Mortality and alcohol consumptionBMJ 1995; 310 doi: https://doi.org/10.1136/bmj.310.6977.470b (Published 18 February 1995) Cite this as: BMJ 1995;310:470
- Richard Doll,
- Richard Peto
- Emeritus professor of medicine Professor of medical statistics and epidemiology ICRF Cancer Studies Unit, Radcliffe Infirmary, Oxford OX2 6HE
EDITOR,—In our prospective study of mortality in relation to use of alcohol 12000 British male doctors who had been born in 1900-30 were asked in 1978 what they then drank, and over the next 13 years one third of them died of various causes.1 After standardisation for age and smoking the mortality from ischaemic heart disease was about one third lower among those who had said that they usually had a few drinks a day than among those who had said that they did not drink at all (P<0.0001).
The higher mortality from ischaemic heart disease among the self reported non-drinkers cannot be attributed to the inclusion of some heavy drinkers, or former heavy drinkers, among them because even the highest category of alcohol use was not associated with any material increase in cardiac mortality. Moreover, for the reasons discussed in our paper, among the “non-drinkers” the proportion thus misclassified is likely to be only a few per cent. David F Marks's suggestion that it might be about 50% is entirely unjustified (especially since the proportion of the deaths among non-drinkers that were attributed to liver disease or cancers of the mouth or throat was only 2% (10/486) and not the 53% (10/19) that Marks calculated),1 and his redrawing of our graphs should be ignored.
A G Shaper suggests that the apparently protective effects of moderate drinking might be chiefly due to confounding by some factor other than alcohol consumption.2 We showed, however, that it was not due to confounding by vascular ill health, smoking, age, or social class and cited evidence that it was unlikely to be due to confounding by diet. Differences between drinkers and non-drinkers in factors such as physical exercise or blood pressure would have to be implausibly extreme to account for a reduction of one third in cardiac mortality. In view of the known beneficial effects of alcohol on some haematological factors that may in turn affect the incidence of heart disease, confounding would seem to be a less likely explanation of our findings than a real protective effect of moderate alcohol use.