Authors' reply

We agree that our findings need to be replicated in other populations and climates. Salib et al may have misinterpreted our analysis of suicides during the 2003 heatwave, as our finding of −1.8% (95% CI – 17.8 to 18.4) change in suicides from expected is consistent with no change rather than a reduction. We discuss reasons for this lack of effect and point out that the lack of power in this calculation leads to an imprecise estimate.

We disagree with Salib et al's assertion that the effect of high temperature on all-cause mortality (rather than suicide specifically) is a reasonable explanation for our findings. We only examined deaths from suicide and undetermined intent, so it is not possible for other causes of death to have ‘confounded’ our results. We considered carefully which confounders to include in our models. Individual-level confounders, for example the effect of individual stress on the hypothalamic-pituitary-adrenal axis, are irrelevant in a time-series analysis as they do not vary day to day across a population. Sunshine hours were sufficiently accounted for by including a term for hours of daylight. We think it unlikely that any of the other potential confounders mentioned by Salib et al could be sufficiently associated with both temperature and suicide to explain our findings. Also, humidity, rainfall and unusual weather conditions (e.g. thunderstorms) tend to vary regionally more than temperature, meaning that exposure misclassification would be a problem in a countrywide analysis. The role of solar winds in the aetiology of suicide is highly speculative.

Higher temperatures affect mortality through a range of mechanisms (Bouchama & Knochel, 2002). Cardiovascular and respiratory deaths during periods of high temperature are caused by physiological changes, including increased coagulation, dehydration and increased cardiovascular output - particularly important in the elderly or those with pre-existing disease. A range of antipsychotic drugs are known to inhibit sweating and therefore thermo-regulation. Recent work has shown that deaths from respiratory and external causes are particularly increased at high temperatures (Hajat et al, 2007). Further research is needed on the pathophysiology of heat, but it is clear that persons with mental illness remain a high-risk group for heatwave mortality (Kovats & Ebi, 2006).