The higher occurrence of cardiovascular diseases in winter is well known, and several explanatory mechanisms have been suggested based on increased blood pressure, haematological changes and respiratory infections. Most investigations have used ecological data such as daily temperatures recorded at weather stations and mortality in the general population. Cause-specific mortality is the outcome measure most commonly used. Local myocardial infarction community registers would offer an ideal database, but may suffer from inadequate statistical power. Hospital discharge records, linked with out-of-hospital deaths, provide a powerful tool for detecting even weak effects of temperature. The association of coronary heart disease and temperature is usually U-shaped, mortality being lowest within the range 15-20 degrees C and higher on both sides of this. The increase in mortality on the colder side is in the region of 1% per 1 degree C fall in temperature, but the increase on the warmer side may be very steep. The exact location of the minimum temperature and the magnitude of the effect can vary between countries. In Finland the winter excess mortality from coronary heart disease has been levelling off during recent decades, but it still represents approximately 6% of annual deaths due to this condition.