- Top of page
- Research Methods and Procedures
Objectives: To examine the power of the combined measurements of body mass index (BMI) and waist circumference (WC) for the prediction of abnormality in coronary heart disease risk factors and to determine whether the additional measurement of WC is predictive in older men and women.
Research Methods and Procedures: 1190 men and 751 women of the Baltimore Longitudinal Study of Aging were dichotomized into younger (<65 years) and older (65+ years) age groups. Coronary risk factors in the realms of glucose/insulin metabolism, blood pressure, and plasma lipids were assessed. The relationship of BMI and WC, singly and combined, to 10 risk factors for coronary heart disease was examined.
Results: In younger and older men and women, BMI and WC are highly correlated (0.84 to 0.88). BMI and WC are also significantly correlated to all 10 coronary risk factors in younger men and women and to 8 of the 10 in the older men and women. Both partial correlation and logistic regression analyses revealed a modest but significant improvement in the prediction of coronary risk in younger men and women by WC after controlling for the level of BMI. There was no improvement in the older subjects.
Discussion: WC adds only modestly to the prediction of coronary risk in younger subjects once BMI is known, and adds nothing to the production of risk in older subjects.
- Top of page
- Research Methods and Procedures
In 1998, both the World Health Organization (WHO) (1) and the National Heart, Lung, and Blood Institute (NHLBI) (2) issued comprehensive recommendations for classifying abnormalities in body weight and body fat distribution. Both reports recommended body mass index (BMI) and waist circumference (WC) as measures of obesity and fat distribution. Both reports summarized the evidence relating these measures to health risk. The WHO report did not attempt to quantify risk for simultaneously measured BMI and WC. The NHLBI report does consider the impact of the two variables combined. In table ES-4 of that report, it is indicated that WC does not modify the risk classification in subjects who are underweight or normal weight and that WC does not modify the classification of risk in subjects in the higher obesity categories of BMI, that is, in those with BMI of 35 kg/m2 or greater. However the NHLBI table does indicate a higher degree of risk if critical WC levels are exceeded in men and women whose BMI lies in the overweight zone (25.0 to 29.9 kg/m2) and in the obesity class I zone (30.0 to 34.9 kg/m2). However, this increased risk is characterized only qualitatively. Thus, if the WC exceeds the critical cut-points of 102 cm for men or 88 cm for women, then the risk in overweight individuals is described as “high,” in contrast to “increased” in individuals below those WC cut-points. Similarly, in obesity class I individuals, a “high” WC confers a “very high” risk compared with a high risk if the WC is not increased. These are obviously qualitative descriptors; uncertainty exists about the exact increase in risk conferred by a large waist.
The WHO report presents a more complex WC classification than does the NHLBI report in that two cut-points are indicated for men (94 and 102 cm) and two for women (80 and 88 cm). Thus, individuals fall into one of only two zones by the NHLBI classification but into one of three zones by the WHO report.
A natural question arises: is either set of standards applicable to older individuals? Zamboni et al. (3) indicated that the amount of visceral adipose tissue related with WC was significantly higher in older than in younger individuals. Also, Borkan et al. (4) showed that the intra-abdominal fat area measured by computed tomography was greater in older than in younger men, although weight was 8.2 kg greater in younger than in older men. Molarius and Seidell (5) emphasized the need to examine the possibility that there could be age-related differences in the contribution of the pattern of fat distribution to risk factors. However, the WHO and NHLBI reports barely address the question of the applicability of BMI and WC standards for older persons.
We (6) (7) recently reported that at least with respect to the effects of BMI and WC on the traditional coronary heart disease risk factors, these anthropometric variables, considered independently of each other, continue to be associated with harmful effects, even in men and women over 65 years old, although at a lower level of significance than in younger individuals.
The other question that remains is whether measurement of WC gives additional information to the estimation of coronary risk that is predicted by BMI alone in younger and older men and women.
In this report, we have tested the hypothesis that although BMI and WC are highly correlated with each other, the quantification of risk is improved if both variables are considered together. Furthermore, we hypothesized that this conclusion would also apply to older individuals, albeit at a weaker level than in younger individuals.
- Top of page
- Research Methods and Procedures
Previous reports on the BLSA population showed that BMI and WC independently were significant predictors of coronary risk factors in the blood pressure, glucose insulin, and plasma lipid domains in younger and older men and women (6) (7). Furthermore, the gradation of BMI into normal, overweight, and obese zones according to NHLBI and WHO recommendations (1) (2) was supported. Also, the sex-specific WHO cut-points for WC that provided three zones (NHLBI standards provided two zones) were also found to be applicable to the risk factors in the four age/sex categories.
Both WC and BMI have been shown to be related to cardiovascular disease risk factors (17) (18) (19) (20) (21). Although the waist-to-hip ratio had been the favored index of an unfavorable pattern of fat distribution, in recent years WC has become the preferred measure of body fat distribution (22) (23) (24). However, the important question of whether assessment of WC adds to the estimation of coronary risk predicted by BMI alone remains. Certainly, if WC is added to BMI for assessment of risk, as Booth et al. (25) found, then the percentage of individuals in a population classifiable as being overweight by BMI criteria alone will be significantly increased. As we have shown (see Figure 1), there are younger and older men and women whose BMI is in the normal zone who fall in the intermediate WC zone (1), while those with BMIs in the 25- to 29-kg/m2 range (now identified as overweight) will fall within all three WC zones. However, the question still remains whether WC gives additional information to the estimation of coronary risk to that predicted by BMI alone in younger and older men and women.
In the present study, this question was examined by two statistical methods: partial correlation, in which WC and BMI and the 10 coronary risk factors were analyzed as continuous variables and logistic regression, in which these same variables were stratified into well-defined categories. As noted in the Research Methods and Procedures section, the lower WHO cut-points for WC were used. Table 6 summarizes these analyses along with the results of the simple bivariate correlation of WC and the 10 risk factors (unadjusted for BMI).
Table 6. Summary of the statistical significance of the relationship of WC to 10 coronary risk factors by three methods of analysis (see Discussion section)
| ||Bivariate||Partial correlation||Logistic analysis||Bivariate||Partial correlation||Logistic analysis||Bivariate||Partial correlation||Logistic analysis||Bivariate||Partial correlation||Logistic analysis|
|Systolic blood pressure||†||†||†|| || || ||†||*|| ||*|| || |
|Diastolic blood pressure||†||†||†||†|| || ||†||*||*||*|| || |
|Fasting glucose||†||†||†||†||*|| ||†||*|| ||†||*|| |
|2-hour glucose||†||†||*||†||*||†||†||*|| ||†|| || |
|Fasting insulin‡||†||†||†||†|| || ||†|| ||†||†|| || |
|HOMAIR‡||†||†||†||†||*||*||†||*||†||†|| || |
|Total cholesterol||†||†|| || || || ||†||*|| || || || |
|Triglyceride||†||†||†||†|| || ||†||†||*||†||*|| |
|HDL cholesterol||†||†||*||†|| || ||†||*||†||†|| || |
|LDL cholesterol||†||†|| ||*|| || ||†||*|| || || || |
To place this study in the context of previous publications, we have summarized the data from 25 studies that, at a minimum, had simultaneous measures of fatness (usually BMI) and of fat distribution (usually anthropometric measurements that included the WC) and also presented correlation coefficients of these measures with one or more of the traditional coronary risk factors (19) (24) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37) (38) (39) (40) (41) (42) (43) (44) (45) (46) (47) (48). We are preparing a detailed analysis of these reports but will limit the present discussion to a summary of that analysis. For clarity of presentation, indices of fatness will be referred to as BMI and indices of fat distribution will be referred to as WC.
Five questions were addressed:
. Does fatness or fat distribution pattern correlate more strongly with other risk factors? In all, the data provide 178 correlations among BMI, WC, and other risk factors. The importance of these indices is shown by the fact that r values were statistically significant in 150 of the 178 correlations. However, there was no clear superiority of BMI or WC in the overall strength of the associations.
. Which of the risk factor domains (blood pressure, glucose/insulin, or lipids) are most strongly related to the anthropometric measures? Correlations with total and LDL cholesterol were decidedly the least consistent: 51% of the r values were not statistically significant. In contrast, only 8% of the correlations with blood pressure, 5% to 6% of the glucose and insulin correlations, and 8% of the triglycerides and HDL cholesterol correlations failed to reach statistical significance. Only in the triglycerides/HDL domain was there clear superiority of the WC measurement over BMI. There was no clear superiority of WC or BMI in any of the other domains.
. Is there an age difference in the strengths of associations? The present report is the only study that reports both younger and older men and women in the same population. As we have shown, the correlations are very much stronger in the younger individuals.
. Is there a sex difference in the strengths of associations? There is an interesting sex difference in the strength of the BMI and WC associations with three lipid moieties—triglycerides, HDL, and LDL cholesterol, but this is true only in the older individuals, the correlations in older men being stronger than those in older women. In 19 of the 22 comparisons in the literature, r values were higher in men.
. Does the measurement of the fat distribution pattern add strength to that of BMI alone in the prediction of abnormality in the associated risk factors? This question was approached in the present study and in 9 of the other 25 reports by conducting partial regressions of WC (controlling for BMI) on the coronary risk factors (24
). The results of these partial correlations are highly age-dependent. In younger men, 32 of the 37 analyses (86%) and in younger women 12 of the 15 analyses (80%) showed WC to remain a statistically significant correlate. In contrast, only 5 of 20 (25%) and 4 of 17 (24%) of the analyses in the older men and women achieved statistical significance. Thus, it is clear that WC is quite a consistent independent influence on coronary risk factor in younger but not in older men and women.
In summary, WC and BMI independently are significantly related to the risk factors in all four demographic groups in nearly all cases. When BMI is brought into the analysis, WC remains a significant predictor for most of the variables in younger men and women, but significance is almost entirely lost in older men and women. The logistic regression technique confirms this conclusion in that significance of WC is almost entirely lost in older men and, indeed, in older women, no risk factor remains significantly related to WC. Thus, the effects of obesity on coronary risk factors are captured almost entirely by the measurement of BMI in older individuals, but, in younger men and women, WC still adds significantly to the assessment of risk by BMI alone. The explanation for this age difference is not clear. Mykkänen et al. (35) discuss some possibilities extensively. They note that equal BMIs in younger and older adults represent different degrees of fatness; lean body mass, especially, muscle mass, and bone mass decrease with aging, whereas fat mass increases. Thus, equal body weights in young and old individuals represent different degrees of fatness. BMI then could be expected to increase in predictive power with age. Furthermore, the relationship of WC to intra-abdominal fat changes with age. The relative distribution of subcutaneous to intra-abdominal fat probably changes, and abdominal wall laxness may increase with age so that a simple measurement of WC, although still predictive in itself of other risk factors, may not be as reliable a measure in older individuals.
It must be noted that these conclusions are based on studies in a Europid population and comparable studies in other racial/ethnic groups are needed. In addition, a truism of cross-sectional studies in elderly individuals is that only survivors can be evaluated; the truly predictive power of WC needs to be quantified in prospective studies. It also must be emphasized that we used only lower cut-points for BMI (25 kg/m2) and WC (94 cm for men and 80 cm for women) due to the small number of subjects in the obese and high WC groups. Because the variables examined are risk factors for coronary heart disease, direct analyses of the predictive power of WC (with and without BMI) on the development of coronary heart disease would be instructive. In larger populations, it should be possible to define more accurately than we were able to do, the age range at which the predictive power of WC is lost. We empirically chose 65 years old as a definition of an older population, but the effects of WC in, for example, decades of the lifespan, would be important. Finally, the high degree of correlation between BMI and WC in this population introduced the problem of colinearity. Colinearity inflates the estimates of variance in both the continuous and categorical analyses when BMI and WC are entered into the regression models. The four correlation coefficients between BMI and WC in younger and older men and women in our study ranged from 0.84 to 0.88. These values agree closely with those reported in three other studies (49) (50) (51). However, because the correlations between BMI and WC were so similar in younger and older men and women, we believe the conclusion of the relative importance of WC still applies.