Health Stream Article - Issue 56 - December 2009

Endemic Waterborne Disease Study

A recently published study from the US has found that point-of-use water treatment reduced the incidence of gastrointestinal illness in people aged 55 years and older by 12% although the tap water supply met current US water quality standards (1). The results of the study once again raise the issue of endemic waterborne disease due to pathogens in treated drinking water, and the adequacy of current methods of defining drinking water safety.

Endemic disease refers to the ongoing 'background' occurrence of illness in a population over time in contrast to the short peaks of disease associated with point source outbreaks. A range of surveys and research studies have estimated that the background rate of gastroenteritis in developed nations is around 0.5 to 1.0 episodes of illness per person per year on average (2). Such illnesses are assumed to be predominantly of infectious origin and to arise from pathogens transmitted by the faecal-oral route. Possible modes of transmission include drinking water, food, recreational water, person-to-person contact, animal-to-person contact and faecal contamination of environmental surfaces. The number of cases of illness associated with endemic gastroenteritis far outweighs cases attributed to recognised outbreaks.

The degree to which treated drinking water meeting prevailing quality standards contributes to endemic gastrointestinal illness has been a matter of debate and research for more than two decades. This debate was stimulated by increasing recognition that viruses and protozoa from source waters could still be present and potentially infectious in treated water which met conventional standards for absence of coliform bacteria. In addition to the hypothesis of infection due to faecal pathogens originating from source waters and not adequately removed by treatment, alternative or additional hypotheses have also been advanced relating to ingress of faecal pathogens in the distribution system, growth of opportunistic pathogens in the distribution system, and accumulation of pathogens in biofilms followed by sloughing events releasing them into the water.

A number of epidemiological studies of various designs have addressed this issue (3). The methodologically strongest studies have been randomised trials which used point-of-use devices to provide additional treatment to tap water for one group of consumers who were then compared to another group drinking conventional tap water. Prior to the most recent study described in this article, two randomised studies had been conducted in Canada, one in Australia and one in the US. The Canadian studies suggested a significant contribution to community gastroenteritis from waterborne pathogens, however these studies were unblinded, with participants being aware of whether or not they received a water quality intervention. This may have influenced the perception or reporting of illness differentially between the groups.

The Australian study and the US study used a blinded design where households were supplied with real or sham water treatment devices, and both participants and researchers were unaware of the water treatment status of each household until after the results had been analysed. Both studies found no evidence of waterborne gastrointestinal disease, with a non-significant 1% to 2% difference in disease rates between real and sham treatment groups. The Australian study was conducted in Melbourne in a system with a fully protected surface catchment, high quality source water and minimal treatment (disinfection only). The US study was conducted in Davenport, Iowa in a system with a poor quality surface water source and well operated conventional filtration and disinfection. All of these previous intervention trials focused on family groups containing children, as gastroenteritis is most common among young children and thus inclusion of this age group provides greater statistical power to detect differences in disease rates. In contrast, the most recent US study applied the randomised double-blinded design to examine the effect of point-of-use water treatment on gastrointestinal illness rates in adults aged 55 years and over.

The study participants comprised 988 individuals who were resident in 714 households in Sonoma County, California. Some of the participants were recruited from an existing cohort study of physical performance and age-related changes, while others were recruited by mail from a random sample of county residents meeting the study age requirements. Households were excluded if they included any employee of the Sonoma County Water Agency or any other water district, if any member was unwilling to sign a consent form for installation of the water treatment device, if they included any immuno-compromised individuals, or if they consumed less than 75% of their in-home drinking water from the household tap. The water treatment device was a counter-top model with a design similar to those used in the previous Australian and US randomised trials. The active devices comprised a 1 micron absolute filter (rated to remove protozoal cysts) and a UV unit (rated to inactivate bacteria and viruses). The sham devices were identical in outward appearance but did not affect the water quality.

The water supply for Sonoma County is derived from the Russian River which is subject to urban discharges and agricultural run-off. To obtain high quality water, a bank filtration system is used to draw water indirectly from the river. Water is pumped from a series of six collector wells drilled into the aquifer adjacent to the river. The water is naturally filtered as it is drawn from the river through layers of sand and gravel to the base of the wells which extend 50-60 feet below the level of the streambed. The water is then chlorinated before distribution to several towns through a system of underground aqueducts and storage tanks. The paper does not discuss water quality parameters, however media articles reporting the outcome of the study note that no positive results were found in the water supply during the 2008/09 year from 477 tests for E. coli and total coliforms. In the previous US study in Davenport a major water quality monitoring study was undertaken in parallel with the epidemiological study, however it appears that no additional water quality monitoring was conducted in association with the Sonoma trial.

The study was conducted from 2001 to 2005, with individual households participating for two 26-week periods with a 2-week washout period in between. A crossover design was used with households receiving the opposite type of water treatment unit in the second 26-week period from the type assigned in the first period (i.e. real followed by sham, or sham followed by real). This design provides increased statistical power as it allows individuals to be compared with themselves as well as comparing the real treatment group with the sham treatment group. Participants recorded gastrointestinal symptoms in health diaries. The primary health outcome measured was highly credible gastrointestinal illness (HCGI), defined as any of the following four conditions, preceded by at least 6 HCGI-free days: (i) vomiting, (ii) watery diarrhoea, (iii) soft diarrhoea and abdominal cramps, and (iv) nausea and abdominal cramps. Daily longitudinal prevalence of HCGI, defined as total days of illness divided by total days in the study, was used as the secondary outcome in statistical analysis.

A quality assurance review in the middle of the study (conducted by an unblinded researcher not involved with participants or data analysis) revealed that a batch of 180 water treatment units had been wrongly labelled during manufacture. This error resulted in 157 households having the same type of treatment device throughout the study, rather than crossing over to the opposite type as intended for the second observation period. It was decided to exclude data from these households from analysis and the enrolment period was therefore extended to compensate for this data loss. The final sample consisted of 557 households (770 individuals) and 83% of these completed both observation periods. The dropout rate was not significantly different between the two device types. About 10% of participants were aged 85 years or older at enrolment while the remainder were fairly evenly distributed between 55-64, 65-74 and 75-84 age groups. The majority of people reported themselves to be in excellent (32%) or good health (54%) at enrolment.

A generalised estimating equation (GEE) was used to estimate the effect of real versus sham devices on health outcomes over the whole study population and generalised linear mixed models (GLMMs) were used to analyse the effects in individuals. The incidence rate ratios for HCGI for active versus sham treatment were 0.88 (95% CI 0.77-1.00) and 0.85 (95%CI 0.76-0.94) for GEE and GLMM models respectively. In other words, there was a 12 to 15% lower rate of reported HCGI (after adjustment for relevant covariates) when people were using real water treatment devices compared to sham devices. The results of analysis for the secondary outcome, daily longitudinal prevalence of HCGI, were similar.

The gastrointestinal illnesses experienced by study participants appeared generally to be mild with an average of less than 2 days of work missed per year due to such illness, and less than one medical consultation per 3 years on average for gastrointestinal illness. A total of 13 deaths occurred during the study (7 while using active water treatment devices and 6 while using sham devices) but none were attributed to gastrointestinal illness. One instance of hospitalisation due to infectious gastroenteritis was recorded (type of device not given) out of 69 hospitalisations in total (35 while using active device, 34 while using sham device).

There were no significant differences in the amounts of water reportedly consumed from real or sham devices, and participants were effectively blinded to the types of device assigned to them during the study. The number of reported episodes of HCGI declined in the second cycle compared to the first cycle. This has been reported in previous intervention studies and may represent a decline in enthusiasm by participants. However this effect was similar regardless of the order of device assignment (i.e. real-sham or sham-real). When sub-groups were analysed separately it was found that the effect of real water treatment devices was different between the genders, with a protective effect apparent only in men; relative risk of 0.76 (95%CI 0.60-0.95) for real versus sham treatment devices in men and relative risk of 0.99 (95%CI 0.85-1.17) for real versus sham treatment devices in women using adjusted GEE analysis. There were no clear trends in risk levels relative to age group, total daily water consumption, length of time in the study or employment status.

Overall this was a well-conducted study with effective randomisation, a low dropout rate and successful blinding. The apparent protective effect of real water treatment devices on the study population as a whole was modest (about 12%) and in some analysis models reached statistical significance at the p=0.05 level. However, the fact that the protective effect was confined to men is certainly puzzling. This observation implies that a proportion of HCGI experienced by men in the study was attributable to pathogens in tap water, but HCGI occurrence in women was not significantly influenced by the same tap water. There is some evidence of gender differences in occurrence of gastrointestinal diseases however the available data seem to suggest that women are more susceptible than men rather than less susceptible. Surveys in several countries have indicated that women report diarrhoeal illness more frequently than men. US studies have also indicated women are more likely than men to be hospitalised or to die from such illnesses. A review of waterborne outbreaks in developed nations also noted that women predominate among identified cases and fatalities, although data were limited and the imbalance may be due to chance (4). There seems to be no obvious reason to expect that older women would be less susceptible than older men to infection by the range of enteric pathogens potentially transmitted by a chlorinated drinking water supply such as that in Sonoma County.

(1) Colford JM Jr, Hilton JF, Wright CC et al. (2009) The Sonoma Water Evaluation Trial: A Randomized Drinking Water Intervention Trial to Reduce Gastrointestinal Illness in Older Adults. American J Public Health 99(11):1988-1995.
(2) The calculated rate of gastroenteritis depends on the symptom definition. Usually a single episode of diarrhoea or vomiting is not considered to constitute gastroenteritis.
(3) See Health Stream Issue 43 article US Waterborne Disease Estimate for a brief summary of a Journal of Water and Health 2006 Supplement which reviewed studies on endemic waterborne disease.
(4) Safe Drinking Water: Lessons from Recent Outbreaks in Affluent Nations By SE Hrudey and EJ Hrudey. ISBN: 1 84339 042 6 IWA Publishing.

© Copyright Water Quality Research Australia Limited http://www.wqra.com.au/
Health Stream articles may be reproduced and communicated to third parties provided WQRA is acknowledged as the source. Literature summaries are derived in part from copyright material by a range of publishers. Original sources should be consulted and acknowledged.