A study into the possible
association between step changes in water hardness and incidence of
cardiovascular disease in the community
Purpose of the report
The aim of this study was to investigate the association between
drinking water hardness and cardiovascular mortality in areas that had
experienced a stepped change in water hardness, calcium or magnesium
levels in England and Wales.
Data collection: Mortality data for cardiovascular diseases in England
and Wales between 1981 and 2005 were obtained from the Office of
National Statistics. Such data included date of death, age at death,
gender, cause of death and postcode of last residence. Time series data
detailing drinking water concentrations of total hardness, calcium,
magnesium and sodium were requested from each company. Concentration
data were requested for both pre and post change time periods.
For each area a linear or Poisson regression of log counts on monthly
mean maximum daily temperature, its square, the previous months
temperature, and variable accounting for the flu epidemic was fitted
allowing for a quadratic time trend and the number of days in the
month. The change in water hardness was included using an indicator
variable. The coefficient of an indicator variable, as with all Poisson
regression models, is the risk ratio.
We found no evidence of an association between step changes in drinking
water hardness, either increases or decreases, and cardiovascular
mortality. We undertook several sensitivity analyses to examine the
impact of the model definition and variables upon the results obtained.
The results were robust to the inclusion of areal population estimates,
the type of model used and the form of the dependent variable. We also
undertook a power calculation to examine whether the non significance
of the association could be explained by a lack of power of this study
to adequately examine the hypothesis.
Only 14 areas were identified in England and Wales that experienced a
step change in drinking water hardness over the period 1981 to 2005. Of
these, few were areas that had undergone substantial changes with large
populations. Overall, drinking water supply characteristics were mostly
stable in England and Wales, particularly with regard to magnesium. As
the course of the study progressed, it became apparent that the quality
and availability of essential data including water hardness
concentration, areal extent and population were variable. This
variability could not have been predicted in advance.
This study found no evidence of an association between step changes in
drinking water hardness, calcium or magnesium and cardiovascular
mortality. As we only had one geographical area with a substantial
magnesium change, the application of this time series approach in other
settings may contribute to the state of knowledge regarding this
hypothesis. If the time series approach were to be replicated in other
countries and populations, it should only be applied to populations
with accurate historical water quality records, mortality and
Copies of this report may be available as an Acrobat pdf download under the 'Post 2000 Reports' heading on the DWI website.