Effects of single and integrated water, sanitation, handwashing, and nutrition interventions on child soil-transmitted helminth and giardia infections: A cluster-randomized controlled trial in rural Kenya

Background Helminth and protozoan infections affect more than 1 billion children globally. Improving water quality, sanitation, handwashing, and nutrition could be more sustainable control strategies for parasite infections than mass drug administration, while providing other quality of life benefits. Methods and findings We enrolled geographic clusters of pregnant women in rural western Kenya into a cluster-randomized controlled trial (ClinicalTrials.gov NCT01704105) that tested 6 interventions: water treatment, improved sanitation, handwashing with soap, combined water treatment, sanitation, and handwashing (WSH), improved nutrition, and combined WSH and nutrition (WSHN). We assessed intervention effects on parasite infections by measuring Ascaris lumbricoides, Trichuris trichiura, hookworm, and Giardia duodenalis among children born to the enrolled pregnant women (index children) and their older siblings. After 2 years of intervention exposure, we collected stool specimens from 9,077 total children aged 2 to 15 years in 622 clusters, including 2,346 children in an active control group (received household visits but no interventions), 1,117 in the water treatment arm, 1,160 in the sanitation arm, 1,141 in the handwashing arm, 1,064 in the WSH arm, 1,072 in the nutrition arm, and 1,177 in the WSHN arm. In the control group, 23% of children were infected with A. lumbricoides, 1% with T. trichiura, 2% with hookworm, and 39% with G. duodenalis. The analysis included 4,928 index children (median age in years: 2) and 4,149 older siblings (median age in years: 5); study households had an average of 5 people, <10% had electricity access, and >90% had dirt floors. Compared to the control group, Ascaris infection prevalence was lower in the water treatment arm (prevalence ratio [PR]: 0.82 [95% CI 0.67, 1.00], p = 0.056), the WSH arm (PR: 0.78 [95% CI 0.63, 0.96], p = 0.021), and the WSHN arm (PR: 0.78 [95% CI 0.64, 0.96], p = 0.017). We did not observe differences in Ascaris infection prevalence between the control group and the arms with the individual interventions sanitation (PR: 0.89 [95% CI 0.73, 1.08], p = 0.228), handwashing (PR: 0.89 [95% CI 0.73, 1.09], p = 0.277), or nutrition (PR: 86 [95% CI 0.71, 1.05], p = 0.148). Integrating nutrition with WSH did not provide additional benefit. Trichuris and hookworm were rarely detected, resulting in imprecise effect estimates. No intervention reduced Giardia. Reanalysis of stool samples by quantitative polymerase chain reaction confirmed the reductions in Ascaris infections measured by microscopy in the WSH and WSHN groups. Trial limitations included imperfect uptake of targeted intervention behaviors, limited power to detect effects on rare parasite infections, and that it was not feasible to blind participants and sample collectors to treatment status. However, lab technicians and data analysts were blinded to treatment status. The trial was funded by the Bill & Melinda Gates Foundation and the United States Agency for International Development. Conclusions Integration of improved water quality, sanitation, and handwashing could contribute to sustainable control strategies for Ascaris infections, particularly in similar settings with recent or ongoing deworming programs. Combining nutrition with WSH did not provide further benefits, and water treatment alone was similarly effective to integrated WSH. Our findings provide new evidence that drinking water should be given increased attention as a transmission pathway for Ascaris. The trial protocol and detailed methods are published [28]. The trial was registered at ClinicalTrials.gov, identification number: {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01704105″,”term_id”:”NCT01704105″}}NCT01704105. The study protocol was approved by the Committee for Protection of Human Subjects at the University of California, Berkeley (protocol number 2011-09-3654), the Institutional Review Board at Stanford University (IRB-23310), and the Scientific and Ethics Review Unit at the Kenya Medical Research Institute (protocol number SSC-2271). Innovations for Poverty Action enrolled participants, implemented the intervention delivery, and collected the data. Mothers provided written informed consent for themselves and their children. Clusters of eligible pregnant women were randomized by geographic proximal blocks into 1 of 8 study arms: water treatment (chlorine treatment of drinking water); improved sanitation(provision of toilets with plastic slabs and hardware to manage child feces); handwashing with soap; combined water treatment, sanitation, and handwashing (WSH); improved nutrition (infant and young child feeding counseling plus small-quantity lipid-based nutrient supplements [LNSs]); combined WSH and nutrition (WSHN); a double-sized active control; and a passive control. The trial included a passive control arm to test if promoter visits alone (active control) had an effect on the trial’s primary outcomes diarrhea and growth; children in the passive control arm were purposively excluded from parasitology measurement (Fig 1). STH, soil-transmitted helminth. We conducted a cluster-randomized trial because there could have been behavior and infectious disease interactions between neighboring households. Villages were eligible for selection into the study if they were rural, the majority of the population lacked access to piped water supplies, and there were no other ongoing WSH or nutrition programs. Within selected villages, a census was conducted to identify eligible pregnant women in their second or third trimester who planned to continue to live at their current residence for the next year. Since interventions were designed to reduce child exposure to pathogens through a cleaner environment and exclusive breastfeeding, we enrolled pregnant women to allow time for intervention delivery to occur prior to or as close to birth as possible. After the census, clusters were formed from 1–3 neighboring villages and had a minimum of 6 pregnant women per cluster after the enrollment survey (each village could only be assigned to 1 cluster). Enrolled study compounds were thus a small proportion of the total number of compounds residing in each cluster. Children born to enrolled pregnant women were considered “index” children. We measured parasite infections approximately 27 months post-enrollment (which equates to a minimum of 24 months of intervention exposure since intervention hardware was delivered <3 months after enrollment). Outcomes were assessed among index children, including twins, as well as among 1 older child in the index child’s compound to understand the effect of the interventions on both preschool-aged and school-aged children. The older child was selected by enrolling the youngest available child within the age range of 3–15 years old, with priority for a sibling in the index child’s household. A survey at enrollment measured household socioeconomic characteristics and demographics (including maternal age, maternal education, electricity access, type of floor, and number of people in the household), as well as water, sanitation, and handwashing infrastructure and behaviors (including type of water source, reported water treatment, defecation location, type of toilet, and presence of water and soap at a handwashing station). In addition, at study enrollment we measured Giardia, Entamoeba histolytica, and Cryptosporidium spp. among children residing in study compounds between 18 and 27 months of age (the projected age range for index children at the end of the study) to assess baseline prevalence of these pathogens. STHs were not measured at enrollment among these proxy children because it was not logistically feasible to deworm infected children at baseline. We also collected 100-ml samples from primary drinking water sources accessed by study households and household stored drinking water (if available). We transported the samples on ice to field labs and enumerated Escherichia coli in each sample by membrane filtration followed by culture on MI medium. A few weeks after enrollment, clusters were randomly assigned to intervention/control arms at the University of California, Berkeley, by an investigator independent of the field research team (BFA) using a random number generator. Groups of 9 geographically adjacent clusters were block-randomized into the 6 intervention arms, the double-sized active control arm, and the passive control arm (the passive control arm was not included in the parasite assessment). Participants and other community members were informed of their intervention/control group assignment after the baseline survey. Blinding (masking) of participants was not possible given the nature of the interventions. Data and stool sample collectors were not informed of cluster assignment, but could have inferred treatment status by observing intervention hardware. Lab technicians were blinded to intervention status. Two authors (AJP and JS) independently replicated the statistical analyses while blinded to intervention status. Intervention delivery occurred within 3 months after enrollment. In the water intervention arms (water treatment, WSH, and WSHN), community health promoters encouraged drinking water treatment with chlorine (liquid sodium hypochlorite) using either manual dispensers installed at the point of collection (community water source) in study villages or bottled chlorine provided directly to households every 6 months. In the sanitation intervention arms (sanitation, WSH, and WSHN), households in study compounds received new latrines, or existing latrines were upgraded and improved by installing a plastic slab that included a lid. All households in sanitation arm study compounds were provided with a child potty for each child <3 years as well as a “sani-scoop” to remove animal and human feces from the compound. Households were encouraged to use latrines for defecation and for disposal of child and animal feces. In the handwashing intervention arms (handwashing, WSH, and WSHN), study compounds were provided with 2 handwashing stations—near the latrine for handwashing after defecation and near the cooking area for handwashing before preparing food. Stations included dual foot-pedal-operated jerry cans that could be tipped to dispense either soapy water or rinse water. Households were responsible for keeping the stations stocked with rinse water, and community health promoters refilled soap regularly. In the nutrition intervention arms (nutrition and WSHN), small-quantity LNSs were provided to children 6–24 months of age. Children received monthly rations of LNSs for addition to their other food twice per day. Nutrition messaging included promoting dietary diversity during pregnancy and lactation, early initiation of breastfeeding, exclusive breastfeeding at age 0–6 months, continued breastfeeding through age 24 months, timely introduction of complementary foods, dietary diversity for child feeding, and child feeding during illness. Intervention delivery was at the cluster level for the water intervention (all compounds in villages assigned to the cluster had access to the chlorine dispensers), at the compound level for sanitation and handwashing (non-study compounds in the cluster did not receive handwashing stations or improved toilets), and at the child level for the nutrition intervention (only index children and their siblings under 24 months received LNSs). Community health promoters were nominated by mothers in the community and trained to provide intervention-specific behavior change activities and instructions on hardware use and provision of nutrition supplements. They were also trained to measure the mid-upper arm circumference of the index children to identify and provide referrals for potential cases of severe acute malnutrition. Each intervention consisted of a comprehensive behavior change package of key messages; visual aids in the form of flip charts, posters, and reminder cue cards; interactive activities with songs, games, or pledges to commit to practice target behaviors; and the distribution of arm-specific hardware, products, or supplements. Households in the active control group received visits from promoters to measure child mid-upper arm circumference and provide malnutrition referrals, but did not receive any intervention-related hardware or messaging. Promoters were instructed to visit households monthly. Key messages and promoter materials are available at https://osf.io/fs23x/. Adherence to the interventions was measured during unannounced household visits after 1 year and 2 years of intervention exposure (S1 Text). Stool samples were collected from index children and older children in sterile containers and transported on ice to the closer of 2 central field labs located in Kakamega and Bungoma. Field staff revisited households up to 3 times to collect stool samples. A. lumbricoides, T. trichiura, and hookworm eggs were immediately enumerated (same day) by double-slide Kato—Katz microscopy with 41.7-mg templates. Both slides created from each stool sample were counted by a trained parasitologist, and 2 different parasitologists counted each slide from the same sample. A supervisor with expertise in STH egg identification reviewed 10% of all slides, and any discrepancies were corrected. STH egg counts were averaged for analysis if both slides from 1 stool sample were positive; if 1 slide was negative, the count for the positive slide was used for analysis. Two aliquots of stool (1 mixed with ethanol) were transported on dry ice to the Eastern and Southern Africa Centre of International Parasite Control laboratory at the Kenya Medical Research Institute in Nairobi, Kenya, for further analysis. One aliquot was analyzed by monoclonal enzyme-linked immunosorbent assay (ELISA) (Giardia II, Alere International, Galway, Ireland) for the presence or absence of G. duodenalis cysts. Samples were measured by ELISA in duplicate; if there was a discrepancy between duplicates, the sample was rerun. DNA was extracted from the other aliquot (preserved in ethanol) for stool samples collected from children in the control, WSH, and WSHN groups. Four quantitative polymerase chain reaction (qPCR) assays were run in duplicate on each sample to detect the following targets: N. americanus, A. duodenale, T. trichiura, and A. lumbricoides (see S1 Text for further details) [30]. STH and Giardia infections were prespecified outcomes in the parent WASH Benefits trial prior to the start of data collection; see Fig 3 in Arnold et al. [28]. Parasite infections were measured after 2 years of intervention exposure. The main indicators of parasite infections were the prevalence of each individual STH infection, any STH infection, and Giardia infection among index and older children from the same compound. Additional indicators of parasite infections included intensity of Ascaris, Trichuris, and hookworm measured in eggs per gram (epg) of feces; intensity binary category of Ascaris infection, measured as low intensity (1–5,000 epg) or moderate/high intensity (>5,000 epg), following World Health Organization (WHO) cutoffs; prevalence of coinfection with 2 or 3 STHs; and prevalence of coinfection with Giardia and any STH. The trial’s original protocol included E. histolytica and Cryptosporidium spp. as additional protozoan endpoints. At enrollment, Giardia prevalence was 40% among 535 children 18–27 months old in study compounds, while Cryptosporidium spp. prevalence was 1% and E. histolytica prevalence was 0%. We determined that the extremely low baseline prevalence of E. histolytica and Cryptosporidium spp. made these trial endpoints futile due to limited statistical power, and since each required a separate assay on the ELISA platform, the study’s steering committee decided to not test for them at follow-up. All households in all clusters enrolled into the main trial were invited to participate in the measurement of parasite infections. The main trial was powered for a minimum detectable effect of 0.15 in length-for-age Z score and a relative risk of diarrhea of 0.7 or smaller for a comparison of any intervention with the double-sized control group, assuming a type I error (α) of 0.05 and power (1 − β) of 0.8, 10% loss to follow-up, and a 1-sided test for a 2-sample comparison of means (the main trial statistical analysis plan was later changed to employ 2-sided tests). This led to a planned design of 100 clusters per arm and 10 index children per cluster. Given this design and a single post-intervention measure, we estimated that the trial’s sample size would be sufficient at 80% power with a 2-sided α of 0.05 to detect a relative reduction of 18% in infection prevalence of any parasite (2-sided tests were planned due to a lack of evidence that all interventions would have a protective effect). Our minimum detectable effect calculations assumed 50% prevalence in the control arm, a village intraclass correlation (ICC) of 0.14, 2 children measured per enrolled household (index child plus an older sibling), and 70% successful stool collection and analysis. For perspective, this minimum detectable effect is much smaller than typical effect sizes reported in meta-analyses of the association between improved water, sanitation, and handwashing and helminth/protozoan infections (e.g., odds ratios between 0.46 and 0.58 for sanitation facilities and helminth infections) [2,31]. All statistical analyses and comparisons between arms (water treatment, sanitation, handwashing, WSH, nutrition, and WSHN compared to active control) were prespecified prior to unblinding of investigators, and the analysis plan was published with a time stamp on the Open Science Framework (https://osf.io/k2s47/). Replication scripts and data are also provided at the same link. Our alternative hypothesis for all comparisons was that group means were not equal (2-sided tests). We estimated unadjusted and adjusted intention-to-treat effect differences between study arms using targeted maximum likelihood estimation with influence-curve-based standard errors that treated clusters as independent units and allowed for outcome correlation within clusters [32,33]. Our parameters of interest for dichotomous outcomes were prevalence ratios (PRs) (prevalence in the intervention group divided by the prevalence in the control group). Our parameter of interest for helminth intensity was the relative fecal egg count reduction. We calculated the relative reduction using both geometric and arithmetic means. We did not perform statistical adjustments for multiple outcomes to preserve interpretation of effects and because many of our outcomes were correlated [34]. We estimated adjusted parameters by including variables that were associated with the outcome, to potentially improve the precision of our estimates. We prescreened covariates (S1 Text) to assess whether they were associated (p-value < 0.2) with each outcome prior to including them in adjusted statistical models [35]. We conducted subgroup analyses to explore effect modification on Ascaris and Giardia infection presence by the following factors: index child status (LNSs were given only to index children and siblings under 24 months), consumed deworming medicine in past 6 months (Ascaris only), consumed soil in past week (index children only), >8 people in compound, and time since defecation before stool collection. Statistical analyses were conducted using R version 3.3.2 (https://www.r-project.org).