Independent and combined effects of improved water, sanitation, and hygiene (Wash) and improved complementary feeding on early neurodevelopment among children born to hiv-negative mothers in rural zimbabwe: Sub study of a cluster-randomized trial

Background Globally, nearly 250 million children (43% of all children under 5 years of age) are at risk of compromised neurodevelopment due to poverty, stunting, and lack of stimulation. We tested the independent and combined effects of improved water, sanitation, and hygiene (WASH) and improved infant and young child feeding (IYCF) on early child development (ECD) among children enrolled in the Sanitation Hygiene Infant Nutrition Efficacy (SHINE) trial in rural Zimbabwe. Methods and findings SHINE was a cluster-randomized community-based 2×2 factorial trial. A total of 5,280 pregnant women were enrolled from 211 clusters (defined as the catchment area of 1–4 village health workers [VHWs] employed by the Zimbabwean Ministry of Health and Child Care). Clusters were randomly allocated to standard of care, IYCF (20 g of small-quantity lipidbased nutrient supplement per day from age 6 to 18 months plus complementary feeding counseling), WASH (ventilated improved pit latrine, handwashing stations, chlorine, liquid soap, and play yard), and WASH + IYCF. Primary outcomes were child length-for-age Zscore and hemoglobin concentration at 18 months of age. Children who completed the 18- month visit and turned 2 years (102–112 weeks) between March 1, 2016, and April 30, 2017, were eligible for the ECD substudy. We prespecified that primary inferences would be drawn from findings of children born to HIV-negative mothers; these results are presented in this paper. A total of 1,655 HIV-unexposed children (64% of those eligible) were recruited into the ECD substudy from 206 clusters and evaluated for ECD at 2 years of age using the Malawi Developmental Assessment Tool (MDAT) to assess gross motor, fine motor, language, and social skills; the MacArthur–Bates Communicative Development Inventories (CDI) to assess vocabulary and grammar; the A-not-B test to assess object permanence; and a self-control task. Outcomes were analyzed in the intention-to-treat population. For all ECD outcomes, there was not a statistical interaction between the IYCF and WASH interventions, so we estimated the effects of the interventions by comparing the 2 IYCF groups with the 2 non-IYCF groups and the 2 WASH groups with the 2 non-WASH groups. The mean (95% CI) total MDAT score was modestly higher in the IYCF groups compared to the non-IYCF groups in unadjusted analysis: 1.35 (0.24, 2.46; p = 0.017); this difference did not persist in adjusted analysis: 0.79 (−0.22, 1.68; p = 0.057). There was no evidence of impact of the IYCF intervention on the CDI, A-not-B, or self-control tests. Among children in the WASH groups compared to those in the non-WASH groups, mean scores were not different for the MDAT, A-not-B, or self-control tests; mean CDI score was not different in unadjusted analysis (0.99 [95% CI −1.18, 3.17]) but was higher in children in the WASH groups in adjusted analysis (1.81 [0.01, 3.61]). The main limitation of the study was the specific time window for substudy recruitment, meaning not all children from the main trial were enrolled. Conclusions We found little evidence that the IYCF and WASH interventions implemented in SHINE caused clinically important improvements in child development at 2 years of age. Interventions that directly target neurodevelopment (e.g., early stimulation) or that more comprehensively address the multifactorial nature of neurodevelopment may be required to support healthy development of vulnerable children. The design and methods of SHINE have been previously described [27]; the full protocol and statistical analysis plan are at https://osf.io/w93hy. Briefly, SHINE was a cluster-randomized community-based 2×2 factorial trial testing the independent and combined effects of a WASH intervention and an IYCF intervention on linear growth and hemoglobin at 18 months of age [27]. The study area comprised 2 rural districts of central Zimbabwe, which were divided into 212 clusters, defined as the catchment area of 1–4 village health workers (VHWs) employed by the Zimbabwean Ministry of Health and Child Care. Clusters were randomly allocated to 1 of 4 treatment arms (standard of care [SOC] alone, WASH, IYCF, or WASH + IYCF) at a public event using highly constrained randomization, which achieved balance across arms on 14 measures of geography, demography, water access, and sanitation coverage [28] (described more fully in S1 Text). Due to the nature of the interventions, masking was not possible. Between November 2012 and March 2015, VHWs identified pregnancies through prospective surveillance; women were eligible if they permanently resided in 1 of the rural study clusters, were confirmed pregnant (<14 gestational weeks), and provided written informed consent. Over the recruitment period, the cutoff of gestational age for recruitment eligibility was increased to 18 weeks (August 22, 2013), 24 weeks (January 3, 2014), and any time prior to parturition (October 20, 2014), through trial protocol amendments, to maximize recruitment. Recruitment took place in 211 of the 212 clusters. All women were scheduled to receive 15 VHW visits between enrollment and 12 months postpartum (approximately 1 visit/month). Interventions were informed by extensive formative research and piloting [27,29,30]. Participatory behavior change interventions delivered during these visits were arm-specific and grounded in behavior change theory [27]. The IYCF intervention was based on formative research to identify and target cultural barriers [31]. The WASH intervention was based on the model of planned, motivated, and habitual hygiene behavior and was designed to invoke motivating emotions for hygiene and nurture [32]. At each visit, previous information was reviewed before introducing new information to create a sequenced integrated longitudinal intervention. Between 13 and 17 months postpartum, VHWs undertook monthly visits to provide routine care, deliver intervention supplies, and provide informal reminders to practice relevant behaviors, but formal modules were not delivered. At 18 months postpartum, an intervention review module was delivered to mothers in all trial arms. VHW supervisors assessed timing and fidelity of implementation during scheduled visits and spot-checks (conducted every 3 months, or more often if VHW performance was not optimal). The content of VHW visits and the commodities provided were arm-specific. VHWs delivered a breastfeeding intervention during 4 visits between late pregnancy and 3 months postpartum [27], and promoted family planning, immunization, and prevention of mother-to-child-transmission (PMTCT) of HIV. VHWs delivered all SOC interventions plus modules promoting safe disposal of feces, handwashing with soap, protection of infants from geophagia and animal feces ingestion [27,29], chlorination of drinking water, and hygienic handling and preparation of complementary food. A Blair ventilated improved pit latrine was constructed within 6 weeks of enrollment by builders hired from the study communities; the study provided all materials and labor. Two handwashing stations were installed by 32 weeks’ gestation with monthly delivery of liquid soap until 18 months postpartum. A plastic baby mat and play yard were delivered at 2 months postpartum to protect children from geophagia. Chlorination solution (WaterGuard, Nelspot, Zimbabwe) was delivered monthly between 4 and 18 months postpartum for water treatment [27]. VHWs delivered all SOC interventions plus modules promoting nutrient-dense, diverse infant diets using locally available foods processed to facilitate mastication and swallowing, and frequent responsive feeding during illness. Small-quantity lipid-based nutrient supplement (SQ-LNS; Nutriset, Malaumay, France) was provided from 6 to 18 months postpartum; caregivers added a 20-g sachet daily to complementary food [27,30]. SQ-LNS was an supplement and was not intended to replace other food. VHWs delivered all SOC, WASH, and IYCF interventions. After trial completion, a latrine was constructed for families in the SOC and IYCF arms. Research nurses, separate from the intervention teams, made home visits at baseline (2 weeks after consent), at 32 weeks’ gestation, and at 1, 3, 6, 12, and 18 months postpartum to assess maternal and household characteristics and trial outcomes. At baseline, mothers had weight, mid-upper arm circumference (MUAC), and hemoglobin (Hemocue, Ängelholm, Sweden) measured, and were tested for HIV using a rapid test algorithm. HIV-positive women were urged to seek immediate antenatal care for PMTCT interventions. The following baseline indices were assessed: household minimum dietary diversity, food insecurity (Coping Strategies Index), household wealth (asset ownership index) [33], and maternal capabilities (perceived physical health, mental health, stress, social support, decision-making autonomy, gender norms attitudes, time use, and mothering self-efficacy), as detailed in the trial design paper [27]. Infant birth date, weight, and delivery details were transcribed from health records. The trial provided Tanita BD-590 infant scales to all health institutions in the study area and conducted training. Gestational age at delivery (prematurity) was calculated from the date of the last menstrual period. Infant weight, length, head circumference, and MUAC were measured at every postnatal visit. Children with acute malnutrition or illness were referred to local clinics. Mothers testing HIV-negative at baseline were retested at 32 weeks’ gestation; those testing HIV-negative during pregnancy were retested at 18 months postpartum. Intervention uptake was assessed at all visits and is reported here, as pre-specified, for the 12-month postnatal visit. Nurses assessed WASH-related behaviors through maternal report (open defecation among household members, treatment of drinking water, disposal of nappy water, and child geophagia) and observation of the latrine (type of latrine, whether the path to latrine was trodden, whether the latrine was used for storage; and whether the latrine was shared with other households), handwashing station (presence of Tippy Taps and whether they were filled with soap and water), and play yard (visible cleanliness). Nurses assessed IYCF behaviors through maternal report of whether the child was still breastfeeding; the mother’s understanding of how to feed a child during illness; 24-hour recall of infant minimum dietary diversity and consumption of iron-rich, animal-source, and vitamin-A-rich foods; and 24-hour recall of infant SQ-LNS consumption. The ECD substudy was conducted, among a subgroup of children enrolled in SHINE, to assess the impact of the SHINE interventions on child development at 24 months of age. No additional interventions were provided after 18 months. Children were eligible for the ECD substudy if they had the trial primary outcomes (linear growth and hemoglobin) measured at 18 months of age, and turned 2 years of age (allowable range 102–112 weeks) between March 2016 and April 30, 2017. The substudy initiation date was staggered across research sites from March 1 to March 15, 2016, so that the first visits could be directly supervised by the trial psychologist (JC). Written informed consent for the ECD substudy was obtained where possible during the 18-month SHINE visit. For those who had already had their 18-month visit, families were contacted through their VHW, by phone call, or through a home visit to establish their interest in joining the ECD substudy; written informed consent was then obtained prior to conducting the ECD assessment. We compared baseline characteristics across the 4 treatment arms among 3 groups of children who completed the trial 18-month visit: those who were not eligible for the ECD substudy, those who were eligible and recruited into the ECD substudy, and those who were eligible but not recruited into the ECD substudy. All assessments were conducted in the home over a period of 2–3 hours by 1 of 11 SHINE research nurses who completed 3 weeks of residential training in ECD assessments. All children enrolled into the ECD substudy were assessed, but children who scored “moderate to severe” on the Washington screen for disability [34] were excluded from analysis and referred for appropriate services. All assessment tools used for measuring outcomes were directly observed and were done on 1 occasion, with the child at the age of 24 months. The Malawi Developmental Assessment Tool (MDAT) measures child development in 4 domains: (1) gross motor coordination (36 items), (2) fine motor coordination (36 items), (3) language (36 items), and (4) social (30 items), with a total of 138 items. The fine motor, language, and social domains also measure components of cognitive development [35]. Following translation, back translation, and piloting, minor adaptations were made to kit items, particularly those that the child had to name or identify as part of the language assessment. The MacArthur–Bates Communicative Development Inventories (CDI) [36] assesses child language according to maternal report, including a vocabulary and grammar checklist. The test was specifically adapted for Shona speakers using a detailed protocol approved by the CDI team [37,38]. The adaptation protocol included interviewing 30 mothers of children from 18 to 30 months of age in Shona-speaking families to identify common words, piloting this list of over 350 words with 30 additional mothers of similar-aged children, identifying those words that correlated with age (p > 0.1), and selecting a range of 100 words found to be easy (70%–100% said their child knew the word), moderate (40%–70% said their child knew the word), and hard (only 20%–40% said their child knew the word). This vocabulary checklist was then piloted with another 30 mothers, and inter-rater reliability was tested to ensure 97% agreement between testers on the same list with the same mother [38]. This process was conducted only in Shona-speaking households. We included the CDI grammar checklist for 2-year-olds to increase the number of items specifically targeting language development, which changes dramatically at this age. The A-not-B test assesses object permanence and working memory [39]. This task requires the child to watch as a treat is hidden under 1 of 2 bowls (A or B); after a brief delay, the child is asked to find the treat under one of the bowls and in doing so, to remember (through object permanence) which one of the bowls it was hidden under (A or B). The exercise is repeated 10 times, switching which bowl the treat is hidden under according to a strict protocol for all 10 episodes to check that the child has no perseveration error. Children not completing a full set of 10 tests were excluded from analysis. The self-control task [40] we used assesses impulsivity. The child is required to watch as a treat is promised to them, but they have to wait for 2 minutes to take it. The test is first conducted with a covered treat, then with an uncovered treat. Self-control was defined as a child who waited for 2 minutes. We conducted and scored the test in a similar way to that done in Uganda [41]. We prespecified that the primary inferences would be based on children of mothers testing HIV-negative during pregnancy; these results are presented in this paper. Results among children born to HIV-positive mothers will be reported separately. The prespecified primary outcomes of the ECD substudy were total MDAT score (out of 138), MDAT gross motor score (out of 36), MDAT fine motor score (out of 36), MDAT social score (out of 30), MDAT language score (out of 36), MacArthur–Bates CDI vocabulary checklist (total number of words known out of 100), A-not-B score (out of 10), and the proportion of children with self-control. The prespecified secondary outcome was the proportion of children who used imperatives or the progressive tense, plurals, or combined 2 words as assessed using the MacArthur–Bates CDI grammar checklist. We undertook several validation and quality control procedures. Nurses underwent 6-monthly refresher training and standardization (using non-SHINE children), undertaking an ECD assessment that was observed and double-marked by a gold-standard assessor; percentage agreement had to be >85% to pass the standardization. At each standardization (3 in total), nurses were asked to measure 1 child twice (once in the morning and once in the afternoon), for which intra-class correlations for each test were as follows: MDAT, 0.88 (95% CI 0.82, 0.94); MacArthur–Bates CDI, 0.94 (95% CI 0.90, 0.96); A-not-B test, 0.85 (95% CI 0.80, 0.90); and self-control task, 0.80 (95% CI 0.76, 0.85). Supportive supervision of ECD assessments was undertaken during monthly field visits, with corrective or reinforcing feedback provided to nurses. Finally, 5% of assessments in the field were video-recorded. These assessments were then reviewed and double-marked by a psychologist with expertise in all tests (JC) and a pediatrician with advanced training in child neurodevelopment and Shona language proficiency (GK). Percentage agreement was 93% for MDAT fine motor, 90% for MDAT language, 97% for A-not-B, and 91% for the self-control task. Only the nurse’s score was used in the final analysis. All analyses were intention-to-treat at the child level. For tests with continuous outcomes (MDAT, MacArthur–Bates CDI, and A-not-B test), the absolute difference in mean score between treatment groups was estimated. For tests with dichotomous outcomes (self-control and grammar), the relative risk (RR) of passing the test was estimated in comparing treatment groups. Although the study was not powered to detect a statistical interaction between the IYCF and WASH treatments, it was estimated for each outcome. We accounted for the interaction in the model if it was significant (p < 0.05, according to the Wald test) or had a sizeable point estimate (i.e., difference in mean score > 0.25 SD for continuous outcomes; RR > 2 or <0.5 for dichotomous outcomes). Otherwise, we used a regression model with 2 terms to represent the treatment arms; we estimated the effect of IYCF by comparing the 2 IYCF arms with the 2 non-IYCF arms and estimated the effect of WASH by comparing the 2 WASH arms with the 2 non-WASH arms. If interaction was significant, we used a regression model with 3 terms to represent the 4 treatment arms. We used generalized estimating equations that accounted for within-cluster correlation to estimate effect size, unadjusted for other covariates, with an exchangeable working correlation structure [39]. A log-binomial specification was used to facilitate estimation of RRs. We compared baseline characteristics between arms while handling within-cluster correlation using multinomial and ordinal regression models with robust variance estimation, and Somers’ D for medians. We used Stata (version 14.1) for all analyses. Adjusted analyses controlled for prespecified baseline covariates (as in our statistical analysis plan), which were initially assessed in bivariate analyses to identify those with an important association with the outcome (for dichotomous outcomes: p < 0.2 or RR > 2.0 or < 0.5; for continuous outcomes: p < 0.2 or difference > 0.25 SD). Selected covariates were entered in a multivariable regression model; a forward stepwise selection procedure was implemented, with p < 0.2 required for a variable to enter the model. A per-protocol analysis was conducted to examine intervention effects when delivered at high fidelity (prespecified for WASH + IYCF arm as receiving all 10 core modules; for other arms predefined as receiving all modules scheduled at the same time-points when WASH + IYCF core modules were delivered). A prespecified subgroup analysis by child sex was planned if the interaction terms were p < 0.05. In a prespecified sensitivity analysis, children of women who seroconverted to HIV after pregnancy were excluded. Other ECD intervention trials and comparisons of preterm versus term children have reported effect sizes of 0.3–0.4 standard deviations for similar ECD outcomes [42,43]. Accordingly, we calculated our sample size requirements to detect a 0.2 standard deviation shift with >80% power and a type I error rate of 5%, assuming an ICC of 0.07, 10 children per cluster, 33 clusters per arm, and a total of 132 clusters. We therefore aimed to recruit at least 1,320 children. The Medical Research Council of Zimbabwe and the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health approved the study protocol (Zimbabwe: MRCZ/A/1675; Johns Hopkins University: IRB#4205). The ECD substudy protocol was included as an amendment to the main SHINE trial protocol, submitted to and approved by the 2 institutional review boards. The SHINE statistical analysis plan included the prespecified ECD outcomes. These documents can be found in S1 Text and at https://osf.io/w93hy. An independent data and safety monitoring board comprising 2 physicians from Zimbabwe and a statistician from the UK (listed in Acknowledgments) reviewed interim adverse event data in the main trial between enrollment and 18 months of age, but not in the ECD substudy since no interventions were provided between 18 and 24 months of age. The trial was registered at ClinicalTrials.gov ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT01824940″,”term_id”:”NCT01824940″}}NCT01824940).