点击显示 收起
1 From the Department of Human Nutrition, University of Glasgow.
See corresponding article on page 743.
In this issue of the Journal, Zhu et al (1) present the findings of a cross-sectional study, which support the consistent conclusion from a large body of literature that waist circumference is at least as strong as is body mass index (BMI; in kg/m2) in predicting cardiovascular disease (CVD) risk (2). The new element in this article is the presentation of figures of waist circumference that give degrees of risk equivalent to those of the conventional BMI cutoffs of 25 and 30. For this reasonably representative healthy, nonpregnant, white population aged 20–90 y, the presence of one or more CVD risk factors was associated with a waist circumference of 90 cm for men and of 83 cm for women to a similar extent as was a BMI of 25; the presence of one or more CVD risk factors was associated with a waist circumference of 100 cm for men and of 93 cm for women to a similar extent as was a BMI of 30. Logically, albeit confusingly, the odds ratios on which these figures were based used a BMI of 23.1 for men and of 21.8 for women as the references—these figures provided the lowest CVD risk estimation from having 1 of the 4 cutoffs used.
Anthropometric measures can in principle be useful for 3 rather separate purposes.
First, for epidemiologic and health-economic planning, simple survey data can be used to estimate the health risks of a population and their costs or resource implications. The value of population-directed preventive measures could be estimated from the expected effect on BMI or waist, if a reliably effective intervention were to be developed. The need is often restricted to one disease outcome—CVD. For these purposes, an accurate estimate of population risk is required. Misclassification of individuals is unimportant, and small subgroups at the extremes of the anthropometric measures or health risk are of relatively little importance. For this purpose, waist circumference measured in a standardized way is clearly superior to BMI. Receiver operating characteristic analysis is appropriate, and the correlation with CVD risk is tighter across the range of the vast majority of the population. Data are still sparse for some populations and racial groups.
A second valid use of anthropometric measures is the triggering of clinical decisions as part of management algorithms. The same principles apply for computation of risk in health insurance. Here the need is to quantify the health risk of an individual, and there is particular interest in those with extreme values of anthropometric measures. For this purpose, in the surgery consulting room or the insurance office, there are practical problems with waist circumference. Clothing should be removed from the midriff, which can present problems of cultural sensitivity. With extreme obesity, there is a limit to waist circumference imposed by gravity, such that further increases in weight—even enormous increases—are not reflected by waist circumference because fat mass hangs toward the knees or ankles. Waist circumference is one of the most reproducible anthropometric measures of girth, but, even in relatively normal ranges, the SE of estimating waist circumference is greater than of weight or height (3). Thus, for clinical decision making, the most appropriate measure is BMI, assuming that regularly calibrated scales are used and that height, particularly, is measured carefully. Consulting rooms abound with floppy stadiometers and with staff who do not recognize the Frankfort plane. To monitor change, only weight is required.
Third, the original reason for developing waist circumference as a health indicator arose from the need to improve health promotion by providing the general public with "action levels" (4). Action level 1 was intended as a first warning above which individuals need to take action to prevent further weight gain because multiple health hazards start to develop. Waist circumferences of 94 cm for men and of 80 cm for women were proposed on the grounds that these values would identify virtually everyone with a BMI 25 and that most of the persons with a false-positive result (those with a BMI that was actually < 25) would have high waist-to-hip ratios.
Action level 2 (waist circumferences of 102 cm for men and of 88 cm for women) would identify almost everyone with a BMI 30 (ie, a high health risk); again, persons with a false-positive result would have a high waist-to-hip ratio and should still be advised to seek professional help. For these specific health-promotion needs directed at the general public, it is appropriate to set relatively low action levels with the aim of delaying the progression of risk. False-positive results are not a serious problem provided nobody is completely misclassified, and waist circumference does this job very well. There are practical advantages in that most people (men at least) already know their waist circumference, whereas computation and conceptualization of BMI is problematic for the general public, even when graphic aids are used.
The accepted but erroneous thinking in the 1980s was that BMI gave an indication of total body fat content and that the waist-to-hip ratio reflected fat distribution. More recently, it has become clear that hip circumference has some value in predicting health, independent of waist (5,6). The waist-to-hip ratio is an artificial, derived term, with no biological meaning: it does not usefully reflect fat distribution or health better than waist alone. Waist circumference, perhaps surprisingly, is the best simple anthropometric measure of total body fat, is better than BMI (7), and is also the best simple indicator of intraabdominal fat mass (8,9). It would be useful to distinguish these components, but this is not possible without resorting to the use of scanning methods.
The paper by Zhu et al is valuable in pinpointing the degree of risk in whites with different waist circumferences. To know the risk equivalence of a BMI of 25 or 30 could be of value to planners and insurance companies. It would have been most logical to have based the action levels on a direct risk analysis of the kind used in the article by Zhu et al but using the actual incidence of CVD, not the risk imputed from selected risk factors. For health promotion, the original action levels seem to suit many subpopulations and to reflect the risks of many health outcomes. Zhu et al only considered CVD risk, for which the original action levels appear to be a little generous to men but are slightly lower than necessary for women. However, there are many other expensive, debilitating, and more clinically immediate, potentially preventable consequences of overweight. To base action levels solely on CVD risk would be to ignore most of our patients’ symptoms.
REFERENCES