- M E J Lean, professor of human nutritiona,
- T S Han, studenta,
- C E Morrison, coprincipal investigator, Glasgow arm of MONICA projectb
- aDepartment of Human Nutrition, University of Glasgow, Royal Infirmary, Glasgow G31 2ER
- bDepartment of Public Health, University of Glasgow
- Correspondence and reprint requests to: Professor Lean.
- Accepted 1 May 1995
Objective: To test the hypothesis that a single measurement, waist circumference, might be used to identify people at health risk both from being overweight and from having a central fat distribution.
Design: A community derived random sample of men and women and a second, validation sample.
Setting: North Glasgow.
Subjects: 904 men and 1014 women (first sample); 86 men and 202 women (validation sample).
Main outcome measures: Waist circumference, body mass index, waist:hip ratio.
Results: Waist circumference >/=94 cm for men and >/=80 cm for women identified subjects with high body mass index (>/=25 kg/m2) and those with lower body mass index but high waist:hip ratio (>/=0.95 for men, >/=0.80 women) with a sensitivity of >96% and specificity >97.5%. Waist circumference >/=102 cm for men or >/=88 cm for women identified subjects with body mass index >/=30 and those with lower body mass index but high waist:hip ratio with a sensitivity of >96% and specificity >98%, with only about 2% of the sample being misclassified.
Conclusions: Waist circumference could be used in health promotion programmes to identify individuals who should seek and be offered weight management. Men with waist circumference >/=94 cm and women with waist circumference >/=80 cm should gain no further weight; men with waist circumference >/=102 cm and women with waist circumference >/=88 cm should reduce their weight.
Body mass index is commonly used to identify those with a health risk from being overweight
This study shows that waist circumference may be a simpler measure for identifying need for weight management
Most men with waist circumference >/=102 cm and women with waist circumference >/=88 cm were appreciably overweight or had a high waist:hip ratio and should be urged to lose weight
Waist circumference 94-102 cm in men and 80-88 cm in women should be a warning to avoid weight gain
About half of all British adults have a body mass index (weight (kg)/(height (m)2)) of >25, while almost 15% have an index of >30, and these proportions are rising.1 2 Given the lack of success in the management of obesity and increasing associated health costs,3 4 greater emphasis on prevention is needed, particularly in young people, who often have little contact with health services. While many health professionals are now familiar with the acceptable range for body mass index (20-25),2 5 most members of the public cannot readily calculate their index to establish their own risk or need for weight management. Charts developed by the Health Education Authority are helpful but are still not understood by many.6 Height must be measured accurately as small errors in the denominator are exaggerated by squaring.
The major metabolic cardiovascular risk factors (high blood pressure, plasma lipids, insulin resistance) all aggregate independently with both body mass index and waist:hip ratio7 8 and improve with weight loss.9 10 11 12 13 The circumference of the waist relates closely to body mass index and is also the dominant measurement in the waist:hip ratio, which reflects the proportion of body fat located intra-abdominally, as opposed to subcutaneously,7 and waist circumference is the best indicator of changes in intra-abdominal fat during weight loss.14
We evaluated waist circumference as a simple predictor of health risk from being overweight and through the central distribution of fat to indicate levels at which individuals should take action.
We randomly recruited 904 men and 1014 women, aged 25 to 74 years, from the general population of north Glasgow between January and August 1992, excluding only those who were chair bound.
We recruited separately 86 men and 202 women by advertising locally to test the proposed “action levels” (levels at which individuals may be at risk from being overweight) derived in the determination study.
All measurements were made by trained observers with standard techniques15: weight by digital scales (Seca, Germany) to within 100 g, without heavy clothing; height barefoot by portable stadiometer (Holtain, Crymych, United Kingdom) to within 0.5 cm; circumferences to within 1 mm with plastic tapes calibrated weekly, with waist mid-way between the lowest rib and the iliac crest with the subject standing at the end of gentle expiration, and hips at the greater trochanters. We used the same methods to determine and validate studies by different researchers.
METHODS OF ANALYSIS
Combination with indices in the range height*RF 0.1* to height2 did not improve the correlations of waist circumference alone with body mass index (r=0.89; P<0.001 for both sexes). According to the criteria of Khosla and Lowe,16 height was therefore not used with waist circumference for further analyses.
We determined by cross tabulation between variables17 two action levels for waist circumference for weight management to identify most subjects with a body mass index >/=25 (action level 1) and >/=30 (action level 2), while including a minimum of subjects who had lower body mass index and low waist:hip ratio to maximise sensitivity and specificity.18 High body mass index was defined at two levels as >/=25 or >/=30 for both men and women.6 On the basis of consensus from previous studies,9 10 11 12 13 high waist:hip ratio was defined as 0.95 for men and 0.80 for women and low waist:hip ratio as below these cut offs.
The mean age (range 25 to 74 years), body mass index, and hip circumference were similar for men and for women (table I). However, men were heavier and taller and had a larger waist circumference and waist:hip ratio than women. In both sexes waist:hip ratio correlated positively with body mass index (men, r=0.63; women, r=0.39; both, P<0.001) (table II). Waist circumference correlated weakly with height in men (r=0.19, P<0.05) but not women (r=0.06, P=0.06). On average, men with a waist circumference above action level 1 were 2 cm taller than those below, accounting for 0.7 cm difference in waist circumference, but action levels for waist circumference determined for four different height categories (not shown) did not differ.
The figure shows distributions of the waist circumference related to body mass index divided by the proposed action levels. Table III shows the numbers of subjects in different categories of waist circumference, body mass index, waist:hip ratio. “True positive” subjects were those with high body mass index and those with lower body mass index but high waist:hip ratio; “true negative” subjects were those with low body mass index and those with high body mass index but low waist:hip ratio. “False positive” subjects were those with waist circumference above the action level but with low body mass index and waist:hip ratio; “false negative” subjects were those with waist circumference below the action level but with high body mass index and waist:hip ratio. These numbers were used to determine the sensitivity and specificity for waist circumference as an indicator of need for weight management (table IV).
False negative subjects, who would be missed by health education programmes that use waist circumference as an indicator of need for weight management, would represent less than 1.5% of the population. False positive subjects (1.7% of the population) would be included inappropriately in health promotion directed at those with high waist circumference. Among the subjects who were falsely positive at action level 1, men had a body mass index of 23.2-24.8 and a waist:hip ratio of 0.91-0.95, and women had a body mass index of 23.0-24.9 and a waist:hip ratio of 0.77-0.80. Among the subjects who were falsely positive at action level 2, men had a body mass index of 26.9-29.7 and a waist:hip ratio of 0.88-0.94, and women had a body mass index of 28.7-29.6 and a waist:hip ratio of 0.79-0.80. The misclassified subjects were therefore close to the levels at which body weight control would be recommended, and the hazards of health promotion based on waist circumference action levels would be minimal.
The validation sample was younger than the determination sample, but had similar ranges of anthropometric measurements (table I). The action levels for waist circumference showed similarly high sensitivity (>94%) and specificity (>97%) for identifying high risk subjects in need of weight management (tables III and IV). Waist circumference and height were not significantly related in either men (r=0.07; P=0.54) or women (r=-0.13; P=0.07).
The influences of body mass index and waist:hip ratio on metabolic and cardiovascular disease are multiplicative,19 20 so weight loss should be urged for all those with a high body mass index and can also be justified for those with a lower body mass index but high waist:hip ratio. Consensus emerging from prospective studies suggests cut offs of waist:hip ratio 0.95 in men and 0.80 in women—as values above which health risk increases appreciably. Metabolic risk factors, particularly serum concentrations of triglycerides and high density lipoprotein cholesterol, improve most with weight loss in men with a waist:hip ratio >/=0.9521 22 23 and women with a waist:hip ratio >/=0.80 (table V).10 11 12 In our study 48% of the women and 38% of the men had waist:hip ratios above these figures (table II); just over half had a body mass index 25 or above. The interaction between body mass index and waist:hip ratio offers an opportunity through measurement of waist circumference to target individuals at highest risk to maximise the benefits of interventions.24 25
Many analyses have found strong positive correlations between waist circumference and body mass index. This relation alone permits only limited prediction of body mass index: in our study, using waist circumference to identify those with body mass index >/=25 would have led to misclassification of about 4% of men and 7% of women and to identify those with body mass index >/=30, 10% of men and 11% of women. Health promotion targeted by waist circumference at those with body mass index >/=25 would miss about 21% of the total at risk population—that is, those with an index >/=25—and about 10% of those identified would be targeted inappropriately. However, most of those with high waist circumference but body mass index below these conventional cut offs have a high waist:hip ratio, which still justifies weight management. Most of those with higher body mass index but waist circumference below the action level have low waist:hip ratio, which would indicate a lower health risk, and less benefit from slimming. The action levels for waist circumference derived here, based on both body mass index and waist:hip ratio, are robust in that they led to misclassification of only 1.5% of the overweight men and women.
The simplicity of measurement and its relation to both body weight and fat distribution are major advantages for waist circumference over body mass index and waist:hip ratio. Self measurement and reporting of waist circumference has been reported to be acceptable in recent epidemiological studies,26 27 but better information will be needed about possible self reporting bias and about ability to monitor changes with weight management. Waist circumference is more strongly associated with metabolic function, however, than with waist:hip ratio in adults22 and in children28 and predicts myocardial infarction.20 The proposed action levels match the results of Chan et al, who found progressively increasing relative risk of developing non-insulin dependent diabetes in men as waist circumference rose from 73.7-87.6 cm to 91.7-96.5 cm (relative risk 2.2) and to >102.0 cm (12).27 Pouliot et al observed exponential increases in cardiovascular risk factors with waist circumference above 87 cm in men and 78 cm in women (which correspond to waist circumference of action level 1) and further risk factor increases with waist circumferences above those of action level 2.29 Proof of the value of waist circumference action levels in predicting health risks will require longitudinal follow up of morbidity and mortality. Longitudinal data from the Framingham study suggests that waist predicts mortality better than other anthropometric measures.30
In conclusion, the action levels for waist circumference (measured by using bony landmarks mid-way between the iliac crests and the lowest ribs) that have been identified in our study could form the basis on which health promotion might raise awareness or urge action on weight reduction. The lower action level of waist circumference (94 cm for men and 80 cm for women) represents a threshold above which health risks are increased—particularly for young men.27,31 Further weight gain and rise in waist circumference from action level 1 towards action level 2 should be discouraged. The upper action level (102 cm for men and 88 cm for women) correspond with the point at which symptoms of breathlessness and arthritis begin to develop from overweight, and the health risks are such that medical consultation and weight loss should be urged.
We thank Dr James Currall (Department of Computing Service, University of Glasgow) for statistical guidance. Data for the determination sample were kindly provided by the Glasgow arm of the MONICA project, which was funded (as was CEM) by the Chief Scientist's Office, Scottish Home and Health Department. TSH was supported by the Department of Human Nutrition Discretionary Funds, University of Glasgow.