Effects of improved water, sanitation, and hygiene and improved complementary feeding on environmental enteric dysfunction in children in rural Zimbabwe: A cluster-randomized controlled trial

Background Environmental enteric dysfunction (EED) may be an important modifiable cause of child stunting. We described the evolution of EED biomarkers from birth to 18 months in rural Zimbabwe and tested the independent and combined effects of improved water, sanitation, and hygiene (WASH), and improved infant and young child feeding (IYCF), on EED. Methodology and findings The Sanitation Hygiene Infant Nutrition Efficacy (SHINE) trial was a 2×2 factorial cluster-randomised trial of improved IYCF and improved WASH on child stunting and anaemia at 18 months of age. 1169 infants born to HIV-negative mothers provided plasma and faecal specimens at 1, 3, 6, 12, and 18 months of age. We measured EED biomarkers that reflect all domains of the hypothesized pathological pathway. Markers of intestinal permeability and intestinal inflammation declined over time, while markers of microbial translocation and systemic inflammation increased between 1–18 months. Markers of intestinal damage (I-FABP) and repair (REG-1β) mirrored each other, and citrulline (a marker of intestinal epithelial mass) increased from 6 months of age, suggesting dynamic epithelial turnover and regeneration in response to enteric insults. We observed few effects of IYCF and WASH on EED after adjustment for multiple comparisons. The WASH intervention decreased plasma IGF-1 at 3 months (β:0.89, 95%CI:0.81,0.98) and plasma kynurenine at 12 months (β: 0.92, 95%CI:0.87,0.97), and increased plasma IGF-1 at 18 months (β:1.15, 95%CI:1.05,1.25), but these small WASH effects did not translate into improved growth. Conclusions Overall, we observed dynamic trends in EED but few effects of IYCF or WASH on biomarkers during the first 18 months after birth, suggesting that these interventions did not impact EED. Transformative WASH interventions are required to prevent or ameliorate EED in low-income settings. A detailed description of the SHINE trial design and methods has been published [45], and the protocol, behaviour change modules, and statistical analysis plan are available at https://osf.io/w93hy. SHINE was a 2×2 factorial cluster-randomised trial assessing the individual and combined effects of improved IYCF and improved WASH on child stunting and anaemia at 18 months of age (ClinicalTrials.gov {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01824940″,”term_id”:”NCT01824940″}}NCT01824940). A total of 211 clusters, defined as the catchment area of 1–4 Village Health Workers (VHWs) employed by the Ministry of Health and Child Care (MoHCC), were allocated to one of four intervention groups using highly constrained randomization: standard-of-care (SOC); IYCF; WASH; or IYCF+WASH. Between 22 November 2012 and 27 March 2015, pregnant women living in these clusters were enrolled following written informed consent. VHWs delivered treatment group-specific behavior change interventions and commodities during 15 home visits, between enrolment and 12 months postnatal. Between 13–17 months, VHWs made monthly visits to provide routine care, deliver intervention commodities in active groups, and encourage participants to practice behaviors relevant to their study arm. At 18 months a review module, which reiterated key messages, was implemented in all arms. Commodities and behavior change communication messages delivered in the four treatment groups were: SOC: promotion of exclusive breastfeeding to 6 months [49], and uptake of maternal-child health services. WASH: SOC interventions plus a ventilated improved pit (VIP) latrine, two handwashing stations, plastic mat and play yard (North States, Minneapolis, MN); monthly delivery of soap and chlorine solution (WaterGuard, Nelspot, Zimbabwe) with promotion of safe disposal of feces, handwashing with soap, protection of infants from geophagia, chlorination of drinking water, and hygienic preparation of complementary food. IYCF: SOC interventions plus 20g small-quantity lipid-based nutrient supplement (SQ-LNS) to be fed to the infant daily from 6–18 months; education and counseling to feed the infant nutrient-dense locally available food, feeding during illness, and dietary diversity. WASH+IYCF: All SOC, WASH, and IYCF interventions. A latrine was constructed in SOC and IYCF arms after completion of the trial. Due to the nature of the interventions, masking was not possible. Research nurses made two home visits during pregnancy and five postnatal visits at infant ages 1, 3, 6, 12, and 18 months. At baseline, maternal education and age, household wealth, access to water and sanitation, and food insecurity were assessed; mothers were tested for HIV via rapid test algorithm and HIV-positive women were urged to seek immediate care for prevention of mother-to-child transmission. Infant birth date, weight, and delivery details were transcribed from health facility records; the trial provided Tanita BD-590 infant scales to all health institutions in the study area and trained facility staff. Gestational age at delivery was calculated from the date of the mother’s last menstrual period. Given the household-based trial interventions, visits were not conducted if the mother had moved from the household where she consented, except for the 18-month visit (trial endpoint) when follow-up was conducted anywhere within Zimbabwe. Indicators of uptake of the interventions were collected at all visits and reported here for the 12-month postpartum visit. Part-way through the trial (from mid-2014 onwards) mother-infant pairs were invited to join a substudy to investigate biomarkers of EED. Women were informed about the EED substudy at their 32-week gestation visit and those with live births were enrolled at the 1 month postnatal visit, or as soon as possible thereafter. We pre-specified that primary trial inferences would be based on findings among infants born to mothers who were HIV-negative during pregnancy, because of the likely impact of HIV exposure and infant cotrimoxazole prophylaxis on underlying causal pathways [33]. Results from HIV-unexposed children are reported in this paper; EED results among HIV-exposed infants will be reported separately. From children in the EED substudy, additional specimens were collected at each postnatal visit, including stool (passed on the morning of the research visit and collected by the mother into a plain container) and blood, collected by venepuncture into an EDTA tube. In the field laboratory, EDTA tubes were centrifuged to collect plasma, which was stored at -80°C. Stool specimens were transported in a cold box to the field laboratory, aliquoted into plain tubes and stored at -80°C. Samples were subsequently transferred to the Zvitambo Laboratory in Harare for long-term storage at -80°C until analysis. We selected a range of biomarkers that would characterize the domains of EED. These were previously described, together with the rationale for each [33], with the following modifications since publication of that methods paper [50]. First, plasma citrulline was added as a biomarker of small intestinal damage, since lower circulating concentrations reflect reduced enterocyte mass [51]. Second, EndoCAb was dropped as a measure of microbial translocation due to technical problems with the commercial assay [50]. Third, kynurenine:tryptophan ratio was added as an emerging marker of inflammation in children with EED [37], in place of alpha-1-acid glycoprotein which had a high coefficient of variation in our laboratory. Lactulose-mannitol testing was undertaken at all time-points except 1 month of age, because of concerns about interrupting early exclusive breastfeeding. The test has previously been described in detail elsewhere [33]. Briefly, after a 30-minute fast infants ingested 2mL/kg body weight (maximum 20mL) of a sterile solution containing 250 mg/mL lactulose and 50 mg/mL mannitol. A urine bag was placed, and all urine passed over a 2-hour period was collected, preserved with chlorhexidine and transported in a cool box to the field laboratory, where it was measured, aliquoted into plain tubes and stored at -80C. The final list of EED biomarkers and their respective domains are summarized in Table 1. All assays were undertaken by laboratory scientists masked to the trial intervention arm. Plasma samples were tested at the Zvitambo laboratory by Enzyme Linked Immunosorbent Assay (ELISA) according to manufacturers’ instructions for CRP (limit of detection (LOD) 0.01ng/mL), soluble CD14 (LOD 125pg/mL), IGF-1 (LOD 0.026ng/mL) (all from R&D Systems, Minneapolis, MN, USA); and I-FABP (LOD 47pg/mL); Hycult Biotechnology, Uden, The Netherlands. Stool samples were tested at the Zvitambo laboratory by ELISA according to manufacturers’ instructions for neopterin (LOD 0.7nmol/L; GenWay Biotech Inc, San Diego, CA, USA), myeloperoxidase (LOD 1.6ng/mL; Immundianostik, Bensheim, Germany), A1AT (LOD 1.5ng/mL; BioVendor, Brno, Czech Republic), and REG‐1β (LOD 0.625ng/mL; TECHLAB Inc, Blacksburg, VA, USA). Plasma citrulline (LOD 100ng/mL), kynurenine (LOD 40ng/mL), and tryptophan (200ng/mL) were assayed by ultrahigh-performance liquid chromatography tandem mass spectrometry with electrospray ionization (Waters, Wilmslow, U.K.) at Imperial College, London. Urine samples were tested on a Shimadzu Prominence liquid chromatograph with a Restek Ultra Amino 3μm 150/2.1mm column and tandem Sciex QTRAP5500 mass spectrometer with a Turbo V ion source and TurboIonSpray probe at Pain Care Specialists of Oregon, USA. Both analytes were quantified against lab-made calibrators at ng/mL levels of 3250, 2500, 1250, 750, 350, 100, 10, and 1. The calibrator was confirmed each day by lab-made quality control samples at 150 and 750ng/mL. Lactulose for calibrators and control were obtained from Spectrum Chemical (LOD 1ng/mL). Mannitol was obtained from Tokyo Chemical Industry (LOD 1ng/mL). Both internal standards were obtained from Sigma-Aldrich. Citrulline and KTR were not measured at 18 months (no funding for samples). We also calculated the environmental enteropathy score (EE score) proposed by Kosek et al. [52], as a composite index reflecting the severity of intestinal enteropathy and treated it as a continuous variable in our analyses. The substudy was based on a sample size of 250 children per trial arm with longitudinal assessments of EED, after allowing for missed samples and loss to follow-up. Assuming an average of 2–3 infants per cluster, type I error of 5%, and a coefficient of variation across clusters of 0.25 provides >80% power to detect a difference of at least 0.18 standard deviations between WASH and non-WASH arms, or between IYCF and non-IYCF arms. All analyses were intention-to-treat at the child level. Independent intervention effects of IYCF and WASH, as well as IYCF-by-WASH interaction effects, were evaluated by fitting separate regression models with each biomarker as the dependent variable. Independent variables included IYCF and WASH coded as dummy variables and an IYCF-by-WASH interaction term. If the interaction was not significant (p>0.05) and if the interaction term was <0.25 standard deviations in absolute magnitude, main intervention effects were tested to compare IYCF to non-IYCF and WASH to non-WASH groups in separate models. Except for citrulline and EE score, tobit regression was used to account for unobserved biomarker values below the LOD using the package AER [53], and sandwich standard error estimation was used to account for cluster membership using the sandwich package [54], both in R version 3.5.3. For citrulline and EE score, no infants were below the LOD, so linear regression models were fitted by generalized estimating equations (GEE), with an exchangeable correlation to account for cluster membership, using the geepack package [55]. All biomarker values were natural log-transformed, except for EE score because of zero values. A separate model was fitted for each study visit. For each biomarker, the Holm method was used to account for multiple comparisons across study visits. The robustness of model results to influential observations was assessed by re-fitting each model after 95% winsorization of log-transformed biomarker values or raw EE scores [56]. Regression coefficients are reported as the ratio of biomarker concentrations between intervention and control arms, or as the difference between arms for EE scores. To graphically present the longitudinal biomarker trajectories in each intervention group, we fitted generalized additive models of biomarker concentration against infant age with the use of cubic splines with 3 knots for smoothing. Women gave written informed consent to join the trial and additional consent for their infant to join this substudy. Both the SHINE trial and this substudy were approved by the Medical Research Council of Zimbabwe and the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health.