Independent and combined effects of improved water, sanitation, and hygiene, and improved complementary feeding, on child stunting and anaemia in rural Zimbabwe: a cluster-randomised trial

Background: Child stunting reduces survival and impairs neurodevelopment. We tested the independent and combined effects of improved water, sanitation, and hygiene (WASH), and improved infant and young child feeding (IYCF) on stunting and anaemia in in Zimbabwe. Methods: We did a cluster-randomised, community-based, 2 × 2 factorial trial in two rural districts in Zimbabwe. Clusters were defined as the catchment area of between one and four village health workers employed by the Zimbabwe Ministry of Health and Child Care. Women were eligible for inclusion if they permanently lived in clusters and were confirmed pregnant. Clusters were randomly assigned (1:1:1:1) to standard of care (52 clusters), IYCF (20 g of a small-quantity lipid-based nutrient supplement per day from age 6 to 18 months plus complementary feeding counselling; 53 clusters), WASH (construction of a ventilated improved pit latrine, provision of two handwashing stations, liquid soap, chlorine, and play space plus hygiene counselling; 53 clusters), or IYCF plus WASH (53 clusters). A constrained randomisation technique was used to achieve balance across the groups for 14 variables related to geography, demography, water access, and community-level sanitation coverage. Masking of participants and fieldworkers was not possible. The primary outcomes were infant length-for-age Z score and haemoglobin concentrations at 18 months of age among children born to mothers who were HIV negative during pregnancy. These outcomes were analysed in the intention-to-treat population. We estimated the effects of the interventions by comparing the two IYCF groups with the two non-IYCF groups and the two WASH groups with the two non-WASH groups, except for outcomes that had an important statistical interaction between the interventions. This trial is registered with ClinicalTrials.gov, number NCT01824940. Findings: Between Nov 22, 2012, and March 27, 2015, 5280 pregnant women were enrolled from 211 clusters. 3686 children born to HIV-negative mothers were assessed at age 18 months (884 in the standard of care group from 52 clusters, 893 in the IYCF group from 53 clusters, 918 in the WASH group from 53 clusters, and 991 in the IYCF plus WASH group from 51 clusters). In the IYCF intervention groups, the mean length-for-age Z score was 0·16 (95% CI 0·08–0·23) higher and the mean haemoglobin concentration was 2·03 g/L (1·28–2·79) higher than those in the non-IYCF intervention groups. The IYCF intervention reduced the number of stunted children from 620 (35%) of 1792 to 514 (27%) of 1879, and the number of children with anaemia from 245 (13·9%) of 1759 to 193 (10·5%) of 1845. The WASH intervention had no effect on either primary outcome. Neither intervention reduced the prevalence of diarrhoea at 12 or 18 months. No trial-related serious adverse events, and only three trial-related adverse events, were reported. Interpretation: Household-level elementary WASH interventions implemented in rural areas in low-income countries are unlikely to reduce stunting or anaemia and might not reduce diarrhoea. Implementation of these WASH interventions in combination with IYCF interventions is unlikely to reduce stunting or anaemia more than implementation of IYCF alone. Funding: Bill & Melinda Gates Foundation, UK Department for International Development, Wellcome Trust, Swiss Development Cooperation, UNICEF, and US National Institutes of Health. The design and methods of SHINE have been reported previously,2 and the protocol and statistical analysis plan are available online. Briefly, SHINE was a cluster-randomised, community-based 2 × 2 factorial trial in two contiguous rural districts in Zimbabwe (Chirumanzu and Shurugwi). The districts have a 15% prevalence of antenatal HIV8 and a high prevalence of schistosomiasis, but a very low prevalence of soil-transmitted helminths.9 Rotavirus vaccination was introduced during the trial from May, 2014. Households are usually single-family dwellings surrounded by farm land. Before the trial, mean distance between households was 82·6 m,10 and population density was 18·6 people per km2. Clusters were defined as the catchment area of between one and four village health workers (VHWs) employed by the Zimbabwe Ministry of Health and Child Care. Urban and uninhabited areas were excluded. Between Nov 22, 2012 and March 27, 2015, VHWs prospectively surveyed new pregnancies, established the date of last menstrual period, and referred pregnant women to SHINE research nurses, who enrolled eligible women. Women were eligible for inclusion if they permanently resided in a study cluster and were confirmed pregnant. During the recruitment period, the cutoff of gestational age for eligibility was gradually loosened (from 14 weeks’ gestational age to just before parturition) to maximise recruitment (appendix). The Medical Research Council of Zimbabwe and the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health approved the study protocol. All participants provided written informed consent. Clusters were allocated (1:1:1:1) to one of four treatment groups: standard of care, infant and young child feeding (IYCF), WASH, or IYCF plus WASH. LHM, the study’s senior statistician, used a constrained randomisation technique11 to identify 5000 allocation schemes that achieved balance across the groups for 14 variables related to geography, demography, water access, and sanitation coverage, and also met bias and validity specifications (appendix). From these, ten allocations were randomly selected. The final allocation was selected at a public randomisation event attended by elected representatives of the study districts. Masking of participants and fieldworkers was not possible because of the obvious visual differences between interventions, but investigators were blinded to treatment groups until the final analysis of each prespecified outcome. Interventions were informed by extensive formative research and piloting.12, 13, 14, 15 Behaviour-change modules were delivered by group-specific VHWs, who underwent training for 20 days to deliver standard of care, 30 days to deliver WASH, 32 days to deliver IYCF, and 35 days to deliver IYCF plus WASH. All enrolled women were scheduled to receive 15 behaviour-change modules with specific messages and interactive tools from enrolment until 12 months after the birth of their children (roughly one visit per month). Other family members were also encouraged to participate. A sequential integrated longitudinal intervention was delivered, and at each visit, previous information was reviewed before new information was introduced. Previously missed modules were delivered before any new material. Between 13 months and 17 months postnatal, VHWs continued monthly visits providing routine care. During these visits, VHWs informally encouraged participants to practise behaviours relevant to their treatment group, although formal modules were not implemented. At 18 months, a review module was implemented in all treatment groups before the mother completed the trial. In the standard-of-care group, VHWs promoted exclusive breastfeeding to 6 months,16 advised on neonatal care, and promoted uptake of Ministry of Health and Child Care services, including antenatal care, immunisations, and family planning. In the WASH group, VHWs delivered all the standard-of-care messages, plus information about safe disposal of faeces, handwashing with soap at key times, protection of infants from geophagia and ingestion of animal faeces, chlorination of drinking water (especially for infants), and hygienic preparation of complementary food. Additionally, ventilated improved pit latrines were constructed and two handwashing stations were installed by community builders supervised by Ministry of Health and Child Care environmental health technicians within 6 weeks of enrolment. A plastic mat and play yard (North States, Minneapolis, MN, USA) were delivered to the home by a trial logistician at 2 and 6 months postnatal, respectively, and VHWs made monthly deliveries of soap from the time of the handwashing module (roughly 30 weeks antenatal) and chlorine from the time of the water treatment module (4 months postnatal) until 18 months postnatal. In the IYCF group, VHWs delivered all the standard-of-care messages plus information about the importance of nutrition for infant health, growth, and development; feeding nutrient-dense food and 20 g of a small-quantity lipid-based nutrient supplement17 (Nutriset, Malaumay, France) daily from age 6–18 months; processing foods to facilitate mastication and swallowing; feeding during illness; and dietary diversity. Participants also received monthly deliveries of 30 sachets of the small-quantity lipid-based nutrient supplement from VHWs from infant age 6 months to 18 months postnatal. In the combined group, participants received all standard-of-care, WASH, and IYCF interventions. Ventilated improved pit latrines were built for participants in the standard-of-care and IYCF groups after trial completion. Research nurses made home visits at baseline (roughly 2 weeks after mothers provided consent), 32 weeks’ gestation, and 1, 3, 6, 12, and 18 months post partum to assess maternal and household characteristics and trial outcomes. Intervention uptake was assessed at all visits and reported here as prespecified for the 12-month postnatal visit. At baseline, mothers’ height, weight, mid-upper-arm circumference, and haemoglobin concentrations (with Hemocue, Ängelholm, Sweden) were measured. They were also tested for Schistosoma haematobium (by urinary microscopy) and HIV (via the rapid test algorithm). HIV-positive women were urged to seek immediate antenatal care for prevention of mother-to-child transmission. Other maternal and household characteristics assessed included dietary diversity, food insecurity, household wealth, and maternal capabilities.18 Infant birth date, weight, and delivery details were transcribed from health facility records. We provided Tanita BD-590 (Arlington Heights, IL, USA) infant scales to all health institutions in the study area and trained facility staff in use of the scales. Gestational age at delivery was calculated from the date of the mother’s last menstrual period. Infant weight, length, head circumference, and mid-upper-arm circumference were measured at every postnatal visit (appendix). At the 18-month postnatal visit (ie, the trial endpoint), haemoglobin concentrations were measured, and length was calculated as the median of three measurements. Nurses were standardised against a gold standard anthropometrist by measuring non-study children from the community during a quality-control session held every 6 months. Infant diarrhoea (three or more loose or watery stools in 24 h), dysentery (stool with blood or mucus), and acute respiratory infection (fast or difficult breathing) were assessed by 7-day maternal recall at postnatal visits. Infants with acute malnutrition or illness were referred to clinics. At the 18-month postnatal visit (ie, the trial endpoint), mothers and infants were visited anywhere in the country. However, in view of the household-based nature of interventions, intermediate visits were done only when the mother was available in the household where she consented. Serious adverse events and adverse events were ascertained during data collection visits and by VHWs, and were referred to a senior research nurse who collected details. Events were reviewed by the study physician (AJP) to establish relatedness to trial interventions before reporting them to responsible institutional review boards. An independent data safety and monitoring board comprising two physicians from Zimbabwe and a statistician from the UK reviewed interim adverse event data. We prespecified that primary trial inferences would be based on findings among infants born to mothers who were HIV negative during pregnancy. Primary outcomes were mean length-for-age Z score and haemoglobin concentration at 18 months (target age 76–80 weeks, allowable range 76–130 weeks) in all enrolled participants (intention-to-treat analysis). Secondary outcomes were mean weight-for-age Z scores, weight-for-length Z scores, mid-upper-arm circumference-for-age Z scores, and head circumference-for-age Z scores; the proportion of infants who were stunted (ie, length-for-age Z score less than −2), severely stunted (ie, length-for-age Z score less than −3), anaemic (ie, haemoglobin concentration <105 g/L), severely anaemic (ie, haemoglobin concentration <70 g/L), underweight (ie, weight-for-age Z scores less than −2), and wasted (ie, weight-for-height Z scores less than −2); mean prevalence of diarrhoea, dysentery, and acute respiratory infection based on 7-day maternal history at infant age 12 months and 18 months; and cumulative mortality up to age 18 months. The original sample size calculation was 4800 women to allow for 15% exclusion because of maternal HIV and 20% loss because of fetal and infant deaths, withdrawal, and loss to follow-up. Actual recruitment was 10% higher than the sample size calculation to ensure sufficient power for sensitivity analyses. With a minimum of 816 HIV-unexposed infants per group at 18 months, the trial was powered to detect a 0·2 difference in length-for-age Z scores, a reduction of 8 percentage points in stunting, and a 2·6 g/L shift in haemoglobin for the marginal effect of either intervention, with 90% power and type 1 error of 5%.2 This calculation was based on an assumed coefficient of variation of the true proportions of 0·43, and an effective loss of 33% of sample size because of variability in cluster size. All analyses were done on an intention-to-treat basis at the child level. For primary analyses, we used generalised estimating equations that accounted for within-cluster correlation and contained two dummy variables representing the main effect of the IYCF intervention (the two IYCF groups compared with the two non-IYCF groups) and the WASH intervention (the two WASH groups compared with the two non-WASH groups), unadjusted for other covariates, with an exchangeable working correlation structure.11 Although the study was not powered to detect a statistical interaction between the two interventions, we estimated these interactions for each outcome. When the interaction was significant (ie, p<0·05 according to the Wald test) or had a sizeable point estimate (ie, relative risk [RR] >2 or <0·5 when comparing ratio-of-ratios, or difference-of-differences >0·25 SDs when comparing continuous outcomes), results were based on a regression model with three dummy variables to represent IYCF, WASH, and IYCF plus WASH compared with standard of care instead of the model of two terms. In adjusted analyses we controlled for prespecified baseline covariates, which were initially assessed in bivariate analyses to identify those with an important association with the outcome (ie, p<0·2 or RR >2·0 or <0·5 for dichotomous outcomes, and p<0·2 or difference >0·25 SDs for continuous outcomes). Selected covariates were entered in a multivariable regression model; a forward stepwise selection procedure was implemented with p<0·2 to enter. A log-binomial specification was used to facilitate estimation of RRs. Depending on the analysis, other methods for comparison of groups while accounting for within-cluster correlation included multinomial and ordinal regression models with robust variance estimation, and Somers' D for medians. In a per-protocol analysis, we examined the effect of the interventions when behaviour-change modules were delivered at high fidelity (which was predefined for the IYCF plus WASH group as receiving ten core modules and for the other study groups as receiving all modules scheduled at the same timepoints when IYCF plus WASH core modules were delivered). Several sensitivity analyses were prespecified: exclusion of mothers enrolled before Nov 1, 2013, to account for initial delays in latrine construction; exclusion of children born to women who were HIV negative during pregnancy but HIV positive at 18 months; and restriction of analyses to children in whom primary outcomes were measured during tight infant age windows (ie, at age 76–80 weeks and 76–100 weeks). A prespecified subgroup analysis of primary outcomes by infant sex was planned if the interaction terms were significant. We used Stata (version 14.1) for all analyses. This trial is registered with ClinicalTrials.gov, number NCT01824940. The study funders approved the trial design, but had no roles in data collection, analysis, or interpretation, or writing of the report. The corresponding author had full access to all study data and had final responsibility for the decision to submit for publication.