1. Objectives

To evaluate the effect of interventions to improve water quality and supply (adequate quantity to maintain hygiene practices), provide adequate sanitation and promote handwashing with soap, on the nutritional status of children under the age of 18 years and to identify current research gaps

2. How studies were identified

The following databases were searched in July 2011, with an update search in June 2012:

• Cochrane Public Health Group Special Register
• MEDLINE
• MEDLINE In-Process
• Web of Science (included SCI-EXPANDED, SSCI, CPCI-S and CPCI-SSH)
• EMBASE
• Econlit
• mRCT
• Global Health
• Greenfile
• CAB abstracts
• Grey Literature
• CNKI-CAJ (China National Knowledge Infrastructure for Chinese-language databases)

Reference lists were handsearched and subject experts and relevant organisations were contacted. Relevant journals and conference proceedings were also searched

3. Criteria for including studies in the review

3.1 Study type

Randomized, quasi-randomized and non-randomized controlled trials; controlled before and after studies (cohort or cross-sectional); interrupted time series; and historically controlled studies

3.2 Study participants

Children under 18 years of age

3.3 Interventions

• Interventions aimed at improving the microbiological quality of drinking water (via treatment systems, protected distribution or improved storage), compared to usual practice or a less rigorous intervention
• Interventions aimed at introducing a new or improved water supply or improved distribution, or both, compared to usual practice
• Interventions aimed at introducing or expanding the coverage and use of facilities designed to improve sanitation (e.g., flush toilets, pit latrines, septic tanks, sewer systems), compared to usual practice
• Interventions aimed at the promotion of handwashing with soap after defecation, disposal of child faeces and prior to preparing and handling food, compared to usual practice or no handwashing promotion
• Any combination of the water, sanitation and hygiene (WASH) interventions listed above

3.4 Primary outcomes

Child nutritional status as measured by anthropometry

• Weight-for-age (underweight)
• Weight-for-height (wasting)
• Height-for-age (stunting)

Secondary outcomes included child nutritional status as measured by anthropometry: weight, height, mid-upper arm circumference, skinfold thickness, percent body fat, birth weight, and body mass index (BMI); and child nutritional status as measured by nutrient status: haemoglobin, serum ferritin, soluble transferrin receptor, serum retinol, serum zinc, urinary iodine, and clinical signs of nutrient deficiency.

4. Main results

4.1 Included studies

Fourteen trials, enrolling 22,241 children, were included in this review:

• Five studies were cluster-randomized and nine studies were non-randomized
• All children were aged under five years at baseline and studies ranged from six to 60 months in duration
• Anthropometric data was reported in 10 trials, and authors supplied raw anthropometric data from a further trial
• Three studies reported on interventions to improve the quality of water, one study reported on a sanitation intervention, three studies reported on interventions to improve hygiene, three studies reported on a dual water quality and hygiene intervention, one study reported on a dual intervention on water quantity and sanitation, one study reported on an intervention including water quality, quantity and hygiene, one study reported on an intervention including water quantity, sanitation and hygiene elements, and one study included all four WASH elements (water quality, quantity, sanitation and hygiene)

4.2 Study settings

• Bangladesh (2 studies), Cambodia, Chile, Ethiopia, Guatemala (2 studies), Kenya, Nepal, Nigeria, Pakistan (3 studies), and South Africa
• All studies were conducted in low- or middle-income countries in rural, peri-urban and urban locations

4.3 Study settings

How the data were analysed
Quantitative meta-analysis was restricted to the five cluster-randomized controlled trials from which raw individual participant data (IPD) was obtained for additional IPD analyses. The effect of WASH interventions was compared to control/no intervention on weight-for-age, weight-for-height and height-for-age, height and weight. Analyses were adjusted for clustering, baseline anthropometry and age, duration of treatment and sex.

To investigate heterogeneity, subgroup analysis by sex and age group (two years and under, and two to five years) was conducted. Formal interaction analyses were used in IPD meta-analysis for the same variables. Other potential subgroup analyses (duration of intervention, country setting, community location, age group over five years) were precluded due to a lack of diversity in the identified studies

Results
Weight-for-age Z-scores (WAZ)
In the three non-randomized studies that reported on WAZ, there was no effect of intervention. Meta-analysis was conducted on data from all five cluster-randomized controlled trials, with no difference observed between intervention and control groups, mean difference (MD) in WAZ of 0.05, 95% CI [-0.01 to 0.12], p=0.13; 4627 children. In IPD meta-analysis, the effect was also non-statistically significant (MD 0.10 WAZ; 95% CI [-0.04 to 0.25]; five trials/5386 children).

Weight-for-height Z-scores (WHZ)
No effect on WHZ was found in the two non-randomized studies that reported on this outcome. In meta-analysis of data from the cluster-randomized controlled trials, no evidence of a difference in WHZ was found between intervention and control groups (MD 0.02 WHZ, 95% CI [-0.07 to 0.11], p=0.72; five trials/4622 children). IPD meta-analysis also showed no effect of intervention (MD 0.10 WHZ 95% CI [-0.09 to 0.23], 5 trials/5375 children).

Height-for-age Z-scores (HAZ)
Four non-randomized trials reported on HAZ; three trials reported no effect of WASH interventions, while one trial of 2476 children reported an increase in HAZ in the intervention group compared to control group (MD 0.22 HAZ, 95% CI [0.11 to 0.33]). A borderline statistically significant effect of WASH interventions on HAZ was found in meta-analysis of the cluster-randomized controlled trials (MD 0.08 HAZ, 95% CI [0.00 to 0.16], p=0.05; 5 trials/4627 children. IPD meta-analysis produced a statistically significant effect of WASH on HAZ (MD 0.11 HAZ, 95% CI [0.03 to 0.18]; 5386 children).

Subgroup analyses for primary outcomes
In subgroup analyses using aggregated data from the five cluster-randomized controlled trials, no effect of age group (two years and under versus two to five years) on WAZ and WHZ was identified, and no evidence of an effect of sex on WHZ was observed. However, while there was no evidence of an intervention effect on WAZ among boys (MD 0.00 WAZ, 95% CI [-0.07 to 0.08], an effect was observed among girls (MD 0.11 WAZ; 95% CI [0.01 to 0.21]. Similarly, while there was no evidence of an effect on HAZ among boys (MD -0.01 HAZ, 95% CI [-0.09 to 0.08]), there was an effect among girls (MD 0.14 HAZ, 95% CI [0.04 to 0.25]). No intervention effect was found on HAZ for children under two years of age, but was increased among those aged over two years (MD 0.06 HAZ, 95% CI [0.00 to 0.12]).

In formal interaction analyses using IPD data, the apparent differences observed in subgroup analyses were not upheld. However, a significant interaction between WASH interventions and sex for WHZ (p=0.032) suggested that the intervention had a greater effect among girls. Interaction analysis between age group and treatment for WAZ (p=0.002) suggested that the intervention had a larger effect on weight gain among children aged over two years. A significant interaction was identified between age group and treatment for WHZ (p<0.001), which suggested that the intervention effect on weight gain was greater for those children aged over two years. WASH interventions may also have a greater effect on HAZ in children aged two years and under (p<0.001 for interaction between age group and HAZ).

Secondary outcomes
No difference in weight between intervention and control groups was identified in meta-analysis or IPD analysis. Although no difference in height was found in aggregate data meta-analysis, IPD meta-analysis identified a statistically significant effect of WASH interventions on height, MD 0.53 cm, 95% CI [0.20 to 0.86]. No evidence of an effect of WASH interventions on mid-upper arm circumference or BMI was observed. In one cluster-randomized controlled trial of 461 children, the haemoglobin concentrations of children in the two intervention arms was significantly lower than that in the control arm.

5. Additional author observations*

The overall quality of the identified trials was low. Meta-analysis was only performed on data from five cluster-randomized controlled trials of relatively short duration (nine to 12 months), in which interventions only included water quality improvement and handwashing. Therefore, the estimates presented in the current review may not reflect the effect of other WASH interventions on child nutritional status outcomes. Further ongoing research, including five large randomized controlled trials, will likely modify the effect estimates of WASH interventions presented in the current review, and will also increase the level of confidence in the resulting effect estimates.

Future trials should investigate the mechanism of action of the WASH interventions, long-term adherence to interventions, and the optimal timing and duration of interventions required for the greatest benefits in child nutritional status.