BACKGROUND: There is increasing recognition of the importance of obesity as a cause of death but it is uncommon for obesity to be certified on death certificates. We considered it useful to study what doctors actually do in respect of certification of obesity and to study trends, if any, in certification practice. METHODS: Analysis of two datasets that include all certified causes of death ('mentions'), not just the underlying cause-the Oxford record linkage study 1979-2006 and English national mortality data 1995-2006. RESULTS: Underlying-cause mortality identified only a quarter (26% in Oxford, 25% in England) of all deaths with obesity as a certified cause. The longstanding Oxford dataset showed that there were significant changes over time in the percentage of certificates, with mention of obesity, that were coded with obesity as the underlying cause. Changes coincided with times of national change in selection and coding rules for underlying cause mortality. In the recent English dataset from 1995-2006, mention-based death rates rose by an average annual rate of 7.5% [95% confidence intervals (CI) 6.1-8.8] for men and by 4.0% (2.3-5.7) for women. Analysis of mortality based on underlying cause alone would have missed this rise. We report on diseases commonly certified alongside obesity on death certificates in England. CONCLUSION: There is an emerging trend of increased certification of obesity as a cause of death in England. The use of underlying-cause mortality statistics alone fails to capture the majority of obesity deaths.