Chlorination

Study arms included (1) water treatment: chlorination with sodium dichloroisocyanurate (NaDCC) tablets coupled with safe storage in a narrow-mouth lidded vessel with spigot, (2) sanitation improvements: upgrades to concrete-lined double-pit latrines and provision of child potties and sani-scoops for feces disposal, (3) handwashing promotion: handwashing stations with a water reservoir and a bottle of soapy water mixture at the food preparation and latrine areas, (4) combined water treatment, sanitation and handwashing (WSH), (5) nutrition improvements including exclusive breastfeeding promotion (birth to 6 months), lipid-based nutrient supplements (6–24 months), and age-appropriate maternal, infant, and young child nutrition recommendations (pregnancy to 24 months), (6) nutrition plus combined WSH (N+WSH), and (7) a double-sized control arm with no intervention (Fig 1). Further details of the interventions have been provided elsewhere [22 (link)].
The WSH interventions aimed to reduce children’s early-life exposure to fecal pathogens. Bangladeshi households are clustered in compounds shared by extended families; in our study, the compound containing the household where the index child lived (“index household”) had an average of 2.5 households (range: 1–11). The interventions targeted the compound environment as we expected this to be the primary exposure domain for young children [27 (link)]. Interventions were delivered at index child, index household and compound levels (Fig 2). The nutrition intervention targeted index children only. The water and handwashing interventions were delivered to the index household. The sanitation intervention provided upgraded latrines, potties and scoops to all households in the compound; as the shared compound courtyard serves as play space for children, we aimed to improve sanitary conditions in this environment with compound-level latrine coverage. Because of the eligibility criterion of having a pregnant woman, enrolled compounds represented approximately 10% of compounds in a given geographical area; as such, we did not provide exclusive community-level latrine coverage.
The delivery of interventions was initiated around the time of index children’s birth. Local women hired and trained as community health promoters visited intervention arm participants on average six times per month to deliver intervention products for free, replenish the supply of consumables (chlorine tablets, soapy water solution, nutrient supplements), resolve hardware problems and encourage adherence to the targeted WSH and nutrition behaviors; health promoters did not visit control arm participants (S4 Text). The health promotion visits and supply of consumable intervention products spanned the full study duration, including the period of the STH assessment. All interventions had high user adherence throughout the study as measured by objective indicators (S4 Text). Further details of intervention adherence have been previously reported [28 (link)–30 (link)].

Our primary outcome was the seven-day period prevalence of caregiver-reported diarrhea in index children (6–18 mo at enrollment). We conservatively sized the study to detect a difference in the two-day prevalence of diarrhea due to safe storage plus chlorination over safe storage alone. We assumed 14% two-day diarrhea prevalence in the control group based on data from a large-scale study in rural Bangladesh (SHEWA-B) [40 (link)], 11.6% prevalence in the safe storage group based on 30% diarrhea reduction due to safe storage [30 (link)] and 55% of participants storing water in the home [12 (link)], and 9.1% prevalence in the combined intervention group based on 35% diarrhea reduction due to safe storage plus chlorination [29 (link)]. Assuming one child of eligible age per household, an intra-cluster correlation coefficient (ICC) of 0.13 for repeated observations within a child based on the SHEWA-B study, 5% drop-out and a one-sided α of 0.05, we calculated that 575 participants visited five times would provide 84% power to detect the difference between 11.6% and 9.1% diarrhea prevalence. We enrolled 600 households in each study arm; we conducted five visits during the dry season and five additional visits during the monsoon season to ensure sufficient power to individually detect a health difference in either season.
We conducted all statistical analyses using STATA software (version 12.1, STATA Corp., College Station, TX). We calculated disease prevalence ratios (PR) between pairs of study arms using generalized linear models with a log link, a binomial error distribution, and robust standard errors to account for clustering due to longitudinal sampling and multiple children per household when there was more than one eligible child [41 (link)]. In the case of loss to follow-up, all observed data for a given child prior to leaving the study were used in the analysis. We investigated effect modification by two pre-specified characteristics by including interaction terms in the regression models: season (dry vs. monsoon) and child age (6–12 mo, 13–18 mo and > 18 mo at enrollment). We calculated the ICC for repeated measures within children using one-way ANOVA analysis with the loneway function in STATA.
Our secondary outcome was fecal contamination of stored water, defined as the proportion of samples with an E. coli count exceeding the WHO thresholds of no risk, low risk and moderate risk. We compared stored as well as source water quality across study arms using chi square tests (or Fisher’s exact test in the case of sparse data) for the proportion of samples in these risk categories and conducted subgroup analyses with season.
All analyses were conducted by the original assigned groups in an intention-to-treat analysis. The complete data management process and statistical analyses for the primary and secondary outcomes were independently replicated by two investigators (AE and AMN) to ensure identical, replicable results. Investigators had no access to outcome data until field activities were complete. CONSORT guidelines were followed [42 (link)].

The rearing experiments described here were conducted at the Institute of Marine Research’s experimental facility in Bergen, Norway. This laboratory has a quarantine facility with permit issued by the Norwegian Food Safety Authority (NFSA) to conduct experiments on marine pathogens and includes chlorination of all waste-water to ensure that no pathogens are released to the natural environment. Lice were cultured using well-established rearing protocols for L. salmonis[30 (link)]. During the experiments Atlantic salmon were kept in running seawater and fed commercial fish feed ad libitum. This experiment was conducted in accordance with Norwegian legislation for use of animals in research, and was approved by the Norwegian Animal Research Authority (research permit nr. 2009/186329).
Lice were cultured by infecting Atlantic salmon (Salmo salar) with copepodids from Atlantic and Pacific strains of L. salmonis. The Atlantic strain (LsAtl) was established by mixing approximately equal numbers of copepodids from the previously described LsGulen and LsOslofjord strains [30 (link)]. The Pacific strain (LsPac) was established from copepodids derived from adult female L. salmonis collected at a commercial salmon farm located close to the town of Campbell River (British Columbia, Canada). The Pacific lice were transported to Norway in thermal flasks containing full strength salinity seawater. A permit to import these lice was obtained from the NFSA (Additional file 1). Unfertilized L. salmonis from the LsPac F1 and the LsAtl F2 generations were sorted according to sex at the pre-adult II stage and used immediately (LsAtl) or kept separate (LsPac) on previously uninfected fish until used in the crossing experiment.
To obtain hybrid strains we crossed LsPac females with LsAtl males and LsAtl females and LsPac males using a similar tank rearing system to that described by Hamre and Nilsen [31 (link)]. Briefly, each cross consisted of two (in one instance one) females and one male which were placed on a single uninfected Atlantic salmon that was isolated in its own tank. This allowed definite control of the parent material and offspring generation. As the unfertilized females were in different stages of development (pre-adult II and adults), the time required to obtain fertilized egg strings from the two hybrid strains were dissimilar. After 15 (LsPac females and LsAtl males) and 35 (LsAtl females and LsPac males) days, females bearing egg strings, and the males used to fertilize them in the single fish tanks, were harvested. These sampled adults were stored in individual tubes containing 100% ethanol, and were indexed so that both parents and egg strings could be subsequently matched. The egg strings from the sampled females were incubated in separate incubators after removal of approximately 3-5mm of both egg strings which was stored in 100% ethanol. The parents and the 3-5mm section of egg strings were genotyped (described below) to validate the crosses by parentage assignment before the resulting copepodids were used to infect previously uninfected Atlantic salmon to produce the LsAtlPac (copepodids from LsAtl females and LsPac males) and LsPacAtl (copepodids from LsPac females and LsAtl males) F1 hybrid strains. The F1 hybrid strains were reared separately in replicate fish tanks (Figure 1).
The F1 generation hybrids were allowed to develop, fertilize and reproduce naturally on Atlantic salmon in their respective tanks. Approximately 3 months after infection with F1 hybrid copepodids, eggstrings from F1 hybrid females that had been fertilized by F1 hybrid males were harvested and incubated in separate containers for LsAtlPac and LsPacAtl. Copepodids arising from these egg strings were thereafter used to infect groups of previously uninfected Atlantic salmon in order to establish the LsAtlPac F2 and LsPacAtl F2 hybrid generation in two replicate tanks for each strain. The resulting F2 hybrid generations were allowed to develop until the pre-adult stage to allow sex determination before the experiments were terminated. Numbers of adults contributing to each generation were recorded. In addition, the numbers of copepodids that were used to propagate the F2 generation was estimated by counting an aliquot, and the numbers of pre-adults harvested from the F2 generation upon termination of the experiment, were determined by counting and sex determining all the lice present.

Chlorination experiments were conducted in 10 mM phosphate buffer
(pH 7) in amber glass bottles at room temperature (23 ± 2 °C)
for 5–48 h. The commercial solution of sodium hypochlorite
was standardized by measuring the absorbance of the hypochlorite anion
at 292 nm (ε = 362 M^–1 cm^–1).^{23 (link)} The chlorination of cysteine, glutathione,
and p-phenolsulfonic acid was conducted individually
by applying 250 μM each compound and 0.5–2.5 mM initial
chlorine. The chlorination of the SRFA extract (5 mg/L dissolved organic
carbon, DOC) was performed by applying 5 mg/L as Cl₂ of
initial chlorine. The residual chlorine was measured using the DPD
colorimetric method²⁴ and quenched using
a 2-fold molar excess of Na₂S₂O₃.
The chlorinated solutions of cysteine and glutathione were directly
analyzed on SFC-QTOF without further enrichment, while the chlorinated p-phenolsulfonic acid and SRFA were analyzed after freeze-drying
enrichment.