Observational epidemiologic studies of endemic waterborne risks: cohort, case-control, time-series, and ecologic studies

Observational studies have assessed endemic waterborne risks in a number of countries. Time-series analyses associated increased water turbidity with increased gastroenteritis risks in several public water systems. Several cohort studies reported an increased risk of gastroenteritis in populations using certain public or individual water systems. Although several case-control studies found increased waterborne risks, they also found increased risks associated with other exposures. An increased risk of campylobacteriosis was associated with drinking untreated water from non-urban areas and some tap waters; other signiﬁcant risks included contaminated poultry and foreign travel. Increased risks of cryptosporidiosis and giardiasis were associated with drinking water in some populations; other risk factors included foreign travel, day care exposures, and swimming. These observational studies provide evidence that some populations may be at an increased risk of endemic or sporadic illness from waterborne exposures, but not all studies found an increased risk. Differences in waterborne risks may be due to differences in water quality. System vulnerabilities and contamination likely differed in the areas that were studied. The information from these studies may help inform estimates of waterborne illness for the US population but is inadequate to estimate a population attributable risk.


INTRODUCTION
For a discussion of the various study designs, readers are referred to a previous article (Craun et al. 2006a) in this special issue of the Journal of Water and Health.

COHORT STUDIES
Studies in Canada, France, and Norway assessed waterborne AGI risks. Studies in the United States and New Zealand assessed waterborne giardiasis risks. Participants were selected on the presence or absence of certain characteristics, a specific event, or their exposure status (e.g. chlorinated or unchlorinated drinking water, water source, or coliform bacteria levels in water). Disease incidence was determined during the follow-up period, and disease rates (i.e. rate difference (RD) and rate ratio or relative risk (RR)) were compared for the exposed and unexposed groups. The risk attributable to water may be computed if the study is not confounded or affected by one or more sources of bias (Rothman & Greenland 1986;Monson 1990).
Canada Raina et al. (1999) studied AGI incidence among 442 Southern Ontario families. For 195 of the families, at least one well-water sample was found to exceed the Ministry of Environmental standards for coliform bacteria. One contact person in each household completed a daily family diary describing illness events; AGI was defined as an episode in which diarrhea with or without vomiting was reported.
Interviewers also collected information about well-water consumption, water treatment, and other risk factors. During a site visit, the well depth and distance from a septic tank was obtained. During the one-year follow-up period, 213 gastrointestinal (0.40 per person-year) and 945 non-gastrointestinal (1.78 per person-year) episodes were reported. At least one episode of AGI was reported for 31.6% of persons with Escherichia coli positive wells  After adjusting for other risk factors and confounders, a significant association was observed between E. coli positive wells and illness. The association was modified by the distance from the household septic tank to the well.
When the septic tank was greater than 20 meters from an E.
coli positive well, an increased AGI risk was found (OR ¼ 2.16; 95% CI ¼ 1. 04-4.42). When the well was within 20 meters, the risk was not increased. The investigators suggested that study participants whose septic tank was close to their well may have had higher levels of immunity to AGI because of consistent exposure to waterborne pathogens.
During April to July 1995, Strauss et al. (2001) conducted a prospective study of 235 households (647 individuals) randomly selected from four rural hamlets in eastern Ontario. A self-administered questionnaire, a diary of self-reported symptoms, and two drinking water samples provided information about the cohort. Although a high prevalence of the private, individual wells (20% of households) were contaminated with coliforms or E. coli above the standards, the study found no statistically significant associations between these indicator bacteria and AGI.
The case definition for AGI was similar to that used by Payment et al. (1991): either (1) vomiting or liquid diarrhea, or (2) nausea or soft, loose diarrhea combined with abdominal cramps.
In a year-long study, Isaac-Renton et al. (1996) found

United States
An association between giardiasis and drinking water exposure was suggested in a survey of intestinal parasites in two Washington counties. Harter et al. (1982) reported a 7.1% Giardia prevalence among 518 children, one to three years of age; a higher prevalence was associated with unfiltered surface water. Only one of 37 (2.7%) children residing in a home using filtered surface water was found to be infected with Giardia compared with 10 of 175 (6.9%) children residing in homes using unfiltered surface water.
An increased prevalence of infection was also associated with drinking untreated surface water from streams or lakes during recreational activities. A survey of 383 Utah National Guard members found that 15% had symptoms suggestive of giardiasis (Laxer 1985). The guardsmen were at risk of contracting giardiasis by drinking contaminated water during field exercises in Utah; 62% of the men who had symptoms drank untreated water from lakes, streams, and a cattle watering trough.
In Vermont, Birkhead & Vogt (1989) (Zmirou et al. 1987;Fattal et al. 1988;Meara 1989;Raina et al. 1999). noted that, in their study, older persons and a longer duration of residence were independently associated with contaminated well water and that the incidence of AGI was significantly lower in older persons. Long-term older residents had more frequent exposure to contaminated water and, thus, a greater opportunity to develop protective immunity. In Norway, Kuusi et al. (2003) found increased AGI risks among children, but not adults, who drank water from individual water systems; using chlorinated water was protective for children. Kuusi et al. also noted that children may be more susceptible to illness while adults may have developed some protective immunity from previous infections.
Early studies in the United States suggested that endemic giardiasis was associated with the consumption of untreated surface water and water from private, individual systems. Studies in Vermont and New Zealand found that an increased risk of giardiasis was associated with community systems that do not filter surface water sources.   et al. (1983) found that consumption of untreated water, nursery school exposure for children, and foreign travel to developing countries was associated with a higher risk of acquiring Giardia infection. Chute et al. (1985Chute et al. ( , 1987 (1993) found that drinking water from shallow wells was a significant risk factor; other risk factors were contact with a person in day care and swimming in a lake, pond, stream or river.

Australia
In an exploratory study in South Australia, Weinstein et al. (1993) found that only water-related exposures (i.e. consumption of rain and spring water) were significantly associated with cryptosporidiosis. with diarrhea. Other risk factors were similar for the two cities. In Adelaide, an increased risk was associated with drinking unboiled water from a river or lake in rural Australia (adjusted OR ¼ 3.1; 95% CI ¼ 1.5 -6.5). In Melborne, a small, non-statistically significant risk was associated with drinking unboiled water from a river or lake in rural Australia (adjusted OR ¼ 1.5; 95% CI ¼ 0.8-2.7).  hominis infections (previously genotype 1 and primarily of human origin), significant risk factors were travel abroad and changing the diapers of children less than 5 years of age. For C. parvum (previously genotype 2), eating raw vegetables and eating tomatoes were protective whereas touching farm animals was associated with increased risk.
In the final multivariate model which included all casepatients, a small but statistically significant risk was the number of glasses of unboiled water consumed at home (adjusted OR ¼ 1.135 per glass; 95% CI ¼ 1.010 -1.265).
However, in the univariate analysis, tap water was not associated with an increased risk and use of a water filter or bottled water was not associated with a decreased risk.
Hunter et al. noted the possibility of recall bias and suggested that tap water does not appear to be of major importance as a cause of sporadic cryptosporidiosis in this population.

Summary of case-control studies
An increased risk of campylobacteriosis was associated with drinking water from individual wells, untreated water from lakes, rivers, and streams, water from non-urban areas, or rainwater collected from roofs,. An increased risk was associated with tap water in Quebec Province but not in New Zealand. In Denmark, an increased risk was associated with household drinking water that had a bad taste or smell. In England, an increased risk of campylobacteriosis was associated with bottled water.
An increased risk of giardiasis was found among populations consuming untreated water, either surface or ground, or unfiltered municipal surface water. Other risk factors, especially person-to-person transmission and foreign travel may also be important, and drinking water may not be a significant risk factor in all areas. For example, Esrey et al. (1989) found that personal hygiene and person-to-person transmission in Lesotho, South Africa, were more important than water source in the transmission of giardiasis.
A study in seven states of the United States found that municipal water systems or well water was not associated with cryptosporidiosis risk among immnuocompetent persons; international travel, contact with cattle, contact with persons 2 -11 years of age with diarrhea, and freshwater swimming were important risk factors. In the San Francisco Bay Area, no significant association with drinking water was found for immnuocompetent persons; however, tap water consumption at the highest exposure categories was an important risk for persons with AIDS.
In one study in England, consumption of tap water was not a major source of endemic cryptosporidiosis; significant risk factors included travel abroad, changing diapers of children, and touching farm animals. However, in another study which excluded persons who had traveled abroad, the consumption of tap water was a significant risk factor. This is the first study to show that drinking tap water is a significant risk factor for sporadic cryptosporidiosis among the general population, regardless of the water source and its treatment.
In Australia, tap water from two large public systems with different quality water sources was not a significant risk factor for cryptosporidiosis; however, public swimming pools and person-to person transmission were important sources of infection. In several studies, the consumption of untreated water from a lake, river, or spring and rainwater was significantly associated with cryptosporidiosis.
In summary, tap water may be a significant risk factor for cryptosporidiosis among some immnocompetent as well as immunocompromised populations, and increased risks of campylobacteriosis and giardiasis may be associated with certain types of water systems. Some studies found no increased risks, and in one study, bottled water was implicated as a risk factor for campylobacteriosis. The findings of case-control studies emphasize that the importance of endemic waterborne risks depend on the water system type and vulnerability to contamination and, for some illnesses, the immune status or age of the population. Other risk factors and exposures may also be important and thus, should be considered when evaluating waterborne risks. control for potential confounding. A negative association was found between infection and the percentage of the population receiving water from a public water system (incidence RR ¼ 0.93; 95% CI ¼ 0.91-0.95). These findings suggest that a public water system may protect against infection. However, positive associations were found with the average length of water pipe per person (incidence RR ¼ 1.11; 95% CI ¼ 1.08 -1.15) and ruminant density (incidence RR ¼ 1.12; 95% CI ¼ 1.09-1.14). These findings suggest that livestock contamination may occur in the water distribution system. Several factors might contribute to contamination including leakage of pipes and low water pressure. In Sweden, about 20% of all water produced is lost through leakage, and in sparely populated areas long pipelines can lead to unstable water pressure resulting in contamination from backsiphonage or cross-connections.

United States
An ecologic study in Los Angeles County (Sorvillo et al. 1994) found no difference in the prevalence of cryptosporidiosis among AIDS patients in areas with filtered and unfiltered surface water. However, these results should be interpreted with caution in light of a recently conduced case-control study in San Francisco (Aragon et al. 2003).

United Kingdom
In Blackpool, Wyre, and Fylde, England, prospective and retrospective ecologic studies were conducted of incidence cases of cryptosporidiosis reported between 1987-92 and 1992-93 (Fewtrell & Delahunty 1995). In the retrospective study, most cases were reported in children under the age of four, and no significant differences in incidence were found when cases were evaluated according to water supply zones. The prospective study also found no differences in the incidence of cryptosporidiosis among water zones, and interviews identified recreational water and contact with farm animals as important risk factors.

Gastroenteritis
Germany Dangendorf et al. (2002) conducted a retrospective study of AGI risks in the North Rhine-Westphalia area which is characterized by different drinking water systems (surface or groundwater sources). The study excluded AGI due to Salmonella and Shigella because cases are primarily associated with travel and contaminated foods. Geo-statistical analyses revealed spatial variations in the incidence of AGI, and the amount of drinking water from surface or groundwater was correlated with age-standardized incidence rates.
Districts supplied by surface water had a significantly lower illness incidence. Dangendorf et al. felt that the results reflected the high quality of drinking water provided by the surface water systems. Water treatment included carbon filtration and disinfection, whereas three of the four groundwater systems in the study were untreated.

United States
Aeromonas is ubiquitous in water and some strains have been shown to possess virulence traits. Borchardt et al.

Time-series studies
The time-series study, a variation of the cohort approach using ecologic methods, considers a series of observations at successive points in time. In the 1980 s, the time-series approach was widely used to study the epidemiology of air pollution (Beaudeau, 2003 water quality was investigated. A winter rise in gastroenteritis was evident among all three health outcomes. Turbidity and fecal coliforms counts in all three watersheds also exhibited seasonal variation. Seasonal, long-term effects, and day-ofthe-week effects were controlled.
No association was found between AGI risk and fecal coliform levels or rainfall. However, statistically significant turbidity-AGI relationships were found among multiple age groups for all three health outcomes. In general, the probability of AGI increased as turbidity increased. The turbidity-AGI relationships were strongest among 2 -18 year olds and 18 -65 year olds, and the following lag times: 3-6, 6 -9, 12 -16, and 21-29 days. These lag times are consistent with the incubation periods of waterborne bacterial and protozoan pathogens. Depending on the area of Vancouver, turbidity variations explained 0.8-2.1% of emergency AGI-related physician visits and 0.2 -1.3% of AGI-related hospitalizations.

United States
Associations between drinking water turbidity and AGI were also reported in Milwaukee (Morris et al. 1996, Naumova et al. 2003 and Philadelphia (Schwartz et al. 1997, Schwartz & Levin 1999 found a positive association between age and emergency room visits and hospitalizations due to AGI in persons 65 years of age and older in Milwaukee before the outbreak.

Summary of time-series studies
Increased risks of severe AGI events were associated with turbidity in both filtered and unfiltered surface water sources that meet current water quality standards. Because the nature of the turbidity and its relationship with pathogens may depend both on source water quality and its treatment, extrapolation of these findings to other water systems should be made with caution.

OTHER STUDIES OF INTEREST
Of interest, but not reviewed in this article, are several other studies that should be mentioned. Moe et al. (1991) conducted a study of diarrheal disease risks associated with bacterial indicators in drinking water in the Philippines. Results of this study and a case-control study in urban Zambia (Nchito et al. 1998) are a reminder that waterborne contamination continues to be a substantial problem posing quite large risks in the developing world, especially for children.
Studies by Egorov et al. (2002) and Semenza et al. (1998) are a reminder that endemic illness may be associated not only with source water contamination and inadequate water treatment but also with water distribution system contamination. Egorov et al. associated an increased AGI risk with the deterioration of drinking water quality in the distribution system. Although water leaving the treatment plant at Cherepovets, Russia, was adequately disinfected, chlorine residuals declined rapidly within the system. An interquartile range decrease (0.22 mg/l) in free chlorine in the system was associated with an increased risk associated endemic waterborne risks with distribution system contamination, and in the United States, this source of contamination is increasingly associated with reported waterborne outbreaks (Craun et al. 2006b).

CONCLUSIONS AND RECOMMENDATIONS
An increased risk of endemic waterborne disease (campylobacteriosis, giardiasis, cryptosporidiosis, and AGI) has been epidemiologically associated with some drinking water sources and their treatment. These studies provide evidence that some populations using public or non-public drinking water systems may be at increased risk of endemic or sporadic illness. However, not all of the studies found an increased risk of illness. The differences in illness risk may be associated with different exposures due to various watershed characteristics and water system vulnerabilities in the areas that were studied. Observational studies offer little information about AGI risks in public systems, and it is not possible to use the current information from these studies to help estimate the population attributable risk for endemic waterborne AGI in the United States.
These studies suggest that more information is needed to characterize the risks associated with the various water systems serving the US population. Additional cohort and case-control studies in the United States can expand the database of information about risks associated with different water sources and treatment. In particular, we recommend that additional case-control studies be considered using the FoodNet population. Although the FoodNet population is not representative of the US population, it does represent a large population with exposure to potentially different public water systems. After assessing the range of water sources, treatment, and water quality in the seven states, the feasibility should be assessed of conducting a case-control study of AGI.
Alternatively, investigators can consider a nationwide study that is representative of the raw water sources and water treatment for public water systems in the United States.
Additional case-control studies should also be considered for those diseases caused by pathogens that are important causes of waterborne outbreaks. Additional ecological studies may offer additional hypotheses for further study, and time-series analyses can help clarify turbidity -AGI relationships in various public water systems. Other water quality parameters should be considered for time-series analyses.
Endemic risks associated with distribution system deficiencies should be better characterized in the United States and differentiated from endemic risks associated with source water quality and its treatment.