotion such as reduced risk of noncommunicable diseases or allergy/asthma. Additionally, as for all meta-analyses, decisions must be made about which data to include, for example when multiple interventions are used or the same outcome is assessed in different ways and this can potentially influence findings. We used a systematic approach and discussed these decisions to avoid introducing bias.

In conclusion, EBF promotion interventions were successful at improving EBF. There were modest reductions in respiratory infection and diarrhea with no effects on infant and child growth or mortality. However, even modest decreases in infections, which are associated with significant morbidity and mortality especially in low-income settings, could translate to significant public health benefits and reduced healthcare expenses. Future studies investigating the effects of longer versus shorter EBF durations on other outcomes, and including cost-effectiveness analysis of EBF versus BF promotion could provide further insight into this issue.

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

S.D., F.J.F., L.J.M., A.N., A.Z.K., and H.S. have no conflicts of interest to declare. M.F. has received an unrestricted donation for research on infant nutrition from Philips (not related to the current manuscript); she is Assistant Officer for Nutrition at the Royal College of Paediatrics and Child Health UK, a member of the Infant Nutrition working group at the European Food Safety Authority (EFSA) and General Secretary of the European Society for Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN).

The authors received no financial support for this research.

Dr. Sarah Dib ran the final study search, assessed articles for inclusion, undertook quality appraisal, extraction of data, and data analysis, drafted the initial manuscript, and critically reviewed and revised the manuscript. Ms. Frankie Joy Fair developed and reviewed the study protocol, checked a random sample of included and excluded studies, cross-checked all extracted data within the analyses and assisted with the data analysis, and critically reviewed and revised the manuscript. Dr. Lucy Jane McCann assessed articles for inclusion, cross-checked all extracted data, and critically reviewed and revised the manuscript. Ms Antonia Nicholls conceptualized the study, developed the study protocol, ran the initial search, assessed articles for inclusion in the initial search, and critically reviewed and revised the manuscript. Prof Anastasia Kalea developed the study protocol and critically reviewed and revised the manuscript. Prof Hora Soltani developed and reviewed the study protocol, checked a random sample of included and excluded studies, cross checked, and critically reviewed and revised the manuscript. Prof. Mary Fewtrell conceptualized the study, developed the study protocol, undertook quality appraisal of included studies, discussed discrepancies in study inclusion, checked a random sample of included and excluded studies, and critically reviewed and revised the manuscript. All authors approved the final manuscript as submitted and agreed to be accountable for all aspects of the work.

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