Intended for healthcare professionals


Mortality and alcohol consumption

BMJ 1995; 310 doi: (Published 04 February 1995) Cite this as: BMJ 1995;310:325

Non-drinkers shouldn't be used as baseline

  1. A G Shaper
  1. Emeritus professor of clinical epidemiology Royal Free Hospital School of Medicine, London NW3 2PF

    EDITOR,—The interpretation of the finding that light or moderate drinkers have a lower mortality than non-drinkers remains controversial.1 The original cohort of 34000 male doctors was recruited in 1951, and information on alcohol consumption was obtained from the surviving third of the men some 27 years later (1978) and from the surviving quarter 40 years later (1991). The authors make no comment on the possible implications of these large gaps in time or on the issue of survivorship.

    The British regional heart study has shown that middle aged male non-drinkers are likely to be exdrinkers, are older, and have higher rates of a wide range of diseases and of drug treatment than light or moderate drinkers.2 Non-drinkers should not be used as a baseline against which to measure the effects of alcohol consumption. The proportion of British doctors who are ex-drinkers cannot be assessed on the basis of those few non-drinkers who spontaneously mentioned previous drinking, and further estimates of previous drinking are based on survivors 40 years after recruitment, who are obviously likely to be healthier than those who died. There is a clear tendency between 1978 and 1990 for men drinking >/=15 units a week to move towards lighter or non-drinking status rather than continuing with a stable intake (table II in the paper), and Wannamethee and I have shown that diminishing alcohol intake with increasing age is closely associated with increasing ill health and drug treatment.3

    Non-drinking doctors have a higher mortality than drinkers even when divided into those with and without “previous disease.” Because the shape of the two curves is similar Richard Doll and colleagues conclude that previous disease has little relevance to the relation observed between alcohol intake and mortality, although they have not adequately examined the characteristics of non-drinkers in either group.

    In the study of British doctors deaths from ischaemic heart disease showed no significant trend with alcohol intake, although non-drinkers had a somewhat higher mortality than those drinking 1–14 drinks a week. An early report from the British regional heart study also showed no significant relation between alcohol intake and the incidence of heart attacks, although men drinking 1–2 British units daily had the lowest incidence.4 This group contained the lowest proportion of current smokers and had the lowest mean blood pressure and body mass index and the highest levels of physical activity in leisure time. Doll and colleagues have not examined the issue of such advantageous characteristics, although they might be more relevant than the direct effects of alcohol. In a 9.5 year follow up of the cohort in the British regional heart study men drinking 2–6 British units a day showed a non-significant reduction in the risk of death from ischaemic heart disease compared with occasional drinkers, little reduction in total cardiovascular mortality, and no reduction in total mortality.5

    Doll and colleagues' conclusion that the present guidelines should acknowledge the important disadvantages to health of total abstinence goes beyond the information available in their study. It implies that middle-aged and elderly British men should drink alcohol to reduce their risk of death from all causes. Surely it would be preferable to attempt to reduce the population risk of premature death by paying attention to smoking, diet, and physical activity?


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.