Introduction

It has long been accepted that overweight people have a reduced life span. Last year a major US study reported on a cohort followed from 1960 to 1972, using data from the American Cancer Society. ¹ Now a second study of comparable size, using the same source, has analyzed the results from a cohort followed from 1982 through 1996. The new findings, summarized here, are similar to those from the first study, and leave no doubt that the risk of death from all causes increases through the range of moderate to severe overweight, for both men and women, in all age groups.

Method

The study subjects were drawn from the American Cancer Society's Cancer Prevention Study II, which is being conducted in over 1.1 million people over 30 years from all 50 United States. In 1982 participants completed a questionnaire concerning their personal demographic, environmental, dietary and health affairs. Answers on weight and height allowed the body-mass index (BMI) to be calculated at entry - weight in kilograms divided by the square of the height in meters. Deaths occurring between enrollment and the end of 1996 were ascertained and verified via the National Death Index. Deaths from all causes were of prime interest, but deaths due to cardiovascular disease, cancer, and all other causes were also categorized.

Subjects were classified into 12 categories according to their BMI, ranging from <18.5, 18.5-20.4, 20.5-21.9, and so on, to >40.0. Subgroups were also built according to current or former smoking vs. non-smokers, with and without a history of diseases (these included cancer, heart disease, stroke, respiratory disease, current disease of any sort, or weight loss of >10lb in the previous year).

Out of over one million participants (457,785 men and 588,369 women), 20.1% died during the 14 years of follow-up. Age-adjusted death rates were calculated for each BMI category and smoker/history of disease subgroups. Relative risks were estimated using a Cox proportional-hazards model adjusting for exact age at enrollment, education, physical activity, alcohol use, marital status, use of aspirin, fat and vegetable intake, and estrogen-replacement therapy.

Results

Curves of the relative risks for BMI categories yielded the expected broad U-shaped profile, with the nadirs for the different subgroups in the 23.5 to 28 range. For healthy nonsmokers the nadir for men was at a BMI of 23.5 to 24.9, and for women at a BMI of 22.0 to 23.4. This particular subgroup showed the greatest association between obesity and death, with relative risks of 2.7 (men) and 1.9 (women) for BMIs above 40. (A relative risk of 1.0 was assigned to the nadir for each subgroup.) This is shown in the figure:

For the other three subgroups (nonsmokers with history of disease, smokers with and without history of disease), relative risks were about 1.6 and 1.4 for BMIs over 40 in men and women, respectively.

The association between obesity and death in nonsmoking healthy subjects was stronger in whites than in blacks, who constituted <3.5% of the population. For those with BMI levels over 40, the relative risk for white women was 2.0, compared with 1.2 for black women. There were insufficient deaths in black men at this BMI to yield useful data, but the relative risk for black men with a BMI over 32 was 1.35, compared with 1.65 for white men.

In those who had never smoked and had no history of disease, a high BMI was most predictive of death from cardiovascular disease, especially in men. The relative risks of death from this cause were significantly increased at BMIs over 25 in women and 26.5 in men.

The relation between BMI and risk of death from cancer was positive, without the increase in risk at the leaner end (low BMIs) seen with deaths from other causes; in women, the relative risk of death from cancer reached 1.8 at BMIs over 40.

Analysis of different age subgroups failed to reveal important age-related differences in risk factors for mortality from all causes for the healthy nonsmokers.

Comment

Previous studies have also found that obesity is less strongly associated with mortality in black women. The authors question whether the central and abdominal distribution of fat in black women, as compared with the distribution in white women, may have a weaker effect on atherogenic risk factors e.g. lipid levels, peripheral insulin resistance. The finding, however, is at odds with the higher mortality from diabetes and heart disease reported in black women. An accompanying editorial to this article suggests that black women may have under-reported current disease in their questionnaires, and that the potential confounding effects of this were not corrected in the analyses. ²

The finding that age plays no relevant role in the mortality risk of overweight is at variance with the results of a large German study.

The present study shows that the relative risk of death is significantly raised for men if their BMI is 26.5 or over, and for women if their BMI is 25.0 or over. Nearly a third (32.6%) of adults in the USA have BMIs between 25.0 and 29.9; a further fifth have BMIs of 30.0 or more.² In other words, over 50% of USA adults are overweight enough to be at increased risk of death.

The Canadian Task Force on Preventive Health Care has made a sensible suggestion: "If adult patients have gained 10 lb (4-6 kg) in weight, or 2 in. (5 cm) in waist circumference, their physicians should counsel them to make small but permanent adjustments in physical activity and eating patterns". ³ Although the USA appears to have a serious overweight problem, other affluent countries are not far behind, and all family physicians would do well to follow the Canadian Task Force strictures.