Methods

Results

During 16 years of follow-up (687 519 person-years), we identified 3507 deaths, including 751 CVD deaths and 1748 cancer deaths. At baseline (1986), higher levels of WC, WHR, and waist-to-height ratio were associated with lower levels of physical activity and a lower likelihood of postmenopausal hormone use but with a greater likelihood of clinical hypertension and hypercholesterolemia (Table 1). The different anthropometric measures of abdominal obesity were highly correlated (Table 2). In addition, these measures were all significantly correlated with BMI, although the correlation for WHR (r=0.33) was not as strong as those with WC (r=0.81) and waist-to-height ratio (r=0.83).

After adjustment for age, smoking, and other covariates, both increasing WC and WHR were strongly associated with a graded increase in all-cause mortality (Tables 3 and 4⇓). In general, the association for WC became stronger after adjustment for BMI. Similar results were observed in sensitivity analysis when BMI was fitted as continuous variable or when a nonlinear term of age (age²) was added to regression models. Spline regression showed linear association of both WC and WHR with all-cause, CVD, and cancer mortality (Figure 1), and probability values for test for departure from linearity were not significant (P>0.05). In stratified analyses, the associations of WC and WHR with mortality were not appreciably different between never and ever smokers, between younger (≤55 years of age) and older (>55 years of age) women, and between premenopausal and postmenopausal women. Even among women who were neither overweight nor obese, both WC and WHR were strongly associated with mortality. Similar associations were observed between waist-to-height ratio and mortality. After adjustment for BMI and other covariates, RR and 95% confidence interval (CI) across the lowest to the highest quintile of waist-to-height ratio were 1.00, 1.09, 1.09, 1.22, and 1.54 (95% CI, 1.32 to 1.78) for all-cause mortality; 1.00, 0.81, 0.78, 1.20, and 1.52 (95% CI, 1.11 to 2.09) for CVD mortality; and 1.00, 1.33, 1.21, 1.28, and 1.56 (95% CI, 1.26 to 1.93) for cancer mortality (all P<0.001 for trend).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. RRs of all-cause, CVD, and cancer mortality according to WC (A) and WHR (B). RRs are adjusted for age (5-year interval), quintiles of physical activity, alcohol consumption (0, 0.1 to 4.9, 5.0 to 14.9, ≥15.0 g/d), family history of MI, smoking status (never, past, current [1 to 14, 15 to 24, 25+ cigarettes a day]), menopausal status and hormone use (premenopausal status, postmenopausal status, hormone [never use, past use, current use]), and BMI (<18.5, 18.5 to 24.9, 25 to 29.9, 30 to 34.9, ≥35 kg/m²). Solid black line represents point estimate; dashed lines are 95% CIs.

The association of hip circumference with CVD and cancer mortality was J shaped after adjustment for age, smoking status, physical activity, and other lifestyle and medical covariates so that the lowest mortality was observed among those in the third quintile (Table 5). After adjustment for WC, greater hip circumference was associated with lower all-cause and CVD mortality but not cancer mortality.

We next examined the joint effects of abdominal (WC and WHR) and overall (BMI) adiposity using the currently recommended cutoffs for BMI (≥25 or ≥30 kg/m²), WC (≥35 in [88 cm]), and WHR (≥0.88) (Figure 2). Because smokers tend to weigh less but have much higher mortality rates than nonsmokers, which can lead to the artifact of an apparent elevation in risk of mortality among leaner individuals, we restricted the analyses to never smokers. In addition, because a very low BMI in some older adults may reflect ill health that we could not completely control for even after excluding prevalent cancer and heart disease, we excluded underweight women (BMI <18.5 kg/m²). The highest mortality risk was observed among those who had both abdominal and overall obesity. However, even among normal-weight women, abdominal obesity was associated with significantly higher risk for CVD mortality: RR associated with larger WC ≥88 cm [35 in] was 3.02 (95% CI, 1.31 to 6.99) and for WHR >0.88 was 3.45 (95% CI, 2.02 to 6.92). In addition, within each category of WC or WHR, higher BMI was associated with increased total, CVD, and cancer mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Multivariate RRs of all-cause, CVD, and cancer mortality according to joint classification of BMI and WC (A) or WHR (B) among never smokers. RRs are adjusted for age (5-year interval), quintiles of physical activity, alcohol consumption (0, 0.1 to 4.9, 5.0 to 14.9, ≥15.0 g/d), family history of MI, and menopausal status and hormone use (premenopausal status, postmenopausal status, hormone use [never use, past use, current use]).

Discussion

In the present study of a large prospective cohort of US women, we observed a monotonic dose–response relationship between abdominal adiposity, as measured by WC, WHR, or waist-to-height ratio, and the risk of death among both younger (<55 years of age) and older (≥55 years of age) women and among both never and ever smokers. This positive association persisted in normal-weight women.

The prospective design and high rate of follow-up in this study minimized the possibility of recall bias or bias resulting from loss of follow-up. Furthermore, the large size of the study increased the precision of the RR estimates, and the extensive information on potential confounders allowed us to control for confounding in detail. One potential limitation is that the anthropometric measures of adiposity in the present study were by self-report. However, the validity of self-reported weight and waist and hip circumferences compared with technician measurements was high in this population of health professionals.^15,23 Because of the prospective design of this study, misclassification would be nondifferential and expected to bias the risk estimate toward the null. A major concern in analyzing the relationship between obesity and subsequent mortality is the problem of reverse causation (ie, weight loss can be the result, rather than the cause, of underlying illness).²⁴ To address this concern, we excluded participants with existing CVD and cancer at baseline.

Although a number of epidemiological studies have demonstrated that measures of abdominal adiposity significantly predict chronic diseases such as CVD and diabetes mellitus independently of overall body adiposity,⁷ the associations of these measures with premature death have not been widely studied, and previous findings^8–12 have been inconsistent. The inconsistencies of findings may be related to differences in study populations and sampling, measures of abdominal adiposity, and analytic approaches. Our findings are in general consistent with those supporting an independent contribution of body fat distribution to mortality and the importance of abdominal adiposity in predicting mortality in women. In the Iowa Women’s Health Study,⁹ both WC and WHR were significantly associated with mortality, particularly coronary heart disease mortality. Likewise, in both a Danish cohort¹⁰ and a Swedish cohort¹² of middle-aged and older women, abdominal adiposity was strongly and positively associated with all-cause mortality after adjustment for BMI. The follow-up time was short, and only findings related to WC were reported in the Danish cohort, whereas only findings for WHR were reported in the Swedish cohort. In a recent study of elderly women (75 years of age) in the United Kingdom,⁸ WHR but not WC was positively related to mortality in nonsmoking women only, mainly because of cardiovascular mortality. By contrast, in a relatively small Dutch cohort of elderly women,¹¹ neither WHR nor WC was significantly associated with all-cause mortality. Data on the association of abdominal adiposity and fat distribution with overall cancer mortality are sparse. In the Iowa study,⁹ neither WC nor WHR was significantly associated with cancer mortality after adjustment for multiple risk factors. Association of measures of abdominal adiposity with cancer mortality was not reported in most studies of abdominal adiposity and mortality.^8,10–12 With >600 000 person-years of follow-up and 3507 deaths, ours is one of the largest studies with the longest follow-up duration on abdominal adiposity and all-cause and cause-specific mortality.

The metabolic effects of abdominal adiposity are well established. Greater abdominal adiposity is closely associated with adverse metabolic profiles such as insulin resistance, dyslipidemia, and systematic inflammation, which play essential roles in the pathogenesis of CVD, diabetes mellitus, and certain cancers.^2,3 Of note, the association of abdominal obesity with these adverse metabolic profiles persisted among normal-weight women.²⁵ For instance, higher WC was significantly related to higher total cholesterol, low-density lipoprotein cholesterol and triglyceride levels, higher systolic and diastolic blood pressures, and higher fasting glucose levels among normal-weight women.

There have been continuous debates on whether WC or WHR is a better measure of abdominal adiposity in predicting health risks in epidemiological studies. WC as a measure of both subcutaneous and visceral fat can be measured easily. However, WC also is correlated with body frame size; thus, WHR often is used instead.²⁶ In the present study, both measures were significantly associated with all-cause, CVD, and cancer mortality. WHR did not seem to provide substantially better prediction than WC. Because WHR requires measures of both waist and hip circumference and is more difficult to interpret, WC is more applicable in clinical practice.

The inverse association of hip circumference with mortality (mainly CVD mortality) after adjustment for WC is of interest. It suggests that given a certain level of WC, a greater hip circumference, may be associated with lower CVD mortality. Such an association may be explained by larger peripheral fat mass associated with greater hip circumference.⁷ Recent studies suggested that the femoral-gluteal fat depot, for a given amount of abdominal fat, plays a protective role by acting as a “sink” for circulating free fatty acids²⁷; the femoral-gluteal region is more likely to effectively take up free fatty acids from the circulation and is less likely to release them readily.⁷ Because of the increased free fatty acid uptake in the femoral-gluteal region resulting from a larger hip circumference, detrimental ectopic fat storage in tissues or organs (ie, the liver, skeletal muscle, and pancreas) and subsequent adverse metabolic effects may be prevented.²⁸ Indeed, studies using dual-energy x-ray absorptiometry or computed tomography to estimate fat and muscle content at the legs found that subcutaneous fat at the legs was associated with a more favorable cardiovascular risk profile (for a given amount of abdominal fat).^29–32

Our study population consisted of registered nurses. The relative homogeneity in occupation and education can reduce confounding by socioeconomic factors. Because the study population was predominantly white, our findings require confirmation in other ethnic groups. In a recent study among Chinese women,³³ a significant positive association between WHR and mortality also was observed.

Conclusions

Our data indicate that abdominal adiposity is significantly and positively associated with all-cause and cause-specific mortality. Greater WC or WHR is associated with increased CVD mortality even in normal-weight women. Although maintaining a healthy weight should continue to be a cornerstone in the prevention of chronic diseases and premature death,³⁴ it is equally importanty to maintain a healthy waist size and prevent abdominal obesity.