Waist-to-height Ratio and Coronary Artery Disease in Taiwanese Type 2 Diabetic Patients

Authors

  • Chin-Hsiao Tseng

    Corresponding author
    1. National Taiwan University College of Medicine, Taipei, Taiwan
    2. Division of Endocrinology and Metabolism, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan
    3. Department of Medical Research and Development, National Taiwan University Hospital Yun-Lin Branch, Yun-Lin, Taiwan
      (ccktsh@ms6.hinet.net)
    Search for more papers by this author

(ccktsh@ms6.hinet.net)

Abstract

Waist circumference (WC), waist-to-hip ratio (WHR), waist-to-height ratio (WHeiR), and BMI are indicators for obesity. This study examined the usefulness of these indicators for coronary artery disease (CAD) in Taiwanese type 2 diabetic patients. A total of 1,345 (646 men and 699 women) patients aged 63.3 ± 11.5 years were studied. CAD was defined by history or Minnesota-coded electrocardiogram. The relative importance was evaluated by the magnitude of adjusted odds ratio per 1-s.d. increment, the decrease in −2 log likelihood after adding the index to the logistic model, the c-index, and the Akaiki's information criterion (AIC). Results showed that the four indices were highly intercorrelated and except BMI for men, all indices differed significantly between patients with and without CAD in either sex. In logistic regressions, the respective adjusted odds ratios for WC, WHR, WHeiR, and BMI for every 1-s.d. increment were 1.209 (1.010–1.448), 1.109 (0.935–1.316), 1.231 (1.027–1.474), and 1.207 (0.997–1.461) for men; and were 1.176 (0.995–1.390), 1.105 (0.923–1.322), 1.280 (1.079–1.518), and 1.277 (1.083–1.507) for women. Only WHeiR was significant for both sexes and it also showed the greatest decrease in −2 log likelihood, the largest magnitude of odds ratio, and the smallest AIC while compared with the other indices in either sex. It is concluded that WHeiR has the superiority of independent association with CAD and the highest magnitude of association than WC, WHR, and BMI in both sexes. The usefulness of WHeiR should not be neglected in clinical practice.

Introduction

In Taiwan, cardiovascular disease is the most common cause of death in diabetic patients (1). Obesity is a well-recognized risk factor for hypertension, diabetes, dyslipidemia, and cardiovascular disease (2,3,4). BMI is commonly used as an indicator for generalized obesity (5), but it is not a good indicator for total or cardiovascular mortality in men with a history of cardiovascular disease in a recent longitudinal follow-up of the Physicians' Health Study in the United States (6). Waist circumference (WC), waist-to-hip ratio (WHR), and waist-to-height ratio (WHeiR) are indicators for central obesity (7,8,9,10). A study by Huang et al. in Taiwan has demonstrated the association between these four obesity indices and cardiovascular risk factors including blood pressure, fasting plasma glucose, and lipid disorders in the general population (11). However, central obesity with intra-abdominal fat accumulation is better correlated with insulin resistance and cardiovascular risk factors in other ethnicities (8,12,13,14). A later study using similar dataset as Huang et al. (11) but with larger sample size confirmed that central obesity indicated by WHeiR was better associated with the cardiovascular risk factors in Taiwanese women (15). However, whether these anthropometric factors are associated with coronary artery disease (CAD) in the Asian populations remains to be determined. Moreover, controversies regarding the best indicators for obesity remain. The purpose of this study was to evaluate the association between the obesity indices including BMI, WC, WHR, and WHeiR and CAD in type 2 diabetic patients and to compare the relative importance among the indices.

Methods and Procedures

Study subjects

The study was approved by an ethics committee and the subjects participated voluntarily. More than 96% of the Taiwanese population is covered by a compulsory national health insurance. A total of 256,036 diabetic patients who were using this health insurance were assembled from 1995 to 1998 to investigate a series of epidemiologic issues (1). Baseline data were collected by questionnaires on the onset symptoms and confirmation of diabetic diagnosis from 93,484 patients of the original cohort (16). At random, 4,164 from the main cluster of 93,484 patients were selected and invited to participate in a health examination. A total of 1,441 patients participated from March 1998 to September 2002. No significant differences in age or sex were noted among the main national sample and those who participated in the health examination.

The classification of type 1 diabetes was based on either one of the following two criteria: (i) diabetic ketoacidosis at the onset of diabetes; and (ii) the patients required insulin injection within 1 year of diabetes diagnosis. If a patient was not diagnosed as type 1, he/she was viewed as a patient of type 2 diabetes. Patients with type 1 diabetes and those aged <18 years were excluded. The exclusion of patients with type 1 diabetes was because of the possibility of different pathogenesis of CAD from that of patients with type 2 diabetes. The reason for excluding patients aged <18 years was that the definition of obesity in children might be different from that in adults. A total of 1,345 (646 men and 699 women) type 2 diabetic patients, aged 63.3 ± 11.5 years, were recruited.

Diagnosis of CAD

Diagnosis of CAD was on the basis of the following criteria: (i) definite history of acute myocardial infarction (self-reported with previous diagnosis made by a physician); (ii) definite history of angina pectoris with documented electrocardiographic findings and under specific therapy (self-reported history of chest pain with confirmed diagnosis by an electrocardiogram done previously by a physician and was being treated with medications including sublingual nitroglycerine, coronary vasodilators, or antiplatelet agents (e.g., aspirin, ticlopidine, dipyridamole, or clopidogrel)); (iii) patients who had received a placement of coronary stents, percutaneous transluminal coronary angioplasty, coronary artery bypass graft, or had been positive for a coronary angiography examination, a treadmill exercise test, or a radionuclide test; and (iv) for those without the above-mentioned medical history, resting electrocardiogram was performed and coded according to the Minnesota codes, and CAD was defined by the Minnesota codes of coronary probable (1.1, 1.2, 7.1) and coronary possible (1.3, 4.1–4.3, 5.1–5.3) (ref. 17).

Anthropometric factors

Body height in centimeters (cm) was measured by having the subjects stand with their heals, buttocks, and heads against a wall. A flat object was placed on top of the subjects' head, and their height was marked on a tape measure affixed to the wall. Body weight was measured in kilograms (kg) with a standard portable scale. Body weight and body height were measured with light clothes and bare feet. BMI was calculated as body weight in kilogram divided by the square of body height in meters.

WC was measured according to the recommendation of the World Health Organization (18). The subjects stood with feet 25–30 cm apart with weight evenly distributed. Measurement was taken midway between the inferior margin of the last rib and the crest of the ileum in a horizontal plane by the measurer sitting by the subject and fitting the tape snugly but not compressing soft tissues. Hip circumference was measured around the pelvis at the point of maximal protrusion of the buttocks. Circumference was measured at the end of a quiet expiration of the subject to the nearest 0.1 cm. WHR and WHeiR were calculated by dividing the WC by the hip circumference and the body height, respectively.

Measurements of blood biochemistry

Blood samples were collected in the morning after fasting for at least 12 h. Fasting plasma glucose and serum total cholesterol and triglyceride were measured using an automatic biochemistry analyzer (Cobas Mira S; Roche Diagnostica, Basel, Switzerland) with reagents obtained from Randox Laboratories (Antrim, UK) (19,20). Hemoglobin A1c was measured using boronate affinity chromatography with reagents obtained from the Primus Corporation (Primus CLC385, Kansas City, MO) (19,20).

Measurements of other covariates

The patients' age, sex, diabetic duration, smoking status, hypertension treatment, and systolic and diastolic blood pressure were recorded or measured. Diabetic duration was defined as the time period in years between the time being recruited and the time diabetes was diagnosed. Blood pressure was measured on the right arm after 20 min rest in a sitting position using a mercury sphygmomanometer. Hypertension was defined as systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg and/or being under treatment with antihypertensive agents.

Statistical analyses

Data were expressed as mean (s.d.) or percentage. A P < 0.05 was considered statistically significant.

The baseline characteristics and the four obesity indices were compared between patients with and without CAD by Student's t-test for continuous variables and by χ2-test for categorical variables in separate sexes. Pearson correlation coefficients between the four indices and the continuous covariates were calculated for different sexes. The prevalences of CAD by tertiles of the four indices were tested by linear test for trend in men and women separately. Logistic regression models were created to estimate the odds ratios for CAD either unadjusted or after adjustment for age, diabetic duration, smoking, hemoglobin A1c, hypertension, total cholesterol, and triglyceride for each of the four indices. To compare the relative importance among the four indices, the odds ratios were expressed as per 1-s.d. increment, and the decrease in the −2 log likelihood resulting from the addition of each of the indices to the model containing all covariates was calculated. A larger decrease in −2 log likelihood indicates a larger increase of variance explained by adding the anthropometric parameter to the model (21). To assess the accuracy of the multivariate logistic models, c-index (the area under the receiver-operating characteristic curve) was used to evaluate the discriminatory power (predictive accuracy) and the Hosmer-Lemeshow test to evaluate the goodness-of-fit. C-index ranges from 0.5 (no discrimination) to 1.0 (perfect discrimination) and a P value of >0.05 in the Hosmer-Lemeshow test indicates a good fit of the model. The Akaike's information criterion (AIC = −2 × model log-likelihood + 2 × number of model parameters) was computed, and the model with the lowest AIC was considered the best fit. Models that differed in AIC by <2 units were considered indistinguishable (22).

Results

Table 1 compares the baseline characteristics between patients with and without CAD in separate sexes. The obesity indices (except BMI in men) and age, hypertension, systolic blood pressure and diastolic blood pressure were all significantly higher in patients with CAD in either sex. In the diabetic women, diabetic duration and triglyceride were additionally different significantly.

Table 1.  Comparisons between patients with and without coronary artery disease (CAD)
inline image

Table 2 shows the correlation coefficients between obesity indices and continuous covariates. Age, systolic blood pressure, and diastolic blood pressure were significantly correlated with all four indices in both sexes. The other covariates showed inconsistent correlation.

Table 2.  Correlation coefficients between waist circumference (WC), waist-to-hip ratio (WHR), waist-to-height ratio (WHeiR) and BMI and continuous covariates
inline image

Table 3 shows the prevalences of CAD for the tertiles of the four indices. Except for BMI in men, all indices showed significantly increasing trend of CAD with increasing tertiles in either sex (P trend <0.05).

Table 3.  Prevalence of coronary artery disease by tertiles of obesity indices in separate sexes
inline image

Table 4 shows the unadjusted and adjusted odds ratios for every 1-s.d. increment of the indices. The s.d. for WC, WHR, WHeiR, and BMI for men was 9.82 cm, 0.10, 0.06, and 3.54 kg/m2, respectively; and was 10.38 cm, 0.10, 0.07, and 3.73 kg/m2, respectively, for women. All unadjusted odds ratios were significant except that for BMI in men. In the adjusted models, WC and WHeiR were significant for men and WHeiR and BMI were significant for women. Therefore, only WHeiR was consistently significant for both sexes after adjustment. The magnitude of association expressed by odds ratios for every 1-s.d. increment was largest for WHeiR among the four indices. While examining the decrease in −2 log likelihood, adding WHeiR to the model containing the covariates showed the greatest decrease in either sex. C-index was largest in the model with WHeiR for both men and women, though the magnitude of difference from the other indices was small. The Hosmer-Lemeshow test indicated a good fit for all models. The AIC for WHeiR was the smallest indicating the best fit of the model in either sex. The difference in AIC between the model with WC and that with WHeiR was <2 in men, indicating that the difference in the fitting was indistinguishable between these two models.

Table 4.  Logistic regression models estimating odds ratios for coronary artery disease for the four indices of obesity for every 1-s.d. increment in the diabetic men and women
inline image

Discussion

This study extended the observation in Taiwan that obesity was associated with cardiovascular risk factors (11,15) to the association with CAD in the diabetic patients. All four indices were highly intercorrelated (Table 2) and were significantly associated with CAD (except for BMI in men) in either sex before adjustment (Tables 1, 3, and 4). However, when confounders were adjusted, only WHeiR consistently showed significant association in both sexes (Table 4). Furthermore, the magnitudes of odds ratios for every 1-s.d. increment were largest for WHeiR (Table 4). The relative importance of WHeiR was also demonstrated by the greatest decrease in −2 log likelihood, the largest c-index, and the smallest AIC while compared with models including other indices in either sex (Table 4). In secondary analyses, when adjusted odds ratios were calculated based on unit increment, the significant indices for men (i.e., WHeiR and WC) and women (i.e., WHeiR and BMI) did not change. The respective adjusted odds ratio for WC (every 1-unit increment), WHR (every 0.1-unit increment), WHeiR (every 0.1-unit increment), and BMI (every 1-unit increment) was 1.021 (1.001–1.042), 1.133 (0.923–1.391), 1.462 (1.050–2.036), and 1.054 (0.999–1.113) for men; and was 1.016 (1.000–1.032), 1.101 (0.925–1.310), 1.420 (1.114–1.810), and 1.068 (1.021–1.116) for women.

The relative importance of WHeiR was also supported by a recent meta-analysis showing WHeiR as the best and BMI the poorest discriminator for cardiovascular risk factors including hypertension, type 2 diabetes, and dyslipidemia in either sex (13). Furthermore, BMI in combination with indicators of central obesity did not improve, but might decrease, the discriminatory power (13). The usefulness of BMI has been challenged because it cannot differentiate between excessive fat and excessive muscle mass in individuals having the same BMI. In a recent longitudinal study in the United States, BMI was not associated with total or cardiovascular mortality in men with cardiovascular disease (6). Neither was it associated with CAD in diabetic men in this study (Table 4). However, BMI was independently associated with CAD in the diabetic women (Table 4), suggesting a different effect of BMI in different sexes. The utility of both WHeiR and BMI is limited by the requirement of two measurements (i.e., WC and body height for WHeiR, and body weight and height for BMI) and the calculation of the ratios. However, the measurement of WHeiR requires only tape measures without weighing scale as required for BMI, and its calculation is simpler than that of BMI.

Various methods have been used to quantify the abdominal subcutaneous and visceral fat, including the most deliberate technique of dual-energy X-ray absorptiometry, magnetic resonance images, and the computed tomography (23,24). Clinically, these techniques are not practical and therefore WC, WHR, and WHeiR are used as surrogates. WC and WHR correlate well with the image techniques (7,25), but WC was better correlated with intra-abdominal fat (8) and cardiovascular risk factors (26). Indeed, when comparing WC and WHR in multiple logistic models, WC was superior to WHR in the association with CAD in men, though both WC and WHR were not statistically significant in women (Table 4). Although the correlation between WHeiR and image techniques has not been examined, it should be good taking into account the high correlation between the three indices of central obesity (Table 2) and the good correlation of WC and WHR with the image studies (7,25).

WHR is susceptible to measurement errors because it takes two measurements, and it might be difficult to accurately locate the waist and hip, especially in obese subjects (3,27). The measurement of hip circumference is embarrassing and not acceptable in some societies, especially when the measurers are of opposite sex. Furthermore, WHR might not reflect the change in body size well because it remains the same if both WC and hip circumference change proportionately.

Although WC has the advantages of being easy to measure using a cheap tape, and being easy to be understood as a concept of obesity than anthropometric ratios, the usefulness of WC (and also WHR) was limited by the different cutoffs for different sexes and ethnicities for optimal discrimination of diabetes (3,10) and hypertension (3). On the other hand, WHeiR is a better measurement of the relative fat distribution among subjects of different age, sex, and statures (9,12). Ashwell and Hsieh recommended a cutoff of WHeiR at 0.5 for both sexes, for children and adults, and for different ethnicities (9). In secondary analysis, the optimal cutoff of WHeiR by maximizing the sum of sensitivity and specificity was close to 0.5 before adjusting for confounders and was 0.4 after adjustment for age, diabetic duration, smoking, hemoglobin A1c, hypertension, total cholesterol, and triglyceride for either sex (data not shown). Therefore, it is easy for clinicians to remember the message of “Keep your waist circumference to less than half your height” (9). We would also suggest a further lowering of the WC to 40% of the body height in high-risk patients with diabetes.

Body height is predictive for diabetes and hypertension (28,29), and percent body fat is an independent risk factor for CAD (30). Individuals with shorter stature have significantly higher percent body fat than taller people at similar BMI (31). Therefore, people with the same WC might not have the same percent body fat if they have different body height. WHeiR, taking into account the effects of WC and the influence of body height on the body fat composition, might be a useful indicator for central obesity and cardiovascular disease. Epidemiologic studies also suggest that WHeiR is a more useful indicator for cardiovascular risk factors (9,15) and for urinary albumin excretion rate in patients with type 2 diabetes (32). Because WHeiR is positively correlated with age (Table 2), the association between WHeiR and CAD could be due to the lower body height in patients of older age. However, such an association might not be explained by age alone because the association remained significant after adjusting for age and other confounders (Table 4).

Some limitations deserved mentioning. First, the validity of WC, WHR, and WHeiR as indicators for abdominal fat has not been confirmed by comparing with more sophisticated techniques in the Taiwanese. Second, prospective studies are required to confirm the cause-effect relationship between the obesity indices and CAD in our population. Third, the validity of CAD definition was not known and the generalization of the association between the obesity indices and CAD to Taiwanese nondiabetic subjects should be reconfirmed.

In conclusion, this study demonstrated a close association between obesity and CAD in Taiwanese type 2 diabetic patients. Among the four indices of obesity (WC, WHR, WHeiR, and BMI) only WHeiR shows consistent and independent association with CAD in both the diabetic men and women. Although WC is an additional risk factor for men and BMI for women, this study provides evidence for the superiority of WHeiR as a risk factor for CAD.

Acknowledgment

I thank the following institutes for their continuous support on the epidemiologic studies of diabetes mellitus and arsenic-related health hazards: the New Century Health Care Promotion Foundation; the Department of Health (DOH89-TD-1035; DOH97-TD-D-113-97009), the National Taiwan University Hospital Yun-Lin Branch (NTUHYL96.G001); and the National Science Council (NSC-86-2314-B-002-326, NSC-87-2314-B-002-245, NSC88-2621-B-002-030, NSC89-2320-B002-125, NSC-90-2320-B-002-197, NSC-92-2320-B-002-156, NSC-93-2320-B-002-071, NSC-94-2314-B-002-142, NSC-95-2314-B-002-311, and NSC-96-2314-B-002-061-MY2), Taiwan.

Disclosure

The author declared no conflict of interest.

Ancillary