Diagnostic accuracy of different body weight and height-based definitions of childhood obesity in identifying overfat among Chinese children and adolescents: a cross-sectional study

In large-scale population surveys and public health screening, anthropometry indices [i.e., body mass index (BMI), weight-for-height, waist circumference] are commonly used as surrogates for body fat [1, 2]. Various definitions of childhood obesity have been developed using these anthropometry measures. BMI is one of the most highly recommended and widely used tools for defining childhood obesity [3‐5]. Three commonly used tools are World Health Organization (WHO) BMI-based [6], International Obesity Task Force (IOTF) BMI-based [7], and Centers for Disease Control and Prevention (CDC) BMI-based references [8]. In 2005, the Working Group on Obesity in China (WGOC) developed a BMI-based definition for childhood obesity in Chinese school-aged children [9]. Through these definitions, the cut-off points for childhood obesity are determined using a statistical approach, where an arbitrarily chosen value from a reference distribution defines childhood obesity [7, 10, 11]. The relationship between BMI, body fat percentage (%BF), and health risk is not well established for children and adolescents, including those in Hong Kong.

Relative weight-for-height is used to define childhood obesity in some countries, especially in Asia [12]. In this approach, cut-off points for being obese is defined as a set percent over a standard weight for a given height in the individual’s age and sex population. The standard weight is usually determined as the mean or median determined from a normal reference distribution for the population, and the most commonly used cut-off point for obesity is 110 % or 120 % standard weight-for-height [13]. For the past 20 years in Hong Kong, childhood obesity has been defined as a body weight of over 120 % of the median weight for height according to the local reference (1993 HK reference) [14], where reference values for height are available only for up to 175 cm in boys and 165 cm in girls. Overweight is not defined in the standard.

At the population level, %BF is most commonly evaluated using BMI. However, international studies have found that BMI cut-offs do not provide an equivalent measurement of %BF and are associated with health risks across different ethnic groups [15]. One possible reason to explain this result is the differences in body composition due to ethnicity. Compared with their African-American and European counterparts, Asian children and adolescents have higher %BF at the same BMI level [16‐18], raising the possibility of underestimating excess body fat in Asian groups when using BMI as indirect measure. Such situation is undesirable because it will influence the access of public health services of those who most need intervention. Therefore, large-scale studies need to determine the relationship between anthropometry measurements and %BF, including Hong Kong Chinese children and adolescents. The purposes of the current study was to investigate the diagnostic accuracy of BMI-based definitions and 1993 HK reference in screening excess body fat among Hong Kong Chinese Children and adolescents.

The classification of childhood obesity is complicated. The diagnostic performance of simple and low-cost anthropometry-based indices in screening excess body fat is of practical importance in public health surveys and in clinical evaluations for the health practitioner. Various anthropometry index definitions were developed to define childhood obesity [25, 26]. In the current study, we validated BMI and PWH as predictors of excess fat with the use of %BF evaluation by BIA as the criterion. More than 2000 Hong Kong Chinese children and adolescents aged 9 to 19 years old were recruited for the study. The results of the study suggest that 1) local PWH-based references for obesity had higher sensitivity than BMI-based references for obesity; 2) none of BMI-based references for obesity achieved optimal rates of sensitivity.

Ko et al. investigated the prevalence of obesity among Hong Kong adolescents using four diagnostic references: the 1993 HK reference, the IOTF-BMI reference, the CDC-BMI reference, and the WGOC-BMI reference [27]. The 1993 HK reference gave an exceedingly high figure for obesity compared with BMI-based references [27]. Yeung and Hui observed that the prevalence of childhood obesity in Hong Kong for 2006 was 22.67 % and 16.88 % for boys and girls, respectively, using the 1993 HK reference. When the cut-offs determined by IOTF were used, the prevalence rates of childhood obesity for boys and girls were 6.33 % and 3.08 %, respectively [28]. Moreover, the prevalence rates of overweight using the cut-offs determined by IOTF were similar to those for obesity using the 1993 HK reference [28]. In the current study, the 1993 HK reference obesity rate was higher, but similar when OW and OB rates are combined in all BMI-based references. The results are consistent with the findings of previous studies. Moreover, in the 1993 HK reference, reference values for height are available only for up to 175 cm among boys and 165 cm among girls. So et al. found that the body height of Hong Kong children and adolescents aged 6–18 years increased significantly from 1993 to 2005/2006 for both genders [29]. In the current study, we found that 12.0 % boys and 6.1 % girls were not evaluated by the 1993 HK reference. Thus, the 1993 HK reference for obesity should be adjusted to cover populations with excessive body height. Moreover, the IOTF-BMI reference may underestimate or overestimate obesity in comparison to national definitions [30]. Among Asian children and adolescents, the IOTF-BMI reference for obesity gave a lower rate for childhood obesity than those by other BMI references [31]. Similar findings are shown by the current study. IOTF-BMI reference has lower obesity rate than those by other BMI based references in both genders (Table 2), the difference may be attributed by the different reference population, smoothing methods and approached methods to BMI cut-offs of different obesity references.

In the current study, BMI and PWH showed strong positive correlations with %BF which was measured by BIA in both genders. The results were consistent with the findings in previous studies [32‐35]. Despite the use of correlation coefficients to estimate the degree of closeness of the linear relationship between variables, correlation coefficients cannot detect diagnostic performance of anthropometry-based indices in screening excess body fat.

Among the BMI-based references for obesity (IOTF, WHO, CDC, and WGOC) analyzed in the current study, none of the references achieved optimal rates of sensitivity (0.325–0.502 for boys and 0.128–0.278 for girls). However, specificities were high (>0.9) for both genders. The result is consistent with the finding of previous studies [25, 31‐34, 36]. However, large discrepancies between sensitivity and specificity rates in different studies were found. The differences in sensitivity and specificity rates may be explained by the reference population, adopted cut-offs from various references, methods used to assess %BF, and the cut-offs for excess fat [31, 33].

In previous studies, researchers found that local PWH-based references for obesity had higher sensitivity than BMI-based references for obesity [32, 36]. A similar finding was also evident in the current study. The result suggests that the 1993 HK reference may currently better reflect the true prevalence of obesity among Hong Kong children and adolescents. In addition, when BMI-based references for overweight and obesity were used to assess subjects with excess fat, higher sensitivities and lower specificities were found in both genders in comparison with those of BMI-based reference for obesity. Similar results were reported by other studies [25, 36, 37]. As shown in Table 3, we observed similar sensitivity and specificity rates in screening excess body fat between the 1993 HK reference for obesity and BMI-based references for overweight and obesity. The rates for overweight and obesity determined by the BMI-based references were similar with the rate for obesity by the 1993 HK reference for both genders. Thus, BMI cut-offs for obesity should be decreased to accurately estimate excess body fat among Hong Kong Chinese children and adolescents.

In addition, the current study found significant differences in sensitivity between genders. For each reference for obesity and for each reference for overweight and obesity, boys have higher sensitivity than girls. Moreover, girls had lower diagnostic agreements between %BF reference for excess fat and all BMI-based and WC-based references than boys. Similar results have been found by previous studies [25, 31, 35, 37]. However, inconsistent results were found by other studies [33, 36, 38]. As mentioned above, complex factors, such as ethnicity, age ranges of the population, different cut-offs used to define excess fat, and the method in assessing %BF, may influence the diagnostic performance of different references in screening excess %BF.

In general, the 1993 HK reference for obesity is superior compared with BMI-based references for obesity in screening excess %BF among children and adolescents. Moreover, using the cut-offs for overweight, the diagnostic performance of BMI-based references in screening excess body fat can be improved. WC-based references for overweight and obesity may be used with local reference data similar with the 1993 HK reference and the BMI-based references for overweight and obesity. However, among girls, diagnostic performances of BMI-based references, WC references, and the 1993 HK reference are poorer than expected, i.e., lower sensitivity with poorer diagnostic agreement. Therefore, we attempted to identify optimal cut-offs of different anthropometry-based indices in screening excess %BF for both genders.

The current study has several limitations. One of the limitations in the present study is the difference in definitions for age groups, which may influence diagnostic accuracy of different definitions for childhood obesity in screening overfat. We did not factor in the stage of sexual maturation in our assessment of %BF, and the factor may increase possibilities of over- and under-estimation of overweight and obesity among children and adolescents. Moreover, although we used health-related cut-offs to define excess fat [24], the %BF cut-offs have raised criticism on their reference population, sample size, as well as selected health outcomes and risk measures [24]. To evaluate these cut-offs, others health-related criteria, such as blood pressure, blood cholesterol and blood sugar, may be taken into consideration among Hong Kong Chinese children and adolescents for further study. The choice of same %BF cut-off to define overweight or obesity considered as another limitation of this study. The purpose of the study of using this cut-off in the paper is for comparison among different weight and height-based classification criteria. Ideally, overweight and obesity should be defined as different %BF cut-offs. %BF cut-off to define overweight or obesity may increases the practicality of applying of these weight and height-based classification criteria in real world setting. To our knowledge, there is no unambiguous prior statement of %BF cut-offs to define overweight or obesity. Therefore, further studies should be conducted to identify %BF cut-offs to define overweight or obesity using others health-related criteria. In addition, our sample was not a nationally represented sample of Hong Kong children and adolescents. Further studies are necessary to be conducted using a nationally represented sample to confirm the results of the study. Another limitation is the nature of the criteria measures employed in this study. Previous studies found that foot-to-foot BIA scales give a lower mean and higher intra-individual variation in %BF than criteria measure, such as dual-energy X-ray absorptiometry (DXA), hydrostatic weighing [39, 40]. The method is appropriate for estimating %BF in population rather than individual [39, 40]. Foot-to-foot BIA scales may under- or overestimate %BF depending on the size and gender of individual being measured, therefore, some participants in current study may misclassify as “normal” or “overfat”.

In conclusion, prevalence rates of childhood obesity or overweight varied widely according to the diagnostic references applied. The diagnostic performance for weight and height-based references for obesity is poorer than expected for both genders among Hong Kong Chinese children and adolescents. In order to improve the diagnosis accuracy of childhood obesity, either cut-off values of body weight and height-based definitions of childhood obesity should be revised to increase the sensitivity or should be targeted toward defining local population and ethnic-specific cut-offs for childhood obesity using health-related excess body fat.