Relationship of total body fatness and five anthropometric indices in Chinese aged 20|[ndash]|40 years: different effects of age and gender
S-F Lei,M-Y Liu,X-D Chen,F-Y Deng,J-H Lv,W-X Jian,H Xu,L-J Tan,Y-J Yang,Y-B Wang,S-M Xiao,X Sun,C Jiang,Y-F Guo,J-J Guo,Y-N Li,Y-J Liu,H-W Deng
DOI: https://doi.org/10.1038/sj.ejcn.1602345
2005-01-01
European Journal of Clinical Nutrition
Abstract:Obesity, defined as an excess of body fat, is an increasing health problem. Excess body fat mass is strongly associated with metabolic and vascular diseases (Han et al., 1996; Lindsay et al., 2001; Thomas et al., 2004). However, direct and accurate measurement of adiposity relies on complex technologies such as dual energy X-ray absorptiometry (DEXA), resulting in relatively high cost. Anthropometric indices (such as body mass index (BMI), waist circumference (WC), hip circumference (HC), waist-to-hip ratio (WHR) and conicity index (CI)) are commonly used to assess fat mass because they are relatively straightforward and cheap to obtain (Taylor et al., 2000; Lindsay et al., 2001; Janssen et al., 2002). BMI is the commonly used index in prediction of total body fat (TBF). However, its ability in accurately predicting TBF remains controversial because BMI cannot distinguish fat mass and lean mass, and, more importantly, the accuracy of its prediction is prone to be disturbed by the differences of sample in adipose distribution and body build (Frankenfield et al., 2001; Janssen et al., 2002). WC, HC, WHR and CI are four useful and accurate tools to evaluate the distribution of fat mass or the central adiposity (Taylor et al., 2000; Janssen et al., 2002). Additionally, relatively few data on the prediction of fat mass with anthropometric indices were accumulated in Chinese (Davies et al., 2001; Lanham et al., 2001; Chang et al., 2003; Jia et al., 2003). Furthermore, from these limited data in Chinese, the accuracy of the prediction was still inconsistent (Chang et al., 2003). Hence, the present study was performed to assess the ability of the combined use of the above five anthropometric indices to predict TBF in a very large Chinese population. As there are ethnic-specific energy expenditure and the body build among various ethnic populations (Deurenberg et al. 1998; Duncan et al. 2004; Rush et al. 2004), the data on the relationship between anthropometric indices and fat mass accumulated in Caucasians may not be suitable for Chinese. Compared with Caucasians, few data both on the relationship between anthropometric indices and body fat measured by sophisticated measures, such as DEXA, were accumulated in Chinese (Ko et al., 2001; Zhang et al., 2004). Therefore, it is necessary to perform the ethnic-specific relationship in Chinese. The current study was undertaken with three major aims: (1) evaluating the effect of age and sex on TBF, anthropometric indices and their relationship; (2) investigating the correlation between TBF and anthropometric indices; (3) evaluating the ability of predicting TBF using five anthropometric indices and developing eight age- and gender-specific equations to predict TBF by the above five anthropometric indices in Chinese. where RCPCs are the regression coefficients for four PCs, respectively; EPCs BMI, EPCs WC, EPCs HC, EPCs WHR and EPCs CI are the corresponding Eigenvectors of BMI, WC, HC, WHR and CI for four PCs; SBMI, SWC, SHCI, SWHR and SCI are the corresponding standard values [(measured value-mean measured value)/mean measured value] for five anthropometric index. Age was significantly correlated with TBF, %TBF, and five anthropometric indices either among the whole males or among the total females, but was not significantly associated with them in each 5-years age group in both males and females (data not shown). As shown in Table 1, there was an increasing trend of TBF, %TBF and five anthropometric indices in successively older age groups (For example, the average TBF was 13.23.31, 13.33.23, 14.64.20 and 16.73.66 in four age groups in females.). Compared with females, the males generally had larger average BMI, WC, HC, WHR and CI, but had lower average TBF and %TBF (Table 1). Except for age, statistically significant differences between males and female were found for the other studied indices by Student's t-test (P<0.0001). Correlations between TBF and five anthropometric indices were generally higher than those between %TBF and the studied anthropometric indices in each age group (Table 2). So the following analyses were focused on the prediction of TBF with five anthropometric indices, because the high correlations resulted in more accuracy of prediction. Except for some correlations in 25–29 years age groups, TBF and %TBF were significantly correlated with five anthropometric indices with the correlation coefficients ranging from 0.07 to 0.87 (Table 2). In each age group, there was a decreasing trend of correlations between TBF and BMI, WC, HC, WHR and CI, generally with the highest correlation coefficients between TBF and BMI (For instance, the correlation coefficients between TBF and BMI, WC, HC, WHR and CI in 20–24 years age female group were 0.82, 0.71, 0.69, 0.34 and 0.22.). There were no evident differences of correlations among different age groups in the males, but in females in the 25–29 years age group, the correlation coefficients were evidently smaller than those in other age female groups. In addition, in the same age group, there was intuitively gender-specific correlation, especially in 25–29 years age group. As shown in Tables 3 and 4, the four PCs interpreted over 99% of the total variation of five relative anthropometric indices by PCA, with over 53% of the total variation accounted by PC1. The PC1 is highly and positively correlated with all the indices in all age and gender groups according to their Eigenvectors, but the other three PCs are inconsistently related with the studied indices. Regression analyses (Table 5) showed that the four PCs combined explained a greater variance (R2=55.2–80.8%) of TBF than did BMI alone (R2=40–74.9%) (e.g., in the 20–24 years group, the proportions were 66.6% by BMI, and 74.9% by the four PCs.). Although the four PCs combined or BMI alone were strong correlates of TBF, about 20–60% of the variance in TBF remained unexplained by the four PCs combined or BMI alone. Figure 1 intuitively demonstrated that the simple correlation coefficients between TBF and the predicted TBF by combined PCs or BMI in regression analysis were very high in both females and males at all the age groups (r=0.64–0.9), whereas the correlations of TBF and the predicted TBF by the four PCs combined were higher than those of TBF and the predicted TBF by BMI alone, indicating that using the four PCs predict more precise TBF than using BMI alone. We also developed eight age- and gender- specific equations to predict TBF using measured values of five anthropometric indices (Table 6), which are detailed in Statistical analysis. To investigate the external validation of predicting TBF, we determined the equations to predict body fat in half of the randomly chosen sample and tested the prediction equations in the other half. The prediction equations in half of the randomly chosen sample were very similar with those in whole sample (data not shown), and then the predicted TBF using these equations in the other half were highly correlated with the measured TBF (r=0.73–0.86). Our major findings in this study are that there is an increasing trend of total fat mass and five anthropometric indices in successively older age groups; that age and sex have the important effects on influencing the relationship of TBF and the studied anthropometric indices; and that the accuracy of predicting the TBF using five anthropometric indices is higher than using BMI alone. Compared with Caucasians, few data on the relationship of anthropometric indices and body fat mass were reported in Chinese. A conclusion from a meta-analysis (Deurenberg et al., 1998) has shown that the differences in energy balance as well as in body build may result in their population-specific relationship. Therefore, to increase our understanding on the ethnic-specific correlation and prediction of body fat mass with five anthropometric indices and to develop eight age- and gender- specific equations for prediction of TBF in Chinese, we performed the present study in a very large population of both sexes. Most of the previous studies have concentrated on the prediction of %TBF with anthropometric indices (Goran et al., 1996; Pietrobelli et al., 1998; Lindsay et al., 2001), whereas the present study focused on the prediction of TBF. Two reasons may justify our choice. First, the simple correlations of anthropometric indices are greater with TBF than with %TBF in our large Chinese sample of two sexes, which are in agreement with those from previous studies in Caucasians (Goran et al., 1996; Pietrobelli et al., 1998; Lindsay et al., 2001), and thus the high correlations resulted in more accuracy of prediction. Second, TBF is regarded as an absolute measure that reflects overall size as well as adiposity and %TBF serves as a relative measure of adiposity; %TBF is of clinical importance as a generally accepted criterion of diagnosing obesity (Heyward et al., 1992; Blew et al., 2002; Eto et al., 2004); moreover, %TBF is easily calculated from an estimate of total body fat and total weight. Therefore, the prediction of TBF in the present study is the basis of predicting %TBF. Based on the above reasons, we focused on investigating the prediction of TBF, rather than %TBF, with anthropometric indices. Our results support earlier work indicating that age has a differential effect on body fat mass and five anthropometric indices (Horber et al. 1997). Our results are similar with those from previous studies (Horber et al., 1997; Jackson et al., 2002) that there is an increasing trend of total fat mass and five anthropometric indices in successively older age groups. Potential reason is that individuals tend to gain fatness with aging because of substantial increase of adipocyte size and number (Evans and Campbell, 1993). Therefore, to try our best to decrease the disturbance of age, it is appropriate to divide our sample into four 5-years age groups in both sexes, because age was significantly correlated with total fat mass and five anthropometric indices in all sample, but was not associated with them in each divided age group. Our results indicate that the four PCs from five anthropometric indices are better predictors of TBF than is BMI alone. The evidence may come from that the combined use of the four PCs values substantially increases the explained variance (R2) of TBF, and that the simple correlation coefficients are higher between TBF and the predicted TBF by the four PCs than between TBF and the predicted TBF by BMI alone (Figure 1). BMI is the one most commonly and widely used tool for assessing body fatness. However, use of the BMI predictor of body fatness has been criticized (Smalley et al., 1990; Roubenoff et al., 1995), because of a relatively low correlation coefficient between BMI and fat mass, and low proportion of fat mass variance explained by BMI from multiple regression analysis (Frankenfield et al., 2001). Adipose distribution may partially disturb the accuracy of assessing body fatness by BMI. Four indices of WC, HC, WHR and CI are useful and accurate tools to evaluate the distribution of fat mass or the central adiposity (Taylor et al., 2000). However, those indices are highly related and thus they cannot simply be modeled in multiple regression analysis to predict TBF. Therefore, a PCA has been employed in this study with an attempt to extract useful information from five related indices (BMI, WC, HC, WHR and CI). To the best of our knowledge, this study is the first effort to simultaneously use five anthropometric indices to predict TBF in Chinese adults of both sexes. With the aim of predicting TBF by a simple method, we develop eight age- and gender-specific equations using the measured values of five anthropometric indices in the present study, which have practical importance for routine clinical evaluation. Moreover, the validation of predicting TBF is very high, because the correlation coefficients between TBF and the predicted TBF by the four PCs using multiple regression analysis range from 0.75–0.9 (Figure 1). The combined use of five anthropometric indices has certain advantage over assessment of fat mass using DEXA. Measurement of anthropometric indices are cheaper and technically easier than assessment of adiposity. Therefore, it is of very important clinical implication to accurately assess fat mass using simple anthropometric indices, especially in a developing country like China, because the direct measurement of adiposity incurs relatively high cost. This is the primary motive in performing the present study in a Chinese population. We have confirmed that the BMI and other four anthropometric indices are highly correlated with TBF or TBF% (Pietrobelli et al., 1998; Lindsay et al., 2001) with important effects of gender and age influencing their relationship. However, the correlation coefficients in our Chinese sample are generally smaller than those in Caucasians (Pietrobelli et al., 1998; Lindsay et al., 2001). For example, Lindsay et al. (2001) found that the linear correlation coefficients between BMI and percent fat were above 0.8, but these ranged from 0.45 to 0.75 in this Chinese sample. The differences in body build and fat mass distribution may account for the above observation. Body build tends to vary among different ethnic groups (Duncan et al., 2004). Generally, relative sitting height tends to be higher in Asian ethnic groups, which may result in a relatively low BMI (Deurenberg et al., 1999; Gurrici et al., 1999). Furthermore, when adjusting for body build, most of the ethnic-specific differences were eliminated associated with %TBF prediction equations for bioelectrical impedance analysis (BIA) in Chinese, Malay, and Indian Singaporeans (Deurenberg et al., 2002). Race differences in fat distribution are also evident. There is evidence that Asian children and adolescents have a greater central fat mass than that of Europeans and other ethnic groups (Duncan et al., 2004). When body size (weight and height) was controlled, He et al. (2002) inferred that Asian girls had greater relative truncal or central fat mass. Therefore, these differences in body build and fat distribution may result in the difference in relationship between anthropometric indices and TBF. Additionally, the differences in body build and fat mass distribution may also explain the gender-specific relationship between anthropometric indices and TBF. These results support our belief that gender- and race-specific definitions of adiposity might be required. However, the study has some limitations. First, although DEXA is becoming increasingly popular for the measurement of soft tissue composition as well as bone mineral density, it cannot distinguish between visceral and subcutaneous fat mass. Second, the present results from a normal healthy population may not be applied to an overweight and obese population or other patients with chronic diseases and conditions. How to apply the present data to these populations awaits further studies. Third, the samples are aged between 20 and 40 years. Therefore, these results may have inability to generalize to older or younger populations. In conclusion, the present study explored the complex relationship between the five anthropometric indices and total body fatness in a very large population of both sexes. The obtained results add to our understanding of the relationship of anthropometric indices and adiposity, as well as the effects of age and sex on their relationship. We thank three anonymous reviewers for their comments to improve this paper. The study was partially supported by a key project grant (30230210), a general grant (30470534) from National Science Foundation of China, three projects from Scientific Research Fund of Hunan Provincial Education Department (02A027, 03C226, 04B039), and a grant from Natural Science Foundation of Hunan Province (04JJ1004). Investigator H.W.D. was partially supported by grants from Health Future Foundation of USA, grants of National Health Institute (K01 AR02170-01A2, R01 GM60402 and 5R01 AR050496-02).