Study population

Two thousands seven hundred and eighty women aged 37–74 and 1160 men aged 49–94 were recruited by sending invitation letters to residential buildings, posting local advertisements, giving health talks, and through referrals from local communities in urban Guangzhou, Guangdong Province, China between October 2005 and June 2010. The recruitment procedure is described in more detail in [24]. We excluded participants with previously confirmed severe medical diseases such as cancer, stroke, and heart failure, along with those who were using medication known to affect lipid metabolism. All of the participants gave written informed consent and the study was approved by the Ethics Committee of School of Public Health of Sun Yat-sen University.

Data collection

Staff members with relevant medical knowledge screened the potential participants using a telephone survey. Eligible participants were invited to local health centers or the School of Public Health at Sun Yat-sen University. Face-to-face interviews were conducted using structured questionnaires to collect participants' socio-demographic information, their own and their families' medical histories, medication information, smoking status, and alcohol consumption. Anthropometric measurements, physical examinations, BIA-body fat, and blood pressure measurements were conducted at the same time. A 24-hour physical activity questionnaire containing 19 items was used to estimate daily physical activity [24]. And the physical activity was calculated by the combination of the metabolic equivalent score (MET, kcal.kg⁻¹.h⁻¹) for each type of physical activity after multiplied by its duration per day (h/d) [25]. Venous blood was collected after overnight fasting (≥10 h), and serum was separated within 2 hours (at 4°C) and stored at −80°C before the analysis.

Anthropometric measurements

Anthropometric indices were measured with participants dressed in light clothing and barefoot. BMI was calculated as weight (in kilograms)/height (in meters) squared. Waist and hip circumferences were measured at the level of the umbilicus, and at the level of the maximum girth between the iliac crest and the crotch, respectively. Each parameter was measured twice and a third measurement was carried out if the difference ≥2 cm. WHR was calculated as the ratio of the waist-to-hip circumferences. WHtR was calculated by dividing waist circumference by height. The absolute and the percentage of fat over the whole body and trunk were assessed by a bioelectrical impedance analysis (BIA) system using a Tanita TBF-418B (Tanita Corporation, Tokyo, Japan). The in-vivo reproducibility of the BIA system was 1.54%, 1.85%, 2.50%, and 2.79%, for the %BF, FM, %TF, and TFM, respectively.

Blood pressure measurement

Two blood pressure measurements were performed at an interval of at least 10 minutes by a mercury sphygmomanometer while the participants were in a sitting position after ≥10 min. of rest. The average of the two values was used unless the deviation of the two measurements for systolic pressure was ≥4 mmHg or diastolic pressure ≥3 mmHg, in which case a third measurement was required.

Serum analysis

Serum total cholesterol (TC) and triglycerides (TG) were measured by the enzymatic colorimetric method (Human, Wiesbaden, Germany). Direct methods were applied to assess HDL cholesterol (HDLc) and LDL cholesterol (LDLc) (Seikisui, Tokyo, Japan). The glucose oxidase method was used to detect glucose (Glu) (Homa, Beijing, China), and serum urine acid (UA) was measured by the enzymatic colorimetric method (Fenghui, Shanghai, China). All of these measures were performed by a Hitachi 7600–010 automatic analyzer (Hitachi, Tokyo, Japan). The variation coefficient for the serum measurements was 2.17% (at 5.03 mmol/L TC), 2.86% (at 1.14 mmol/L TG), 3.47% (at 1.70 mmol/L HDLc), 4.67% (at 2.65 mmol/L LDLc), 2.52% (at 4.45 mmol/L Glu), and 5.46% (at 303.5 mmol/L UA).

Definition of cardiometabolic abnormalities

Hypertension was defined as the current use of antihypertensive medication, a systolic blood pressure of ≥140 mm Hg, or diastolic blood pressure of ≥90 mm Hg [26]. Dyslipidemia was defined according to the NCEP ATP III criteria [27]. Accordingly, participants with one or more of the following results were considered dyslipidemic: plasma cholesterol (TC) ≥6.22 mmol/L (240 mg/dl), triglycerides (TG) ≥2.26 mmol/L (200 mg/dl), LDLc ≥4.14 mmol/L (160 mg/dl), or HDLc <1.03 mmol/L (40 mg/dl). Hyperuricemia was diagnosed by serum uric acid ≥420 μmol/l (7.0 mg/dl) in men, or ≥350 μmol/l (6.0 mg/dl) in women [28]. Diabetes was defined as a fasting plasma glucose concentration ≥7.0 mmol/l (126 mg/dL) or treatment for diabetes, and impaired fasting glucose (IFG) was defined as a plasma glucose concentration between 5.6 and 6.9 mmol/l (100–125 mg/dl) [29].

Following the IDF criteria, metabolic syndrome was defined by central obesity (defined as waist circumference ≥90 cm in men or ≥80 cm in women, or a BMI >30 kg/m), plus any two of the following four factors: (1) TG concentration ≥1.7 mmol/L (150 mg/dL); (2) HDLc concentration <1.03 mmol/L (40 mg/dL) in men or <1.29 mmol/L (50 mg/dL) in women; (3) systolic blood pressure ≥130 or diastolic blood pressure ≥85 mm Hg or treatment for previously diagnosed hypertension; or (4) a fasting glucose concentration ≥5.6 mmol/L (100 mg/dL) or treatment for diabetes [30]. Given that the objective of this study was to establish the optimal measurements for the evaluation of MetS, the obesity requirement for this definition was omitted.

Statistical analysis

The data were analyzed separately by sex. We examined the distribution of adiposity measures and other characteristics using the Kolmogorov-Smirnov test in Table 1. The results of continuous variables were presented as “mean ± standard deviation (SD)” when they followed a normal distribution, or mean (interquartile range) when they asymmetrically distributed. Categorical data were expressed as frequencies. The area under the receiver's operating characteristic curve (AUC) and the 95% confidence intervals (CIs) were used to compare the predictive ability of adiposity measurements of metabolic abnormalities. The AUC is a measure of the degree of separation between case and control subjects. The obesity measurements were separately categorized into four quartiles. The odds ratios (OR) and their 95% CIs for the presence of cardio-metabolic disturbances, compared using the highest to the lowest quartile of each obesity index, were estimated by logistic regression using the “enter” procedure with adjustments for age, physical activity, smoking, and alcohol status. All of the analyses were carried out using the SPSS statistical package (version 13.0, SPSS Inc, Chicago, IL), and P-values less than 0.05 (two-tailed) were considered to be significant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the participants^*.

https://doi.org/10.1371/journal.pone.0070893.t001

Body mass index

It is accepted that abdominal fat is an important determinant of obesity-associated risk factors, possibly more so than the degree of excess weight as measured by BMI [12]. This is confirmed by our findings, which indicated that BMI was inferior to abdominal measures in predicting dyslipidemia, diabetes/IFG, hyperuricemia (in women), and metabolic syndrome, as demonstrated in previous studies [31], [32], [33]. The high metabolic and inflammatory activity of visceral fat deposits within the abdominal cavity, compared with subcutaneous deposits in other parts of the body, provide a plausible explanation for the superiority of abdominal obesity measures in predicting metabolic risk [34].

Nevertheless, in our data, BMI exhibited the best performance in identifying hyperuricemia in men and was equal to WHtR in predicting hypertension in both sexes. The robust discriminatory power of BMI in terms of hypertension was demonstrated in a previous study of 8940 Chinese adults in which BMI had a closer association with hypertension than measures of central obesity in both women and men [10]. Similar associations have also been observed in studies conducted in other countries, such as Korea [19], Spain [35], and the UK [11]. Increased BMI is associated with increased serum glucose, insulin, aldosterone, and renin levels along with increased sympathetic tone. All of these factors are thought to increase blood pressure by increasing vascular volume or peripheral resistance [36]. However, accumulating evidence points to visceral obesity as the most important risk factor for hypertension [37]. To date, few studies have evaluated the use of different obesity measures for detecting the presence of hyperuricemia. In our data, abdominal obesity was a better predictor of hyperuricemia in women, whereas BMI was a better predictor in men. The reason for this sex difference is unclear, although differences in anatomy, physiology, metabolism, and sex hormones may offer a partial explanation.

Waist-to-height ratio, waist circumference, and waist-to-hip ratio

Several studies on Asian [3], [4], [5], [6] and Caucasian [32], [38] populations have found WHtR to be superior to WC in identifying cases with cardiovascular risk factors. Meta-analyses provide further confirmation of those findings [33], [39]. In agreement with these findings, our data demonstrated that WHtR had better predictive performance than WC for most conditions, even after controlling for BMI. WHtR may be a better predictor of metabolic risk because it takes height into account. An association between short height and risk of coronary heart disease has been previously reported [40]. Henriksson et al. found that body height had an inverse relation to serum cholesterol and non-HDL cholesterol in middle-aged men, which was independent of BMI and WHR [41]. Thus, it may be important to correct WC for height.

WHR has also been proposed as a good predictor of cardiovascular events [42] and of cardiometabolic risk factors among Tehranian adult men [8]. However, of the three abdominal measures, we found it to have the weakest association with most risk conditions, consistent with previous studies [32]. Differences in subject populations (excluding or including high-risk participants), site of waist measurement (midway between the lower rib and iliac crest, at the narrowest noticeable point, or at the level of the umbilicus), statistical methods (such as adjustment for age and other factors), and the definition of risk factors might collectively lead to these discrepancies. In addition, the validity of WHR has also been questioned as it is more susceptible to measurement error than the other indices. Furthermore, a non-obese and an obese individual could theoretically have the same WHR, as the ratio can remain constant with changes in weight [43]. However, it was noteworthy that WHR's discriminatory performance was superior to that of WC for most conditions after adjusting for BMI, which may be due to the BMI-independent inverse association between HC and cardiometabolic risks (Figure S3).

Hip circumference

Our results indicate that a larger HC was associated with a lower risk of metabolic disturbances after adjusting for BMI (Figure S3). In line with our findings, previous studies evaluating the separate contributions of waist and hip circumferences to metabolic risk factors have also demonstrated the apparent protective effect of a large HC [44], [45]. Similar associations were also found in the context of future diabetes [46], incidence of first myocardial infarction [42], combined cardiovascular disease, and mortality associated with these diseases [47]. A beneficial adipokine profile together with the long-term entrapment of excess fatty acid might be the main reason for the protective properties of gluteofemoral fat [48].

Bioelectrical impedance measures of adiposity

Our observations add to the findings of previous studies, which have found no advantage in using BIA measures of body fat rather than simple anthropometric measures in detecting obesity-related complications and metabolic risk factors [11], [19], [21], [49]. However, other studies have shown a higher predictive capability of direct body fat measures for those associations [13], [50], [51], [52]. BIA has been found to be a reliable method for assessing BF and %BF when a protocol is followed that controls for factors that may affect the measurement [53], [54]. Notable exceptions include the assessment of individuals with abnormal hydration and extreme weight ranges (BMI>34 kg/m²) [53]. We repeated all of our analyses excluding participants with a BMI ≥28 kg/m² (n = 204 in women and n = 78 in men), and the results were remarkably similar (results not reported). Additionally, the intra-individual variability of BIA in our data ranged from 1.54–2.79%, lower than the error rate reported in previous studies (2–3.5%) [54]. Thus, our findings are robust and not influenced by differential BIA measurements among obese individuals.

Limitations

First, our study is limited by its cross-sectional design, which precludes causal inferences. Further longitudinal analyses are needed to provide stronger evidence of these associations. Second, people from southern China tend to be shorter and have a lower BMI and WC than those from northern China [55]. Our participants were not randomly selected, which might attenuate their representativeness. Our sample included a low proportion of obese participants, with only 3.0% of women and 2.5% of men having a BMI >29.9 kg/m². The association between obesity and its related risk factors might be underestimated due to the limited variability in the obesity indices. Therefore, caution should be taken in extrapolating our results to the general population and other ethnic groups. Further investigation with a larger sample of obese people is needed.

Conclusions

Our study favors the use of anthropometric measures of abdominal obesity, especially WHtR, for assessing cardiometabolic disturbances. BMI may be used as an alternative obesity measure for detecting people suffering from hypertension. Further prospective studies are needed before definite conclusions can be made regarding the best predictor of future cardiovascular events.