1. Point-of-Use Drinking Water Treatment in Cambodia: A Randomized, Controlled Trial of Locally Made Ceramic Filters Joe Brown University of Alabama, New College | Departments of Biological Sciences and Civil and Environmental Engineering Mark Sobsey University of North Carolina School of Public Health, Department of Environmental Sciences and Engineering 2008 International Symposium on Household Water Management

2. Outline Overview of ceramic filters Study design Data collection and analysis Results Brief discussion

3. Ceramic Water Purifier (CWP) Potters for Peace design from Latin America Uses gravity-driven porous ceramic microfiltration, with rice husk burnout material Food-grade plastics, local materials and manufacture 10L 1-3 L/hr

5. Study overview Intended to independently assess the CWP (CWP1, CWP2) in field use in Cambodia via RCT A rural/peri-urban village on the Bassac River near Phnom Penh –Prek Thmey, Kandal –Criteria for village selection were proximity to lab, poor quality drinking water & no treatment, high diarrheal disease burden, emerging market for CWPs Just downstream from the Phnom Penh wastewater canal Education and training done with RDI, no messages about diarrhea to bias study – promote clean, beautiful, and tasty drinking water

6. The 2 interventions The CWP currently promoted in Cambodia by the NGO Resource Development International (CWP1) –Uses Ag coating –2-3 liter/hour flowrate –Sales of approximately 20,000 per year in Cambodia –Retail cost of US$7 (unskilled labor rate $1- $2.50/day) A modified version we developed in our lab using an alternative ceramic mixture, also made in the RDI factory (CWP2)

7. Data collection overview Identification of 300 households in Prek Thmey –NGO survey Recruitment of 180 eligible households in cluster- randomized order –Criteria: child under 5, willingness to participate, never owned CWP, no bottled water Baseline data collection: WSH, water quality, covariates Randomization to 1 of 3 groups: CWP1, CWP2, control Follow all 180 households for 22 weeks –Water quality data, including: E. coli/100ml from untreated and treated household water samples –Scoring for diarrheal disease for all family members based on 7-day recall, and other health data –Other detailed data on filter use, WSH, and other factors

8. The study

9. Determine eligibility of village and households based on NGO-led community survey

10. The study Map all houses, work with community leaders, WQ survey

11. The study 7 9 4 1 8 6 11 5 1 3 2 10 1 12 Visiting eligible households in cluster-randomized order

12. The study Baseline data collection: 2 visits over 4 weeks

13. The study Randomization to one of three groups of 60 households each: CWP1, CWP2, control CWP1 CWP2 CWP1 CWP2 CON

14. The study Follow for 18 weeks with biweekly follow up CWP1 CWP2 CWP1 CWP2 CON

15. Prek Thmey

21. Health data analysis Data on diarrheal disease longitudinal prevalence (7 day recall, binary) computed for each group using Poisson extension of GEE: log-risk regression using longitudinal data Stratified estimates of longitudinal prevalence ratios Controlling for clustering within households and in individuals over time Controlling for other variables using an a priori 10% change in effect estimate criterion Regression models applied to determine important predictors of diarrheal illness, including water quality

22. Results: overview CWP1 & CWP2 reduced E. coli concentrations in treated water by a mean 2.0 log 10 (99%) –Filter effectiveness up to 99.9999% –Similar to boiled water samples (98% reduction) Two-thirds of effluent water samples from filters were <10 E. coli/100ml (low risk) –Similar to data for stored boiled water Filters associated with a mean 40% reduction in diarrheal disease in users versus non-users –Positive but weak relationship between E. coli concentrations and diarrheal disease

23. E. coli in drinking water stratified by group

24. Health effects: diarrheal disease CWP1, all persons: LPR = 0.51 (0.41-0.63) CWP1, under 5s: LPR = 0.58 (0.41-0.82) CWP2, all persons: LPR = 0.58 (0.47-0.71) CWP2, under 5s: LPR = 0.65 (0.46-0.93)

25. Diarrhea LP, 7-day recall

26. Other associations Diarrheal disease and measured covariates

27. Principal findings Use of the CWP1 or CWP2 can reduce diarrheal prevalence by approximately 49% and 42% respectively A weak but positive association was observed between diarrhea and increasing levels of E. coli in drinking water, consistent with other studies (data not shown) Diarrheal disease outcomes also associated with sanitation, hygiene, socio-economic measures No significant difference was observed between filters but study was not powered to sort this out

28. Study limitations Not blinded: subject to bias in health data collection Does not capture full seasonality –No data from January/February Little data on long term effectiveness –This was the focus of a previous study we did No data on actual human pathogens

29. Acknowledgements Uon Virak, Choun Bunnara, Lim Kimly, Michelle Molina, Oum Sopharo, Song Kimsrong, and Van Sokheng for conducting interviews and analyzing water samples Financial support: US Environmental Protection Agency and the National Science Foundation Laboratory space and logistical support was generously provided by Resource Development International-Cambodia Mickey Sampson Jan-Willem Rosenboom Ministry of Rural Development, Kingdom of Cambodia Douglas Wait and the UNC Environmental Microbiology and Health group

30. Mean diarrhea prevalence proportion over 18 week intervention period a Prevalence proportion ratio (PPR) (95% CI) ControlCWPCWP-2CWPCWP-2 All persons0.150.0740.0900.51 (0.41-0.63)0.58 (0.47-0.71) Age d <5 years 5-15 years ≥16 years 0.23 0.13 0.12 0.14 0.079 0.045 0.19 0.078 0.091 0.58 (0.41-0.82) 0.62 (0.43-0.90) 0.37 (0.26-0.52) 0.65 (0.46-0.93) 0.48 (0.31-0.75) 0.57 (0.42-0.76) Sex Male Female 0.12 0.17 0.076 0.072 0.081 0.096 0.61 (0.44-0.83) 0.44 (0.33-0.58) 0.60 (0.43-0.83) 0.57 (0.44-0.75) a. Nine sampling rounds, June-October 2006. Figures represent the proportion of individuals reporting diarrhea in the previous 7 days. b. The PPR was computed via Poisson extension of generalized estimating equations (GEE), adjusting for clustering of the outcome within households and within individuals over time. c. 95% confidence interval. d. Age in years at the time of the first household visit.

31. Mean dysentery prevalence proportion over 18 week intervention period a Prevalence proportion ratio (PPR) (95% CI) ControlCWPCWP-2CWPCWP-2 All persons0.0120.00470.0110.39 (0.20-0.77)0.95 (0.55-1.7) Age d <5 years 5-15 years ≥16 years 0.025 0.0082 0.0095 0.0079 0.0061 0.0027 0.017 0.012 0.0083 0.27 (0.091- 0.85) 0.52 (0.10-2.7) 0.29 (0.11-0.80) 0.82 (0.35-1.9) 1.5 (0.40-5.5) 0.87 (0.37-2.0) Sex Male Female 0.0074 0.017 0.0044 0.0049 0.0098 0.012 0.49 (0.17-1.5) 0.31 (0.13-0.73) 1.2 (0.50-2.9) 0.68 (0.32-1.4) a. Nine sampling rounds, June-October 2006. Figures represent the proportion of individuals reporting diarrhea in the previous 7 days. b. The PPR was computed via Poisson extension of generalized estimating equations (GEE), adjusting for clustering of the outcome within households and within individuals over time. c. 95% confidence interval. d. Age in years at the time of the first household visit.

32. Results: water quality Method Water sourcenMean untrtd (log 10 units) Mean treated (log 10 units) LRV mean 95% CILRV std dev LRV variance CWP1 All4853.61.52.12.0-2.21.21.4 Rain water3683.51.42.12.0-2.21.21.4 Surface water1023.41.42.11.9-2.31.31.6 Well water773.91.52.42.2-2.61.0 Other/not known0------ CWP2 All4963.31.32.01.9-2.11.11.2 Rain water3273.31.32.01.9-2.11.11.2 Surface water1163.11.02.11.9-2.31.01.1 Well water1093.31.32.01.8-2.21.11.2 Other/not known0------ Boiling All2823.51.51.91.7-2.01.31.7 Rain water1373.51.61.91.6-2.11.31.6 Surface water643.31.12.22.0-2.51.21.4 Well water743.31.81.51.2-1.81.31.8 Other/not known593.51.52.01.7-2.41.31.7