53 updated the review in 2012. 78% (n=25) were cluster randomised controlled trials, or randomised controlled trials. Six studies used a controlled before-and-after design, and one was a non-randomised controlled trial.
The interventions were implemented in 16 low and middle-income countries. 59% (n=19) were from South Asia: nine from Bangladesh, six from India, two from Pakistan and two from Nepal. Three studies were from two African countries, seven were from Mexico, Central and South America and the Caribbean, and three further studies were carried out in Vietnam, China and Iran.
About half of the studies focused on general health and nutrition behaviours. Several studies explicitly mentioned hand washing and hygiene (n=7), child development and stimulation (n=9), growth monitoring (n=4), responsive feeding (n=6), complementary feeding (n=9), breastfeeding (n=5) and maternal mental health (n=3). Three interventions provided supplementary food to all groups and five to some experimental groups. Two further studies provided micronutrient supplements or Vitamin A to selected groups. The remaining 22 studies did not involve supplementary feeding or micronutrients.
The interventions and their effect on child growth are described in Tables 2.1-2.3, grouped by behaviour change approach. I identified five broad types of behaviour change: health education (Table 2.1), behaviour change and communication (Table 2.2), studies using a mixture of behaviour change methods, multi-component interventions with more than two distinct components including behaviour change and non-behaviour change approaches, and cognitive behavioural therapy (Table 2.3).
Health education and behaviour change communication were the dominant approaches, used in 13 and 11 studies respectively. Four studies used a mixture of behaviour change methods including health education with positive deviance and community mobilisation. Three multi- component interventions used health education in combination with activities such as deworming, immunisations, food provision, community mobilisation, growth monitoring and psychosocial stimulation; two also used conditional cash transfers to incentivise behaviour change. One final study used cognitive-behavioural therapy for maternal depression.
54 Table 2.1 Description of health education interventions
Study no First author and date Country Study design Child age Growth Outcomes 1 Aboud et al (2008) Bangladesh Randomised Controlled Trial 12-24
months
Attained and gained weight Intervention
Results and study grade
Responsive feeding: n=32 village-clusters with existing parenting groups were randomly selected and randomised to: intervention (6 sessions about maternal responsive feeding and child self-feeding, n=102 mother-child pairs) or control (6 regular nutrition education and complementary feeding sessions, n=100 mother-child pairs). Sessions held by local trained peer educators. Data collected at baseline, 2 weeks and 5 months post-intervention. Researchers were blinded to condition. 10% of each group lost to follow-up.
Intervention group was significantly heavier (d=0.28, p=0.0021) and had greater weight gain (d=0.48, p=0.002) than controls. Study grade: high
2 Aboud et al (2009) Bangladesh Randomised Controlled Trial 8-20 months Weight gain, WAZ1 Intervention
Results and study grade
Responsive feeding: n=37 village-clusters with existing women’s groups were randomised to: intervention (n=19 clusters, n=108 children) or control (n=18 clusters, n=95 children). Both groups received five sessions of nutrition education. The intervention group received six extra sessions on child self-feeding and maternal verbal responsiveness during feeding and a booster session six weeks before endline. Data collected at baseline, two weeks and five months post-intervention. Researchers were blinded to condition.
No impact on weight gain or WAZ at post-test (both groups WAZ=-1.93) or five months (intervention=-1.87, control=-1.86). Potential for control group contamination. Study grade: moderate
3 Ahmed et al (1993) Bangladesh Controlled before and after study <19 months WAZ1, HAZ2 Intervention
Results and study grade
Hygiene education: n=185 households with children <19 months from five rural villages with a high prevalence of poor hygiene, diarrhoea and malnutrition were assigned to the intervention. Five matching villages (socio-demographics, hygiene and childcare factors) were controls (n=185 households). The 7-month 'clean life campaign' involved ground sanitation, personal and food hygiene, delivered by health workers and volunteer mothers twice weekly to 3-5 mothers.
14% reduction in severe underweight in the intervention group, significantly greater than controls, adjusted for socio-demographics (p<0.05). The intervention group also had significantly lower HAZ scores than controls at end-line. Limitations: lack of sample size calculation, no adjustment for clustering, results not fully reported, questionable suitability of HAZ as an outcome and potential data collection bias. Study grade: very low
55
4 Bhandari et al (2001) India Randomised Controlled Trial 6-12 months Weight and length gain, WAZ1, HAZ2 WHZ4 Intervention
Results and study grade
Responsive and supplementary feeding: n=418 infants from a south Delhi slum were recruited using household survey data. Children were stratified by weight-for-height (≤80% or >80% NCHS median) and randomised to: (1) monthly food supplementation and nutritional
counselling (2) monthly nutritional counselling (3) monthly home visit (4) control (no intervention). Groups 1-3 were visited twice a week for a morbidity assessment. Intervention lasted 8 months.
Small impact of education and food versus control (+250g); no impact of education only versus control. No impact on length/LAZ or WHZ of either intervention. Study grade: high
5 Bhandari et al (2004) India Randomised Controlled Trial <18 months WAZ1, HAZ2, weight and length gain Intervention
Results and study grade
Nutrition and health education: n=8 rural communities were pair-matched on household characteristics and randomised to health and nutrition education (n=552 households) or no intervention (n=473 households). The intervention involved trained health workers delivering locally developed nutrition and hygiene counselling through monthly home visits from birth to 12 months, growth monitoring every three months and immunisations. Measurements taken at baseline and every 3 months until 18 months of age.
The intervention group had a small but significantly greater length gain than controls (0.32cm, p=0.036) after adjusting for maternal employment, weight, length and breastfeeding status at 6 months. The effect was greater for males (0.51cm). There was no effect on LAZ, WAZ or weight gain. Study grade: high
6 Bowen et al (2012) Pakistan Cluster randomised controlled trial <30 months WAZ1, HAZ2, BMI Z-score3 Intervention
Results and study grade
Hand washing and water treatment: n=47 urban neighbourhoods with ≥1 hour of running water per week and a child <5 years randomised to one of five groups: two water treatment groups (flocculent disinfectant or sodium hypochlorite), a soap and hand washing promotion and disinfectant group, soap and hand washing only, or no intervention. Soap and hand washing groups received ‘instruction and
encouragement’ and materials by field workers during twice weekly home visits. This study followed-up the two soap and hand washing groups and the control groups (n=461 households). Children aged 5-7 years, <30 months during intervention.
No group differences for HAZ, WAZ or BMI Z-score. Limitations: possible bias as data collectors may have been aware of group allocations, study powered to detect change in developmental scores but not anthropometry, limited water access may have undermined intervention. Study grade: very low
7 Elizabeth & Sathy (1997) India Controlled before and after study 6-24 months Weight and height gain
Intervention Psychosocial/developmental stimulation and food supplementation: n=332 underweight children from 10 deprived areas were randomised to (1) nutrition education (breastfeeding, weaning, diet during illness), strengthened primary healthcare (e.g. deworming, medication) and food supplementation; n=118 (2) intervention 1 plus individualised child stimulation, play therapy and motor coordination tasks, daily living training and psychosocial inputs; n=127 or (3) no intervention; n=87. Interventions took two years: weekly for 3 months, fortnightly for 3
56
Results and study grade
months, and monthly for 18 months.
Significant increase ‘normal’ weight in all groups, but more so in the intervention groups; significant increases in ‘normal’ height were limited to the intervention groups. Limitations: no sample size calculation, insufficient data reporting and potential bias of data collectors. Study grade: very low
8 George et al (1993) India Randomised Controlled Trial <60 months Weight gain, WAZ1 Intervention
Results and study grade
Growth monitoring: n=12 non-adjoining poor rural villages were pair-matched (caste, road access, distance to health clinics, crops) and randomised to growth monitoring or non-growth monitoring interventions (n=550 children per group). Both groups received fortnightly home visits by trained local women for health education, received immunisations, weekly clinics for curative care, deworming, and materials for home gardens. The growth monitoring group also had their growth measured monthly and mothers received guidance about use of the growth chart. Measurements taken every 4-5 months until 60 months.
After 30 months of intervention there were no group differences (both groups improved by ≥0.2 WAZ). Limitations: results not fully reported. Study grade: moderate
9 Hamad et al (2011) Peru Randomised controlled trial <5 years WAZ1, HAZ2, BMI for age3 Intervention
Results and study grade
Health education: Microcredit groups (each 15-20 members) were randomised to: (1) health education based on Integrated Management of Childhood Illness modules or (2) no intervention (n=1855). Loan officers who led microcredit groups were trained to deliver education sessions over 8 monthly meetings.
No group differences for child anthropometry. Limitations: reluctance of loan workers to deliver health education, high variability in loan worker skills, no sample size calculation, inconsistent statistical reporting (e.g. the number of children per group at end-line), unclear if data collectors were blinded to condition, potential doubt over generalisability to malnourished population. Study grade: very low
10 Salehi et al (2004) Iran Controlled before and after study <59 months WAZ1, HAZ2, WHZ4, arm circumference Intervention
Results and study grade
Nutrition education: n=960 Qashqa’i tribe families randomly selected from 48 sub-tribes of Iran. N=406 children were randomised to intervention, n=405 to control. The intervention was a one-year community-based education programme, tailored to families from researcher observation of food preparation and cooking methods. Measures at baseline and 3-months post intervention.
Increases in WAZ, HAZ, WHZ and arm circumference were significantly greater in the intervention group than controls (by 0.45, 0.41, 0.27 SDs and 0.5cm respectively), although both groups showed significant improvements. Limitations: no sample size calculation, no adjustment for higher percentage of malnutrition in intervention group at baseline, or obvious confounders (e.g. tribal group), or multiple comparisons. No baseline arm circumference given, WHZ was normal in both groups at baseline.
Study grade: low for WAZ and HAZ, very low for WHZ and arm circumference
57
Intervention
Results and study grade
Nutrition education: n=28 municipal health centres were paired on socio-economic factors and child malnutrition then randomised to intervention or control. The intervention gave additional training to doctors on Integrated Management of Childhood Illness modules; doctors at control group centres received no extra training. The first 12-13 children <18 months attending for consultations with 33 doctors were recruited (n=218 intervention; n=206 controls). Child growth was measured at home visits 180 days post-consultation, data collectors were blinded to condition. Children who were hospitalised were excluded.
No overall effect on growth. Sub-group of intervention children aged 12-17 months had significantly higher WAZ and WHZ scores than controls; no effect on LAZ in this sub-group. Limitations: possible lack of generalisability because mothers were already motivated to seek care; no information given about reasons for children attending clinics; children not malnourished at baseline; questionable validity of LAZ outcome for short intervention. Study grade: very low
12 Vazir et al (2013) India Cluster randomised controlled trial 3 months WAZ1, HAZ2, WHZ4 Intervention
Results and study grade
Complementary and responsive feeding, psychosocial stimulation: 60 villages in rural Andhra Pradesh non-randomly selected and grouped into threes matched on population size, maternal literacy and birth weight. Village trios were randomised to: (1) standard care (2) standard care and complementary feeding education or (3) intervention 2 and guidance about responsive feeding and child development (n=200 mother-child pairs per arm). Interventions included 30 home visits by trained village women over 12 months. Data collectors blinded to treatment condition.
Simpler education group (group 2) had a 79% reduced stunting risk at 15 months than controls; there was no impact of health education + stimulation. No impact of either intervention on WAZ or WHZ. Limitations: results may not be generalisable to small villages, the higher percentage of people from tribal groups in group 3 may have influenced results, food insecurity limited adherence to feeding advice, Z- scores not fully reported, inconsistency (same health education component had differential effects on linear growth between groups 2 and 3); overall weight gain higher in control group than intervention groups, but not mentioned in text (p<0.052).
Study grade: low for WAZ, moderate for WHZ, high for LAZ.
13 Walker et al (1991) Jamaica Randomised Controlled Trial 9-24 months WAZ1, HAZ2, arm circumference Intervention
Results and study grade
Psychosocial stimulation and supplementary feeding: n=129 stunted children in a poor area of Kingston were identified via household surveys, age-stratified (>16 months or ≤16 months) and randomised to: (1) no intervention (n=33) (2) supplementary food (n=32) (3) stimulation via weekly home-visits by health workers to help mothers structure play sessions with toys and cognitive stimulation (n=30) (4) interventions 2 and 3 (n=34). A further group of non-stunted children were matched to every 4th intervention child (age, sex and location). Measures were taken at baseline, 6 months and 12 months post-intervention.
Stimulation had no impact on growth. Supplemented groups had significantly greater weight and length gains (adjusted for age) than other stunted groups at 6 months (p<0.01). Mean HAZ increased in all groups (0.7cm in stunted non-supplemented, 1.1cm in stunted
58
supplemented groups). Non-stunted children significantly taller than all other groups at 12 months. Limitations: no sample size calculation and potential contamination of control group. Study grade: low
1
WAZ – Weight-for-age Z-score
2
HAZ/LAZ – Height-for-age or length-for-age Z-score
3
BMI – Body Mass Index
59 Table 2.2 Description of behaviour change and communication interventions
Study no First author and date Country Study design Child age Growth Outcomes 1 Aboud & Akhter (2011) Bangladesh Cluster Randomised Controlled Trial 8-20 months Weight and length gain Intervention
Results and study grade
Responsive feeding and stimulation, with or without micronutrient supplementation: n=302 mother-child pairs randomised to: (1) control group: 12 health and nutrition education sessions by a health worker over three months (n=110) (2) The same 12 sessions plus 6 sessions with a peer-educator (trained local woman), including modelling and coached practice in self-feeding, and verbal responsiveness during play (n=92) or (3) intervention (2) plus daily micronutrient sprinkles (n=100). Measurements at baseline, post-test and 3-month follow-up. Data collectors blinded to condition.
No impact of education only on WAZ compared to controls. There was a small impact of education + micronutrients on WAZ (d=0.15) and weight gain (d=0.38) compared to the education only group. There was no impact of either intervention on length gain. Limitations: loss to follow-up was higher for those with lower baseline home environment scores; results not fully reported. Study grade: moderate
2 Arifeen et al (2009) Bangladesh Randomised Controlled Trial 7 days-59 months HAZ1, WHZ2 Intervention
Results and study grade
Integrated Management of Childhood Illness (IMCI), July 2001-June 2007: n=20 first-level government health facilities randomised to: (1) IMCI health worker training, health system strengthening, family and community activities (e.g. theatre groups to communicate IMCI messages) and usual care or (2) usual care. N=4400 children were randomly selected (n=220 per cluster)
Stunting declined significantly faster in the intervention group than the control (percentage point difference 7.3%). There was no impact on wasting. Study grade: high
3 Bhandari et al (2003) India Randomised Controlled Trial 3-6 months HAZ1, WHZ2 Intervention
Results and study grade
Exclusive breastfeeding promotion: n=8 communities were pair-matched (prevalence of child stunting, wasting, recent morbidity, mortality and socioeconomic status) and randomised to: (1) control group or (2) education to promote exclusive breastfeeding for 6 months and complementary feeding thereafter. Messages conveyed to caregivers of children <2 years via government health workers and specially trained health workers at monthly meetings, plus additional meetings for message repetition. Measurements at 3 and 6 months for n=1115 infants born 9 months after health worker training (n=552 intervention, n=473 control).
There were no growth differences between groups at 3 or 6 months. Limitations: potential bias as mothers recalled breastfeeding status for 4, 5 and 6 months at 9 month visit. Study grade: moderate
4 Brown et al (1992) Bangladesh Controlled before and after study 6-12 months WAZ3, arm circumference Intervention Nutrition education: n=3 villages were identified for intervention and n=62 weaning age children were selected using census data; n=55
60
Results and study grade
children of the same age were selected from 5 other villages (1 hour walk away) as controls. The intervention was 5-months of
complementary feeding messages delivered by volunteers via modelling techniques (e.g. home demonstrations of how to enrich foods), encouragement to continue breastfeeding, advice about feeding frequency and hygiene. Children were 9-18 months at end-line. Mean WAZ significantly higher (0.46) and arm circumference significantly greater (0.3cm) in the intervention group than the control. Limitations: no sample size calculation, potential bias as intervention implementers also collected data.
Study grade: moderate for WAZ, low for arm circumference
5 Hamadani et al (2006) Bangladesh Randomised Controlled Trial 6-24 months HAZ1, WHZ2,WAZ3 Intervention
Results and study grade
Psychosocial/developmental stimulation: n=20 community nutrition centres randomised to: (1) standard care - the Bangladesh Integrated Nutrition Programme (n=102) or (2) standard care plus weekly group meetings and home visits for 1 year to improve mother-child
interaction and provide developmentally appropriate activities, led by local 'play-leaders', using stories, songs and books (n=104) or (3) control group: n=107 normal weight children, matched to every 2nd child in groups 1 and 2 (age, sex and village) recruited from community nutrition centres. Interventions took 2 years.
No intervention effect for weight or height indicators in adjusted analyses; there was a significant increase in wasting in all groups. Limitations: no sample size calculation and potentially underpowered; results not fully reported. Study grade: low
6 Langford et al (2011) Nepal Non-randomised controlled trial 3-12 months HAZ1, WHZ2, WAZ3 Intervention
Results and study grade
Hand washing and hygiene: n=8 Kathmandu slum settlements were divided into Northern and Eastern locations and randomised to intervention or control (no intervention). N=45 children were randomly selected from intervention areas and n=43 from control areas using household survey data. The intervention: 6-months of hand washing promotion to change attitudes and social norms and create demand for good hygiene. Methods included a community play, posters and discussions, and daily home visits by ‘community motivators’ for two weeks, decreasing to once a week. Community motivators also held fortnightly mother’s meetings to promote hand washing and provided soap. Child growth was measured weekly.
No impact on child growth: WAZ and WHZ worsened faster in the intervention group (not significantly). Limitations: pre-existing group differences not accounted for, limited access to water and cost may have undermined hand washing, possible bias in self-reported behaviour, intervention may have been too brief to reduce stunting, no adjustment for clustering, unclear why WHO growth standards not used. Study grade: very low
7 Lutter et al (2008) Ecuador Controlled before and after study 9-14 months Weight and linear growth, WLZ2 Intervention Information, education and communication with food supplementation: n=10 primary health clinics were selected for intervention, and 6
for control. Both areas were eligible for intervention but a phased-roll out was planned. Intervention: Ecuador’s National Food Nutrition Programme, targeted at infants and young children in poor areas to improve feeding behaviours and dietary quality. Key components:
61
Results and study grade
information, education and communication, health worker training in nutrition counselling, community participation and provision of micronutrient fortified food. Health workers made weekly home-visits to children. Children were measured at baseline and after 11 months (N=338 intervention; n=296 controls).
Significant intervention impact on weight gain compared to controls (0.38kg, p=0.029). No intervention impact on linear growth (near significant for children 12-14 months at enrolment (p=0.08). No impact on WLZ. Limitations: low study power due to 50% loss to follow-up, this was associated with lower baseline WLZ and results may represent healthier children; characteristics of initial refusals not described, potential bias in data collection by health workers implementing the programme; unclear if potential contamination of control group; children not wasted at baseline in either group. Study grade: low for weight gain, very low for linear growth and WLZ
8 Roy et al (2005) Bangladesh Randomised Controlled Trial 6-24 months WAZ3, WAM4, Weight gain, Intervention
Results and study grade
Nutrition education with and without supplementary feeding: n=282 underweight children randomised to: (1) intensive nutrition education