%20-%20https://sketch.com%20--%3e%3ctitle%3e3ECBBF50-FD08-412A-943D-2CFC5E025A16%3c/title%3e%3cdesc%3eCreated%20with%20sketchtool.%3c/desc%3e%3cg%20id='Stylesheet'%20stroke='none'%20stroke-width='1'%20fill='none'%20fill-rule='evenodd'%3e%3cg%20transform='translate(-926.000000,%20-668.000000)'%20id='BUTTONS'%3e%3cg%20transform='translate(100.000000,%20597.000000)'%3e%3cg%20id='BTN-/-RIGHTS'%20transform='translate(816.000000,%2065.000000)'%3e%3cg%20id='ICON-/-Rights'%20transform='translate(10.000000,%206.000000)'%3e%3cpath%20d='M13.5752841,0%20L1.34943182,0%20C0.59375,0%200,0.586666667%200,1.33333333%20L0,22.6666667%20C0,23.4133333%200.620738636,24%201.34943182,24%20L17.6505682,24%20C18.40625,24%2019,23.3866667%2019,22.6666667%20L19,5.36%20L13.5752841,0%20Z%20M13.5752841,1.89333333%20L17.0838068,5.36%20L13.5752841,5.36%20L13.5752841,1.89333333%20Z%20M1.34943182,22.6666667%20L1.34943182,1.33333333%20L12.1988636,1.33333333%20L12.1988636,5.36%20C12.1988636,6.10666667%2012.7926136,6.69333333%2013.5482955,6.69333333%20L17.6235795,6.69333333%20L17.6505682,22.6666667%20L1.34943182,22.6666667%20Z'%20id='Shape'%20fill='%23595959'%20fill-rule='nonzero'%3e%3c/path%3e%3ccircle%20id='Oval'%20stroke='%23595959'%20stroke-width='1.5'%20fill='%23595959'%20cx='9.5'%20cy='13.5355339'%20r='5.5'%3e%3c/circle%3e%3cpolyline%20id='Path-2'%20stroke='%23FFFFFF'%20stroke-width='1.5'%20transform='translate(9.500000,%2013.035534)%20rotate(45.000000)%20translate(-9.500000,%20-13.035534)%20'%20points='8%2016.0355339%2011%2016.0355339%2011%2010.0355339'%3e%3c/polyline%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Obesity is a complex heterogeneous disorder which is strongly associated with multiple metabolic dysfunctions including dyslipidaemia, impaired glucose tolerance and arterial hypertension(Reference Grundy1, 2). Although previous studies have documented the association between obesity and metabolic diseases, there is increasing evidence suggesting that the presence of excess fat in specific regional body areas is a more significant indicator than total fat volume(Reference Zhang, Rexrode and van Dam3–Reference Despres and Lemieux6). The excess accumulation of central fat, even without manifestations of general obesity, is closely associated with metabolic disturbances that indicate increased risks of CVD and diabetes(Reference Esteghamati, Khalilzadeh and Anvari7–Reference Nguyen-Duy, Nichaman and Church9) whereas peripheral subcutaneous fat may be relatively benign with respect to metabolic disease risks(Reference Van Pelt, Evans and Schechtman10–Reference Snijder, Visser and Dekker12). For instance, the Quebec Family Study showed that larger hip circumference was inversely associated with several cardiovascular risk factors, including high TAG and insulin concentrations(Reference Seidell, Perusse and Despres13). Therefore, collection of information on fat regional distribution patterns may add more insights in understanding the considerable metabolic variation among individuals.
Women are more likely to store fat in the lower body (around the hips and bottom, or gynoid region) and are more susceptible to age-related fat distribution change than men(Reference Van Pelt, Jankowski and Gozansky14). The CVD risk is much lower in premenopausal women than in age-matched men, although their overall adiposity is much greater than men's(Reference Wingard, Suarez and Barrett-Connor15). However, with the remarkable fat redistribution that occurs after middle age, or especially after the menopause, the risk of CVD increases sharply in women(Reference Ley, Lees and Stevenson16, Reference Pascot, Lemieux and Lemieux17). This body composition change includes an increase in total-body and central-area (android region) fat, and a decrease in the proportion of gynoid fat. However, little is known about how and to what extent regional fat and its distribution pattern are associated with CVD progression for women. Furthermore, disparities in such associations might exist among ethnic groups(Reference Boutcher, Dunn and Gail Trapp18). Compared with their Western counterparts, Asian women have a greater increment of upper-body fat store with age, which is ∼0·33 kg/10 years in Asian women v. ∼0·07 kg/10 years in Caucasian or African-American women(Reference Wu, Heshka and Wang19). However, evidence regarding the association of regional fat depots with the outcomes of metabolic disorders in this subgroup is limited(Reference Deurenberg-Yap, Chew and Deurenberg20).
To elucidate these issues, the present study aimed to explore the association of regional body fat, indicated as android fat and gynoid fat from dual-energy X-ray absorptiometry (DXA), with metabolic risk factors in a community-based study of Chinese women. We are particularly interested to know if accurate measurement of site-specific body fat may provide a more thorough understanding of metabolic disease characteristics in Chinese women.
Materials and methods
Participants
A total of 627 women aged from 18 to 79 years old were voluntarily recruited through leaflets and posters between November 2008 and May 2009. Measurements were conducted at the Obesity and Body Composition Research Center of Zhejiang University School of Public Health. The study was designed to establish a community-based database on obesity-related chronic disease. Individuals with known malignancies, thyroid disease, pituitary disorders or hypogonadism were not eligible. Fourteen women with missing DXA information and another four women with missing data from the blood test were also excluded. In the end, 609 women were included in the analysis. In addition, none of the women had ever received hormone replacement therapy. The study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University. Written informed consent was obtained from each participant.
Variable definition
Comprehensive questionnaires on lifestyle, medical history and menopause were conducted face-to-face at the research centre. In terms of smoking status, participants were classified into two categories: (i) smokers, who are either current smokers or past smokers; and (ii) non-smokers, who never smoked. Drinking level was categorized as heavy, moderate and never. Heavy drinkers were defined as those who drank beer, wine or hard liquor no less than one time per day during the past month. Moderate drinkers were those who consumed an alcoholic beverage less than once per day during the past month. Occupational physical activity was classified into four categories: (i) ‘low activity’ such as office work; (ii) ‘moderate activity’ such as driver; (iii) ‘high activity’ such as farming or steelmaking; and (iv) ‘other’, which refers to participants with missing information on occupational activity(Reference Lin, Zhang and Song21, Reference Wang22). Leisure-time physical activity (LTPA) was categorized as none, 0 < LTPA < 150 min/week, or LTPA ≥ 150 min/week(Reference Garber, Blissmer and Deschenes23). Menopause was defined if there had been complete cessation of menses for more than 12 months.
Anthropometric measures and blood pressure
Anthropometric measurements were obtained based on a standardized protocol. To prevent any hypoglycaemia accidents, all anthropometric measurements were conducted after a simple breakfast including a piece of bread and a bottle of yoghurt. Height was measured without shoes to the nearest 0·1 cm, while weight with only light clothing was measured to the nearest 0·1 kg (Detecto, Webb City, MO, USA). BMI was calculated as body weight (in kilograms) divided by the square of height (in metres). Blood pressure (BP) was measured using a standard mercury sphygmomanometer after the participant had rested for at least 10 min. All values were recorded as the mean of three measurements.
Laboratory measurements
All blood samples were obtained from the participants after a requested 12 h fast. Fasting plasma glucose (FPG) concentrations were determined using the hexokinase method, TAG concentrations using the enzymatic method and HDL-cholesterol (HDL-C) concentrations were determined using an enzymatic homogeneous assay. All reagents were obtained from Roche (Roche Diagnostics, Indianapolis, IN, USA).
Fat measurements
DXA (software version 11·40·004; GE Lunar Prodigy, Madison, WI, USA) was used to measure fat mass (FM), percentage body fat (%BF), android fat and gynoid fat through whole-body scans. %BF was calculated as the ratio of FM (kg) to body weight (kg) multiplied by 100. For the android region, the lower boundary was the top of the pelvis as the line of demarcation. The upper boundary was placed above the pelvis line of demarcation at a position that was equivalent to 20 % of the distance between the pelvis and femoral neck. Lateral boundaries were the lines for the arms when in normal position for a whole-body scan. The gynoid region was defined with the upper boundary positioned below the pelvis cut line by 1·5 times the height of the android region. The lower boundary was positioned such that it was equal to two times the height of the android region. The lateral boundaries were the outer leg lines of demarcation. Central fat distribution pattern was assessed by the android:gynoid fat ratio, as shown in Figure 1(Reference Novotny, Going and Teegarden24). DXA was calibrated daily using a standard phantom provided by the manufacturer. Measurements were maintained within the manufacturer's precision standards of ≤0·8 %.
Scans and regions by dual-energy X-ray absorptiometry for bone (a) and body composition (b); android, top section; gynoid, bottom section
Metabolic risk factors
Metabolic risk factors were defined as outcome variables including elevated TAG, reduced HDL-C, elevated BP and elevated FPG according to the recently harmonized criteria to diagnose metabolic syndrome(Reference Alberti, Eckel and Grundy25): (i) raised TAG: ≥150 mg/dl (1·7 mmol/l); (ii) reduced HDL-C: <40 mg/dl (1·03 mmol/l) in men and <50 mg/dl (1·29 mmol/l) in women; (iii) raised BP: systolic BP ≥130 mmHg or diastolic BP ≥85 mmHg; and (iv) raised FPG: ≥100 mg/dl (5·6 mmol/l).
Statistical analysis
Variables were presented as means and standard deviations for continuous variables and as numbers and percentages for categorical variables. BMI-adjusted partial Pearson correlation analyses were performed to examine the associations of fat measures with continuous values of metabolic risk factors. Multivariate logistic regression models were used to estimate the odds ratio for each metabolic risk factor according to a 1 sd change in fat measurement. To ensure comparability, fat measurements were expressed as a Z-score in the analysis. A Z-score is calculated as an observed value minus the sample's mean value, divided by the sample's standard deviation. In the first logistic regression analysis (model 1), the associations of each fat depot with metabolic risk factors were explored after adjusting for age, menopause, smoking, drinking, occupational physical activity and LTPA. Model 2 and model 3 were subsequently adjusted for BMI, and BMI and %BF, respectively, on the base of model 1. Gynoid and android fat were further mutually adjusted for each other in model 4 to determine the independent effect of body composition parameters on metabolic risk factors. Because menopause was an important confounding variable, logistic regression analyses were repeated stratified by menopause status. In addition, the tertiles of android fat and gynoid fat were calculated to estimate their associations with the risk for having at least two metabolic risk factors. The variance inflation factor and eigenvalue were used to examine the collinearity of the variables in the models. The statistical software package SPSS version 16·0 for Windows was used for all analyses. All statistical tests were two-tailed, and P < 0·05 was considered significant.
Results
The characteristics and body composition of the 609 women included in the analyses are shown in Table 1. According to Chinese criteria, 3 % of the participants were diagnosed as underweight (BMI < 18·5 kg/m2), 52 % as normal (BMI = 18·5–23·9 kg/m2), 45 % as overweight (BMI = 24·0–27·9 kg/m2) and none as obese (BMI ≥ 28·0 kg/m2). The prevalence of high TAG, low HDL-C, high BP and high FPG was 25·5 %, 42·7 %, 42·4 % and 32·7 %, respectively. Participants with at least two risk factors accounted for 43·8 % of the study population.
Characteristics of the study participants: Chinese women (n 609) aged 18–79 years, November 2008–May 2009
Values are presented as means and standard deviations or as numbers and percentages.
LTPA, leisure-time physical activity (≥150 min/week); DXA, dual-energy X-ray absorptiometry; FM, fat mass; %BF, percentage body fat (FM in kg/body weight in kg × 100); HDL-C, HDL-cholesterol; BP, blood pressure; FPG, fasting plasma glucose.
The coefficients of BMI-adjusted partial Pearson correlation of fat measures with risk factors are listed in Table 2. %BF, android fat and android:gynoid fat ratio were all positively correlated with age, whereas gynoid fat was negatively correlated. The coefficient of android fat with gynoid fat was low (r = 0·18; P < 0·01), but both of them had a moderate correlation with %BF (r = 0·62 and r = 0·55, respectively; P < 0·01). Android fat and android:gynoid fat ratio were positively correlated with TAG, systolic BP, diastolic BP and FPG, and negatively correlated with HDL-C. In contrast to this, the correlations between gynoid fat and metabolic risk factors were negative. %BF did not have any significant correlations with metabolic risk factors.
BMI-adjusted partial Pearson correlation coefficients of adiposity measures from DXA with various variables in Chinese women (n 609) aged 18–79 years, November 2008–May 2009
DXA, dual-energy X-ray absorptiometry; %BF, percentage body fat; HDL-C, HDL-cholesterol; BP, blood pressure; FPG, fasting plasma glucose.
*P < 0·05 and **P < 0·01.
The odds ratios and 95 % confidence intervals by 1 sd increase of each regional fat measure for metabolic risk factors are shown in Table 3. After adjusting for age, BMI, menopausal status and lifestyle-related covariates (i.e. smoking, drinking, occupational physical activity and LTPA), gynoid fat was associated with decreased odds for all risk factors (all P < 0·01) and android fat was associated with increased odds of low HDL-C (P < 0·01). Android:gynoid fat ratio was independently associated with increased odds for all risk factors (all P < 0·01). The results were similar even when additionally adjusted for %BF in model 3. In model 4, when android fat and gynoid fat were included in the same model, gynoid fat was still associated with decreased odds for all metabolic risk factors (all P < 0·001).
Odds ratios and 95 % confidence intervals of metabolic risks with 1 sd increase of each adiposity measurement in Chinese women (n 609) aged 18–79 years, November 2008–May 2009
HDL-C, HDL-cholesterol; BP, blood pressure; FPG, fasting plasma glucose; %BF, percentage body fat.
All logistic regression models adjusted for covariates: age, menopausal status, smoking status, drinking status, occupational physical activity and leisure-time physical activity.
The Z-score of each adiposity measurements was calculated as (observed value – sample mean)/sd.
Participants were further stratified by menopause status. Android fat and android:gynoid fat ratio were significantly higher in postmenopausal women than in premenopausal women (P < 0·01), whereas gynoid fat was not different between the two groups (P > 0·05). After adjusting for age, BMI, %BF and lifestyle covariates, the significant difference in android fat between pre- and postmenopausal women was diminished (P > 0·05). Android:gynoid fat ratio was still significantly higher in postmenopausal women, and gynoid fat was significantly lower in premenopausal women than in postmenopausal women (P < 0·01). The logistic models were rerun stratified by menopause status and the results were similar to those in non-stratified analysis (data not shown).
Figure 2 shows the adjusted odds for having at least two metabolic risk factors according to tertile of android fat and gynoid fat in women. The risks for having at least two risk factors increased remarkably among individuals with less gynoid fat and more android fat (P for trend < 0·01). Women in the lowest tertile of gynoid fat had a threefold higher risk for having at least two risk factors than those in the highest tertile (P < 0·01), when their android fat amount was in the lowest tertile. Interestingly, even when their android fat level was high (moderate or highest tertile), women in the highest tertile of gynoid fat still had lower odds for having at least two risk factors compared with those in the lowest gynoid fat tertile (P for trend < 0·01). In the multicollinearity test of the models, the variance inflation factors between adiposity measures were all <7 and the eigenvalues were all >0, both indicating that multicollinearity was not a problem in the logistic regression models.
Adjusted odds ratios for having at least two metabolic risk factors according to tertiles (T1, T2 and T3, where T1 is the lowest and T3 the highest) of android fat and gynoid fat in Chinese women (n 609) aged 18–79 years, November 2008–May 2009. Covariates adjusted for in the analysis include age, BMI, menopausal status, smoking status, drinking status, occupational physical activity, leisure-time physical activity and percentage body fat
We also did the same analysis in men (n 396) in the present study. After adjusting for age and other covariates (i.e. smoking, drinking, occupational physical activity and LTPA), %BF, android fat, gynoid fat and android:gynoid fat ratio were all associated with increased odds for high TAG, high BP and high FPG (all P < 0·05). When additionally adjusted BMI and %BF in the models, android fat and gynoid fat were positively associated with high TAG, whereas android:gynoid fat ratio was positively associated with high TAG and high FPG (P < 0·05). When including android fat and gynoid fat in the same model, only android fat maintained a positive association with high TAG (P < 0·05); gynoid fat had no associations with any of the metabolic features (all P > 0·05; data not shown).
Discussion
In the current study, android fat and android:gynoid fat ratio were both significantly associated with increased odds of metabolic risk factors, whereas gynoid fat showed a decreased risk even after adjusting for android fat. In addition, compared with women in the lowest tertile of gynoid fat, those in the highest tertile had significantly lower odds of having at least two metabolic risk factors. The independently inverse association of gynoid fat with metabolic risks, in the present study, may be of great importance in Chinese women in the prevention of obesity-related diseases.
There has been much debate about which regional adiposity depot is more associated with morbidity and mortality. Previous studies demonstrated that greater waist circumference and smaller hip circumference both confer metabolic or cardiovascular risks(Reference Shen, Punyanitya and Chen26, Reference Heitmann and Lissner27). However, these simple measurements cannot distinguish between fat mass and lean mass. Moreover, anthropometric measurements are subject to intra- and inter-examiner variations. Recently, more accurate methods including DXA have been of great interest. It should be noted that the measurement by DXA is of an area, rather than of one selected level as in computerized tomography, waist circumference or hip circumference, which may provide more information to understand the associations of regional fat with potential health risks(Reference Ito, Nakasuga and Ohshima28–Reference Wang, Thornton and Kolesnik30).
Previous studies conducted in Asian women demonstrated that visceral fat, rather than waist circumference, was a major determinant of metabolic risks(Reference Hyun, Kim and Jang31). In addition, body fat distribution, independent of general obesity, was believed to be more crucial in determining potential atherosclerotic risk factors among the healthy Chinese population(Reference Wu, Yao and Lu32). In the present study, DXA-derived android fat amount, especially android fat distribution pattern, indicated as android:gynoid fat ratio, was positively associated with metabolic abnormalities in Chinese women. Android fat region derived from DXA consists of all the fat in the upper abdomen area and around some important organs including the liver, pancreas and lower part of the heart. Compared with peripheral subcutaneous adiposity, fat accumulated in the android region is more sensitive to lipolytic stimuli, and produces more pro-inflammatory cytokines, which exerts a greater risk of developing atherothrombotic and inflammatory abnormalities associated with metabolic syndrome(Reference Wajchenberg, Giannella-Neto and da Silva33–Reference Ibrahim35). Previous studies have demonstrated that liver fat, but not visceral fat, is a better marker of the obesity-associated metabolic derangement(Reference Magkos, Fabbrini and Mohammed36, Reference Fabbrini, Magkos and Mohammed37). Adipose tissue accumulated at the pericardial region was also found to induce endothelium dysfunction and stimulate the progression of vascular disease by releasing adipokines which acted on the adjacent organs and coronary vasculature(Reference Mazurek, Zhang and Zalewski38–Reference Lim, Son and Song41). However, studies on the association of android fat depot with metabolic diseases are very limited(Reference Kang, Yoon and Ahn42, Reference Aucouturier, Meyer and Thivel43). In a recent study, android fat amount was more closely associated with metabolic syndrome than abdominal visceral fat derived from computerized tomography in elderly people(Reference Kang, Yoon and Ahn42). Another study conducted in obese children and adolescents also reported the harmful effects of android fat distribution on insulin resistance(Reference Aucouturier, Meyer and Thivel43). These results suggest that fat accumulation in the android area should be noticed in the pathogenesis of metabolic disturbance in addition to visceral fat.
Compared with the well-established association of central fat with disease risks, there are only a few studies regarding the association of the lower-body fat depot, which is very predominant in women, with metabolic diseases. Among the limited relevant studies, most of them reported inverse associations between hip circumference and mortality or heart diseases(Reference Zhang, Rexrode and van Dam3, Reference Heitmann and Lissner27, Reference Lissner, Bjorkelund and Heitmann44–Reference Rexrode, Carey and Hennekens51). Nevertheless, hip circumference was not associated with CHD or other metabolic diseases without adjusting for waist circumference or BMI and the effect of central obesity on mortality risk could be seriously underestimated without adjusting for hip circumference(Reference Zhang, Rexrode and van Dam3, Reference Cameron, Magliano and Shaw52). The present study observed that the opposite associations of android fat and gynoid fat were independent of each other. In addition, according to Table 2, gynoid fat was associated with all metabolic markers whereas android fat was associated with only three of them, suggesting that gynoid fat might be a better predictor of decreased metabolic risk factors than android fat. These results agree with previous studies reporting that gluteofemoral fat, as measured by thigh circumference, hip circumference or leg adipose tissue mass, was protective and associated with lower metabolic risks, including lower LDL-cholesterol, lower TAG and higher HDL-C levels(Reference Manolopoulos, Karpe and Frayn53). The findings of our study may be particularly crucial for Asian women, because their lower-body fat has been found to exert a greater inverse influence on insulin resistance than that in Westerners(Reference Boutcher, Dunn and Gail Trapp18).
Increased prevalence of CVD and metabolic syndrome, particularly after the menopause, has raised curiosity about the underlying factor. Previous studies suggest that the menopause transition is associated with deleterious changes in body fat distribution(Reference Poehlman, Toth and Gardner54). To be specific, an increase in android fat and a reduction in peripheral fat were observed when total fat mass had not significantly increased(Reference Tremollieres, Pouilles and Ribot55). These body composition changes may create a compatible environment for abnormal metabolism and aggravate cardiometabolic risk factors(Reference Dasgupta, Salman and Lokesh56). In our study, postmenopausal women were characterized by an absolute decreased gynoid fat and a relative increased central fat pattern. When adjusting for important variables like age, BMI and %BF, regional fat depots were still independently associated with metabolic risk factors. This finding agrees with previous studies indicating that menopause status and the associated regional fat change are the major predictors of metabolic aberration over age in women(Reference Poehlman, Toth and Gardner54–Reference Dasgupta, Salman and Lokesh56).
The major strength of the present study was that DXA-derived android fat and gynoid fat measures were used. Both of them were closely with traditional fat measures like waist circumference and hip circumference (the correlation coefficient between android fat and waist circumference was 0·90, P < 0·000; that between gynoid fat and hip circumference was 0·82, P < 0·000), whereas they are more accurate indices to understand the association of fat and fat distributions with metabolic or cardiovascular diseases. Most previous epidemiological studies have considered truncal fat as the central adiposity index(Reference Van Pelt, Evans and Schechtman10, Reference Van Pelt, Jankowski and Gozansky14). However, the truncal fat depot can be partitioned into upper-body (android or central) and lower-body (gynoid or peripheral) areas. Mechanistically and theoretically, use of the android fat measure is more precise to manifest the impact of central fat on metabolic disease(Reference Kang, Yoon and Ahn42). Moreover, decreased fat in the gynoid region, which is specific for women to manifest the age-correlated fat redistribution pattern, could exert an independent influence on metabolic disorders. Lower-body obese women, despite having greater upper-body fat mass, were found to have lower NEFA release than their non-obese counterparts(Reference Martin and Jensen57). These specific characteristics are important to prevent ectopic fat accumulation in women and lead to a decreased metabolic disease risk.
Limitations to the present study should be noted. First, because of the cross-sectional nature of the study, it is impossible to confirm a predictive role of regional fat on metabolic risks and subsequent incidence of diseases. Second, we failed to observe similar results in men. This may be attributable to the limited sample size in men or there might be a gender difference in the associations between fat distribution and metabolic risks. Third, BMI was a crude measure of total body size. Studies with a larger sample size are needed to analyse the associations between regional fat and metabolic risks stratified by BMI. In addition, taking the ethnic difference in the associations between regional fat measures and metabolic profile among women into account, findings from the current study may not be generalized to other ethnicities(Reference Hosain, Rahman and Williams58).
Conclusion
Our study showed that there were opposite associations of android and gynoid fat with metabolic risks in Chinese women. Gynoid fat rather than android fat might be more important for inclusion in metabolic disease risk evaluation in female Asians. The findings from the present study suggest that not only central fat but also gynoid fat should be considered in metabolic disease risk-reducing strategies in Chinese women.
Acknowledgements
Sources of funding: This study was supported, in part, by funding from Zhejiang University and the China Medical Board (CMB) (grant 10-014). Zhejiang University and CMB had no role in the design, analysis or writing of this article. Conflicts of interest: All authors have no conflicts of interest in relation to the present study. Authors’ contributions: X.F. and S.Z. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: X.F. and S.Z.; acquisition of data: S.Z.; statistical analysis: X.F. and S.Z.; analysis and interpretation of data: X.F., A.S., Y.Z., X.M., J.J., M.Y. and S.Z.; drafting of the manuscript: X.F. and S.Z.; critical revision of the manuscript for important intellectual content: X.F., A.S., Y.Z., J.J., M.Y. and S.Z.; obtaining funding: S.Z.; study supervision: S.Z.
