Intended for healthcare professionals


Contour control, survival, and quality of life

BMJ 1997; 314 doi: (Published 03 May 1997) Cite this as: BMJ 1997;314:1291

Ideal body weight is far lower than average

  1. Hugh Tunstall-Pedoe, Professor of cardiovascular epidemiologya
  1. a Cardiovascular Epidemiology Unit, University of Dundee, Ninewells Hospital and Medical School, Dundee DD1 9SY

    Body weight is the archetypal risk factor for morbidity and mortality. It is repeatedly measured without going away. It figures voluminously in the press, supporting an inflated consumer industry in dietary and other products. It was adopted early by life insurance companies when little else was measurable. Medical science is bankrupt of effective treatments for obesity, it confuses the public as to whether being overweight is determined by fate or free will, and it keeps changing the units of measurement imperial to metric, body mass index,1 waist to hip ratio,2 and waist circumference3 resulting in familiarity by few British doctors and even fewer patients. The government decrees, Canute-like, that the tide of obesity is to go out4 when, as elsewhere, it is coming in,5 with Britain accumulating, I estimate, some 10 000 metric tonnes of blubber a year. Into this sombre picture comes the paper in this week's BMJ by Shaper and colleagues (p 1311).6 Using data from the British regional heart study, the authors report 15 years' follow up of mortality and cardiovascular morbidity in relation to the body mass index of middle aged British men seen in the late 1970s. What does the paper show?

    The answer on first glance might well be: not much. Those indoctrinated in the perils of obesity may be unimpressed by the gradient of mortality against body mass index. The pattern, as elsewhere, is U shaped. All cause annual mortality was 12/1000 overall, 19 in the leanest group (body mass index < 20), 11 in the best surviving group (body mass index 22-24), and 15 in the obese group (body mass index > 30). These categories can be typified in a man 1.73 metres tall (5 feet 8 inches). His squared height would be 3.0, and a body mass index of 20 would imply a weight of 60 kg (about 9.5 stone), while a body mass index of 30 would imply a weight of 90 kg (about 14 stone). The excess mortality in the obese group was 28% above the average and 39% above the best. The closeness of the overall average mortality (12) to the optimum mortality (11) might suggest that obesity is not contributing to many deaths in the population, but appearances can be deceptive.

    Flat, U shaped, or J shaped curves for all cause mortality are found in relation to other risk factors such as cholesterol7 and alcohol consumption.8 Underweight cigarette smokers help to flatten the mortality curve for obesity. The body mass index is a crude indicator of obesity across individuals. In our sedentary society, barring fluid retention or enforced immobility, we can relate personal variation in weight to changes in subcutaneous or intra-abdominal fat. Extrapolation across the population is less reliable. People of similar height who vary in weight do so in bone and muscle mass as well as fat. Body mass index corresponding to individual lean carcass mass will vary considerably. The study's survival curve is flattest in the middle of the range, where most confusion and inaccuracy will occur in categorising individuals.

    The composite curve of all cause mortality is derived from several components. The curves for cancer deaths and for non-cancer, non-cardiovascular deaths show a negative gradient with increasing body mass index that is strong enough almost to obliterate the effects of the positive gradient for cardiovascular deaths. The findings for cardiovascular deaths are strongly reinforced if morbidity from non-fatal myocardial infarction, stroke, and diabetes are added to the event rates. In addition the paper translates a positive correlation with increasing body weight for several cardiovascular risk factors.

    Based on combined mortality and morbidity and rates of risk factors, the authors suggest that the ideal body mass index should be around 22, so our typical man should weigh 66 kg (about 10½ stone) instead of 77 kg (over 12 stone). They also suggest that national policies should aim to modify the distribution of body weight in the whole population rather than in the proportion of the population labelled as obese. This recommendation is consistent with the epidemiological principle that the proportion of the population in the extreme category is determined by the population mean.9 Reducing numbers at the extreme would require most of the population to change. At present in Britain the population mean weight is increasing, which implies massive percentage increases in those who are obese and very obese.

    What are the implications of increasing body weight in the population? The authors make a convincing case for increased cardiovascular risk factors, morbidity, and mortality, particularly from diabetes. Data on the consequences of change in body weight in individuals are entirely consistent,10 (and, for the reasons given above, the within person mortality gradient will be steeper than the population graph) so there is a plausible cause and effect relation.

    Unfortunately, the same is not true at the low end of the U shaped curve, where the pathways relating low body mass index to excess cancer or respiratory deaths are too poorly understood for us to be sure that weight gain would reduce mortality.

    In their analysis of morbidity the authors did not include musculoskeletal, respiratory, and other conditions whose relation to obesity might have been of interest. The implications from this and similar studies are depressing, but there is an epidemiological paradox. Why is cardiovascular mortality falling in many countries where obesity is increasing? While increasing obesity is a threat to survival and the quality of life, some more powerful trends in risk factors must be acting in mitigation.

    Meanwhile, for those of us not susceptible by sex age to other biological explanations for increasing girth, and change in waist circumference seems the best way of monitoring potentially harmful intra-abdominal fat.2 3 4 5 6 7 8 9 10 If we are spilling over the tops of our old dinner jacket trousers we should be attempting to shrink back into them.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    View Abstract