Long-term impacts of an urban sanitation intervention on enteric pathogens in children in Maputo city, Mozambique: study protocol for a cross-sectional follow-up to the Maputo Sanitation (MapSan) trial five years post-intervention

Abstract

Background We previously assessed the effect of an onsite sanitation intervention in informal neighborhoods of urban Maputo, Mozambique on enteric pathogen detection in children after two years of follow-up (Maputo Sanitation [MapSan] trial, clinicaltrials.gov: NCT02362932). We found significant reductions in Shigella and Trichuris prevalence but only among children born after the intervention was delivered. In this study, we assess the health impacts of the sanitation intervention after five years among children born into study households post-intervention. Methods We are conducting a cross-sectional household study of enteric pathogen detection in child stool and the environment in the 16 MapSan study neighborhoods at compounds (household clusters sharing sanitation and outdoor living space) that received the pour-flush toilet and septic tank intervention at least five years prior or meet the original criteria for trial control sites. We are enrolling at least 400 children (ages 29 days - 60 months) in each treatment arm. Our primary outcome is the prevalence of 22 bacterial, protozoan, and soil transmitted helminth (STH) enteric pathogens in child stool using the pooled odds ratio across the outcome set to assess the intervention effect. Secondary outcomes include the individual pathogen detection prevalence and gene copy density of 27 enteric pathogens (including viruses); mean height-for-age (HAZ), weight-for-age (WAZ), and weight-for-height (WHZ) z-scores; prevalence of stunting, underweight, and wasting; and the 7-day period-prevalence of caregiver-reported diarrhea. All analyses are adjusted for pre-specified covariates and examined for effect measure modification by age. Environmental samples from study households and the public domain are assessed for pathogens and fecal indicators to explore environmental exposures and monitor disease transmission. Discussion We are using stool-based detection of multiple enteric pathogens as an objective measure of exposure to evaluate an existing sanitation intervention. We propose a novel pooled estimate of the treatment effect across a pre-specified outcome set to summarize the overall impact of the intervention on enteric pathogen exposure as the primary trial outcome. Trial Registration This trial was prospectively registered on 16 March 2022 (ISRCTN86084138).

Long-term impacts of an urban sanitation intervention on enteric pathogens in children in Maputo city, Mozambique: study protocol for a cross-sectional follow-up to the Maputo Sanitation (MapSan) trial 5 years postintervention

“…The mixed effects models will be specified as Bayesian hierarchical models with regularising hyperpriors (see data analysis plan for discussion of prior distributions) and sampled using Markov chain Monte Carlo approaches 50 93 96. For binary enteric pathogen detection outcomes, the canonical logistic link function will be used for computational stability and a Bayesian parametric g-formula algorithm will be applied to estimate marginal prevalence ratios (PR) and differences (PD) from the posterior predictive distribution 97.…”

“…The posterior predicted PR and PD distributions for each pathogen will be approximated by the ratio and difference, respectively, of the posterior predicted probability draws under the all-treated and none-treated scenarios. Differences in average gene copy density (scaled by pathogen-specific sample standard deviation to facilitate comparison across pathogens) will be estimated as the measure of effect using linear Bayesian hierarchical models for semicontinuous enteric pathogen quantity outcomes, with non-detects considered true zeros and censoring used to create a zero class (as in Tobit regression; refer to data analysis plan for implementation details) 96 99. As a sensitivity analysis, we will also implement the two-stage parametric g-formula approach of Rogawski McQuade et al to estimate differences in average quantity separately for each pathogen, applying the Benjamini-Hochberg procedure to account for multiple comparisons 11.…”