Sir, the study by Elmer et al (BDJ 2014; 216: E10) claims promise for a new way to assess the effectiveness of fluoridation. However, it relies on data of questionable validity from the Hospital Episode Statistics system (HES).

Other authors found HES data unreliable, missing as many as 80% of actual general anaesthesia hospitalisations (GAs).1,2 Many practitioners did not even report to the HES system. Since Elmer compared just two health regions, and the number of practitioners doing hospital extractions is small, there is a high risk of bias from regional differences in reporting rates.

Referral practices may also vary by region and can be strongly influenced by health policies.3 Modest referral differences can produce large differences in GA rates since GAs are a small fraction of all extractions.

Regional differences in dental care are another potentially confounding factor. The dental care index (ft/dmft) was 35% higher in Elmer's fluoridated West Midlands region than his comparison unfluoridated region North West, suggesting a better level of care in the fluoridated region.4 This contrast was amongst the most extreme in England, and was significant (p = 0.002, based on sub-region variance, 2-tailed t-test).

These problems also apply to a recent PHE fluoridation monitoring report that relied on HES data.5

Elmer says evidence on fluoridation from randomised controlled trials (RCTs) remains 'understandably absent' because RCTs are 'very challenging'. While a community-level fluoridation RCT would be difficult, an individual-level RCT using coded bottled water would not. Pharmaceutical companies do thousands every year. A few RCTs could settle the question of fluoridation's effectiveness. As the authors of the York Review said, existing evidence is 'poor quality' and '... only high-quality studies can fill in the gaps ... Recourse to other evidence of a similar or lower level than that included in the York review..., no matter how copious, cannot do this.'6