Objective: To clarify whether deaths associated with hot and cold days are among the frail who would have died anyway in the next few weeks or months. Design: Time series regression analysis of annual deaths in relation to annual summaries of cold and heat. Setting: London, UK. Participants: 3 530 280 deaths from all natural causes among London residents between October 1949 and September 2006. Main outcome measures: Change in annual risk of death (all natural cause, cardiovascular and respiratory) associated with each additional 1°C of average cold (or heat) below (above) the threshold (18°C) across each year. Results: Cold years were associated with increased deaths from all causes. For each additional 1° of cold across the year, all-cause mortality increased by 2.3% (95% CI 0.7% to 3.8%), after adjustment for influenza and secular trends. The estimated association between hot years and all-cause mortality was very imprecise and thus inconclusive (effect estimate 1.7%, −2.9% to 6.5%). These estimates were broadly robust to changes in the way temperature and trend were modelled. Estimated risk increments using weekly data but otherwise comparable were cold: 2.0% (2.0% to 2.1%) and heat: 3.9% (3.4% to 3.8%). Conclusions: In this London annual series, we saw an association of cold with mortality which was broadly similar in magnitude to that found in published daily studies and our own weekly analysis, suggesting that most deaths due to cold were among individuals who would not have died in the next 6 months. The estimated association with heat was imprecise, with the CI including magnitudes found in daily studies but also including zero.

London annual analysis dataset_BMJ OpenDataset containing natural deaths (number of) and temperature (summary statistic) for London, October 1949 to September 2006.Data aggregated for each12 month period (October to September). Mortality data obtained from Registrar's weekly summaries (1949 to 1975) and Office for National Statistics (1976 to 2006). Temperature data derived from baseline data courtesy of the British Atmospheric Data Centre. Excel 2010 file.London annual analysis core code_BMJ OpenSTATA core code to be used with London annual analysis dataset. STATA file.London annual analysis core code.do