Impact of Infant and Young Child Feeding (IYCF) Nutrition Interventions on Breastfeeding Practices, Growth and Mortality in Low- and Middle-Income Countries: Systematic Review

Undernutrition is associated with 45% of total infant deaths, totalling 2.7 million globally per year. The vast majority of the burden is felt in low- and middle-income countries (LMICs). This review aims to assess the effectiveness of infant and young child feeding (IYCF) interventions. We searched multiple databases including Cochrane Controlled Trials Register (CENTRAL), MEDLINE, EMBASE. Title/abstract screening and full-text screening and data extraction filtered 77 studies for inclusion. Breastfeeding education interventions (n = 38) showed 20% increase in rates of early initiation of breastfeeding, 102% increase in exclusive breastfeeding (EBF) at 3 months and 53% increase in EBF at 6 months and 24% decreases in diarrheal diseases. Complementary feeding education intervention (n=12) showed a 0.41 standard deviation (SD) increase in WAZ, and 0.25 SD in HAZ in food secure setting. Complementary food provision with or without education (n=17) showed a 0.14 SD increase in HAZ and 36% decrease in stunting. Supplementary food interventions (n=12) showed a significant 0.15 SD increase in WHZ. Subgroup analyses showed healthcare professional led interventions were largely more effective, especially on breastfeeding outcomes. We believe this is a comprehensive review of the existing literature on IYCF studies in LMICs. Though breastfeeding education is well supported in its effectiveness on breastfeeding practices, limited evidence exists for growth outcomes. Supplementation interventions seem to have better effects at improving growth. However, more research is required to reach more substantial conclusions.


Introduction
Undernutrition is associated with 2.7 million child deaths worldwide which makes up 45% of all infant mortality, according to the World Health Organization [1]. Lower-and middle-income countries (LMICs) take the overwhelming brunt of the tragedy, accounting for 99% of such instances of infant mortality in Sub-Saharan Africa and South Asia [1]. It is estimated that 250 million children under the age of five are at risk of sub-optimal development and stunting [1]. Infant malnutrition has been

Objectives
The overall objectives are to assess the effectiveness of the following on child health and nutritional status: • Interventions to promote early and exclusive breastfeeding; • Interventions to promote continued breastfeeding; • Interventions to promote appropriate complementary feeding (education or provision of complementary food) during infancy and childhood; • Effectiveness of community-based interventions to prevent moderate and severe acute malnutrition.

Types of Studies and Participants
We included primary studies, including large-scale program evaluations to assess the efficacy and/or effectiveness of interventions. We included randomized controlled trials (RCTs) randomized at the individual or cluster level (i.e., in cluster randomized controlled trial), cross-over study designs. We also included non-randomized studies such as quasi-experimental designs, including controlled before-after studies (CBA), interrupted time series (ITS) studies, natural experiments and regression discontinuity designs. Non-controlled pre-post studies were not included.

Types of Participants
Our target population of study was mothers/nursing mothers of children under 2 years of age living in lower middle-income countries (LMICs) of reproductive age, regardless of comorbid and general health status. Participants could be pregnant while interventions were applied. We also included studies in which only a portion of the study sample fit our inclusion criteria, provided they were compared independently to a control group and analyzed independently. Conversely, if arms of an included study did not fit our inclusion criteria, those arms of interventions were not included in our analysis.

Types of Interventions
This review focused on the effects of the following interventions: 1.
Interventions designed to promote early initiation, exclusive and continued breastfeeding, such as education and support of early and exclusive breastfeeding practices.

2.
Interventions to promote dietary diversification and appropriate complementary feeding. These will be divided in interventions that provide only education and interventions in which the provision of complementary food for healthy individuals takes place regardless of education. 3.
Interventions to prevent moderate and severe acute malnutrition, such as educative interventions and supplementary food for children suffering from MAM or SAM.
Interventions were supervised by healthcare professionals, community health workers and through telephone network platforms e.g., text messages, phone apps. Interventions were compared to control groups receiving no interventions or standard of care.

Types of Outcome Measures
Primary outcomes: Early initiation of breastfeeding (within an hour of birth), exclusive breastfeeding (EBF) at 3 and 6 months of age, infant growth (weight gain (kg), height gain (cm), Z-scores for height-for-age (HAZ), weight-for-age (WAZ), weight-for-height (WHZ), stunting (HAZ < 2), wasting (WAZ < 2), underweight (WHZ < 2)). Our secondary outcomes included neonatal mortality (death from all causes within a month of birth per total live births), infant mortality (death from all causes within 12 months of age per total live births), neonatal sepsis (the proportion of neonates dying due to possible serious bacterial infections among all neonates), incidence of acute respiratory infections (ARI), incidence of diarrheal illness, any adverse events including gastrointestinal symptoms, non-compliance or Nutrients 2020, 12, 722 4 of 21 decreased feeding, cost-effectiveness of the intervention (reported in narrative text) and the psychosocial health of the infant (different scales for psychomotor development, cognitive development, attention,  memory, language).
Studies were excluded if they did not report any of the outlined outcomes.
Every effort was made to contact the relevant organization, institutions and experts for the identification of unpublished and ongoing studies. Google Scholar and the Web of Sciences were searched for citations and the references sections and annotated bibliographies were cross referenced for bonus eligible studies.

Data Collection and Analysis
Data collection and analysis was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [17]. The selected literature was double screened at the title and abstract level by two independent review authors. The resulting full-texts were similarly double-screened by two independent authors with reasons recorded for exclusions. Disagreements were resolved by the consultation of a third senior review author. Duplicates were excluded. Multiple reports of the same study were collated to ensure each study was the unit of interest in this review. The selection process was recorded sufficient to construct a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram (see Figure 1) [18].
Two independent review authors extracted data on data extraction sheets on Microsoft Excel. Independent data extraction sheets were matched, and disagreements were resolved by discussion and third senior review author was consulted, if the need arose. We used a piloted data collection form for study characteristics and outcome data. If any information from the study was unclear or missing or could not be from the publications, the authors were contacted for further clarification.
Data was exported and analyzed on RevMan v5.3. Risk ratios (RR) were used for dichotomous outcomes, mean differences (MD) were calculated for continuous outcomes and standard mean differences were calculated for continuous outcomes reporting the same outcome but measured on a different scale within the same meta-analysis. A 95% confidence interval was used for all estimates of effect. We analyzed outcomes from studies with multiple groups in an appropriate way to avoid double counting of participants by adding them as an appropriately duplicated entry with an added footnote specifying the differences between groups.
Nutrients 2020, 12, 722 5 of 21 abstract level by two independent review authors. The resulting full-texts were similarly doublescreened by two independent authors with reasons recorded for exclusions. Disagreements were resolved by the consultation of a third senior review author. Duplicates were excluded. Multiple reports of the same study were collated to ensure each study was the unit of interest in this review. The selection process was recorded sufficient to construct a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram (see Figure 1) [18].

Assessment of Risk of Bias in Included Studies
Two review authors independently assessed the risk of bias of randomized controlled trials using the Cochrane Collaboration Risk of Bias tool [17] for selection bias, performance bias, attrition bias, reporting bias and other sources of bias. The risk of bias was rated either "high", "low" or "unclear".
For non-randomized controlled trials, controlled before-after studies and interrupted time series, we used EPOC methods (EPOC 2017a) to assess the risk of bias according to the following domains: random sequence generation, allocation concealment, baseline outcome measurements, baseline characteristic, incomplete outcomes, knowledge of allocated interventions adequately prevented during the study, protection against contamination, selective outcome reporting and other sources of bias. The risk of bias was rated either "high", "low" or "unclear".

Assessment of Quality Evidence
A "Summary of Findings" table was constructed for all primary outcomes using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria [19] which involves consideration of within-study risk of bias (methodological quality), directness of evidence, heterogeneity, precision of effect estimates and risk of publication bias. The tool then rated quality of evidence as "high", "moderate", "low" and "very low". Observational studies stood to be upgraded if they exhibited a large magnitude of effect, dose-response relationships and/or effect of plausible residual confounding as they begin at a "low" level of quality of evidence.
Breastfeeding education interventions had the following outcomes GRADEd: rates of early initiation of breastfeeding (within an hour of birth), rates of exclusive breastfeeding (at 3 and 6 months of age) and rates of continued breastfeeding (at 12, 18 and 24 months) Complementary feeding education and provision studies and supplementary feeding interventions had the following outcomes GRADEd: mean Z-score for height-for-age, mean Z score for weight-for-height, mean Z score for weight-for-age, stunting, wasting, underweight.

Subgroup Analysis, Investigation of Heterogeneity and Sensitivity Analysis
We conducted the following subgroup analyses where there were at least three studies in each sub-group, in adherence with Cochrane collaboration recommendations. Subgroups were divided by setting (home-based, facility-based or community-based), timing of intervention (prenatal, postnatal or a combination), duration of intervention (less than or equal to six months or greater than six months), observer (self-reported or clinically observed outcomes), moderators (community health worker (CHW)/volunteers or healthcare professionals) and by training of personnel (with WHO or United Nations International Children's Education Fund (UNICEF) materials or from other sources) for breastfeeding education studies. Complementary feeding and supplementary feeding studies were subgroup analyzed by the level of food security of their settings using information available in the full texts of the articles. Sensitivity analyses were also performed to consider the impact of the following: high risk of bias in allocation concealment methodology, high risk of attrition bias due to lost participants and differences in outcomes due to different methods of calculating cluster adjustment
The minimum population size was 60 infant mother pairs in Ahmed et al. [22] and the maximum population size was 113,816 females in Baqui et al. [27]. The oldest average age among participants was 30.75 months in Younes et al. [94]. The timing of the intervention was prenatal in four studies; postnatal in 17 studies; and both prenatal and postnatal in 17 studies. In 21 studies, the intervention was provided by community health workers. In 10 studies, intervention was provided by health care professionals and 8 studies had study investigators as the providers respectively. Twelve studies' training was sourced from WHO materials. Thirteen studies detailed training which was not based on WHO authored resources. The duration of training varied greatly from a minimum of two days in Arifeen et al. [26] and Sikander et al. [84] to a maximum of nine to 14 days in Nair et al. [67]. The training details are expanded in Table S6 (see Supplementary Data S1). The training methods were not specified in 13 studies, thus, they could not be included in the training/timing of training subgroup analyses. The risk of bias by study is enumerated in Figure 2.
Breastfeeding education had no effect on the incidence of infection, as reported by three studies [26,54,63]. Sensitivity analysis by allocation concealment methodology was not possible as all studies were at a low risk of bias. However, sensitivity analysis for attrition bias removed one study [54] and still showed no effect on the incidence of infections.
Breastfeeding education had no effect on the incidence of infection, as reported by three studies [26,54,63]. Sensitivity analysis by allocation concealment methodology was not possible as all studies were at a low risk of bias. However, sensitivity analysis for attrition bias removed one study [54] and still showed no effect on the incidence of infections.
The complementary feeding education interventions spanned two continents. Asia housed nine studies; three in India [30,31,92], two in China [43,82], two in Pakistan [77,95] and two in Bangladesh [55,76]. Three studies were carried out in South America; two studies were carried out in Brazil [79,93] and one in Peru [74].
The range of the populations was from 212 at its minimum [76] and 1025 at its maximum [31]. The ages of participants ranged from 37 weeks of gestation at the minimum to 15.45 months at the maximum. There was also some variation in the intricacies of the interventions provided. The risk of bias by study is enumerated in Figure 2.
Complementary feeding education interventions caused a 0.41 SD increase in WAZ in food secure settings [43,74,79,95] and a 0.47 SD increase in food insecure settings [76] when compared with control groups (see Table 1).
Complementary feeding interventions had a significant effect in HAZ in food secure settings [43,74,79,95] and caused a 0.26 SD increase in food insecure settings [76] (see Table 1).
Complementary feeding education interventions had no significant effect on WHZ in food secure settings [74,79,95] and a 0.50 SD increase in a food insecure setting [76] when compared with control groups (see Table 1).
Complementary feeding education interventions caused a non-significant 50% decrease in the prevalence of stunting in food secure settings [43,74,77] and a 35% decrease in food insecure settings [30,31,92] when compared with control groups (see Table 1).

Comparison 2b: Complementary Feeding Provision with or without Education vs. Control
Seventeen trials were included that addressed the effect of complementary feed provision studies regardless of the provision of education. HAZ was reported by 12 studies, WAZ was reported by 10 studies, WHZ was reported by 10 studies, stunting was reported by seven studies, wasting was reported by five studies, weight was reported by four studies and height gain by three studies (see Table 1).

Comparison 3: Supplementary Feeding Programs vs. Control
The studies spanned five continents. Africa housed five studies i.e., two in Niger [41,50], two in Malawi [59,89], one in Kenya [90]. Two studies were conducted in Asia i.e., one in Indonesia [47] and one in Vietnam [80]. Two studies were conducted in only North America in Jamaica [40,45]. One study was conducted in only South America in Brazil [35] and Peru [36]. Simondon et al. [84] was conducted in four different countries in Congo, Senegal, Bolivia and New Caledonia (i.e., Africa, Africa, South America and Oceania, respectively).
There was no composite effect of supplementary feeding interventions on HAZ. Supplementary feeding interventions had a significant effect on HAZ in a food secure setting [47] but had no significant effect in food insecure settings [50,59,80,89,90] when compared to control groups (see Table 1).
Supplementary feeding interventions caused a significant decrease (RR 0.61; 95% CI 0.38 to 0.97; two studies; 4757 participants) in infant mortality rates when compared with control groups. All included studies were conducted in food insecure settings [41,50] (see Table 1).
Supplementary feeding interventions had no significant composite effect (MD 0.06; 95% CI −0.01 to 0.12; five studies; 845 participants; GRADE) on weight gain when compared with control groups. Similarly, no effect was seen in a food secure setting (MD −0.07; 95% CI −0.15 to 0.01; one study; 90 participants). However, food insecure settings saw a statistically significant 0.08 kg increase (MD 0.08; 95% CI 0.01 to 0.14; five studies; 755 participants) in mean weight gain (see Table 1).
Supplementary feeding intervention had no significant composite effect (MD 0.13; 95% CI −0.03 to 0.29; five studies; 832 participants) on mean height gain when compared with control groups. Similarly, no effect was detected in food secure nor food insecure settings (see Table 1).

Discussion
This review includes 66 studies of which 38 studies contributed to breastfeeding education comparisons, 17 studies contributed to complementary feed provision regardless of, 12 studies contributed to complementary feeding education only and 13 studies contributed to supplementary feeding comparisons. All primary outcomes were reported except for continued breastfeeding at 12 and 24 months of age. The included studies span 29 countries from four continents, all contributing to a comprehensive review of evidence.
Breastfeeding intervention caused a 20% increase in rates of early initiation of breastfeeding, a 102% increase in the rates of EBF at 3 months and 53% increase in EBF at six months. It was also associated in a 24% decrease in diarrheal disease. Complementary feeding was associated in 0.12 SD increase in height-for-age Z-scores, a 13% decrease in stunting and 11% decrease in wasting. Supplementary feeding caused a 39% decrease in the rates of infant mortality. Other outcomes did not report significant associations.
Subgroup analyses revealed information on the study characteristics that went into producing significant results. Early initiation of breastfeeding was only supported by significant effects in facilityand community-based interventions, postnatal or prenatal/postnatal CHW/volunteer or healthcare professional led, interventions lasting more than six months, trained on other resources. Moreover, strictly prenatal interventions and those lasting less than six months had no effect. That is to say that the moderators' profession and setting do not necessarily matter, but that the duration and timing of intervention are more crucial points in study design aiming to maximize early initiation of breastfeeding. It is worth mentioning that healthcare professional moderated interventions had double the effectiveness of those moderated by CHW/volunteers. Exclusive breastfeeding at 3 months of age was supported by significant effects in both community and facility-based interventions. Of note, facility-based interventions were more than 3 times more effective than the community-based interventions.
Exclusive breastfeeding at 6 months of age was supported by significant effects in all subgroups allowing the conclusion that the outlined subgroups can all be utilized for significant improvements in breastfeeding practices at 3 months. However, it was found that community-based interventions delivered in prenatal and postnatal period for periods less than six months based on WHO/UNICEF materials had more efficacious magnitudes of effect than their subgroup counterparts. This information can be used to design a maximally effective intervention if the goal is maximizing EBF at 3 months.
Complementary feeding education had significant effects on WAZ and HAZ in all food security settings. Complementary feeding education had significant effects on WHZ in only food insecure settings.
Complementary feed provision interventions had significant effects on HAZ in food insecure settings and on stunting in food secure and insecure settings. Supplementary feeding interventions had significant effects on HAZ and WAZ in food secure settings and on WHZ and infant mortality in food insecure settings.
The quality of outcomes was determined with the GRADE method [35]. In breastfeeding education interventions, early initiation of breastfeeding was graded low due to very serious inconsistencies, EBF at 3 months was graded very low due to serious inconsistency and EBF at 6 months was graded very low due to serious risk of bias and very serious inconsistency. Every effort was made to minimize the sources of bias including independent screening and assessment of data.
Our findings appear to agree with other systematic reviews of the relevant literature. Sinha et al. [97] reported a 66% increase in early initiation of breastfeeding in studies conducted in LMICs. The same study also conducted subgroup analyses and found facility-based interventions caused a 12% increase in early initiation of breastfeeding, which compare closely with the 18% increase observed in our subgroup analyses. Sinha et al. [98] also found community-based interventions caused an 86% increase in early initiation of breastfeeding which is significant but much larger than the 17% increase found in our subgroup analyses. It should be noted that the subgroup analyses do not strictly capture the effect of LMICs.
Rollins et al. [1,98] also had a significant 11% increase in rates of early breastfeeding which was in agreement with our findings but to a smaller magnitude.
Balogun et al. [99] was also in agreement, finding a 22% increase in the rates of early breastfeeding. Additionally, subgroup analyses performed by moderators showed healthcare professional led interventions had a 43% increase in early initiation of breastfeeding. These results were significant and in agreement with our subgroup analyses finding a 33% increase in healthcare professional moderated interventions' effect on early initiation of breastfeeding.
Rates of EBF at six months were reported in two systematic reviews [100,101]. Sinha's review found interventions in LMICs increased rates of EBF at six months by 69% [101]. This compares closely as a significant effect with the 53% increase found in our meta-analysis. Additionally, Sinha et al. conducted a subgroup analysis by setting which found facility-based interventions to have a significant 46% increase in EBF at six months [101]. This is again concordant with the 20% increase concluded by our subgroup analysis. Kim et al. included an extensive review of subgroups very similar to ours, allowing for a rich parallel comparison [97]. As a composite outcome, they found a 173% increase in the odds of EBF at six months (compared to the 53% increase found in our meta-analysis) [97]. Subgroup analysis by setting showed community-based interventions had a 177% increase in the odds of EBF at six months (compared to the 90% increase found in our meta-analysis) [97]. Kim et al. also conducted a subgroup analysis by timing of interventions and found prenatal interventions had no effect on EBF at six months, postnatal interventions caused 165% increased odds and pre/postnatal interventions caused an 83% increase [97]. Again, this is in agreement with our findings of prenatal interventions having no effect, postnatal interventions causing an 89% increase and pre/postnatal interventions causing a 71% increase in the rates of EBF at six months. Overall, wherever possible to draw a comparison, Kim et al.'s findings seems to be qualitatively in agreement with our findings [97]. However, Kim et al.'s findings consist of larger magnitudes of effect and are calculated as odd ratios [97].
Complementary feeding studies were divided into those providing only complementary feeding education and those providing only complementary feeding provision. Arikpo et al. [102] were studying only education interventions and reported no effects of interventions on the prevalence of stunting, wasting and underweight. Similarly, our review found no effect of complementary feeding education on stunting and wasting.
Lassi et al. [103] found significant effect of CF on HAZ and WAZ in food secure and food insecure settings which matched the findings of our review. Lassi et al. also found significant reductions in stunting in a food insecure setting, whereas our review included more studies but found no association in both food insecure and food secure settings.
Supplementary feeding programs were extensively analyzed as subgroups in Kristjansson et al.'s review [104]. This allowed for some comparison with RCT only outcomes of Kristjansson et al. [104] since the composite results included a wide variety of study designs. Subgroup analysis by LMICs showed there was no effect on WHZ and increases in weight gain (MD 0.12 kg), height gain (0.27 cm), change in WAZ (MD 0.15) and height gain (MD 0.85). These results are in complete agreement with the results of Kuusipalo et al. [59] as reviewed by us, as these outcomes were only reported by this study. Other supplementary feeding studies in the literature were not comparable, as, for example, Beaton et al. [105] focused on the effects of interval growth, Pollitt et al. [106] focused on the effect on cognitive performance.
The updated and extensive nature of this review offers much utility to policy-making bodies in the interest of maximizing resources for the betterment of infant nutrition. The present review offers such utility in breastfeeding education and practice as well as for the provision of complementary nutrition allowing for maximized growth and minimized states of malnutrition and disease. However, owing to the exclusive inclusion of studies from LMICs, the data presented is not necessarily generalizable outside of such socioeconomic settings.

Conclusions
We believe the present review provides a comprehensive review of the literature and has much to contribute to the literature in the forms of extensive subgroup analyses. The updated and extensive nature of this review offers much utility to policy-making bodies in the interest of maximizing resources for the betterment of infant nutrition in LMICs. The present review offers much utility in breastfeeding education and practice as well as for the provision of complementary nutrition allowing for maximized growth and minimized states of malnutrition and disease in the socioeconomic settings of our interest. The beneficial effects of breastfeeding interventions on future research in the area of supplementary feeding would allow more robust meta-analyses as current data is largely inconclusive. Another point of research could be investigating the cost effectiveness of such interventions as has been alluded to by many of the included studies but only reported by two of the studies.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/12/3/722/s1. Table S1: Breastfeeding education meta-analyses; Table S2: Complementary feeding education and provision meta-analyses; Table S3: Supplementary feeding meta-analyses; Table S4: Single study outcomes; Table S5: Search strategy; Table S6: Characteristics of included studies; Table S7: Summary of findings table of breastfeeding  education; Table S8: Summary of findings table of complementary feeding; Table S9: Summary of findings table of  supplementary

Conflicts of Interest:
The authors declare no conflict of interest.