Hip Circumference and Cardiovascular Morbidity and Mortality in Men and Women

Authors

  • Berit L. Heitmann,

    1. Research Unit for Dietary Studies and the Danish Epidemiology Science Centre, Institute of Preventive Medicine, Copenhagen, Denmark
    2. Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark
    Search for more papers by this author
  • Peder Frederiksen,

    1. Research Unit for Dietary Studies and the Danish Epidemiology Science Centre, Institute of Preventive Medicine, Copenhagen, Denmark
    Search for more papers by this author
  • Lauren Lissner

    Corresponding author
    1. Department of Primary Health Care, Sahlgrenska Academy, Gothenburg University, Gothenburg, Sweden
    2. Nordic School of Health, Gothenburg, Sweden
    Search for more papers by this author

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Research Unit for Dietary Studies at the Institute of Preventive Medicine, Kommunehospitalet, 1399 Copenhagen K, Denmark. E-mail: BLH@ipm.hosp.dk

Abstract

Objective: It has recently been demonstrated that, in middle-aged women, a wide hip circumference is a protective factor for a number of health endpoints in later years. The effect seems to be independent of both overweight and waist circumference. This paper aims to replicate this finding in another population-based sample consisting of women and men.

Research Methods and Procedures: This was a prospective observational study consisting of a random subset of adult Danes. A total of 2987 subjects born in 1922, 1932, 1942, or 1952 and 35, 45, 55, or 65 years of age (at examination in 1987 to 1988) participated in the Danish MONICA (MONItoring trends and determinants of CArdiovascular disease) project, with measurements of height, weight, and hip and waist circumference taken. Through personal identification numbers, incidence of coronary heart disease (CHD) and cardiovascular disease (CVD) until the end of 1998 and all causes of death until 2001 were retrieved from the National Registers of Hospital Discharge. There was an average of 10 years of follow-up for incidence of CVD and CHD and 13 years of follow-up for total mortality.

Results: Large hip circumference, relative to body size and waist circumference, predicted less incidence of CVD, CHD, and total death in women. This was not the case in men; BMI and waist circumference were the strongest independent predictors.

Discussion: A large hip circumference seems to have independent and positive effects on CVD and CHD morbidity and mortality in women, but no protective effect on cardiovascular health in men. However, a borderline significant protective effect on total mortality was observed.

Introduction

The associations between BMI and mortality from both coronary heart disease (CHD)1 and cardiovascular disease (CVD) are well established, as are associations with abdominal obesity measured as waist-to-hip ratio and waist circumference (1). More recently, it has been proposed that hip measurements should be discarded, with a concentration on waist circumference only, because waist circumference provides a more practical correlate of abdominal fat distribution and associated ill health than waist-to-hip ratio (2). However, because hip circumference has been found to be a significant protective risk estimator for both myocardial infarction and CVD, after adjustment for covariation in BMI and waist circumference, we have recently argued that it is premature to conclude that hip circumference measurements are unnecessary in the assessment of obesity-related risk status and health promotion in women (3). This is consistent with recent cross-sectional studies demonstrating separate and opposite associations of waist and hip circumference with lifestyle and other CVD risk factors (4, 5, 6).

However, to our knowledge, no other studies have examined the risk of dying or developing CVD or CHD related to hip circumference in men and women. Therefore, the purpose of this study was to examine the independent effects of hip circumference on risk of developing CVD or CHD and on total mortality among adult Danish men and women.

Research Methods and Procedures

All residents of Denmark have a personal number that includes their date of birth and other vital statistics. A random subset of 4807 subjects was selected from the Central Person Register in 1982 from citizens born in 1922, 1932, 1942, or 1952 and living in 11 municipalities in Copenhagen County (7). Of these, 226 were of non-Danish origin. Among the 4581 Danes invited, 3608 participated in a subsequent health examination in 1982 to 1983. Five years later, in 1987 to 1988, an invitation to participate was sent to the 3608 subjects, and 2987 (83%) attended this second examination, which included hip and waist circumference measurements (8). The nonrespondents have been described in detail elsewhere (9). This study was part of the Danish MONICA project (MONItoring trends and determinants of CArdiovascular disease), an international study conducted under the auspices of the World Health Organization to monitor trends in and determinants for CVD.

All subjects who had their hip and waist circumference, weight, and height measured and had an average of 10 years of follow-up for development of CVD and CHD and 13 years of follow-up for total mortality were included in the study.

Anthropometric Measures

Body weight was measured to the nearest 0.1 kg using a lever balance, with subjects dressed in light clothing or underwear only. Height was measured to the nearest 1 cm; subjects did not wear shoes. Waist circumference was measured midway between the lower rib margin and the iliac crest in the horizontal plane. While the subjects were standing, hip circumference was measured at the point yielding the maximum circumference over the buttocks using a tape measure to measure to the nearest 1 cm. All anthropometric measurements were taken in accordance with World Health Organization standards (10).

Questionnaire Data

Leisure time activity was recorded on a four-point scale ranging from 1) almost completely inactive: reading, television viewing, and cinema; 2) some physical activity (at least 4 h/wk); 3) regular activity; and 4) regular hard physical training for competition.

Subjects recorded their present and previous smoking habits and were classified as either current, former, or nonsmokers.

Endpoints

Subjects were followed until September 18, 2001 for total mortality using information from the Danish Central Person Register. During the 37, 108 person-years of follow-up, 425 had died. Subjects were followed ∼12.5 years. Cause of death was obtained from the National Board of Health using International Classification of Diseases (ICD)-8 and ICD-10 codes. Subjects were followed until December 31, 1998 for incidence of CVD and CHD (fatal and nonfatal). ICD-8 codes 390–458 and ICD-10 codes I00-I52 and I60-I99 were used to evaluate incidence of CVD. For incidence of CHD, ICD-8 codes 410–414 and ICD-10 codes I20-I25 were used. In the morbidity analyses, subjects were followed for ∼10 years.

Statistics

We used a Cox regression model with age as the underlying time axis. Separate analyses were performed for men and women. In a first series of models, hip circumference was categorized into quartiles. With adjustment for BMI, waist circumference, physical activity, smoking status, and, for women, menopausal status at baseline, we compared the risk associated with hip circumference by quartile hip using the lowest quartile as a reference. In a second series of models, we compared the relative/separate associations of BMI, hip circumference, and waist circumference with the endpoints under study. To reduce colinearity among BMI, hip, and waist measurements, we computed the residuals from linear regression of respective hip circumference and waist circumference on BMI. After this, there remained a slight colinearity between the hip and the waist residuals of 0.2 for women and 0.4 for men. The residuals were entered into a continuous effects model including BMI, smoking status, physical activity, and, for women, menopausal status. Linearity of the continuous covariates was verified by nesting the above models into models where the covariates were included using linear splines. The proportional hazards assumption was examined and verified using Schoenfeld residuals. All analyses were done before and after exclusion of subjects with previous myocardial infarction, stroke, or cancer. The analyses were performed using SAS Version 8.2 (SAS Institute, Cary, NC).

Results

Of 2987 subjects, 2960 had complete information on anthropometric measures and questionnaire data. Table 1 shows a summary of their characteristics. During the 13 years of follow-up, 270 men and 155 women died. Exclusion of subjects who had experienced myocardial infarction, stroke, or cancer at or before the baseline examination in 1987 to 1988 reduced the number of cases slightly, but the characteristics of the subjects were essentially similar before and after exclusions (data not shown). On average, both men and women were slightly overweight (BMI = 25.9 and 24.4 kg/m2, respectively). Average waist and hip circumferences were 92.0 and 98.7 cm for men and 78.7 and 98.2 cm for women, respectively.

Table 1. . Characteristics of 35 to 65-year-old Danish men (N = 1514) and women (N = 1446)
 MenWomen
  • Results given are mean ± SD (BMI, hip, waist, and age), as prevalence (smoking and leisure time physical activity), or as numbers (incidences of CHD or CVD and total mortality).

  • *

    Values in parentheses are incidences of CHD and CVD and total mortality after exclusion of previous myocardial infarction, stroke, or cancer.

Hip (cm)98.7 ± 6.498.2 ± 8.5
Waist (cm)92.0 ± 10.478.7 ± 10.4
BMI, kg/m225.9 ± 3.424.4 ± 4.2
Age, years50.7 ± 10.950.1 ± 11.0
Percent current smokers5548
Percent ex-smokers86
Percent nonsmokers3746
Percent sedentary2227
Percent walking5561
Percent sports doing2312
Percent post-menopausal 52
Incident CVD, n313 (263)*161 (140)
Incident CHD, n135 (103)44 (34)
Total mortality, n270 (229)155 (133)

Table 2 gives the number of subjects, fraction of fatal and nonfatal endpoints, level of physical activity, smoking, and BMI for men and women by quartile of hip circumference. As expected, men and women in the highest hip quartile also had higher BMI, waist circumference, and age. Furthermore, they smoked and exercised less, were more frequently menopausal (in women), and encountered more of the endpoints under study. Exclusion of subjects who had experienced myocardial infarction, stroke, or cancer at or before the baseline examination in 1987 to 1988 reduced the number of cases slightly, but the characteristics of the subjects were essentially similar before and after exclusions (data not shown).

Table 2. . Characteristics of subjects given by quartiles of hip circumference and p for trend among Danish men (hip quartile cut-off point 94.0, 98.0, and 102.5 cm) and women (hip quartile cut-off point 92.5, 97.0, and 102.5 cm)
 Hip circumference 
CharacteristicsQ1Q2Q3Q4p for trend
  • Results given are mean ± SD (BMI, hip, waist, and age), as prevalence (smoking and leisure-time physical activity) or as numbers (incidences of CHD or CVD and total mortality).

  • *

    Values in parentheses are incidences of CHD and CVD and total mortality after exclusion of previous myocardial infarction, stroke, or cancer.

Menn = 381n = 402n = 383n = 348 
 Hip, cm91.3 ± 3.096.4 ± 1.1100.4 ± 1.3107.4 ± 4.3 
 Waist, cm82.4 ± 5.688.3 ± 5.794.4 ± 6.4104.2 ± 9.1<0.001
 BMI, kg/m222.6 ± 2.024.7 ± 1.826.6 ± 2.130.2 ± 3.2<0.001
 Age, years49.3 ± 11.150.3 ± 11.250.1 ± 10.752.2 ± 10.5<0.001
 Percent sports participation242722190.05
 Percent current smokers655553450.001
 Percent ex-smokers78980.56
 Percent nonsmokers28373847<0.001
 Percent sedentary221922270.04
 Percent walking545456540.99
 Incident CVD, n62 (52)*78 (60)75 (66)98 (85)<0.001 (<0.001)
 Incident CHD, n24 (18)37 (25)35 (30)39 (30)0.03 (0.02)
 Total mortality, n60 (52)66 (54)63 (55)81 (68)<0.001 (0.02)
      
Womenn = 379n = 373n = 344n = 350 
 Hip, cm89.1 ± 3.194.9 ± 1.399.9 ± 1.5110.0 ± 6.4 
 Waist, cm69.7 ± 5.374.9 ± 4.980.4 ± 6.390.9 ± 9.9<0.001
 BMI (kg/m2)20.5 ± 1.822.8 ± 1.725.0 ± 2.029.9 ± 3.8<0.001
 Age, years47.2 ± 11.148.2 ± 10.751.5 ± 10.453.7 ± 10.4<0.001
 Percent sports participation12131380.04
 Percent current smokers57504439<0.001
 Percent ex-smokers66550.54
 Percent nonsmokers37445156<0.001
 Percent sedentary252628300.07
 Percent walking636159620.76
 Percent post-menopausal44445666<0.001
 Incident CVD, n44 (40)30 (25)44 (38)43 (38)0.34 (0.46)
 Incident CHD, n14 (11)8 (6)11 (7)11 (10)0.92 (0.79)
 Total mortality, n53 (48)35 (31)28 (21)39 (33)0.30 (0.19)

Tables 3 and 4 present the results of the regression analysis. The top parts of the Tables show the risk associated with hip circumference by quartile of hip for men and women while taking covariation from BMI and waist circumference into account. The bottom parts of the Tables compare the predictive values for hip, waist, and BMI by reporting the Wald χ2 test statistics and the associated p values.

Table 3. . Hip circumference as a predictor of CVD, CHD, and total mortality in 1514 Danish men
MenCVD morbidity (n = 313)CHD morbidity (n = 135)Total mortality (n = 270)
  • Quartiles: multivariate models adjusted for BMI, waist circumference, smoking status, and physical activity at baseline.

  • Continuous effects: Wald χ2 and p level indicating independent contributions to predictive model. Multivariate models adjusted for BMI, hip and waist residuals, smoking status, and physical activity at baseline. Hip regression coefficients negative. Waist positive. BMI positive.

  • *

    Values in brackets give the continuous effects after exclusion of subjects with myocardial infarction, stroke, or cancer.

Quartiles   
 Hip Q11 (reference)1 (reference)1 (reference)
 Hip Q21.01 (0.71 to 1.44)0.94 (0.53 to 1.66)0.88 (0.61 to 1.27)
 Hip Q30.89 (0.60 to 1.34)0.77 (0.39 to 1.52)0.72 (0.47 to 1.10)
 Hip Q40.98 (0.59 to 1.64)0.99 (0.45 to 2.16)0.79 (0.46 to 1.36)
    
Continuous effects   
 Hip residuals0.05 (p = 0.80)0.02 (p = 0.9)3.2 (p = 0.07)
 [0.50 (p = 0.50)]*[0.10 (p = 0.70)][3.0 (p = 0.08)]
    
 Waist residuals0.56 (p = 0.45)0.01 (p = 0.9)7.9 (p = 0.005)
 [0.50 (p = 0.50)][0.03 (p = 0.90)][8.4 (p < 0.01)]
    
 BMI16.2 (p < 0.001)5.2 (p = 0.02)6.4 (p = 0.01)
 [15.9 (p < 0.001)][4.3 (p = 0.04)][3.5 (p = 0.06)]
Table 4. . Hip circumference as a predictor of CVD, CHD, and total mortality in 1446 Danish women
WomenCVD morbidity (n = 161)CHD morbidity (n = 44)Total mortality (n = 155)
  • Quartiles: multivariate models adjusted for BMI, waist circumference, smoking status, physical activity, and menopausal status at baseline.

  • Continuous effects: Wald χ2 and p levels indicating independent contributions to predictive models. Multivariate models adjusted for BMI, hip and waist residuals, smoking status, physical activity, and menopausal status at baseline. The effect of BMI on mortality was modeled by linear splines with knots at 20, 25, and 30 kg/m2. The Wald statistics associated with BMI in the model for the total mortality was referred to the χ2 distribution with 4 degrees of freedom. Hip regression coefficients negative. Waist positive.

  • *

    Values in brackets give continuous effects after exclusion of subjects with myocardial infarction.

Quartiles   
 Hip Q11 (reference)1 (reference)1 (reference)
 Hip Q20.62 (0.38 to 1.02)0.52 (0.20 to 1.31)0.70 (0.40 to 1.21)
 Hip Q30.74 (0.44 to 1.22)0.46 (0.18 to 1.21)0.27 (0.13 to 0.56)
 Hip Q40.54 (0.26 to 1.13)0.27 (0.07 to 1.13)0.23 (0.10 to 0.53)
    
Continuous effects   
 Hip residuals4.3 (p = 0.03)3.9 (p = 0.05)15.1 (p < 0.001)
 [3.3 (p = 0.70)]*[3.1 (p = 0.08)][12.7 (p < 0.001)]
    
 Waist residuals1.1 (p = 0.30)1.7 (p = 0.20)12.0 (p < 0.001)
 [0.20 (p = 0.70)][0.9 (p = 0.40)][7.1 (p < 0.01)]
    
 BMI0.3 (p = 0.60)0.3 (p = 0.60)17.6 (p = 0.002)
 [0.40 (p = 0.60)][0.1 (p = 0.90)][17.1 (p < 0.01)]

In women, having wide hips was significantly and inversely associated with total death, as well as developing CVD or CHD; this was suggested by the stronger independent continuous effects of the hip residuals for all endpoints. In men, hip circumference was not statistically associated with the endpoints studied. Rather, BMI and waist circumference were the strongest independent predictors of early mortality or of developing CVD or CHD. In both genders, results were essentially similar, as indicated by the predictive values for hip, waist, and BMI, before and after exclusions of subjects who had experienced myocardial infarction, stroke, or cancer at, or before, the baseline examination. (See exclusion information given in parentheses below.)

With the use of the lowest quartile of hip circumference as a reference, women belonging to the highest quartile showed a 46% reduction in the risk of developing CVD (45% after exclusion of subjects with prevalent myocardial infarction, stroke, or cancer), an 83% reduction in the risk of CHD (51% after exclusions), and an 87% reduction in total mortality both before and after exclusion of subjects with prevalent myocardial infarction, stroke, or cancer. In men, none of the higher hip quartiles was associated with differences in morbidity or mortality compared with the lowest hip quartile before or after exclusion of subjects with prevalent myocardial infarction, stroke, or cancer.

Discussion

We found hip circumference in women to be a stronger independent predictive risk factor for total death and development of CVD or CHD than either BMI or waist circumference. In fact, once the variability associated with hip circumference was accounted for, neither BMI nor waist circumference was statistically related to CVD or CHD. In men, hip circumference was unrelated to these endpoints, whereas BMI and waist circumferences were independently related to all the endpoints studied. These findings suggest that in women, but not in men, large hip circumference may independently convey protection toward later development of CVDs and early mortality. This study supports findings from our previous study in women (3), but is the first to demonstrate, with a prospective design, that hip size does not seem to be related to early CVD in men. It should be noted, however, that we cannot exclude entirely the possibility that wide hips may be protective against early total mortality for men because of the borderline significant trend between the hip residuals and total mortality (p < 0.08 both before and after exclusion of men with prevalent myocardial infarction, stroke, or cancer) or the possibility that the colinearity between the hip and waist residuals influenced the effects. Indeed, the protective effect of wide hips in women may also be relevant for early overall mortality in men, although the clear lack of significant findings for CHD and CVD makes this less likely.

Earlier cross-sectional studies have shown that both men and women with small hips have an excess risk of having diabetes (6, 7), hypertension, or gall bladder disease (7). This is in keeping with findings of separate and opposite associations of waist and hip circumference with both lifestyle and other CVD risk factors (4, 5). We were not able to study diabetes incidence, partly because only about one-half of all diabetics were diagnosed (11) and because data on diabetes cannot be retrieved from the Danish registers. However, this study demonstrates that the results from these cross-sectional observations of risk factors for CVD and early death do not translate into the risk factors for cardiovascular and overall early mortality.

Previous studies have discussed that smaller-than-expected hip circumferences may reflect reduced femoral fat, small pelvic bone structure, or muscle atrophy in the gluteo-femoral region (4). The gender difference shown in this study suggests that either genes or sex hormones may be the operating biological mechanism behind such differences.

A number of studies have shown that waist circumference may be more predictive of health risks and a more practical correlate of abdominal fat distribution than waist-to-hip ratio (2, 12), and, consequently, it has been proposed that hip measurement should be discarded from prediction analyses, with a concentration on only waist circumference. This study suggests that this decision may have been premature, at least for women. In this context, it should be noted that, in this study, the waist-to-hip ratio was not generally less predictive of health risks than waist circumference. Indeed, within each gender and for each of the endpoints, the attributable risk associated with the waist-to-hip ratio and the waist circumference was generally either essentially similar or greater for the waist-to-hip ratio than for waist circumference. As an example, for total death, 31% and 32% of deaths in women were found to be attributable to waist circumference and waist-to-hip ratio, respectively, whereas in men, 15% of deaths were attributable to waist-to-hip ratio and only 8% to waist circumference (data not shown).

In conclusion, a large hip circumference seems to convey strong and independent protection against development of CVD or early mortality in women. Hip size does not seem to be related to CVDs in men. However, it cannot be excluded entirely that wide hips may be protective against early total mortality in men.

Acknowledgment

This study was supported by grants from the Danish Medical Research Council (The FREJA programme), The Ministry of The Interior and Health, and the Augustinus Foundation.

Footnotes

  • 1

    Nonstandard abbreviations: CHD, coronary heart disease; CVD, cardiovascular disease; ICD, International Classification of Diseases.

Ancillary