Does Tap Water Make You Ill?

During the last twenty years a number of research studies have attempted to test the hypothesis “Drinking water that complies with current standards for microbiological quality nevertheless causes a proportion of gastrointestinal (GI) illnesses reported in the community”. This question has been raised again by the 2009 publication of the Sonoma Water Evaluation Trial In this USA study of 988 individuals in 744 properties, half of the participants consumed tap water and half consumed tap water that had been passed through a point-of use (POU) treatment system comprising a 1 micron filter and a UV disinfection unit. During the study period, participants kept diaries of their water consumption and incidence of symptoms of gastrointestinal (GI) illness such as vomiting, nausea, diarrhoea, and abdominal cramps.

In order to avoid reporting bias, a sham POU unit was installed in the premises of the control group members who consumed unmodified tap water. This ensured that the control group and test group were unaware of the sub-group to which they had been assigned. Data analysis revealed that the sub-group drinking the POU treated water reported 12% less GI than was reported by the group consuming unmodified tap water. However, the apparent protection was afforded only to male participants.

This is a surprising result but this type of study has produced contentious results in the past. The prototype of the genre was the 1991 Montreal Study, in which participants consumed unmodified tap water or tap water treated with a reverse osmosis filter POU device. GI illness rates were reported to be one third lower in the group consuming POU treated water. The report’s authors concluded approximately 35% of the reported GI illnesses among those consuming unmodified tap water were water-related and preventable. The results raised questions about the adequacy of drinking water quality standards to prevent water-borne endemic gastrointestinal illness.

The results of the Montreal study have since been questioned. Participants were aware of the sub-group to which they had been assigned; they were also aware of the hypothesis being tested. This led to the suspicion that reporting bias had influenced the results. Furthermore, there was concern that the study did not take adequate account of the high dropout rate among participants. The reporting rate for GI illnesses declined as the study progressed and as participants dropped out. This raised the possibility that participants remaining in the study might have exhibited different perceptions of GI symptoms to those participants who dropped out.

More recently, two studies have been reported which used a blinded design where households were supplied with real or sham water treatment devices and where researchers and participants were unaware of the water treatment status of the premises until after the study had been completed. Both studies reported insignificant differences in GI illness rates between the real and sham water treatment groups. The 2005 Davenport Iowa study in the USA was conducted in a population served by a poor quality river source that received conventional filtration and disinfection. The 2006 Melbourne Water Study examined GI illness rates in a population served by a high quality upland source that received disinfection treatment only.

The question remains as to which of the studies referred to above are relevant to the UK. All of the studies examined whether tap water that complied with current microbiological standards nevertheless caused illness in the population served. Standards for microbiological quality are essentially the same in the USA, Canada, Australia and the UK. Therefore the high compliance rate of UK water supplies is presumably no guarantee of protection from GI illness.

The results of the Montreal study may have been affected by reporting bias but the Sonoma did not suffer from that limitation. However, the water supply source and water treatment provision in the Sonoma study consisted of bankside filtration of river water draining urban and agricultural catchments, followed by chlorination. In the UK this combination of source water and treatment would be considered as posing a Cryptosporidium risk. This might explain an elevated rate for GI illness, although that study’s finding that POU devices protected only male participants does not fit accepted patterns of infection rates. The Davenport and Melbourne studies provide reassurance that well operated treatment that is appropriate to the quality of the available source water produces water that does not cause an identifiable rate of GI illnesses among consumers.

It is understood that there are no plans to mount similar studies in the UK. DWI published a critique of the Montreal Study in 1992. The report included proposals for a UK study that addressed the perceived shortcomings of the Canadian study. The estimated cost of that study in 1992 was £1.6M. At today’s prices, it would be hard to justify and the results could only provide information about risks in a defined water supply setting for a specific group of participants. The money might be better spent in appraisal and monitoring of water suppliers Water Safety Plans that set out, inter alia how the risk that pathogens might contaminate water supplies is characterised and addressed.