Metabolic syndrome (MetS) is a health hazard condition that reflects a constellation of several cardiometabolic risk factors, including excessive abdominal adiposity, high blood pressure and fasting plasma glucose, and abnormal triglyceride and high-density lipoprotein-cholesterol [1
]. Longitudinal studies have demonstrated that MetS makes adolescents higher susceptible to developing MetS, type 2 diabetes mellitus, and cardiovascular diseases in adult life [2
]. Because MetS and its risk components occurring in childhood can be restored to normal [5
], identifying and treating adolescents with this syndrome through an efficient risk screening implement is an advocated strategy for controlling subsequent adverse disease consequence.
Among 5 components of MetS, 3 cardiometabolic dysfunctions must be determined via blood biochemical examination. Because blood assessment is an invasive and more costly test approach, using it as a universal screening tool for MetS may be less feasible among younger children. Therefore, the development and validation of a simple and robust screening instrument for identifying adolescent MetS is a vital work for pediatric public health.
Anthropometric studies have reported that, compared with entire body fatness measured by body mass index (BMI), regional fat distribution is more associated with metabolic disturbances and cardiovascular health risks [6
]. Adiposity investigations into measurement of bodyfat distribution have indicated that body adiposity index (BAI) can directly estimate bodyfat percentage [7
]; body roundness index (BRI) is a predictor of the percentages of bodyfat and visceral adipose tissue [8
]; waist-to-height ratio (WHtR) acts as a better risk marker for cardiovascular risk and shorter lifespan than BMI [9
]; abdominal volume index (AVI) can be employed to estimate overall abdominal volume that theoretically comprises intra-abdominal fat and adipose tissue volumes [11
]; waist circumference (WC) was a stronger predictor of obesity-related cancers [12
]; waist-to-hip ratio (WHR) measures apple- or pear-like body shape [13
]; a body shape index (ABSI) expresses the excess risk from high WC adjusted for the effect of BMI [14
]; and conicity index (CoI) is a double cone-shaped derived indicator for abdominal obesity that can better predict 10-year cardiovascular risk [15
]. In insulin and fat function studies, triglyceride-glucose index (TGI) was used as a surrogate for identifying insulin resistance among healthy subjects [17
], and visceral adiposity index (VAI) was identified as an indicator of visceral adipose function and insulin sensitivity [18
]. Furthermore, lipid accumulation product (LAP) has been recommended as a lipid overaccumulation indicator [19
]. Although these adiposity indices were developed from adult populations, they may be appropriate to be used as a risk screening tool for identifying adolescent MetS, given that obesity has been a major risk for adult MetS [20
Adiposity indices created to measure body adiposity, abdominal obesity, body shape, and visceral fat accumulation are a group of interrelated variables that may have a characteristic aggregation. Principal component (PC) analysis is a statistical method that can convert interrelated variables into reduced independent and interpretable PCs with a specific characteristic through multilinear subspace algorithms, and produce a characteristic-weighted combined score for each retained PC [22
]. This technique has been used to study the clustering of pediatric cardiometabolic risk factors [23
], and is an appropriate method to investigate the characteristic groups of adiposity indicators.
In one school-based longitudinal investigation of the alteration of MetS typology, 16.4%, 5.7%, and 1.1% of adolescents were observed to have new incident, unstable/remitted, or persistent MetS, respectively, as adolescents aged into young adulthood [5
]. This raises the question of what kind of adiposity indicators can effectively identify MetS for children in school settings using criteria for adolescents and young adults. In the present study, data from the adolescent Nutrition and Health Survey in Taiwan (ado-NAHSIT) was used to develop efficient adiposity indicators from 12 characterized candidates for identifying MetS that may occur during the transition from adolescence to young adulthood. Data from the multilevel Risk Profiles for adolescent Metabolic Syndrome (mRP-aMS) study was employed to validate the accuracy and efficiency of determining MetS for the selected adiposity indicators.
displays the distribution of anthropometric characteristics, obesity indicators, and MetS, stratified by sex, for adolescents in the ado-NAHSIT and mRP-aMS studies. In both investigations, sex differences in anthropometric parameters and adiposity indicators were notable. The boys had higher levels of WC, systolic blood pressure, and fasting plasma glucose than the girls, whereas the girls had greater levels of high-density lipoprotein cholesterol than the boys. In the ado-NAHSIT study, 2.37% and 4.11% of girls and boys, respectively, had IDF-adoMetS; and 3.30% and 4.53% of girls and boys, respectively, had JIS-AdMetS. In the mRP-aMS investigation, 1.43% and 3.16% of female and male adolescents, respectively, had IDF-adoMetS; and 2.72% and 3.46% of female and male adolescents, respectively, had JIS-AdMetS.
In the training dataset, the principal component analysis (PCA) procedure converted 12 obesity indicators into a comparable PC structure in each sex, with the first 3 PCs explaining 92.4% and 93.8% of the overall variance for girls and boys, respectively (Table 2
). Among girls, PC1, PC2, and PC3 scores were respectively more correlated with bodyfat-, body-shape-, and lipid-enhanced adiposity indicators (factor loadings: 0.366–0.419, 0.416–0.656, and 0.417–0.650; total variance explained: 52.7%, 20.7%, and 19.0%, respectively). Among boys, PC1 to PC3 were separately more associated with bodyfat-, lipid-, and body-shape-enhanced obesity indicators (factor loadings: 0.365–0.414, 0.314–0.682, and 0.377–0.767; total variance explained: 59.7%, 17.8%, and 16.3%, respectively).
illustrates the covariate-adjusted risks of IDF-adoMetS and JIS-AdMetS associated with single and combined adiposity indicators in the training dataset. Apart from ABSI for the girls, all obesity indicators were associated with a significantly higher risk of developing MetS. Standardized LAP was the strongest risk factor for girls in the IDF-adoMetS and JIS-AdMetS (adjusted OR = 5.5 and 7.9, respectively), while standardized WC was the strongest risk factor for boys in the 2 MetS (adjusted OR = 6.2 and 5.5, respectively).
presents the pCorr and pR2
of single and combined adiposity indicators associated with the number of abnormal IDF-adoMetS and JIS-AdMetS components in ado-NAHSIT participants. For the IDF-adoMetS, AVI, WHR, LAP, and PC1 respectively had the highest pCorr (0.613, 0.443, 0.613, and 0.621 in girls and 0.623, 0.502, 0.648, and 0.622 in boys) in bodyfat-enhanced, body-shape-enhanced, lipid-enhanced, and combined adiposity index groups, after adjusting for the covariates. These obesity indicators separately explained the greatest variability in the number of abnormal IDF-adoMetS components in each group (pR2
, 19.6–38.6% in girls; 25.2–42.1% in boys). Comparable pCorr and pR2
patterns for these adiposity indicators were observed for the JIS-AdMetS.
presents the discriminatory abilities of adiposity indicators in identification of IDF-adoMetS and JIS-AdMetS among ado-NAHSIT adolescents. For both sexes, AVI/WC, WHR, LAP, and PC1 respectively had the greatest AUC for identifying IDF-adoMetS (0.941/0.941, 0.826, 0.942, and 0.939 among girls; 0.955/0.955, 0.898, 0.956, and 0.953 among boys) in bodyfat-enhanced, body- shape-enhanced, lipid-enhanced, and combined indicator groups. Similar results for these 5 indicators were found for the JIS-AdMetS (0.916/0.916, 0.833, 0.921, and 0.918 among girls; 0.922/0.922, 0.871, 0.938, and 0.922 among boys). YI provided comparable information for the two MetS. Using the AVI cutoff points of 13.96 for girls and 16.57 for boys, respectively, this adiposity indicator revealed a superior discrimination in identifying IDF-adoMetS (sensitivity/specificity: 95.7%/86.7% among girls; 100.0%/88.4% among boys) and JIS-AdMetS (sensitivity/specificity: 93.8%/87.4% among girls and 90.7%/88.3% among boys).
illustrates the differences in AUCs for identifying IDF-adoMetS and JIS-AdMetS across 5 superlative adiposity indicators in the training dataset. The AUCs were compatible across AVI, WC, LAP, and PC1. Nevertheless, all were significantly greater than that for WHR for the 2 MetS across both sexes (all p
-values ≤ 0.028).
presents the MetS discriminatory ability for the selected adiposity indicators in mRP-aMS adolescents using the cutoff points determined by the training dataset. Among girls, AVI (0.816) and PC1 (0.826) had the highest YI for identifying IDF-adoMetS and JIS-AdMetS, respectively, and the second highest YI was WC (0.814) and AVI (0.740). Among boys, WC provided the greatest YI for IDF-adoMetS (0.884) and JIS-AdMetS (0.798), and AVI (0.860) and PC1 (0.787) offered the second highest YI for the 2 MetS, respectively. For girls, AVI had a superior identification efficiency in positive test number, in that every 7.4 and 4.3 positive tests of AVI had a correct IDF-adoMetS and JIS-AdMetS identification. For boys, WC had an exceptional detection efficiency in total test number, in that every 32.4 tests of WC had an accurate identification in both MetS.
Except for the ABSI for girls, the single and combined adiposity indicators investigated were all qualified as a risk marker for IDF-adoMetS and JIS-AdMetS after taking multiple confounding effects into account. In the training dataset, AVI/WC, WHR, LAP, and PC1 were, respectively, the principal adiposity indicators for identifying the 2 MetS among bodyfat-enhanced, body-shape-enhanced, lipid-enhanced, and combined indicator groups for both sexes. In the validation dataset, AVI/WC for IDF-adoMetS and PC1/AVI for JIS-AdMetS among girls, and WC/AVI for IDF-adoMetS and WC/PC1 for JIS-AdMetS among boys were validated to have an excellent discrimination for the identification of MetS.
For the diagnosis of MetS, the IDF-adoMetS criteria (central obesity + 2 other abnormal components) are stricter than that for JIS-AdMetS (any 3 or more abnormal components) [20
]; thus, the proportions of MetS for IDF-adoMetS in the 2 study samples were both lower than that for JIS-AdMetS (Table 1
). Adolescence is an important growth stage that can carry health hazards into young adulthood. Our study examined and validated the accuracy of adiposity indicators for determining IDF-adoMetS and JIS-AdMetS. This method can investigate the possible application for obesity indicators in identifying MetS during the transition from adolescence to young adulthood.
The assessed obesity indicators were derived from mathematical models for abdominal fat and adipose tissue volumes [11
], the measurements of bodyfat and visceral adipose tissue using dual energy X-ray absorptiometry and/or magnetic resonance imaging [7
], or their association with visceral fat dysfunction, central lipid accumulation, and cardiometabolic risks [9
]. Nevertheless, their correlations with each other are undiscussed. Our PC analysis indicated that BMI, BAI, BRI, WHtR, AVI, and WC were relatedly clustered in bodyfat-enhanced PC, ABSI, CoI, and WHR in body shape-enhanced PC, and TGI, VAI and LAP in lipid-enhanced PC, respectively. This implies that these adiposity indicators have characteristic aggregation. Additionally, the 3 characterized PC scores were associated with IDF-adoMetS and JIS-AdMetS in a different risk (lipid-enhanced PC score having the highest risk in both MetS, Figure 2
) and with abnormal MetS component numbers in a heterogenous correlation (bodyfat-enhanced PC score having the greatest correlation, Table 3
) among both sexes. These findings suggest that characteristic-specific adiposity indicators should be separately considered in the valuation of their associations with various health risks given their interrelated nature.
This study revealed that all adiposity indicators (ABSI aside) were positively associated with IDF-adoMetS and JIS-AdMetS risks and their abnormal component numbers among girls and boys, and the associations were statistically independent of potential confounders. Obesity is a vital contributor of adolescent MetS [21
]. In a meta-analysis, childhood adiposity was verified as a significant predictor for the risks of abnormal carotid intima media thickness and dysglycemia in adulthood [38
]. Although specific nutritional intake and lifestyles might be valuable for enhancing the accuracy of MetS screening, the adiposity indictor is still a single, simple, robust, and applicable risk-screening tool for identifying adolescent MetS in the community. Even the selection of an appropriate indicator and its associated cutoff point may depend on sex, age, and ethnic/cultural group. ABSI is an indicator that was created to adjust for the effect of BMI [14
]. As observed in this study this index was not substantially associated with the BMI-correlated MetS outcomes.
In the training dataset, we found that bodyfat- and lipid-enhanced adiposity indicators generally had a better discriminating ability for identifying IDF-adoMetS and JIS-AdMetS than did body-shape-enhanced indicators among both sexes (AUCs, 0.830–0.956 vs. 0.489–0.898, Table 4
). The AUCs for the superlative adiposity indicators of bodyfat- and lipid-enhanced groups were both significantly higher than that for the body-shape-enhanced group (Figure 3
). Brazilian, Chilean, and Spanish investigators have recommended the bodyfat-enhanced indicators WHtR, AVI, and BMI as an excellent screening tool for adolescent MetS [39
]. Although no adolescent studies have investigated LAP and VAI, one adult study reported that LAP was the most accurate indicator for determining male and female MetS [43
]. Obesity is a growing global health problem that increases the risk of multiple physical and mental disorders [44
]. These results indicate that bodyfat- and lipid-enhanced adiposity indicators should be intensively applied in risk assessments of the association between pediatric obesity and cardiometabolic diseases.
Studies that have evaluated anthropometric indicators for the identification of adolescent MetS in the community have found that AVI and WC for both girls and boys in Spain, BMI/WC for females and WHtR/WC for males in Chile, and WHtR for both sexes in Brazil have an excellent capacity for discriminating MetS [39
]. However, there was a lack of validation of screening accuracy and efficiency for the indicators in these studies. Using a comparable adolescent population for validation, this investigation demonstrated that AVI/WC for girls and WC/AVI for boys had an excellent discriminating ability for identifying IDF-adoMetS in the community; however, the best indicators in discriminating JIS-AdMetS were PC1/AVI for girls and WC/PC1 for boys. This indicates that the AVI- and WC-included bodyfat-enhanced combined score (PC1) becomes more significant in detecting JIS-AdMetS. Although PC1 was recognized as an excellent adiposity indicator for identifying JIS-AdMetS, it requires intricate calculation. In community practices, our study suggests AVI as a female, and WC as a male, risk screening tool for MetS that can be applied to the transition from adolescence to young adulthood in a Taiwanese population.
In this study, the AVI cutoff points for identifying female IDF-adoMetS and JIS-AdMetS were both 13.96. Evaluated in the validation sample, 90.0% and 81.6% of sensitivity and 91.6% and 92.4% of specificity were observed (YI, 0.816 and 0.740), respectively, for the two MetS. AVI also was recommended as a risk-screening instrument for identifying female IDF-adoMetS in Spanish adolescents; however, the YI of this indicator with cutoff of 10.89 was only 0.70 (sensitivity, 100.0% and specificity, 70.0%) [39
]. In our investigation, the WC cutoff points for determining male IDF-adoMetS and JIS-AdMetS were both 90.5 cm, which is close to the abnormal level of central obesity defined for Asian adults (WC ≥ 90 cm) [20
]. Evaluated in the validation sample, every 32.4 tests of WC were observed to have an accurate IDF-adoMetS and JIS-AdMetS, respectively. In Spain, WC was suggested as an anthropometric discriminator for identifying male IDF-adoMetS, in which the cutoff of 75.0 cm rendered a 0.57 of YI (sensitivity, 100.0% and specificity, 75.0%) [39
]. The discriminating abilities of AVI and WC for identifying IDF-adoMetS were confirmed in a southeast Spanish adolescent population [47
In a clinically established adolescent database, if the blood samples of the participants had not been collected, or blood samples had not been examined for MetS-related variables, an adiposity indicator with very high sensitivity may be used as a tool for discovering adolescent MetS. Furthermore, an adiposity indicator with very high specificity may be employed as an instrument for confirming adolescent MetS. This framework can be applied to clinical settings. However, the appropriate adiposity indicators and their associated cutoff points have to be developed for the study population.
This study had several strengths. First, a large-scale nationally representative sample was used to develop efficient adiposity indicators and the accuracy and efficiency of the selected indicators were validated in another large-scale representative sample. Second, the clustering characteristics of obesity indicators and characterized functions were evaluated and validated. Third, multiple adiposity indicators were concurrently assessed and were used to investigate their discriminating capability for identifying MetS in the transition from adolescence to young adulthood. Fourth, although appropriate adiposity indicators and their cutoff points might vary by study populations, our research methodology and network can be applied to other countries that want to develop their specific obesity indicators. Alternatively, we recognize the limitations of our analyses. First, the recommended adiposity indicators had no causally predictive capability in determining MetS since they were all developed in a cross-sectional nature. Second, the participants used to develop indicators were Taiwanese adolescents, thus our findings may not be directly generalizable to other populations.