Background It has been hypothesised that dissolved minerals in water, particularly calcium carbonate and magnesium carbonate, are protective against heart disease. Animal studies suggest biologically plausible mechanisms for this and statistically significant effects have been presented from several large ecological studies.

Setting The West Midlands Government Office Region in central England, has a large and diverse population which has great variation in water hardness over a relatively small area. Our units of analysis were West Midlands neighbourhoods (lower level super output areas) falling in areas supplied by two water companies. We obtained exposure estimates from water industry measurements taken in 2007.

Methods Using geographical information system software, digitised supply water supply maps were mapped to our neighbourhood geography. Having established the distribution of water hardness through neighbourhoods, we mapped tertiles of hardness and identified neighbourhoods inside them. To these we linked emergency myocardial infarction (MI) admissions for 45–74 year-old residents for a three year period, adjusting for the age distribution of the neighbourhoods, ethnic mix and socio-economic deprivation. We used a negative binomial model to determine the degree of association between water hardness and MI admission counts, adjusting for the other variables.

Results We were able to accurately determine the mineral content of water supplied to 2,925 neighbourhoods with a total population of over 4.5 million. Contrary to other studies, we found just a small but non-significant negative correlation between hardness and MI admissions were seen in men, with an incidence rate ratio (IRR) of 0.97, per tertile, (0.92 – 1.03, p=0.37) and a small, non-significant positive one seen in women, IRR=1.02, (0.93 – 1.12, p=2).

Conclusion We found no evidence of a protective effect of hard water against acute MI admission, contrary to other published studies. This may be because our methods allowed us to better account for possible confounding variables. We go on to hypothesise that as water supply networks developed historically, with soft water from upland areas being used supply the needs of rapidly industrialising areas and other areas left to use local groundwater, some interesting accidental correlations between neighbourhood socio-economic status and water hardness arose which may have confounded previous work. However, mass differential exposure to highly bio-available minerals, especially calcium and magnesium, does raise some interesting further questions about the relationship between water supply and population health, which our methods can be used to investigate further.