Education of family members to support weaning to solids and nutrition in later infancy in term-born infants.

BACKGROUND
Education of family members about infant weaning practices could affect nutrition, growth, and development of children in different settings across the world.


OBJECTIVES
To compare effects of family nutrition educational interventions for infant weaning with conventional management on growth and neurodevelopment in childhood.


SEARCH METHODS
We used the standard strategy of Cochrane Neonatal to search the Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 5), MEDLINE via PubMed (1966 to 26 June 2018), Embase (1980 to 26 June 2018), and the Cumulative Index to Nursing and Allied Health Literature (CINAHL; 1982 to 26 June 2018). We searched clinical trials databases, conference proceedings, and references of retrieved articles. We ran an updated search from 1 January 2018 to 12 December 2019 in the following databases: CENTRAL via CRS Web, MEDLINE via Ovid, and CINAHL via EBSCOhost.


SELECTION CRITERIA
We included randomised controlled trials that examined effects of nutrition education for weaning practices delivered to families of infants born at term compared to conventional management (standard care in the population) up to one year of age.


DATA COLLECTION AND ANALYSIS
Two review authors independently identified eligible trial reports from the literature search and performed data extraction and quality assessments for each included trial. We synthesised effect estimates using risk ratios (RRs), risk differences (RDs), and mean differences (MDs), with 95% confidence intervals (CIs). We used the GRADE approach to assess the certainty of evidence.


MAIN RESULTS
We included 21 trials, recruiting 14,241 infants. Five of the trials were conducted in high-income countries and the remaining 16 were conducted in middle- and low-income countries. Meta-analysis showed that nutrition education targeted at improving weaning-related feeding practices probably increases both weight-for-age z scores (WAZ) (MD 0.15 standard deviations, 95% CI 0.07 to 0.22; 6 studies; 2551 infants; I² = 32%; moderate-certainty evidence) and height-for-age z scores (0.12 standard deviations, 95% CI 0.05 to 0.19; 7 studies; 3620 infants; I² = 49%; moderate-certainty evidence) by 12 months of age. Meta-analysis of outcomes at 18 months of age was heterogeneous and inconsistent in the magnitude of effects of nutrition education on WAZ and weight-for-height z score across studies. One trial that assessed effects of nutrition education on growth at six years reported an uncertain effect on change in height and body mass index z score. Two studies investigated effects of nutrition education on neurodevelopment at 12 to 24 months of age with conflicting results. No trials assessed effects of nutrition education on long-term neurodevelopmental outcomes.


AUTHORS' CONCLUSIONS
Nutrition education for families of infants may reduce the risk of undernutrition in term-born infants (evidence of low to moderate certainty due to limitations in study design and substantial heterogeneity of included studies). Modest effects on growth during infancy may not be of clinical significance. However, it is unclear whether these small improvements in growth parameters in the first two years of life affect long-term childhood growth and development. Further studies are needed to resolve this question.


T A B L E O F C O N T E N T S
feeding practices probably increases both weight-for-age z scores (WAZ) (MD 0.15 standard deviations, 95% CI 0.07 to 0.22; 6 studies; 2551 infants; I = 32%; moderate-certainty evidence) and height-for-age z scores (0.12 standard deviations, 95% CI 0.05 to 0.19; 7 studies; 3620 infants; I = 49%; moderate-certainty evidence) by 12 months of age. Meta-analysis of outcomes at 18 months of age was heterogeneous and inconsistent in the magnitude of e ects of nutrition education on WAZ and weight-for-height z score across studies. One trial that assessed e ects of nutrition education on growth at six years reported an uncertain e ect on change in height and body mass index z score. Two studies investigated e ects of nutrition education on neurodevelopment at 12 to 24 months of age with conflicting results. No trials assessed e ects of nutrition education on long-term neurodevelopmental outcomes.

Authors' conclusions
Nutrition education for families of infants may reduce the risk of undernutrition in term-born infants (evidence of low to moderate certainty due to limitations in study design and substantial heterogeneity of included studies). Modest e ects on growth during infancy may not be of clinical significance. However, it is unclear whether these small improvements in growth parameters in the first two years of life a ect long-term childhood growth and development. Further studies are needed to resolve this question.

Review question
We reviewed the evidence for e ects of nutrition education about appropriate feeding practices during weaning on growth and development in children born at term gestation.

Background
Around the world, over 150 million children are undernourished and over 42 million are overweight and obese. Providing families with appropriate education about feeding practices during weaning may help to optimise nutrition while helping to protect children who are at risk of undernutrition, as well as those susceptible to being overweight and obese.

Study characteristics
We examined research published up to December 2019 and found 21 clinical trials recruiting 14,241 babies. The nutrition education provided in all included studies, whereby analysis could be pooled together, was aimed at reducing the risk of undernutrition in childhood. Five studies were undertaken in high-income countries, but the findings reported could not be included and pooled together in this review.

Risk with nutrition education
Relative effect (95% CI) № of participants (studies)

Certainty of the evidence (GRADE) Comments
Weight-for-age z score at 12 months of age (WAZ 12 months) Scale from 5 to -5 Mean weight-for-age z score at 12 months of age ranged from -1.6 to 0.9 z score MD 0.15 z score higher (0.07 higher to 0.22 higher) -2551 (6 RCTs)

Moderate a
Change from baseline value was used for 1 study. Endpoint values were used for the other 4 studies Height-for-age z score at 12 months of age (HAZ 12 months) Scale from 5 to -5 Mean height-for-age z score at 12 months of age ranged from -2 to -0.5 z score MD 0.10 z score higher (0.02 higher to 0.17 higher) -3208 (7 RCTs) ⊕⊕⊕⊝

Moderate a
Change from baseline value was used for 1 study. Endpoint values were used for the other 6 studies Height-for-age z score at 18 months of age (HAZ 18 months) Assessed with z score Scale from 5 to -5 Mean height-for-age z score at 18 months of age ranged from -2.2 to -0.5 z score MD 0.16 z score higher (0. 10

Description of the condition
The World Health Organization (WHO) defines weaning, or the introduction of complementary feeding, as the period when the child's diet changes from complete breastfeeding to eating normal family food. This transition usually starts at four to six months of age and is finished at around one year (WHO 1988). More broadly, the term is used to describe the period when solid foods are introduced to complement human or formula milk. The decision made by WHO to include everything except breast milk as complementary food is intended to emphasise the importance of exclusive breastfeeding; however this may be misleading. Infants are frequently fed human milk substitutes such as infant formula even from the first week of life. Complementary feeding is generally used to describe giving any nutrient-containing foods or liquids other than breast milk, infant formula, or follow-on formula (Agostoni 2008), and weaning is the process by which such complementary foods are introduced into the infant's diet.
Although WHO, the United Nations Children's Fund (UNICEF), and the American Academy of Pediatrics recommend exclusive breastfeeding for the first six months of life (AAP 2012;Kramer 2002;UNICEF 2005), most guidelines, particularly from high-income countries (World Bank 2015), recommend that weaning should not occur before 17 weeks, should not be delayed beyond 26 weeks, and should be guided by the individual infant's nutritional needs and developmental abilities (Agostoni 2008). Weaning should be timely, safe, and adequate in nutritional content and in the variety of food items o ered, and it should be o ered to the infant at the correct frequency and in an appropriate manner (Weaver 2001). Adequate renal, gastrointestinal, immunological, and neurodevelopmental maturation should have been achieved for the transition from milk to solid foods.
Undernutrition and faltering growth may occur unintentionally due to delayed weaning or weaning with low-energy density foods and may increase the risk of iron deficiency and iron deficiency anaemia in late infancy (Hopkins 2007). Furthermore, inappropriate weaning has been linked to several other health problems, such as increased risk of allergic disorders, dental caries, and poor neurocognitive outcomes.
At the other end of the spectrum, early weaning, particularly with inappropriately high-energy food, can increase the risk of childhood obesity and cardiovascular illness in later life. In high-income countries, where feeding practices are determined mainly by parental beliefs and understanding of infant feeding, observational evidence shows that early weaning to solid foods is significantly associated with overweight or obesity at three years of age (Baughcum 2001;Hawkins 2009).
The nutritional challenges faced by populations in low-and middleincome countries usually di er from those seen in high-income countries. In low-and middle-income countries, gains attained by promoting exclusive breastfeeding for the first six months of life need to be sustained by encouraging appropriate weaning, as it is well recognised that between six and 24 months of age, children are particularly vulnerable to malnutrition due to limitations in the quality and quantity of foods (Lassi 2013). Families are faced with limited availability and access to food along with lack of information about correct choices for weaning. In high-income countries, parents face anxieties and challenges despite adequate availability of food for weaning (Redsell 2010).
Parents make infant feeding choices based on a variety of influences including advice from family members and health professionals, leaflets, magazines, and, increasingly, information from the Internet (Gage 2012). Evidence suggests that compliance with weaning guidelines is low and mothers o en experience conflict in deciding when and how to wean their infants (Arden 2010; Moore 2012). Surveys of parents demonstrate that they feel unsupported and experience anxiety due to a variety of factors such as inadequate knowledge and understanding of the physiological needs of the infant and confusing information from multiple commercially oriented sources, as well as social pressures, controversial cultural patterns and expectations, lack of information about healthy diet, and apprehension about cooking even the simplest weaning foods (Redsell 2010).
Weaning practices impact the long-term eating habits of children. Parental anxieties about infant-feeding also manifest as control of feeding practices and attempts to impose the amount or types of foods the infant eats. Studies show that parents who lack awareness of infant hunger cues are more likely to force their child to eat more (pressure/control feeding) or to refrain from certain foods or to take in limited amounts due to anxieties about weight gain (restriction for weight) (Musher-Eizenman 2007). Such practices have been shown to be associated with food neophobia (avoidance and rejection of novel foods), which is associated with reduced dietary quality and lower nutrient intake in later life (Cassells 2014). Empowering parents with the knowledge to recognise and respond to their infant's hunger cues may reduce the use of controlling feeding practices and may improve lifelong dietary habits.
Despite di erences in opinion and lack of consensus among experts, parents and families need information and support while weaning their infants. Parents are receptive to advice but need better support in accessing and understanding best practices around infant feeding (Redsell 2010). Inadequate nutrition may be caused by limited access to su icient food; however, caregivers may not be able to make the best use of available resources because of lack of knowledge and inappropriate beliefs and advice. Education of caregivers may have an impact and may improve nutritional status among children by empowering parents/ caregivers to provide the best possible diet and to use the most appropriate feeding styles to wean their infants.

Description of the intervention
Nutrition education has been defined as "any combination of educational strategies, accompanied by environmental supports, designed to facilitate voluntary adoption of food choices" (Contento 2010). Educational interventions may be provided to the individual parent or caregiver or may be delivered via community-based programmes, and could include nutritional counselling of caregivers; dissemination of information via verbal, written, or audiovisual aids; and/or any other strategy that provides information about weaning practices to families. Environmental supports may include changes in healthcare and food policies, as well as in social structure in the community, to create a conducive environment for nutrition education, such as arrangement of home visits, a suitable accessible location to carry out group educational activities, and provision of visual aids.

How the intervention might work
Nutrition education is an essential component of health promotion and disease prevention. Several theories of behaviour change, such as the theory of planned behaviour -Ajzen 1980 -and the social-cognitive theory -Bandura 2004 -explain the complex relationship between knowledge, beliefs, and perceived social norms, and how nutrition education can induce behavioural changes in a given set of circumstances. Interventions that provide relevant information and education to parents and caregivers could induce changes in behaviour that may impact nutritional practices, thereby improving nutrition, growth, and long-term metabolic health outcomes among children (Lassi 2013). The nutritional messages o en emphasise the importance of breastfeeding duration, initiation of weaning to food, frequency of feeding, or the composition of food (in terms of protein, energy, and micronutrient content), which will improve nutrition intake and growth. The dietary supply of specific nutrients may influence the maturation of cortical function. Feeding breast milk has o en been associated with better later cognitive outcomes; however, some studies have shown that certain foods provided during weaning are associated with improved outcomes, such as an increase in the Bayley Psychomotor Developmental Index (Morgan 2004), in visual acuity (Ho man 2003), and in higher behavioural indices (Krebs 2006). In older children, nutrition education modifies eating behaviour and optimises growth, and parental education can have a positive impact on child nutrition (Luepker 1996).

Why it is important to do this review
Previous systematic reviews have evaluated the impact of nutrition education and have demonstrated improvement in both weight and linear growth (Dewey 2008; Imdad 2011). However, both of these reviews concentrated on populations in low-and middleincome countries and included non-randomised studies as well as studies that included children older than 12 months of age. This review will collate the current evidence to determine whether use of nutrition educational interventions to support families during the weaning process optimises growth and nutrition among infants born at term gestation in all parts of the world.
The need for educational programmes to improve infant nutrition has been highlighted by several studies (Hoare 2002;Redsell 2010), particularly as infant nutrition is subjected to strong pressures by commercial as well as non-profit motivated self-help groups. The double threat of childhood undernutrition and obesity and their potential long-term impact on health has prompted attention to e ective interventions that improve the nutritional status of children in all parts of the world (Black 2013). Nutrition education has the potential to improve child health at both ends of the malnutrition spectrum. It is imperative that parents and families have access to nutrition education through scientifically correct, culturally sensitive, and economically appropriate advice about healthy diet for infants (Caroli 2012). It is also vital to ensure that such interventions are e ective, as significant resources could be saved by eliminating time-and resource-intensive educational programmes that prove to be of no benefit.

O B J E C T I V E S
To compare e ects of family nutrition educational interventions for infant weaning with conventional management on growth and neurodevelopment in childhood.

M E T H O D S Criteria for considering studies for this review Types of studies
We included published randomised and quasi-randomised trials, including cluster-randomised trials. We described any imbalances in baseline characteristics and outcome measurements between clusters in both groups. We did not include non-randomised trials such as controlled before-and-a er studies. The review was not limited to any particular region or socio-economic category, and we included studies published in any language.

Types of participants
Parents and families of infants born at term gestation (37 to 42 weeks' gestation) and younger than one year of age at recruitment are included.

Types of interventions
We included studies comparing any nutrition educational intervention for parents or families of infants born at term (37 to 42 weeks' gestation) and younger than one year of age at recruitment with conventional management for weaning. We included studies that use any form of nutrition educational intervention such as nutrition counselling, face-to-face sessions, audiovisual packages, support groups, additional input from health visitors or other allied professionals, and any other form of support involving nutrition education provided to families. We looked at nutrition educational messages emphasising the importance of breastfeeding duration, initiation of weaning to food, frequency of feeding, or composition of food (in terms of protein, energy, and micronutrient content). Conventional management was defined as standard clinical support and/or appointments without a nutrition educational focus.

Search methods for identification of studies
We used the criteria and standard methods of Cochrane and Cochrane Neonatal (see the Cochrane Neonatal search strategy for specialised register). We did not limit the search to any particular geographical region, language, or timing of publication.

Electronic searches
We

Searching other resources
We examined the reference lists of included studies and previous reviews, and we examined proceedings of annual meetings of the Paediatric American Societies (1993 to 12 December 2019), the European Society for Paediatric Research (1995 to 12 December 2019), the Royal College of Paediatrics and Child Health (2000 to 12 December 2019), and the Perinatal Society of Australia andNew Zealand (2000 to 12 December 2019). Trials reported only as abstracts were eligible for inclusion if su icient information was available from the report, or from contact with study authors, to fulfil the inclusion criteria.

Data collection and analysis
We used the standard methods of Cochrane Neonatal.

Selection of studies
Three review authors (SO, ZE and TK) screened the title and abstract of studies and potentially-relevant reports identified from the above search. The review authors independently assessed the full articles for all potentially relevant trials and any disagreements were resolved by discussion and input from the fourth author (JD).

Data extraction and management
Three review authors (SO, ZE, and TK) independently extracted data from the full-text articles of included studies using a data collection form for details of design, methods, participants, interventions, outcomes, and educational e ects. We crosschecked information and resolved any discrepancies by discussion until we reached agreement. respective 95% confidence intervals (CIs). We also reported the number needed to treat for an additional beneficial outcome (NNTB) or an additional harmful outcome (NNTH) for analyses with a statistically significant di erence in RD.
For categorical outcomes, we calculated typical estimates for relative risk, RD, NNTB, and NNTH and used 95% CIs.

Unit of analysis issues
The unit of analysis was the participating infant in individually randomised trials. An infant was considered only once in an analysis. We excluded infants with multiple enrolments from analysis unless we obtained data from the report or from investigators related to the first episode of randomisation. If data from the first randomisation could not be identified, we excluded the study, as we were not able to address unit of analysis issues that arose from multiple enrolments of the same infant. We included infants from multiple births.
We intended to conduct intention-to-treat analyses. The participating health organisation was the unit of analysis in clusterrandomised trials. We analysed these trials using 'approximate analyses' to obtain 'e ective sample sizes' as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). The intracluster correlation coe icient (ICC) does vary, depending on geographical area as well as one size of the cluster used. External estimates of the ICC from similar studies done in developing countries range from 0.01 in Shi 2009 to 0.05 in Handa 2018. Hence, we used an ICC of 0.05 to reduce the unit of analysis error as much as possible by reducing the 'e ective sample size'. We did not use a summary measurement from each cluster along with the cluster as the unit of analysis as this would have considerably and unnecessarily reduced the power of studies (Higgins 2019).

Dealing with missing data
If data were missing or were reported unclearly, we requested additional data on important outcomes from trial authors. When data were still missing, we examined the impact on e ect size estimates in sensitivity analyses using the 'best-worst case scenario' technique.

Assessment of heterogeneity
We examined intervention e ects of individual trials and heterogeneity between trial results by inspecting forest plots. We calculated the I statistic for each RR analysis to quantify inconsistency across studies and to describe the percentage of variability in e ect estimates that may be due to heterogeneity rather than to sampling error. Degree of heterogeneity has been classified according to the I statistic as follows: < 25%: none, 25% to 49%: low, 50% to 74%: moderate, 75% or higher: high.
If we detected moderate or high heterogeneity (I > 50%), we explored the possible causes (e.g. di erences in study design, participants, interventions, or completeness of outcome assessments). In addition, we employed a Chi test of homogeneity to determine the strength of evidence that heterogeneity is genuine.

Assessment of reporting biases
If we included more than ten trials in a meta-analysis, we checked a funnel plot for asymmetry to assess potential reporting bias.

Data synthesis
We used the fixed-e ect model in Review Manager 5.3 for metaanalyses (as per Cochrane Neonatal Group recommendations) (RevMan 2014). We used standard methods of the Cochrane Neonatal Review Group to synthesise data using RR, RD, NNTB, NNTH, MD, and 95% CIs. When substantial heterogeneity existed, we tested for potential causes in subgroup and sensitivity analyses.

Subgroup analysis and investigation of heterogeneity
We had planned to perform the following subgroup analyses, if data were available.
1. Infants and families living in middle-and low-income countries. 2. Infants and families living in high-income countries.
However, no study from a high-income country was eligible for inclusion in the meta-analyses. We also performed the following a posteriori subgroup analyses to investigate for heterogeneity.
1. Age of infants when intervention was started (antenatally, during first six months of age, or a er six months of age). 2. Duration of intervention (12 months or longer than 12 months). 3. Delivery of intervention in terms of setting (one-to-one, group, or combination of one-to-one and group) and person delivering the intervention (professional health workers and community workers).

Sensitivity analysis
We performed sensitivity analyses to determine whether findings are a ected by including only studies using adequate methods (low risk of bias), defined as adequate randomisation and allocation concealment, blinding of intervention and measurement, and less than 10% loss to follow-up.

Summary of findings and assessment of the certainty of the evidence
We used the GRADE approach, as outlined in the GRADE Handbook (Schünemann 2013), to assess the certainty of evidence for the following (clinically relevant) outcomes.
1. Growth rates (weight gain, linear growth, and head growth) in the first two years of life; change in weight, height, or head circumference z scores.

Cognitive development based on Bayley Mental Development
Index greater than 70 during follow-up at 12 months. 3. Iron deficiency seen as serum ferritin less than 12 micrograms/ L during follow-up at six months.
Two review authors independently assessed the certainty of evidence for each of the outcomes above. We considered evidence from RCTs as high certainty but downgraded evidence by one level for serious (or two levels for very serious) limitations based upon the following: design (risk of bias), consistency across studies, directness of evidence, precision of estimates, and presence of publication bias. We used the GRADEpro GDT Guideline Development Tool to create a 'Summary of findings' table to report the certainty of evidence.
The GRADE approach results in an assessment of the certainty of a body of evidence according to one of four grades.

Results of the search
We identified 75 studies for full-text screening ( Figure 1). Of these, we excluded 53 studies (see Characteristics of excluded studies  table). We conducted the last search on 12 December 2019.  A total of 14,241 infants were included (7730 in the nutrition intervention group and 6511 in control groups).

Interventions and comparisons
The intervention consists of nutrition education delivered via the following.

Excluded studies
Most of the studies set in high-income countries and those that aimed to investigate the e ectiveness of nutrition education in reducing risk of childhood overweight and obesity were ineligible for inclusion in this review and were excluded. The intervention in Cupples 2010 was peer-to-peer mentoring, and nutrition education was not defined. Daniels 2013 and Daniels 2015 describe results of the Australian NOURISH RCT, wherein mothers were randomised to usual care versus two six-session interactive group education modules that provided guidance on early feeding practices.
However, this study included healthy infants at greater than 35 weeks' gestation (i.e. some preterm infants would have been included). Participants were four months old at baseline. Daniels 2013 reported weight and weight z score, length and length z score, and body mass index (BMI) and BMI z score at 18 months from baseline (i.e. 22 months of age). Daniels 2015 reported outcomes at 3.5 years and at five years, showing no statistically significant di erences between groups for any anthropometric outcomes.
Gross 2016, an RCT conducted among Hispanic/Latina women in New York City, in the USA, included a small number of preterm infants (5 of 266 in the control group and 10 of 263 in the intervention group) and reported outcomes at three months of age only. The intervention was aimed at breastfeeding counselling -not at complementary feeding.
Outcomes in Jonsdottir 2012 in Iceland (serum ferritin and other laboratory indices at six months of age) were not suitable for inclusion in the review.    Six studies reported weight-for-age z score (WAZ) at 12 months (Muhoozi 2018;Olaya 2013;Penny 2005;Roy 2007;Shi 2009;Vazir 2013), and meta-analysis shows that infants in the nutrition education group had higher WAZ at 12 months of age when compared to those in the conventional management group (MD 0.15, 95% CI 0.07 to 0.22; P < 0.0001; 6 studies; 2551 infants; I = 32%) (Analysis 1.4 Figure 4). We downgraded the certainty of evidence to moderate due to limitations in the methods of some of the included studies (Summary of findings 1

Footnotes
(1) Change from baseline measure (change in WAZ between 6 and 12 months of age) due to significant differences at baseline (2) WAZ at 12 months was presented in a graph and values were calculated using the scales provided in Adobe Acrobat software. The SE estimated from these measurements were converted to SD u

Height (length) and height for age
Data from five studies were available to analyse e ects of nutrition education on change in height ( infants in the nutrition education group had higher HAZ at 12 months of age when compared to those in the conventional management group (MD 0.12, 95% CI 0.05 to 0.19; P = 0.0008; 7 studies; 3620 infants; I = 49%) (Analysis 1.6; Figure 5). We downgraded the certainty of evidence to moderate due to limitations in the methods of some of the included studies (Summary of findings 1). The e ect of the intervention on HAZ was reported at 18 months by five studies (Nair 2017;Nikiema 2017;Penny 2005;Roy 2007;Shi 2009). Meta-analysis shows that at 18 months, infants in the nutrition education group also had higher HAZ when compared to those in the conventional management group (MD 0.16, 95% CI 0.10 to 0.22; P < 0.00001; 5 studies; 4813 infants) (Analysis 1.7; Figure 6 I = 67%). We rated this evidence as low certainty due to substantial inconsistency in the magnitude of e ects (I = 67%) and methodological limitations of these studies (Summary of findings 1).

Footnotes
(1) Change from baseline measure (change in HAZ between 6 and 12 months of age) due to significant differences at baseline (2) HAZ at 12 months was presented in a graph and values were calculated using the scales provided in Adobe Acrobat software (Adobe Systems Software Ireland Ltd). The standard errors estimated Figure 6. Forest plot of comparison: 1 Nutrition education versus conventional management, outcome: 1.7 Heightfor-age z score at 18 months of age.

Weight for height
Five studies reported data on weight-for-height z score (WHZ) at 12 months (Muhoozi 2018; Nikiema 2017; Olaya 2013; Roy 2007; Shi 2009). Meta-analysis demonstrates that nutrition education did not show an e ect on WHZ at 12 months (MD 0.05, 95% CI -0.03 to 0.13; P = 0.24; 5 studies; 2831 infants; I = 11%) (Analysis 1.9). We downgraded the certainty of evidence to low due to limitations in the methods of some of the included studies and imprecision with a wide confidence interval. However, WHZ at 18 months was reported by five studies (Nair 2017; Nikiema 2017; Penny 2005; Roy 2007; Shi 2009). Meta-analysis reveals considerable heterogeneity and inconsistency in the direction and magnitude of e ects across these studies (I = 90%) (Analysis 1.10). Hence, we have not reported results of the meta-analysis as we are very uncertain of these results (Summary of findings 1).

Head circumference
Muhoozi 2018 reported that nutrition education did not show an e ect on the head circumference z score at 12 months due to imprecision (MD 0.17, 95% CI -0.04 to 0.38; P = 0.11; 1 study; 467 infants).

Other findings not included in the meta-analysis
Saleem 2014 reported growth parameters and growth parameter z scores, but data were insu icient for inclusion in the metaanalysis. At the end of the study (which occurs 7.5 months a er the intervention), infants in the control group were found to be 0.35 kg lighter and 0.66 cm shorter in terms of mean weight and height. There was no di erence on the proportions of infants, with WAZ more than 2 standard deviations below the World Health Organization (WHO) standard (adjusted odds ratio (ORadj) 0.75, 95% CI 0.40 to 1.79). Saleem 2014 found that infants in the control group were more likely to have HAZ more than 2 standard deviations below the WHO standard (ORadj 8.36, 95% CI 5.6 to 12.42) at the end of the study.
Murthy 2019 reported that nutrition education did not have an e ect on a malnourished infant at one year of age (odds ratio (OR) 0.823, 95% CI 0.590 to 1.147; P = 0.2). However, information was insu icient for inclusion of this study in the meta-analysis.
Rafieya-Kopaei 2019 reported that nutrition education did not have an e ect on the absolute value and z score of growth status Cochrane Database of Systematic Reviews between intervention and control groups (weight, height, and head circumference) at four, six, and eight months of age.

Education of family members to support weaning to solids and nutrition in later infancy in term-born infants (Review)
Palacios 2019 and Watt 2009 were nutrition education studies that were carried out in high-income countries to address obesity and were not included in the meta-analysis. Watt 2009 found that nutrition education did not have an e ect on the di erence in weight and height at 12 and 18 months of age. However, insu icient information was reported to incorporate growth parameter findings into the meta-analysis as growth at six months and z scores were not reported.
Palacios 2019 reported that nutrition education did not have an e ect on the di erences in weight at the end of the trial (four to six months old), as well as changes in weight between the first visit at zero to two months old and the second visit at four to six months of age.

Neurodevelopmental scores in children 12 months of age or older based on validated assessment tools
Muhoozi 2018  14; P = 0.32; 1 study; 412 infants) due to the small sample size. Meta-analysis was not possible as di erent assessment tools were used at di erent ages between both studies.

Duration of exclusive breastfeeding
Rates of exclusive breastfeeding at six months of age were reported by six studies (Bhandari 2004; Bortolini 2012; Nair 2017; Nikiema 2017; Palacios 2019; Rafieya-Kopaei 2019). Meta-analysis from three studies reveals considerable heterogeneity and inconsistency in the direction of e ects across studies (I = 83%) (Analysis 1.14) (Bhandari 2004; Bortolini 2012; Nair 2017). Hence, meta-analysis results are not reported, as we are very uncertain of these results (Summary of findings 1).

Exclusive breastfeeding at six months
Rafieya-Kopaei 2019 reported that the frequency of exclusive breastfeeding at six months was significantly higher in one of the intervention groups (loss-framed) (P = 0.03), but the number of exclusively breastfeeding infants and the measure of e ects were not reported.
In Palacios 2019, there were only six babies who were six months of age at final follow-up -all in the intervention group. Only one of the six babies was exclusively breastfeeding at six months of age.
Although Nikiema 2017 found that nutrition education improved exclusive breastfeeding rates at six months of age (risk ratio (RR) 1.28, 95% CI 1.20 to 1.37; P = 0.020; 3514 visits), this has to be interpreted with caution as the result was reported for the proportion of completed questionnaires for all follow-up visits (3514 visits) rather than for the number of infants (2253 infants).

Exclusive breastfeeding at three months
Morandi 2019 reported that nutrition education did not have an e ect on exclusive breastfeeding at three months of age due to imprecision (odds ratio (OR) 1.34, 95% CI 0.96 to 1.88; P = 0.08; 562 infants).

Compliance with advice regarding the timing of weaning
This outcome was not reported in any of the included studies.

Cognitive ability in children at five, six, or seven years of age based on validated assessment tools such as the Weschler Intelligence Scale for Children (Wechsler 1974) and school examinations
This outcome was not reported in any of the included studies.

Long-term growth: weight, height, skinfold thickness, or body mass index at five, six, or seven years of age
Only Olaya 2013 reported long-term growth at six years. These researchers reported that nutrition education did not have an e ect on the change in height z score (MD -0.22, 95% CI -0.60 to 0.16; 50 infants) and on body mass index (BMI) z score (MD -0.24, 95% CI -0.75 to 0.27; 50 infants) at six years of age, respectively. However, this study did show that nutrition education led to reduction in HAZ (MD -0.75, 95% CI -1.25 to -0.25; 50 infants) at six years of life. The reason for this is unclear, and these results should be interpreted with caution due to the very small sample size.

Serum ferritin (< 12 microg/L) and haemoglobin (< 110 g/L) levels in children six months of age and older (WHO 2011)
Two studies reported anaemia (serum haemoglobin levels < 110 g/L) at 12 months of age. Meta-analysis reveals that nutrition education did not have an e ect on this outcome (Analysis 1.12) (RR 1.16, 95% CI 0.96 to 1.39; P = 0.12; 2 studies; 585 infants; I = 67%). This evidence is of low certainty due to inconsistency in the direction and magnitude of e ects, substantial heterogeneity across studies (I = 67%), small sample size, and methodological limitations of studies. Only one study reported iron deficiency as measured by serum ferritin levels < 12 micrograms/L (Olaya 2013). This study demonstrated that nutrition education did not have an e ect on this outcome due to imprecision (RR 0.92, 95% CI 0.42 to 2.04; P = 0.84; 1 study; 85 infants).

Parental stress when the child is six months of age or older, measured by validated assessment tools such as the Parenting Stress Index (Grotevant 1989)
These outcomes were not reported in any of the included studies.

Infant quality of life when the child is six months of age or older measured by the Infant and Toddler Quality of Life Questionnaire (ITQOL) (Bowling 2004)
This outcome was not reported in any of the included studies.

Prevalence of atopic conditions in childhood
This outcome was not reported in any of the included studies.

Prevalence of food neophobia or 'picky/fussy eating'
This outcome was not reported in any of the included studies.

Death before one year of age
Four studies provided data for death by one year of age (Bhadari 2001;Nair 2017;Roy 2007;Vazir 2013). Meta-analysis shows that nutrition education did not have an e ect on the risk of dying by one year of age (RR 0.69, 95% CI 0.44 to 1.08; 4 studies; 4234 infants; I = 0%) (Analysis 1.13 Figure 7). The certainty of evidence for this finding is moderate due to methodological limitations of these studies (Summary of findings 1).

High-versus low-and moderate-income countries
Our a priori subgroup analysis was done to compare studies conducted in high-versus low-and moderate-income countries. However, for various reasons that are described in the Characteristics of excluded studies table, no studies from a highincome country were eligible for inclusion in meta-analyses.

Growth in the first two years of life
Three studies carried out in high-income countries reported this outcome with conflicting results (Morandi 2019; Palacios 2019; Watt 2009). Information from all three studies was insu icient for inclusion in the meta-analysis.
Palacios 2019 and Watt 2009 reported that nutrition education did not have an e ect on the di erences in growth parameters at six months of age (Palacios 2019), as well as at 12 and 18 months of age (Watt 2009),. However, Morandi 2019 found a reduction in changes in weight and height between six and 12 months of life in the intervention group as compared to the control group (MD -0.22, 95% CI -0.35 to -0.09; P = 0.0007; MD -0.30, 95% CI -0.58 to -0.02; P = 0.03; 1 study; 562 infants, respectively).

Breastfeeding
Similarly, three studies conducted in high-income countries reported breastfeeding outcomes with conflicting results (Morandi 2019; Palacios 2019; Watt 2009). Information was insu icient for inclusion in the meta-analysis.
Morandi 2019 and Watt 2009 reported that nutrition education did not have an e ect on the percentage of infants exclusively breastfeeding at three and four months of age, respectively. In Palacios 2019, there were only six babies who were six months of age at final follow-up. All were in the intervention group, and only one of the six babies was exclusively breastfeeding at six months of age.

A posteriori subgroup analyses
Following our review, we noted that there were di erences in how the studies were carried out. Hence, we performed further subgroup analyses to explore these di erences further.

Age of infant when intervention was started
We split the studies into three groups in terms of when nutrition education was given, ranging from antenatally ( . Subgroup analyses for this aspect were possible only for the following analyses.

Weight for age
For studies in which nutrition education was carried out during the first six months of life, infants in the nutrition education group had higher WAZ at 12 months (MD 0.11, 95% CI 0.02 to 0.21; P = 0.02; 4 studies; 1473 infants; I = 30%) and at 18 months (MD 0.28, 95% CI 0.17 to 0.39; P < 0.0001; 2 studies; 976 infants; I = 0%). Findings at 12 months of age were similar to initial analyses (Analysis 1.4, Figure  4). Analysing the WAZ at 18 months by age of the infant when the  Similar findings were found when analyses included studies in which the intervention was carried out a er the first six months of life with higher WAZ at 12 months (MD 0.18, 95% CI 0.07 to 0.29; P = 0.001; 2 studies; 1078 infants; I = 55%).

Height for age
For studies in which nutrition education was carried out during the first six months of life, infants in the nutrition education group had higher HAZ at 12 months (MD 0.17, 95% CI 0.06 to 0.27; P = 0.002; 4 studies; 1473 infants; I = 12%) and heterogeneity at 18 months was considerable (I = 79%). These findings were similar to those of the initial analyses (Analysis 1.6; Figure 5: Analysis 1.7 Figure 6).
For the two studies in which the intervention was carried out a er the first six months of life, meta-analysis reveals considerable heterogeneity and inconsistency in the magnitude of e ects across studies (I = 81%).
For the two studies in which the intervention was carried out antenatally, infants in the intervention group had higher HAZ at 18 months, which barely achieved statistical significance (MD 0.08, 95% CI 0 to 0.16; P = 0.04; 2 studies; 3226 infants; I = 0%).

Weight for height
For studies in which nutrition education was carried out during the first six months of life, meta-analyses revealed that nutrition education did not have an e ect on the WHZ at 12 months (MD -0.03, 95% CI -0.18 to 0.12; P = 0.34; 2 studies; 684 infants; I = 0%), but infants in the intervention group had higher WHZ at 18 months (MD 0.20, 95% CI 0.09 to 0.32; P = 0.0007; 2 studies; 976 infants; I = 64%).
For the two studies in which the intervention was carried out a er the first six months of life, the intervention group had higher WHZ at 12 months (MD 0.14, 95% CI 0.01 to 0.26; P = 0.03; 2 studies; 1078 infants; I = 0%).
For the two studies in which the intervention was carried out antenatally, meta-analyses revealed that nutrition education did not have an e ect on the WHZ at 18 months (MD 0.03, 95% CI -0.04 to 0.10; P = 0.42; 2 studies; 3175 infants; I = 64%).
Although conclusions drawn from these analyses were similar to those of the initial analyses (Analysis 1.9; Analysis 1.10), worth noting is the change in direction of e ects on the WHZ at 12 months for studies whereby the nutrition education was done in the first six months of life, as well as the lack of statistically significant e ects in the antenatal studies that reported WHZ at 18 months. This may indicate a potential impact of timing of delivery of the intervention on WHZ.

Exclusive breastfeeding at six months
Meta-analysis of the two studies in which nutrition education was carried out during the first six months of life reveals considerable heterogeneity in e ects of nutrition education on exclusive breastfeeding rates at six months (I = 92%), which was similar to that in the initial analysis (Analysis 1.14; Figure 9). 14 Exclusive breastfeeding at 6 months of age.

Study or Subgroup
Bhandari 2004

Death before one year of age
Meta-analysis of the two studies in which nutrition education was carried out during the first six months of life reveals that nutrition education did not have an e ect on the risk of dying before one year of age due to imprecision (RR 0.98, 95% CI 0.32 to 3.01; P = 0.97; 2 studies; 622 infants; I = 0%), which was similar to that in the initial analysis (Analysis 1.13 Figure 7).

Duration of intervention
We divided the studies into two groups, in which nutrition education was delivered for 12 months ( Subgroup analyses for this aspect were possible only for the following analyses.

Weight for age
For studies in which nutrition education was carried out for 12 months, meta-analysis did not show an e ect on the change in weight from six to 12 months of age due to imprecision and substantial heterogeneity (MD 0.03, 95% CI -0.03 to 0.08; P = 0.35; 3 studies; 1834 infants; I = 74%), but infants in the nutrition education group had higher WAZ at 12 months (MD 0.13, 95% CI 0.05 to 0.21; P = 0.001; 4 studies; 1762 infants; I = 55%) and heterogeneity at 18 months was considerable (I = 77%).
For studies in which nutrition education was carried out for longer than 12 months, infants in the nutrition education group had higher WAZ at 12 months (MD 0.21, 95% CI 0.03 to 0.38; P = 0.02; 2 studies; 789 infants; I = 0%) and heterogeneity at 18 months was considerable (I = 84%). These findings were similar to those of the initial analyses (Analysis 1.4, Figure 4: Analysis 1.5 Figure 8).
For studies in which nutrition education was carried out for longer than 12 months, infants in the nutrition education group had higher HAZ at 12 months (MD 0.13, 95% CI 0.02 to 0.24; P = 0.02; 3 studies; 1858 infants; I = 67%) and heterogeneity at 18 months was considerable (I = 80%). These findings were similar to those of the initial analyses (Analysis 1.6; Figure 5: Analysis 1.7 Figure 6).

Weight for height
For studies in which nutrition education was carried out for 12 months, meta-analyses reveal that nutrition education did not have an e ect on the WHZ at 12 months (MD 0.07, 95% CI -0.03 to 0.17; P = 0.27; 3 studies; 1295 infants; I = 23%) but infants in the intervention group had higher WHZ at 18 months (MD 0.41, 95% CI 0.31 to 0.52; P < 0.0001; 2 studies; 1210 infants; I = 69%).
For studies in which nutrition education was carried out for longer than 12 months, meta-analyses reveal that nutrition education did not have an e ect on the WHZ at 18 months (MD 0.04, 95% CI -0.03 to 0.11; P = 0.23; 3 studies; 3552 infants; I = 40%). The conclusions from these analyses are similar to those of the initial analyses (Analysis 1.9; Analysis 1.10).

Exclusive breastfeeding at six months
Meta-analysis of the two studies in which nutrition education was carried out for 12 months show considerable heterogeneity in e ects of nutrition education on exclusive breastfeeding rates at six months (I = 92%), which is similar to findings of the initial analysis (Analysis 1.14; Figure 9).

Death before one year of age
Meta-analysis of the two studies in which nutrition education was carried out for 12 months shows that nutrition education did not have an e ect on the risk of dying before one year of age due to imprecision (RR 0.72, 95% CI 0.14 to 3.62; P = 0.69; 2 studies; 821 infants; I = 0%). For studies carried out for longer than 12 months, similar findings were found (RR 0.69, 95% CI 0.43 to 1.11; P = 0.12; 2 studies; 3413 infants; I = 0%). These findings are similar to those of the initial analysis (Analysis 1.13 Figure 7).

Delivery of the intervention
We decided to investigate the impact of how nutrition education was delivered in terms of the setting (one-to-one session during a home or clinic visit; group session; or combination of group and one-to-one), as well as the person delivering the nutrition education (trained professional versus community worker).

Weight for age
For studies in which nutrition education was carried out in oneto-one settings, meta-analysis found that nutrition education did not have an e ect on the change in weight from six to 12 months (MD 0.04, 95% CI -0.02 to 0.11; P = 0.19; 2 studies; 1437 infants; I = 0%) but infants in the nutrition education group had higher WAZ at 12 months (MD 0.20, 95% CI 0.07 to 0.33; P = 0.002; 3 studies; 874 infants; I = 0%).
Similar findings were found upon analysis of studies in which the intervention was carried out in a group setting with WAZ at 12 months (MD 0.18, 95% CI 0.07 to 0.29; P = 0.001; 2 studies; 1078 infants; I = 55%).
For studies in which nutrition education was carried out in both one-to-one and group settings, meta-analysis shows considerable heterogeneity for WAZ at 18 months (I = 86%). These findings were similar to those of the initial analyses (Analysis 1.4; Figure 4: Analysis 1.5; Figure 8).

Height for age
For studies in which nutrition education was carried out in one-toone settings, considerable heterogeneity was evident in the e ects of nutrition education on change in height from six to 12 months (I = 78%) and in HAZ at 18 months (I = 84%), but infants in the nutrition education group had higher HAZ at 12 months (MD 0.13, 95% CI 0.03 to 0.24; P = 0.001; 4 studies; 1943 infants; I = 51%).
For the two studies in which the intervention was delivered in a group setting, meta-analysis reveals considerable heterogeneity Cochrane Database of Systematic Reviews and inconsistency in the magnitude of e ects across studies (I = 81%).

Education of family members to support weaning to solids and nutrition in later infancy in term-born infants (Review)
For studies in which nutrition education was carried out in both one-to-one and group settings, infants in the nutrition education group had higher HAZ at 18 months (MD 0.10, 95% CI 0.02 to 0.17; P = 0,01; 2 studies; 3160 infants; I = 0%). These findings were similar to those of the initial analyses (Analysis 1.6; Figure 5: Analysis 1.7 Figure 6).
For the two studies in which intervention was delivered in group settings, the intervention group had higher WHZ at 12 months (MD 0.14, 95% CI 0.01 to 0.26; P = 0.03; 2 studies; 1078 infants; I = 0%).
For studies in which nutrition education was carried out in both one-to-one and group settings, meta-analyses reveal considerable heterogeneity in e ects on WHZ at 18 months (I = 90%).

Exclusive breastfeeding at six months
Meta-analysis of the two studies in which nutrition education was delivered in a one-to-one setting shows considerable heterogeneity in e ects of nutrition education on exclusive breastfeeding rates at six months (I = 92%), which was similar to findings of the initial analysis (Analysis 1.14; Figure 9).
Death before one year of age Meta-analysis of the two studies in which nutrition education was carried out in a one-to-one setting did not show an e ect on the risk of dying before one year of age due to imprecision (RR 0.98, 95% CI 0.33 to 3.01; P = 0.97; 2 studies; 622 infants; I = 0%), which was similar to findings of the initial analysis (Analysis 1.13 Figure 7).

Weight for age
For studies in which nutrition education was carried out by a trained professional, substantial heterogeneity is evident in the e ects of nutrition education on change in weight from six to 12 months (I = 74%) but infants in the nutrition education group had higher WAZ at 12 months (MD 0.14, 95% CI 0.06 to 0.22; P =0.0006; 5 studies; 2139 infants; I = 44%) and at 18 months (MD 0.37, 95% CI 0.29 to 0.45; P < 0.0001; 3 studies; 1587 infants; I = 62%). These findings were similar to those of initial analyses (Analysis 1.4 Figure  4: Analysis 1.5 Figure 8).

Weight for height
For studies in which nutrition education was carried out by a professional, considerable heterogeneity is evident for WHZ at 18 months (I = 81%). Conclusions from these analyses were similar to those of the initial analyses (Analysis 1.9; Analysis 1.10).
Exclusive breastfeeding at six months Meta-analysis of the two studies in which nutrition education was carried out by a trained professional shows considerable heterogeneity in e ects of nutrition education on exclusive breastfeeding rates at six months (I = 92%), which was similar to findings of the initial analysis (Analysis 1.14; Figure 9).
Death before one year of age Meta-analysis of the two studies in which nutrition education was carried out by a trained professional did not show an e ect on the risk of dying before one year of age (RR 0.72, 95% CI 0.14 to 3.62; P = 0.69; 2 studies; 821 infants; I = 0%). For studies carried out by a community worker, similar findings were found (RR 0.69, 95% CI 0.43 to 1.11; P = 0.12; 2 studies; 3413 infants; I = 0%). These findings are similar to findings of the initial analysis (Analysis 1.13 Figure 7).

Unit of analysis issues
A total of 12 cluster-randomised controlled trials are included in this review ( We did not use a summary measurement from each cluster and we did not use the cluster as the unit of analysis, as this would have considerably and unnecessarily reduced the power of the studies (Higgins 2019). Instead, we carried out 'approximate analyses' to obtain 'e ective sample sizes' (Higgins 2019). The intracluster correlation coe icient (ICC) does vary depending on the geographical area, as well as on the size of clusters used. External estimates of the ICC from similar studies done in developing countries range from 0.01 in Shi 2009 to 0.05 in Handa 2018. Hence, we used the ICC of 0.05 to reduce the unit of analysis error as much as possible by reducing the 'e ective sample size' (Table 1).
We carried out repeat analyses for outcome measures involving cluster-randomised trials.

Weight for age
Meta-analysis still demonstrate that nutrition education did not have an e ect on the change in weight from six to 12 months (MD -0.01, 95% CI -0.09 to 0.07; 4 studies; I = 44%) (Analysis 2.1).

Cochrane Database of Systematic Reviews
However, the direction of e ect has changed from that seen in the initial analysis (Analysis 1.1).

Height for age
A er approximate analyses were done to account for the unit of analysis error for the clustering e ect, meta-analyses revealed that nutrition education did not have an e ect on the change in height from six to 12 months (MD -0.02, 95% CI -0.20 to 0.17; 4 studies; I = 45%) (Analysis 2.2). Results show a change in both the magnitude and the direction of e ect as compared to the initial analyses (Analysis 1.2).

Weight for height
Similar to the initial analyses (Analysis 1.9), the meta-analysis still demonstrates that nutrition education did not have an e ect on the WHZ at 12 months (MD 0.11, 95% CI -0.02 to 0.23; 5 studies; I = 0%) (Analysis 2.7). A er approximate analyses, there was considerable heterogeneity in the WHZ at 18 months (I = 84%) (Analysis 2.8). A change in the magnitude of e ect led to the lack of e ect seen as compared to the initial analyses (Analysis 1.10).

Death before one year of age
Similar to the initial analysis (Analysis 1.13 Figure 7), the metaanalysis still demonstrates that nutrition education did not have an e ect on the risk of dying before one year of age (RR 0.70, 95% CI 0.37 to 1.29; 3 studies; I = 0%) (Analysis 2.9).

Exclusive breastfeeding at six months
Similar to the initial analysis (Analysis 1.14; Figure 9), there was still considerable heterogeneity about e ects of nutrition education on exclusive breastfeeding rates at six months (I = 79%) (Analysis 2.10).

Summary of main results
Low-to moderate-certainty evidence suggests that nutrition education targeted at improving feeding practices related to weaning improves weight and height in the first two years of life. However, these results must be interpreted with caution due to substantial heterogeneity in the results, which may be due in part to di erences in study design (cluster versus non-cluster randomised controlled trial (RCT)) and in delivery of nutrition education in terms of timing of initiation, duration, setting (oneto-one, group, or combination of both), and the professional experience required to deliver the nutrition education. Limited data are available to conclude which of these factors plays a bigger role in the e ectiveness of nutrition education in improving growth parameters during the first two years of life. Improvement in growth parameters is seen as between 0.12 and 0.16 standard deviation in z scores. This is of unclear clinical significance, as it accounts for a small improvement in weight and height in grams and centimetres, respectively. It is also unclear how this small improvement in growth parameters during the first two years of life impacts longterm growth and health. Further long-term studies are needed to see if this small short-term improvement in growth parameters persists, translating into a bigger di erence in growth over the long term.
Besides this, all studies included in the meta-analysis were carried out in low-income countries and reported mean z scores of growth parameters below 0 at baseline, indicating some degree of malnutrition. Hence, these findings could not be extrapolated to obesity/overnutrition or to countries of high income. The impact of nutrition education on long-term growth is unclear, with only one study reporting the outcome measure with mixed results.
We did not find evidence that nutrition education reduces the prevalence of anaemia or iron deficiency anaemia in infants. Two studies included interventions that focused on increasing the intake of iron-rich and/or meat-based weaning foods, but their combined results reveal that nutrition education did not have an e ect on anaemia. Most of the studies included interventions that promoted exclusive breastfeeding until six months of age. The e ect of nutrition education on exclusive breastfeeding rates at six months remains unclear based on the meta-analysis of current studies. Only two studies investigated e ects of nutrition education on neurodevelopmental outcomes and reported no e ect on cognitive or motor scores. We did not find any studies that investigated e ects of this intervention on long-term, school-age developmental outcomes. Data on death before the first birthday were available from four studies and showed that nutrition education did not have an e ect on death in the first year of life. Other secondary outcomes considered in this review (parent and infant quality of life, risk of atopy, food neophobia) were not reported by any of the included studies.
undernutrition. In high-income countries and increasingly in certain populations in low-and middle-income countries, the target of nutrition education is likely to be the oppositei.e. prevention of overnutrition and amelioration of childhood overweight and obesity. We found several such studies, which were set in countries such as Australia, the USA, and the UK, but their results could not be combined in any meta-analyses due to variations in inclusion criteria and in reported outcomes (see We included studies that provided nutrition education as the only intervention, and we excluded studies that combined education with food supplementation. In addition, we looked at studies in all settings, irrespective of food security in the population. Food security was not described as a potential problem of malnutrition in the included studies, although it is likely that nutrition education alone may not be as e ective in settings where food scarcity is an underlying cause of malnutrition. The World Health Organization defines the problem of the double burden of malnutrition as the coexistence of undernutrition along with overweight and obesity, or diet-related non-communicable diseases (NCDs), within individuals, households, and populations, and across the life course (WHO 2017a), and has identified "doubleduty actions" (i.e. interventions, programmes, and policies that have the ability to simultaneously reduce the risk or burden of both undernutrition and overweight, obesity, or diet-related NCDs) (WHO 2017b). Promotion of appropriate early and complementary feeding in infants is included in the list of potential candidate interventions capable of achieving double duty. This review demonstrates that nutrition education may be an e ective tool for improving growth and nutrition (as measured by weight and height in the first two years of life) with evidence of low to moderate certainty due to di erences in study design and in the delivery of nutrition education. Evidence is insu icient to conclude which aspect of the delivery of nutrition education plays a bigger role in improving growth parameters. Although the clinical significance of the observed improvement in growth parameters is likely to be small, its impact on long-term growth is unclear. Lack of evidence makes it di icult to decide whether such interventions would be e ective in reducing the burden of childhood overweight and obesity or undernutrition in settings where food scarcity is a major problem.

Quality of the evidence
We used the GRADE approach to assess the certainty of evidence for the following outcomes: weight-for-age z score (WAZ) at 12 months, height-for-age z score (HAZ) at 12 and 18 months, prevalence of anaemia at 12 months of age, and death before one year of age (Summary of findings 1). Evidence from RCTs was mainly of low to moderate certainty at best -downgraded due to risk of bias (attrition bias imbalance in baseline demographics) and unexplained heterogeneity.

Potential biases in the review process
Our main concern with the review process is the possibility that findings are subject to publication and other reporting biases. We attempted to minimise this threat by screening the reference lists of included trials and related reviews and searching for proceedings of major international conferences to identify trial reports that are not published or in preparation. Manually imputing data from the published graph in Penny 2005 may have introduced a moderate risk of bias. Sensitivity analyses completed by removing the study data of Penny 2005 did not alter the overall findings of this review.
The precision of some of the analyses may be overestimated, as 'approximate analyses' were used to account for clustering, rather than using a summary measurement from each cluster and using the cluster as the unit of analysis. This is because the latter method would unnecessarily reduce the power of the included study. We noted moderate to considerable heterogeneity and inconsistencies in the included studies in some of the analyses, even a er clustering e ects were accounted for. Subgroup analyses exploring di erent aspects of delivery of nutrition education were carried out to investigate the heterogeneity.
The wide confidence intervals with large standard deviations for z scores of growth parameters reported in some of the included studies may indicate heterogeneity of weight and height of infants recruited in the study ranging from undernutrition to overnutrition. Hence, this may explain some of the heterogeneity seen in the analyses and the small improvement noted in growth parameters.

Agreements and disagreements with other studies or reviews
Another recent Cochrane systematic review examined educational interventions for improving primary caregiver complementary feeding practices for children age two years and younger (Arikpo 2018). This extensive review focused on age at introduction of complementary foods, exclusive breastfeeding for four months, and hygiene practices. Review authors also reported on e ects of the intervention on growth parameters. They reported on e ects of the intervention on duration of breastfeeding and found that combining data from three studies showed a trend towards a higher rate of exclusive breastfeeding for four months or longer in the intervention group (risk ratio (RR) 1.58, 95% confidence interval (CI) 0.77 to 3.22) compared to the control group. We chose to investigate the duration of breastfeeding as per the World Health Organization (WHO) recommendation (exclusive breastfeeding up to six months of age) and found no e ects of the intervention in a meta-analysis of data from three studies. When assessing growth parameters, we did not include absolute values of anthropometric parameters such as weight or height at any time point, as in Arikpo 2018, because the studies were conducted in heterogeneous populations. Arikpo 2018 did not find any e ect of the intervention on weight or height at 6, 12, or 18 months of age. Review authors also reported that nutrition education did not have an e ect on the outcomes of stunting (height-for-age z score (HAZ) ≤ -2 standard deviations (SD)), wasting (weight-for-height z score (WHZ) ≤ -2 SD), and underweight (weight-for-age z score (WAZ) ≤ -2 SD). Growth parameters were the primary outcomes for our review. We did not dichotomise the growth measures but instead analysed them (WAZ, HAZ, and WHZ) as continuous variables. Use of growth parameters as continuous variables rather than as dichotomous variables provides more information and a better overview of the impact of nutrition education on growth. We found that the children of those who received nutrition education had higher HAZ at 18 months of age and higher WAZ and HAZ at 12 months of age. Similar to other reviews, such as Imdad 2011, our analyses demonstrate that significant improvements in growth parameters may be achieved with appropriate nutrition education. However, the clinical significance of the improvement noted in our review is small and may explain the findings of Arikpo 2018 that such improvement in growth parameters may not be su iciently large to bring a statistically significant number of children out of the dichotomous brackets of wasting, stunting, or underweight without a large sample size. The long-term significance of the small improvement in growth parameters seen in our review in the first two years of life is also unclear.
Imdad 2011 is a systematic review of published randomised and quasi-randomised controlled trials conducted to assess the e ectiveness of nutritional counselling alone and of providing complementary foods with or without counselling on mean change in weight and height outcomes. In their analyses, which included studies with nutritional counselling as the only intervention, review authors included eight studies and demonstrated that educating mothers about complementary feeding improved weight gain (weighted mean di erence 0.30 SD, 95% CI 0.05 to 0.54) and height gain (weighted mean di erence 0.21 SD, 95% CI 0.01 to 0.41). These results are in keeping with our findings, although we excluded some of the studies included in these meta-analyses: Guldan 2000 and Kilaru 2005, due to incomplete randomisation, and Santos 2001, as the intervention was delivered to healthcare providers and not directly to families and was not focused on weaning. In addition, both Kilaru 2005 and Santos 2001 reported changes in absolute weight or height only, and hence the outcomes were not suitable for inclusion in this review. In addition to these exclusions, this review is up-to-date with inclusion of studies published since 2011. Similar findings were reported by Dewey 2008, which conducted an extensive review of studies related to a variety of complementary feeding support strategies but did not perform formal metaanalyses of available data.
Lassi 2013 is a systematic review of randomised and nonrandomised trials conducted to assess the e ectiveness of education for complementary feeding and of providing complementary foods for the growth and morbidity of children below two years of age in developing countries. Review authors found that education on complementary feeding alone improved HAZ (standard mean di erence 0.23, 95% CI 0.09 to 0.36) and WAZ (standard mean di erence 0.16, 95% CI 0.05 to 0.27). The impact on height and weight gain was reported to be uncertain. Again, these findings are similar to those reported by our review. However, Lassi 2013 did not specify the age at which growth parameters were measured. Review authors included Guldan 2000, Kilaru 2005, and Santos 2001, which were excluded from our review for the reasons stated previously. Zaman 2008 was not included in our review as the intervention comprised training of healthcare professionals rather than nutrition counselling targeted at families.
We found only four studies that reported exclusive breastfeeding up to six months of age as an outcome. The combined results of three of these studies did not show any e ect of nutrition education.
McFadden 2017 reviewed 100 trials to examine the e ectiveness of di erent modes of o ering supportive interventions to improve breastfeeding rates and included stopping exclusive breastfeeding before six months postpartum as an outcome. Upon combining data from 46 studies including 18,591 women, review authors found that women in the intervention groups were less likely to have stopped exclusive breastfeeding before six months (average RR 0.88, 95% CI 0.85 to 0.92; moderate-certainty evidence).
The interventions in these studies were focused on supporting breastfeeding, and many were started during the antenatal period.
Education that is focused more on weaning and is provided later in infancy may not have the same e ect in increasing rates of exclusive breastfeeding, and we suspect that this and smaller sample sizes are the reasons why the studies included in our review do not show any e ect.
In this review, we have focused on studies with nutrition education as the only intervention and have included only randomised controlled trials and cluster-randomised controlled trials that reported our pre-specified outcomes. We have systematically reviewed studies in both high-and low-and middle-income country settings, as our purpose was to assess the e icacy of nutrition education as a double-duty intervention (WHO 2017b). None of the studies set in high-income countries and no studies that aimed to use nutrition education to target childhood obesity were included in the meta-analyses, as reported outcomes were not in keeping with the outcomes of this review. This highlights the research gap and the need for good quality randomised trials conducted to assess the e ectiveness of nutrition education as an instrument to prevent or reduce childhood obesity. These trials should be undertaken in appropriate settings both in high-income settings and in appropriately targeted populations in low-and middleincome countries because there is a subgroup of populations in low-and middle-income countries that is at increasing risk of overnutrition and its concomitant health risks.

Implications for practice
In 2014, there were over 42 million overweight or obese children and 156 million stunted children worldwide. The WHO has identified double-duty actions that have the potential to reduce the risks or burden of undernutrition and of overweight and obesity, which include actions to optimise early nutrition and promotion of appropriate early and complementary feeding in infants. This review demonstrates that nutrition education delivered to families of infants (between 0 and 12 months of age) may reduce undernutrition as evidenced by low-to moderate-certainty evidence for improvement in weight and height during the first two years of life. Based on the studies reviewed, the clinical significance of the improvement in growth parameters is small and must be interpreted with caution due to inconsistencies among the studies, which may be due in part to di erences in study design and in delivery of nutrition education. Limited data at present do not allow us to conclude which of these factors plays a bigger role in the e ectiveness of nutrition education in improving growth parameters during the first two years of life. Evidence is insu icient to support that such educational interventions would ameliorate the risks of childhood overnutrition and would prevent overweight and obesity among young children.
Possibly due to the small numbers of available studies and participants, this review also does not provide evidence that nutrition education can improve rates of exclusive breastfeeding or reduce the risk of iron deficiency anaemia in infancy. These results should be interpreted with caution due to the small number of studies included in these analyses and to incomplete reporting of these outcomes.

Implications for research
Findings of this review suggest that nutrition education on appropriate weaning/complementary feeding practices in infancy may be an e ective tool for reducing the risks of childhood undernutrition, particularly in low-and middle-income countries.
Although only a small improvement in growth parameters is seen during the first two years of life, it is unclear whether these improvements may continue into later life, leading to a bigger di erence in growth parameters long term. Hence, studies are needed to investigate the impact of early nutrition education on long-term growth. Good quality evidence is insu icient to support the hypothesis that educational interventions could reduce risks of overweight and obesity in children. Randomised controlled trials of nutrition education interventions targeting populations at high risk of overnutrition both in high-income countries and in appropriately selected populations in low-and middle-income countries are required to test the true "double-duty" e icacy of this intervention. Comparison with a parallel group selected by adequate randomisation is crucial for investigating this, as such a trial design will ensure that the study measures e ects of the intervention without confounding due to other ongoing health programmes, changes in health or marketing regulations, or natural changes in population demographics. These studies must be conducted in appropriate settings and must include populations in high-and low-and middle-income countries.
It is also important to determine core outcomes for such trials, so that e ects of nutrition education on the same clinically relevant outcomes can be studied in di erent settings. This would allow for future meta-analyses and systematic reviews and would enable comparability between studies. Such outcomes should include some long-term e ects such as risk of non-communicable disease in later life and measures of infant and parental quality of life. Systematic reviews and meta-analyses investigating e ects of nutrition education on growth must ensure that growth data from studies targeting undernutrition and those addressing childhood obesity are combined into predefined separate analyses, as the intended e ects of one group of studies are likely to be opposite the intended e ects of the others despite common goals of combating the double burden of malnutrition. E ects on other outcomes such as reduction in anaemia, neurodevelopment, and parent and infant quality of life measures would be amenable to a single metaanalysis each, as interventions would be targeted to achieve the same e ect.
Further studies are also warranted to investigate whether appropriately targeted nutrition education can reduce the risk of micronutrient deficiencies such as iron deficiency anaemia.

A C K N O W L E D G E M E N T S
We would like to thank Cochrane Neonatal for help with the literature review. We would also like to thank Dr Vazir and Dr Palacios for providing further information regarding their studies.
The methods section of this review is based on a standard template used by Cochrane Neonatal.

Bhadari 2001 {published data only}
Bhandari N, Bahl R, Nayyar B, Khokhar P, Rohde JE, Bhan MK. Food supplementation with encouragement to feed it to infants from 4 to 12 months of age has a small impact on weight gain. Journal of Nutrition 2001;131 (7)  This is a 4-arm trial: food supplementation, nutritional counselling, no intervention, visitation groups. For this review, we have included the nutritional counselling (intervention) and no intervention (control) groups only Imputing of SD for change in weight and length between 6 and 12 months:

Characteristics of included studies [ordered by study ID]
Mean change in weight and length between 6 and 12 months of age was calculated by subtracting mean weight/length at 26 weeks' gestation from mean weight/length at 52 weeks' gestation for both groups, as given in Tables 3 and 4 The SD for these changes was determined by imputation of the SD for changes from the baseline method given in Chapter 16.1.

of the Cochrane Handbook for Systematic Reviews of Interventions.
First, the correlation of coefficient was calculated by taking the average of correlation coefficients for changes between 26 and 38 weeks' gestation and 38 and 52 weeks' gestation using the data given in Table 3. This average correlation coefficient was then used to impute the SD for the change between 26 and 52 weeks' gestation

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Unclear risk Quote: "children were stratified by weight for height status (≤ 80% and > 80% of the National Center for Health Statistics median for that age) and randomly assigned to one of the four study groups", but methods of sequence generation and randomisation are not given Control group received routine services, which included advice on initiation of complementary feeding at 4 to 6 months, types of foods to be fed, and frequency of feeding, but the focus was on family planning and immunisation Outcomes Weight gain between 6 and 12 months of age Increase in length between 6 and 12 months of age

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Quote: "generated 4 single-digit random numbers using a random numbers table; the first listed community in a pair was allocated to the intervention group if the random number was 0-4 and the second if it was 5-9" Allocation concealment (selection bias) Low risk A statistician not involved with the study performed the randomisation (i.e. there was central allocation using random number tables) Blinding of participants and personnel (performance bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants or of personnel delivering the intervention; however, outcomes included in this review are unlikely to be affected by lack of blinding Blinding of outcome assessment (detection bias) All outcomes

Low risk
There was an attempt to blind personnel who collected outcome data Quote: ''mothers and infants were visited at home by workers who were not involved in the delivery of the intervention at 3,6,9,12,15, and 18 months of age'' Also, outcomes included in this review are unlikely to be affected by lack of blinding Incomplete outcome data (attrition bias) All outcomes High risk The number of infants followed up at 12 months is not reported. However, attrition at both 9 months and 18 months was > 10%. At 9 months, 451 (81.2%) in the intervention group and 403 (85.2%) in the control group were followed up. The numbers followed up were further reduced at 18 months: 435 (78.8%) in intervention group and 394 (83.3%) in control group. There also appears to be imbalance in the reasons for loss to follow-up. There were 26 (

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Unclear risk Quote (Costa 2017): "mothers who agreed to participate were sequentially listed on the basis of their time of delivery, grouped in blocks of five, and their names separated in opaque, sealed envelopes. Two mothers from each block were randomly assigned to the intervention group, and the remaining three mothers were allocated to the control group" However, the method of randomisation is not given in any of the publications Allocation concealment (selection bias)

Low risk As above
Blinding of participants and personnel (performance bias) All outcomes

High risk
The intervention is not amenable to blinding of participants or of personnel delivering the intervention. One outcome considered in this review (serum haemoglobin level) is unlikely to be affected by lack of blinding, but the other (prevalence of exclusive breastfeeding) is highly likely to be affected by knowledge of treatment allocation Blinding of outcome assessment (detection bias) All outcomes High risk This report indicates that outcome assessors were blind Quote: "interviewers who were not involved in the intervention process and who were blind to which group children belonged conducted home visits at 6 and 12 months in order to collect data on the study variables" Incomplete outcome data (attrition bias) All outcomes Low risk 397 (79.4%) of the 500 randomised infants were included in the final analysis. A similar percentage of infants were lost to follow-up in both intervention (37 (19%)) and control groups (66 (22%)) Selective reporting (reporting bias)

Low risk
The protocol for this study is available: some outcomes, such as effectiveness of a nutrition advice programme for occurrence of diarrhoea, respiratory problems, use of dental caries, and hospitalisation, were not reported. However, all proposed outcomes relevant to this review have been reported Other bias Low risk

Study characteristics
Methods RCT Participants Study setting: Hospital de Clínicas de Porto Alegre (Brazil), a public general university hospital where about 3000 to 4000 births take place annually Inclusion criteria: adolescent mothers (younger than 20 years) living with their mothers or not, living in the city of Porto Alegre, with healthy non-twin newborn infants with birth weight ≥ 2500 g, in the rooming-in ward in Hospital de Clínicas de Porto Alegre, who had started breastfeeding Exclusion criteria: pairs that had to be separated due to problems related to the mother or the baby, adolescent mothers who lived with their newborn's paternal grandmother 323 mothers and infants were recruited: 163 (76 (47%) males) in the intervention group and 160 (88 (55%) males) in the control group

Interventions
Intervention group: counselling sessions in the maternity ward and at mother's home in the first 6 months. Session included advice on breastfeeding as per WHO guidelines (in the first 4 months) and complementary feeding as per "Brazilllian Nutririon Guidelines for Children Younger Than Two Years Old" from 4 months of age Three other publications from the same study were identified from the study registration at clinicaltrials.gov, but outcomes of interest were not reported. Review authors contacted the authors of this study but did not receive a response

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Participants were assigned to study groups by block random allocation in groups of 2. Two spheres of similar texture and size, 1 bearing the word "Yes" (assignment to intervention group) and the other bearing the word "No" (assignment to control group), were drawn from a dark bag, and participants were allocated to study groups accordingly Allocation concealment (selection bias) "Yes" (assignment to intervention group) and the other bearing the word "No" (assignment to control group), were drawn from a dark bag, and participants were allocated to study groups accordingly Blinding of participants and personnel (performance bias) All outcomes

High risk
The intervention is not amenable to blinding of participants or of personnel delivering the intervention. The main outcome of duration of exclusive breastfeeding is likely to be affected by bias Blinding of outcome assessment (detection bias) All outcomes Low risk Interviewers who delivered the questionnaire for data collection were blinded to the group to which the mothers belonged Incomplete outcome data (attrition bias) All outcomes High risk 257 (79.6%) of the 323 randomised mothers were included in the final analysis. However, the dropout rate appears to be higher in the control group -35 mothers (21.9%) -as compared to the intervention group -28 mothers (17.5%) Selective reporting (reporting bias) Low risk All outcome measures proposed were reported based on clinicaltrials.gov (NCT00910377) Other bias High risk There appear to be more males (55% vs 47%), breastfeeding duration longer than 6 months appears lower among previous children (44% vs 59%), and breastfeeding duration for the adolescent mother appears lower (12.9 vs 17.7 months) in the control group, which may introduce bias to the analysis de Oliveira 2012 (Continued)  Inclusion criteria: random sub-sample of participants in a large study exploring children's fruit and vegetable acceptance during weaning was selected to participate in the study before initiation of complementary feeding. In the large study, 327 women in the final trimester of their pregnancy and mothers of infants younger than 6 months of age were eligible to participate if they were older than 18 years of age at recruitment and were sufficiently proficient in each country's respective native language to understand the study materials, and if their infant was born after 37 weeks' gestation, without diagnosed feeding problems

Methods
For this study, 146 mothers and infants were recruited: 75 (33 out of 68 (49%) males) in the intervention group and 71 (40 (57%) males) in the control group

Interventions
Intervention group: counselling sessions either at home or at the paediatrician's office up to 4 weeks before initiation of complementary feedings. Session included advice on introduction of vegetables and the technique of exposure feeding for weaning and on vegetable intake and preference Control group: usual care without any specific guidance, instructions, or information on weaning with vegetables Outcomes Taste test with intake (g) of the novel vegetable as the primary outcome Identification Notes

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Women in the final trimester of their pregnancy and mothers of infants younger than 6 months of age Allocation concealment (selection bias) Low risk Women in the final trimester of their pregnancy and mothers of infants younger than 6 months of age Blinding of participants and personnel (performance bias) All outcomes High risk Because of the nature of the intervention, parents in the intervention arm and researchers delivering the intervention were not blind. The same researcher who delivered the intervention also delivered the taste test Blinding of outcome assessment (detection bias) All outcomes

High risk
The researcher present at the taste test was the same individual who delivered the intervention and therefore was not able to be blinded to the condition Incomplete outcome data (attrition bias) All outcomes Low risk 139 (95%) infants completed the study Selective reporting (reporting bias) Low risk All proposed outcomes were reported Other bias High risk A random sub-sample of participants in a large study was used in this study. It is unclear how the sub-sample of participants was selected as this may introduce bias to the study

Study characteristics
Methods RCT Participants Study setting: Dortmund, Germany, including members of a nationwide compulsory health insurance company (NOVITAS Vereinigte BKK, Duisburg, Germany) who reported the birth of a baby Inclusion criteria: mother speaking German and available by telephone, as well as healthy term (> 37 weeks of pregnancy) infant with birth weight > 2500 g 183 mothers and infants were recruited: 142 (68 (48%) males) in the intervention group and 41 (

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias)

Unclear risk This was not mentioned in the article
Allocation concealment (selection bias)

Unclear risk This was not mentioned in the article
Blinding of participants and personnel (performance bias) All outcomes High risk Because of the nature of the intervention, parents in the intervention arm were not blinded Blinding of outcome assessment (detection bias) All outcomes High risk Outcome assessment depended upon maternal recall of infant's diet, whereby the mother was not blinded to the intervention given Incomplete outcome data (attrition bias) All outcomes Low risk All recruited infants completed the study Selective reporting (reporting bias) Low risk All proposed outcomes were reported Other bias High risk Only 37% and 54% of potential users used the telephone hotline and personal telephone counselling, respectively Methods Cluster-RCT Participants Study setting: primary paediatricians in the Veneto region (northeastern Italy) registered in their district list of primary care practitioners interested in research projects Inclusion criteria: healthy full-term newborns whose parents or guardians had given their informed consent at the time of the first routine visit who were followed until the age of 2 years Exclusion criteria: preterm or post-term birth or any congenital disorder, disease, or syndrome 562 infants were recruited: 295 (48% males) in the intervention group and 267 (50% males) in the control group. There is no baseline difference between groups of infants Interventions Intervention group: parents were provided with standardised oral and written nutrition education concerning protective practices at all routine visits scheduled during the children's first 2 years of life (at 1, 3, 6, 12, and 24 months of age). Encouraged behaviours were breastfeeding, feeding on demand, responsive feeding, timely complementary feeding, giving portions based on the child's appetite, alternating protein sources correctly, and playing active games with the child Control group: routine care and follow-up Outcomes Change in weight and length between 6 and 12 months Exclusive breastfeeding at 3 months

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Quote: "Twenty-two accepted and were randomly assigned 1:1 to the intervention or control arm using the random selection function of Microsoft Excel. The eleven 'intervention paediatricians' were asked to recruit at least 30 consecutive newborns during the first six months of the study" Allocation concealment (selection bias) High risk Although paediatricians were randomly allocated to the intervention or control arm, they were aware of their patient's treatment allocation while recruiting the 30 patients from their practice Blinding of participants and personnel (performance bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants and personnel, but outcome measures relevant to this review are unlikely to be affected by lack of blinding Blinding of outcome assessment (detection bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants and personnel, but outcome measures relevant to this review are unlikely to be affected by lack of blinding Incomplete outcome data (attrition bias) All outcomes Low risk 468 of 562 (83%) infants completed the study. There was no major difference in dropout rates across both groups (81% vs 85% in control vs intervention group) Selective reporting (reporting bias) Other bias High risk Mother's professional category was higher in control group as compared to the intervention group at baseline

Study characteristics
Methods Cluster-RCT Participants Study setting: the study was conducted in Kabale and Kisoro districts in southwestern Uganda because of the high prevalence of stunting (UBOS -Uganda Demographic and Health Survey, 2011). Town centres were excluded to minimise differences in socio-economic status and feeding practices. People living in the study area are predominantly small-scale farmers. Both districts are densely populated and are made up of several sub-counties, each consisting of 18 to 25 villages Inclusion criteria: children between 6 and 8 months of age Exclusion criteria: households with a child having a congenital malformation, a physical disorder that would influence assessments and/or nutrient intake, and/or a diagnosis of mental or brain illness, as reported by the mother or by a health worker A total of 511 infants were enrolled, of whom 263 (139 males) were in the nutritional counselling group and 248 (123 males) were in the no intervention group. There were no apparent differences between intervention and control groups at baseline Interventions Intervention group: nutrition education was delivered via 3 group meetings over a period of 6 months to 26 groups of mothers (4 to 12 mothers per group). It was delivered by a trained education team and included 2 behaviour change techniques: Providing information and prompt practice (i.e. demonstrations of preparing food and stimulating children) Control group: routine care Outcomes Weight-for-age z score at 12 months Height-for-age z score at 12 months Weight-for-height z score at 12 months Head circumference z score at 12 months Bayley Scales of Infant and Toddler Development III scales (cognitive) at 12 to 16 months Random sequence generation (selection bias) Low risk Quote. "By use of computer-generated random numbers, villages whose assigned number matched with the random numbers were selected" Allocation concealment (selection bias) Low risk Although investigators were not blinded of the numbers given to each village as the villages were listed alphabetically and were numbered in ascending order, use of computer-generated random numbers to select the village will reduce allocation concealment bias Blinding of participants and personnel (performance bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants or of personnel delivering the intervention; however, outcomes (growth and development) included in this review are unlikely to be affected by lack of blinding of participants Blinding of outcome assessment (detection bias) All outcomes Low risk Growth and development outcome assessors were blinded to the allocation as stated in Muhoozi 2018. Outcome assessors for Atukunda 2019 outcomes (child development at 36 months) were blinded, but all other outcomes measured previously were not assessed by blinded assessors Incomplete outcome data (attrition bias) All outcomes Unclear risk Attrition bias was unclear as proportionate sampling was used to obtain 10 sub-counties (6 out of 19 in Kabale, and 4 out of 14 in Kisoro) to participate in the study. Researchers used a 3-stage procedure to obtain households for the study. Hence, only a sub-sample of a population that has received the intervention participated in the study and had outcome measures obtained. This is an appropriate method for a large-scale community study, However, risk of attrition bias is unclear. The follow-up study (Atukunda 2019) included only patients from January to May 2014. This may be fully representative of the whole recruited sample, as this represents only 30% of the original sample size Selective reporting (reporting bias) Low risk All proposed outcomes were reported Other bias Low risk

Study characteristics
Methods Pseudo-randomised controlled trial Participants Study setting: 2 municipal wards (F North and M East) in the urban slum areas of Mumbai (India) were purposely selected due to their large slum area, high population proportion classified as low-income, and no prior exposure to mMitra. Each ward is typically served by 1 maternity home and 5 or 6 health posts that provide pregnancy and infant health services. Each ward appoints roughly 100 community health workers who make home visits, register pregnant women, and motivate them to seek health care for themselves and their children Inclusion criteria: pregnant women speaking the Hindi or Marathi language were enrolled in the study Exclusion criteria: women without access to a mobile phone at home or not likely to be in Mumbai for 4 to 5 months during the pregnancy and post-delivery periods A total of 2016 mothers enrolled, of whom 1516 were in the intervention group and 500 were in the no intervention group

Interventions
Intervention group: the mMitra package consisted of 145 voice messages designed to be shared from when a woman was 6 weeks pregnant until the infant reached 1 year of age. Messages were delivered 2 times per week during pregnancy; they were clustered at 1 message per day immediately postpartum (P = 0.02) and read newspapers (P = 0.02). Their husbands were also more likely to be literate (0.02) and employed (P = 0.003)" Quote: "This study was funded by Johnson & Johnson. The funders contributed to critical review of the manuscript and the decision to submit the paper for publication"

Study characteristics
Methods Cluster-RCT Participants Study setting: West Singhbhum and Kendujhar -2 adjoining rural districts of Jharkhand and Odisha in eastern India Inclusion criteria: individual participants were pregnant women identified and recruited in study clusters and their children Exclusion criteria: stillbirths and neonatal deaths, infants whose mothers died, those with congenital abnormalities, multiple births, and mother and infant pairs who migrated out of the study area permanently during the trial period 3001 infants from 120 geographical clusters were recruited in the study

Interventions
Intervention group: community-based workers conduct a single home visit to each pregnant woman in the third trimester of pregnancy for counselling on maternal nutrition, followed by monthly home visits to all children younger than 2 years with counselling for growth promotion, and facilitate 2 to 3 participatory meetings with local women's groups per month Outcomes Change in height in the first 18 months of life Height-for-age z score at 18 months of age Weight-for-age z score at 18 months of age Weight-for-height z score at 18 months of age Exclusive breastfeeding at 6 months of age Identification Notes

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Meeting participants put numbered balls corresponding to clusters in each stratum in a local tombola (lottery device), then sequentially allocated each ball (cluster) to the intervention or control arm Allocation concealment (selection bias) High risk Due to the nature of the intervention, participants and the intervention team were not masked to allocation Blinding of participants and personnel (performance bias) Low risk Although participants and personnel were not blinded due to the nature of the intervention, the growth parameters obtained were an objective measure Cochrane Database of Systematic Reviews

All outcomes
Blinding of outcome assessment (detection bias) All outcomes

Low risk
The data collection team and the data manager were masked to allocation Incomplete outcome data (attrition bias) All outcomes

Low risk
The overall follow-up rate was similar across both intervention and control groups at 86% and 85%, respectively Selective reporting (reporting bias) High risk There was diversion from protocol on outcomes reported as the number of secondary outcomes was later reduced on the online trial registration form after feedback from the data monitoring committee. The following outcomes in the published protocol were not reported: change in weight from birth to 18 months of age, and mean mid-upper arm circumference z score at 18 months of age Inclusion criteria: a cohort of pregnant women in their third trimester was prospectively recruited from each cluster. Pregnant women were eligible for inclusion if they had no intention of leaving the study area for the next 2 years and provided informed written consent. Eligible pregnant women were identified through antenatal consultations (women attending their third antenatal visit) and were included until the desired sample size was reached at each health centre Exclusion criteria: only 1 infant from multiple pregnancy was included. Infants with major birth defects were excluded 2253 infants were recruited from 12 clusters of a primary healthcare catchment area

Interventions
Intervention group: individual nutrition counselling was provided to all women attending intervention centres during pregnancy and during the first 18 months of their child's life. Counselling contacts were scheduled to begin during pregnancy and to continue until the child reached 18 months of age. At each contact, health providers used a patient-centred approach to explore the caregiver's and the child's situation and current feeding practices, and to identify their specific needs in terms of nutrition, health advice, and curative care

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk For each pair of health centres, 2 identical pieces of paper were numbered corresponding to each health centre and were put into a basket. A volunteer not involved in the study was asked to choose a paper for the intervention centre Allocation concealment (selection bias)

Low risk As above
Blinding of participants and personnel (performance bias) All outcomes High risk Participants and personnel were not blinded due to the nature of the intervention. However, anthropometric assessments are unlikely to be affected by lack of blinding. On the other hand, the other outcome considered in this review (duration of exclusive breastfeeding) is dependent on parental report and may be affected by lack of blinding Blinding of outcome assessment (detection bias) All outcomes High risk Data collection was performed by trained field workers not involved in intervention delivery who were not blind to the intervention. However, anthropometric assessments are unlikely to be affected by lack of blinding. On the other hand, the other outcome considered in this review (duration of exclusive breastfeeding) is dependent on parental report and may be affected by lack of blinding Incomplete outcome data (attrition bias) All outcomes High risk Overall 67.5% children were lost to follow-up, with 68.6% and 66.3% in the intervention and control arms, respectively. No significant difference was reported between the 2 study arms in the distribution of children lost to follow-up. However, reasons for loss to follow-up were not reported Selective reporting (reporting bias) Low risk There was no deviation from the published protocol Other bias Low risk No difference in baseline characteristics was noted in the intervention and control arms, respectively

Study characteristics
Methods RCT Participants Study setting: Bogota, Colombia Inclusion criteria: term-born infants with birth weight > 2.500 g who were exclusively breastfed at 4 months of age and were still breastfeeding at 6 months of age Exclusion criteria: infants with a haemoglobin concentration < 110 g/L A total of 85 infants were enrolled, of whom 42 (20 males) were in the new complementary feeding guideline (nutritional counselling) group and 43 (21 males) were in the no intervention group

Interventions
Intervention group: nutrition counselling with face-to-face sessions and detailed verbal and written guidance. This nutrition counselling prioritised the importance of continuing breastfeeding alongside complementary feeding, the importance of including red meat as a source of iron to prevent anaemia, and the importance of fruit and vegetables as part of an infant's diet. 28 mothers (74%) randomised to this group followed the recommendations completely during the whole period of the intervention. Reasons given for not following the recommendations included family pressure (17%), grandmother's opinion (50%), and mother's own decision (33%) at 12 months of age Adverse events (narrative report)

Notes
There was commercial support for the study: "Tommee Tipee (United Kingdom) donated the feeding spoons, cups, and beakers used in the study"

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) Low risk Quote: "Randomisation assignments were prepared by using randomised blocks of permuted length by a member of the team who had contact with study subjects" Allocation concealment (selection bias) Low risk Quote: "Stored in sealed opaque envelopes" Blinding of participants and personnel (performance bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants or personnel delivering the intervention. However, anthropometric and laboratory assessments are unlikely to be affected by lack of blinding Blinding of outcome assessment (detection bias) All outcomes Low risk Quote: "It was not possible to blind researchers who collected anthropometric and food-intake data, but laboratory measurements were blinded" Incomplete outcome data (attrition bias) All outcomes Low risk At 12 months' follow-up, attrition rate for anthropometric outcomes was 9.4% but attrition was higher for serum haemoglobin (14%) and serum ferritin (33%) Selective reporting (reporting bias) High risk Retrospectively registered in the ISRCTN registry (ISRCTN57733004) Iron and zinc status, anthropometric parameters, and mothers' opinions are reported, but motor development, energy and nutrient intakes, and serum retinol are not reported in this paper Other bias High risk Baseline differences in groups: "those in the control group were significantly heavier with higher mid-upper arm circumferences, weight-for-age z score, weight-for-length z score, and MUAC z score at baseline (6 months of age)"

Study characteristics
Methods RCT Participants Study setting: participants were recruited from 2 Women, Infants, and Children (WIC) clinics in Puerto Rico and from 4 WIC clinics in Hawaii, USA. WIC clinics were selected based on availability and accessibility to investigators, with the help of the WIC programme at each site.
Inclusion criteria: infants 0 to 2 months old; caregivers must be 18 years of age or older, owners of a mobile phone with unrestricted SMS capability, responsible for infant care, and willing to participate for the full study duration Exclusion criteria: infants with special diets, infants with limited mobility, preterm birth (< 37 weeks), small or large for gestational age (birth weight < 10th or > 90th percentile), inability to consent to participate, unwillingness to be randomised, not being able to read A total of 202 infants were enrolled, of whom 102 (51 males) were in the intervention group and 100 (52 males) were in the no intervention group. There was no difference in baseline demographics. There is no baseline group difference Exclusive breastfeeding data at 6 months were provided by the study author (CP). Data were not included in the meta-analysis as there was a high dropout of 6 and 0 infants in the intervention and control groups with exclusive breastfeeding data Weight at 6 months old: no significant group differences were found in weight status at the end of the trial nor weight changes from visit 1 (0 to 2 months old) to visit 2 (4 to 6 months old). When stratified by site, there was a significant decrease in the proportion of infants classified as underweight during the initial visit compared to the follow-up visit in the intervention group in Puerto Rico. In Hawaii, a significant decrease in the proportion of infants with adequate weight and a significant increase in the proportion of infants with overweight or obesity in both control and intervention groups were noted

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias) computer-generated list of randomisation numbers and corresponding IDs was created by a biostatistician. Investigators were blind to the size of each block. Participants were allocated an ID sequentially as they were recruited, and this ID was matched with the randomised group Allocation concealment (selection bias)

Low risk As above
Blinding of participants and personnel (performance bias) All outcomes High risk Blinding is not possible for this intervention. The only included outcome -exclusive breastfeeding at 6 months of age -is subject to bias Blinding of outcome assessment (detection bias) All outcomes Allocation concealment (selection bias) Low risk Quote: "randomisation was done before formative research to avoid it acting as an intervention" Investigators used coin flipping as the means of randomisation. If used faithfully, this would effectively conceal each upcoming allocation Blinding of participants and personnel (performance bias) All outcomes Low risk Quote: "Families were not told whether they were in the intervention or control group" Blinding of outcome assessment (detection bias) All outcomes Low risk Quote: "Data were collected by project field workers who were not involved in the delivery of the intervention" Incomplete outcome data (attrition bias) All outcomes Low risk At 18 months' follow-up: in the intervention group 171/187 (91.4%) babies were followed up, and in the control group 167/190 (87.8%) babies were followed up Combined 90% of babies were followed up Selective reporting (reporting bias) Low risk Trial registration was not reported in the published paper. However, all proposed outcomes have been reported Other bias High risk Socio-economic differences were noted between families in intervention and control groups (e.g. more educated mother in the intervention group (52%) in comparison to the control group (36%))

Study characteristics
Methods Cluster-RCT Participants Study setting: 6 public health centres located in downtown Isfahan (centre of Iran), which were similar in terms of economic, social, and cultural conditions, were chosen

Bias Authors' judgement Support for judgement
Random sequence generation (selection bias)

Unclear risk No information was provided
Allocation concealment (selection bias)

Unclear risk No information was provided
Blinding of participants and personnel (performance bias) All outcomes

Low risk
The intervention is not amenable to blinding of participants nor personnel delivering the intervention; however, outcomes included in this review are unlikely to be affected by lack of blinding Quote: "Families were not told whether they were in the intervention or control group" Blinding of outcome assessment (detection bias) Inclusion criteria: mothers of infants aged 10 to 20 weeks' gestation who were exclusively or partially breastfed but had not started or had recently started (< 1 week before enrolment) complementary feeding Exclusion criteria: all infants with weight for age < 5th centile as per the WHO growth chart, history of 2 or more hospital admissions, serious congenital anomalies, other chronic conditions impairing feeding, or presence of acute illness and/or severe anaemia that required urgent hospital treatment A total of 212 infants were recruited (118 in the intervention group and 94 in the control group), but baseline data were reported on 110 infants and 84 infants in the 2 groups, respectively

Interventions
Intervention group: 4 visits (baseline and then at 10 weekly intervals) with 10 key messages developed on WHO/UNICEF recommended practices: importance of breastfeeding continuation for 2 years, hygiene, complementary feeding initiation at 6 months, advice on promoting protein and iron-rich complementary foods The intervention continued for 6 months with education messages provided once a week for the first 3 months and then once every 2 weeks for the next 3 months

Footnotes
(1) Change from baseline measure (change in WAZ between 6 and 12 months of age) due to significant differences at baseline (2) WAZ at 12 months was presented in a graph and values were calculated using the scales provided in Adobe Acrobat software. The SE estimated from these measurements were converted to SD u Analysis 1.

Footnotes
(1) Change from baseline measure (change in HAZ between 6 and 12 months of age) due to significant differences at baseline (2) HAZ at 12 months was presented in a graph and values were calculated using the scales provided in Adobe Acrobat software (Adobe Systems Software Ireland Ltd). The standard errors estimated

Footnotes
(1) Change from baseline measure (change in WHZ between 6 and 12 months of age) due to significant differences at baseline

Footnotes
(1) Change from baseline measure (change in WAZ between 6 and 12 months of age) due to significant differences at baseline (2) WAZ at 12 months was presented in a graph and values were calculated using the scales provided in Adobe Acrobat software. The SE estimated from these measurements were converted to SD  Embase: (infant, newborn or newborn or neonate or neonatal or premature or very low birth weight or low birth weight or VLBW or LBW or Newborn or infan* or neonat*) AND (human not animal) AND (randomized controlled trial or controlled clinical trial or randomized or placebo or clinical trials as topic or randomly or trial or clinical trial) CINAHL: (infant, newborn OR newborn OR neonate OR neonatal OR premature OR low birth weight OR VLBW OR LBW or Newborn or infan* or neonat*) AND (randomized controlled trial OR controlled clinical trial OR randomized OR placebo OR clinical trials as topic OR randomly OR trial OR PT clinical trial)

A D D I T I O N A L T A B L E S
Cochrane Library: (infant or newborn or neonate or neonatal or premature or preterm or very low birth weight or low birth weight or VLBW or LBW)