- A Gerald Shaper, emeritus professor of clinical epidemiologya,
- S Goya Wannamethee, British Heart Foundation research fellowa,
- Mary Walker, research administratora
- a Department of Primary Care and Population Sciences, Royal Free Hospital School of Medicine, London NW3 2PF
- Correspondence to: Professor Shaper
- Accepted 11 March 1997
Objective:To determine the body mass index associated with the lowest morbidity and mortality.
Design:Prospective study of a male cohort.
Setting:One general practice in each of 24 British towns.
Subjects:7735 men aged 40-59 years at screening.
Main outcome measures:All cause death rate, heart attacks, and stroke (fatal and non-fatal) and development of diabetes, or any of these outcomes (combined end point) over an average follow up of 14.8 years.
Results:There were 1271 deaths from all causes, 974 heart attacks, 290 strokes, and 245 new cases of diabetes mellitus. All cause mortality was increased only in men with a body mass index (kg/m2) <20 and in men with an index ≥30. However, risk of cardiovascular death, heart attack, and diabetes increased progressively from an index of <20 even after age, smoking, social class, alcohol consumption, and physical activity were adjusted for. For the combined end point the lowest risks were seen for an index of 20.0-23.9. In never smokers and former smokers, deaths from any cause rose progressively from an index of 20.0-21.9 and for the combined end point, from 20.0-23.9. Age adjusted levels of a wide range of cardiovascular risk factors rose or fell progressively from an index <20.
Conclusion:A healthy body mass index in these middle aged British men seems to be about 22.
The body mass index associated with the lowest mortality and the lowest incidence of coronary heart disease, stroke, and diabetes mellitus is not known
In this study of middle aged men the risk of cardiovascular mortality, heart attack, and diabetes increased progressively from a body mass index <20
For a combined end point (heart attack, stroke, diabetes, or death from any cause) the lowest risk was in the range 20-24
Levels of a wide range of cardiovascular risk factors increased progressively from an index of <20
A healthy body mass index in middle aged men seems to be around 22
The ideal, desirable, or healthy body weight is usually defined as that associated with the lowest mortality.1 For adults a body mass index (kg/m2) of 20-27 is widely recommended as the standard weight range within which there is little benefit from further leanness in relation to all cause mortality.2 3 However, criteria based on risk factors and morbidity may be more appropriate in determining healthy body weights.4 5 Data from the British Regional Heart Study has shown the effect of smoking on the U shaped relation between body mass index and all cause mortality commonly found in epidemiological studies.6 Among men who had never smoked the lowest mortality was observed in those with a body mass index of 20.0-21.9. Other reports from this study have been concerned with the independent relation between body mass index and the development of coronary heart disease and non-insulin dependent diabetes mellitus.7 8 This paper aims to determine healthy body weight based on mortality, morbidity, and cardiovascular risk factors in middle aged men drawn from general practice registers in 24 British towns and followed up for some 15 years. It examines the prospective relation between initial body mass index and subsequent risk of death from all causes and the incidence of heart attack, stroke, and diabetes as well as the relations between body mass index and cardiovascular risk factors in these men.
Subjects and methods
The British Regional Heart Study is a large prospective study of cardiovascular disease comprising 7735 men aged 40-59 selected from the age-sex registers of one group general practice in each of 24 towns in England, Wales, and Scotland. The criteria for selecting the town, the general practice, and the subjects as well as the methods of data collection have been reported.9 Men with pre-existing cardiovascular disease or taking regular medication were included in the study. Research nurses administered to each man a standard questionnaire that included questions on smoking habits, alcohol intake, physical activity, and medical history. Several physical measurements were made, and non-fasting blood samples were taken for measuring biochemical and haematological variables including serum lipids and packed cell volume.9 10 Triglyceride and insulin measurements were available for men in 18 towns (7th-24th; n=5675 and n=5661 respectively). We adjusted for the marked diurnal variation in both parameters.11 Details of blood pressure and heart rate and classification methods for smoking status, alcohol consumption, occupation (social class), physical activity, and body mass index have been reported.12 13 14 Body mass index (calculated as weight/height2) was used as an index of relative weight.
All men were followed up for death from any cause, cardiovascular morbidity, and development of non-insulin dependent diabetes from the initial screening in January 1978 to July 1980 up to December 1993, a mean period of 14.8 years (range 13.5-16 years),15 and follow up has been achieved for 99% of the cohort. Information on death was collected through the established “tagging” procedures provided by the NHS registers in Southport (England and Wales) and Edinburgh (Scotland). Criteria for accepting a diagnosis of non-fatal myocardial infarction and deaths from ischaemic heart disease have been reported7 as has the method for ascertaining new cases of non-insulin dependent diabetes.8
We used Cox's proportional hazards model to obtain the relative risks for the seven body mass index groups adjusted for age, smoking, physical activity, social class, and alcohol intake.16 Smoking (five levels), physical activity (six levels), alcohol intake (five levels), and social class (seven levels) were fitted as categorical variables. Body mass index was fitted as six dummy variables for the seven groups. Tests for trend were carried out fitting body mass index in its original continuous form. Indirect standardisation was used to obtain age adjusted rates/1000 person years with the study population as the standard. The analysis of covariance was used to obtain age adjusted mean levels of the cardiovascular risk factors for the seven body mass index groups.
To assess the U shaped relation between body mass index and total mortality we entered body mass index both as a linear and quadratic term in its original continuous form in the model; the analysis indicates a U shaped relation if the quadratic term is significant.
The mean (SD) body mass index was 25.48 (3.22). The men were divided into seven body mass index groups: <20 (n=268), 20.0-21.9 (n=703), 22.0-23.9 (n=1549), 24.0-25.9 (n=2080), 26.0-27.9 (n=1638), 28.0-29.9 (n=858), ≥30 (n=636). Data were not available for three men.
Mortality from any cause
During the mean follow up period of 14.8 years there were 1271 deaths from all causes. These comprised 643 deaths from cardiovascular causes and 628 from non-cardiovascular causes, of which 432 were due to cancer. Figure 1) shows the age adjusted mortality for all causes, cardiovascular disease, cancer, and other non-cardiovascular non-cancer causes. A U shaped relation was seen with all cause mortality with the lowest total mortality in the body mass index groups 22.0-27.9. Mortality was significantly increased in men with an index <20 or ≥30. A test for the U shaped curve fitting a linear and quadratic effect of body mass index showed a significant indication of a U shaped relation (quadratic term; P<0.0001). A positive association was seen between body mass index and cardiovascular mortality (test for trend, P<0.0001). For cancer, mortality was significantly increased in men with a body mass index <20 but thereafter there was no trend. For deaths from non-cardiovascular, non-cancer causes there was a significant inverse trend (P<0.0001). The excess deaths in the leaner men (<22) were largely due to respiratory causes.
Adjustment for lifestyle factors
Body mass index was strongly and inversely associated with cigarette smoking and physical activity and positively associated with social class and alcohol intake.17 The relation between body mass index and mortality was examined with adjustment for age and then in addition for these factors (table 1). Men with a body mass index of 20.0-21.9 were used as the reference group as this group lies at the lower end of the weight range usually regarded as acceptable.3 There was little difference in the age adjusted risk of death from all causes in men with an index of 20.0-29.9. The additional adjustment reduced the increased risk for men with an index <20, although it remained significant, and increased the risk seen in all the heavier groups. Mortality increased slightly at an index of 28 and was significantly increased at an index of >30. For cardiovascular mortality there was a progressive increase in relative risk through all groups after full adjustment (test for trend, P<0.0001). For cancer, after full adjustment mortality remained significantly increased only in men in the <20 group. For other non-cardiovascular deaths, full adjustment had little effect on the age adjusted relative risks (test for trend, P<0.001). Exclusion of deaths which occurred within the first five years of follow up did not greatly affect the patterns of risk observed.
Cardiovascular disease and diabetes
We examined the relation between body mass index and subsequent risk of fatal and non-fatal major coronary heart disease (n=974) and stroke (n=290). Table 2) shows the age adjusted rates/1000 person years for these outcomes and the relative risks adjusted for age and then in addition for alcohol intake, physical activity, smoking, and social class. No adjustment was made for blood pressure or blood lipid concentration as these are mechanisms in the pathway linking body weight and cardiovascular disease.
Coronary heart diseaseIncidence increased progressively with increasing body mass index. After age and the lifestyle factors were adjusted for, the overall trend in relative risk of coronary heart disease was significant, although the risk increased significantly only at an index of 24.0 and above compared with the baseline group.
StrokeThe age adjusted risk was increased but not significantly in lightest (<20) and heaviest men (≥30). This finding was not significant (P=0.06) after adjustment for age and lifestyle factors. The lowest risks were seen in those with an index of 20.0-21.9 and risk tended to increase thereafter.
DiabetesAll men with diabetes at screening (n=121) or who were diagnosed in the same calendar year as screening (n=14) or who had blood glucose concentrations ≥11.1 mmol/l at screening (n=22) were excluded from the analysis. In the 7575 men with no evidence of diabetes at screening there were 245 cases of non-insulin dependent diabetes during follow up. Risk of diabetes increased progressively with increasing body mass index from <20 (test for trend, P<0.0001) and was significantly raised at an index of 26 and above (table 2).
Combined end pointIn all, 2033 men either died or developed one of the end points (development of heart attack, stroke, or diabetes during follow up). After the full adjustment, the lowest risks were seen in men with an index of 20.0-23.9. Risk increased slightly at an index of 24 and was significantly increased at an index of 26 and beyond.
Exclusion of men with known coronary heart disease, stroke and diabetes
There were 604 men who recalled a doctor diagnosing coronary heart disease (heart attack or angina) or stroke or who had evidence of diabetes at screening (see above). Exclusion of these men made little difference to the relations between body mass index and the specific end points. For the combined end point (n=1717) the relative risks (95% confidence intervals) adjusted for age, smoking, social class, alcohol intake, and physical activity for the seven body mass index groups (lowest to highest) were 1.17 (0.88 to1.55), 1.00, 0.97 (0.79 to 1.19), 1.05 (0.86 to 1.27), 1.26 (1.04 to 1.53), 1.36 (1.10 to 1.69), and 1.99 (1.60 to 2.47).
Smoking is an important confounder in the relation between body mass index and mortality.6 We therefore examined the relation between body mass index and mortality from any cause, coronary heart disease, and the combined end point (death, coronary heart disease, stroke, or diabetes) separately by smoking status at screening (never smoked, former smokers, and current smokers) adjusting for age (fig 2). Current smokers showed higher mortality than former and never smokers at all body mass indices, and mortality from any cause was increased in those with an index <20 in all smoking groups. In never and former smokers, mortality increased thereafter (test for trend, P<0.008 and P=0.01 respectively). A shallow U shaped relation was seen in current smokers.
For coronary heart disease events a linear positive association was seen in never smokers and current smokers. In former smokers the lowest rates were seen in the 22.0-23.9 groups. For the combined end point, rates were lowest in the 20.0-23.9 groups in both never and former smokers and increased progressively thereafter (test for trend, P<0.0001). Among smokers the lowest rates were in the 20.0-25.9 groups and increased thereafter. Further adjustment for lifestyle factors made minor differences to the relations within these smoking categories.
Cardiovascular risk factors
We examined the relations between body mass index and cardiovascular risk factors after age was adjusted for (table 3). For most of these factors the levels increased (in the case of high density lipoprotein cholesterol decreased) progressively with increasing body mass index. Mean heart rate was slightly raised in those with a body mass index <20 but lower again in the 20-21.9 group and increased progressively thereafter.
Adjustment for biological factors
The positive relation between body mass index and cardiovascular mortality (table 1) and between body mass index and coronary events (table 2) was attenuated after further adjustment for systolic blood pressure and blood cholesterol concentration, although it remained significant (P=0.03 and P=0.002 respectively). Additional adjustment for high density lipoprotein cholesterol abolished the positive trend for both cardiovascular mortality and coronary events (data not shown).
Concepts of desirable or healthy weight have depended heavily on the relation between body mass index and mortality,1 2 although more recently the associations between body weight and cardiovascular risk profile, morbidity, and diabetes mellitus have been emphasised in considering optimal weight.4 5 18 In this study the well established U shaped relation between mortality and body weight was confirmed, with excess deaths in very lean men largely due to cancer and other non-cardiovascular causes and the excess deaths in the heaviest men predominantly due to cardiovascular disease. The relative risk of both heart attack and of diabetes increased progressively from an index <20 and the lowest risk of stroke was seen in those with an index of 20.0-23.9.
With a combined end point (heart attack, stroke, diabetes, or death) the lowest relative risk was in the 20.0-23.9 groups. All the major risk factors for cardiovascular disease rose progressively from an index <20. These findings strongly suggest that the healthy, biologically normal, or optimum body weight in these middle aged men is towards the lower end of the range which is currently regarded as acceptable.
Issues of adjustment
In most studies examining the relation between body weight and outcome in terms of morbidity or mortality, adjustments have been made in multiple regression models for blood pressure and total cholesterol concentration, and occasionally for other biological factors related to body weight and risk of cardiovascular disease. Any relations observed between body weight and the end points have been considerably attenuated after adjustment, often becoming non-significant. This has then been interpreted as meaning that “body weight does not matter” as these other variables have accounted for the relations observed. Our results were also attenuated after we adjusted for blood pressure, cholesterol, and high density lipoprotein cholesterol. In trying to assess the effects of body weight it seems illogical to adjust for those factors which are almost certainly the mechanisms by which increasing body weight brings about vascular damage.
Previous prospective studies have generally focused on the U shaped relation between body mass index and all cause mortality or on the relation between body mass index and specific end points–for example, coronary heart disease. Few studies have focused on assessing healthy weight based on risk factors or morbidity, or both. In a cross sectional study of 3582 Japanese men (mean body mass index 23.3) and 983 women (mean 21.8) aged 30-59 years, 10 medical problems (hypertension, hyperlipidaemia, hyperuricaemia, ischaemic heart disease, lung disease, anaemia, upper gastrointestinal disease, liver, and renal disease) were selected to determine morbidity.4 The relation between body mass index and a composite morbidity index formed a J shaped curve with the lowest point at an index of 22.2 in men and 21.8 in women. The authors conclude that an index of about 22 seems to be the ideal. In the Framingham offspring study a similar conclusion was drawn after assessing the relation between the scapular skinfold (as a direct measure of adiposity) and several cardiovascular risk factors in 2447 non-smoking men and women aged 20-59 years. Healthy adiposity corresponded to a mean body mass index of 22.6 for men and 21.2 for women.5 Although not specifically aimed at assessing ideal body weight, a Finnish study of about 16 000 men and women aged 30-59 years showed clearly that the main risk factors increased progressively from an index of 20 upwards and that from a level of 22 an increase in body weight equivalent to 1 body mass index unit was related to a 4-5% increase in coronary heart disease mortality.19 In our study an increase in 1 body mass index unit from 20.0-21.9 onwards was associated with an approximately 10% increase in the rate of coronary events and a 10% increase in the combined end point. In the Nurses health study, in women who had never smoked and who recently had stable weight, the lowest mortality was among the leanest women (body mass index <19.0).20 These women were at least 15% below the United States average weight for middle aged women. All cause mortality did not increase substantially until a body mass index of 27, although trends were apparent for coronary heart disease and cancer among women at average weights and among those who were mildly overweight.
Public health aspects
The British government's Health of the Nation strategy set targets for the reduction in the prevalence of obesity (body mass index ≥30) in men and women aged 16-64 years but made no recommendations regarding the distribution of weight in the population and set no standards for healthy body mass indices in the population.21 The recent task forces report suggests targeting adults with a body mass index of 25-30 and expresses “a concern to develop a strategy to prevent the population in general becoming fatter.”22 In 1993, the proportion of men and women in England who were obese was 13% and 16% respectively with a mean body mass index of 25.9 for men and 25.7 for women. As an index of 25-30 is generally regarded as overweight,23 half of the adult population of England is overweight or obese.24
By contrast, in the United States, which has a similar epidemic of obesity, the focus of two recent reports has been on healthy weight rather than on obesity itself.18 25 Both reports recommend maintaining a lean body weight throughout adult life and weight reduction in those who are overweight with or without obesity related disorders. The American Institute of Nutrition recommends a single body mass index criterion of 18-25 for both sexes and suggests that “most people will be healthier towards the lower end of the range.” They proposed a format indicating gradations of risk–for example, body mass index 18-23=lowest risk, 24-25=mild risk, 26-29=medium risk, and ≥30=high risk.25
The American Health Foundation Expert Panel proposes a healthy weight target of a body mass index <25 for adults, representing the upper limit beyond which weight related disease risk becomes a concern and morbidity associated with obesity becomes manifest. In those exceeding the healthy weight target and without a diagnosis of a weight related disease, they propose a healthier weight goal. This represents the amount of weight loss that will reduce disease risk and is roughly two body mass index units (about 6 kg or 1 stone). This modest weight loss is regarded as achievable and maintainable and is more likely to be reached than the healthy weight target.18
Our findings broadly agree with the United States reports, and it is clear that the emphasis must be on maintaining healthy body weight in early adulthood and the prevention of obesity. The British focus on obesity seems to avoid the issue of a healthy weight and to direct attention to the clinical management of obesity.
It is well established that at all body mass indices individuals with visceral obesity (excess deep abdominal fat as indicated by waist-hip ratio or waist circumference) are at highest risk of cardiovascular disease.26 Thus estimates of risk based on body mass index or other crude measures alone may not be sufficient for assessment. Clinical decisions on the importance of body mass index in individuals will depend on the distribution of fat and muscular development as well as on the overall profile of risk.
In industrialised societies increasing body weight is closely related to an increasing incidence of non-insulin dependent diabetes and coronary heart disease and to increasing blood pressure, blood lipid, glucose, and insulin concentrations, urate concentration, and packed cell volume–factors all involved intimately in the development of coronary heart disease. There is also considerable evidence of the benefits of weight reduction on risk factors for cardiovascular disease and diabetes.18 Although the benefits of weight reduction in overweight people for coronary heart disease are still controversial,27 28 29 the importance of maintaining a healthy weight throughout life as a major primary preventive measure against cardiovascular disease and diabetes seems incontrovertible. Within the “normal” range of body mass index (20-27) it is better to be leaner, and the optimal healthy body mass index for adults is about 22. The implications of this conclusion for public health are considerable, and with the rising tide of obesity in the industrialised world deserve to be treated with some urgency.
Funding: The British Regional Heart Study is a British Heart Foundation research group and receives support from the Department of Health and the Stroke Association.
Conflict of interest: None.