Investigating The Denitions of Metabolic Syndrome Among High School Students in Taipei and The Optimal Cutoff Points of Relevant Risk Factors

The effects of different denitions for metabolic syndrome (MetS) on its prevalence were examined, and the differences in the discriminatory power, as well as the optimal cutoff points of relevant risk factors, were analyzed in this study. Methods: 45,756 health checkup data sets from 2011 to 2014 of high school students aged between 15 to 17 years were sourced in Taipei city. The database included the students’ gender, age, height, weight, waist circumference (WC), systolic and diastolic blood pressure, as well as biochemical markers such as triglycerides (TG), high-density lipoprotein cholesterol, and fasting glucose (FG) levels. The ROC curve statistical approach was used to analyze the discriminatory power and optimal cutoff points of the relevant MetS risk factors. percentile,


Introduction
Childhood and adolescent overweight and obesity are major concerns in developed countries [1].
According to the 2010-2011 Nutrition and Health Survey in Taiwan (NAHSIT), the prevalence among Taiwanese adolescents aged between 11 and 18 years of overweight and obesity was 12.4% and 16.8%, respectively [2]. With the number of overweight and obese adolescents on the rise, the prevalence of metabolic syndrome (MetS) has also escalated, and has a positive correlation with obesity [3]. From 1988 to 1994, the prevalence of MetS among American adolescents aged 12 to 19 years was 28.7% in overweight subjects (BMI ≥ 95th percentile), which was signi cantly higher than the prevalence of 6.8% and 0.1% in subjects with a BMI in the 85th to 95th percentile and low-risk subjects (BMI < 85th percentile), respectively [3]. Research has shown that cardiometabolic risk factors identi ed in childhood and adolescence are associated with subclinical atherosclerosis in adulthood [4]. The presence of MetS during childhood exacerbates the risk of having adulthood MetS, cardiovascular disease (CVD), and type 2 diabetes mellitus (T2DM) [5,6]. Therefore, screening of MetS and obesity during childhood and adolescence, as well as further interventions (especially changes in diet and increase physical activity), are important factors to improve the future health of the adult population [7].
MetS comprises of the risk factors for CVD and T2DM [8]. It represents the association between obesity, insulin resistance, hypertension, dyslipidemia, T2DM, and CVD [9]. In 1988, Reaven rst proposed the notion of insulin resistance (IR), and named the cluster of risk factors for CVD and diabetes as "Syndrome X" [10]. In 1998, the World Health Organization (WHO) formally named the syndrome as MetS and de ned its criteria [11]. Afterwards, various de nitions were proposed, such as that of the European Group for the Study of Insulin Resistance (EGIR) in 1999 [12]; the National Cholesterol Education Program Adult Treatment Panel III(NCEP ATP III) in 2001 [13]; the International Diabetes Federation (IDF) [14] and the American Heart Association/National Heart, Lung, and Blood Institute (AHA/NHLBI) [15] in 2005. Early studies suggested IR as the primary cause of MetS [16], but subsequent studies deduced that the interaction between obesity, IR, and in ammation plays a crucial role in the development of MetS [17].
Moreover, MetS is in uenced by other metabolic and pathological factors such as in ammatory factors, adipocytokines, cortisol, oxidative stress, vascular factors, inheritance, and lifestyle factors [9].. The uni ed criteria for diagnosing MetS in adults have been determined [8] and successfully applied in clinical practice and research [9]. However, due to the growing concern about MetS in children and adolescents, many studies have attempted to de ne MetS in these groups [9]. At present, as there are no uni ed diagnostic criteria, the di culty in de ning the condition may be associated with physiological changes during growth, racial differences, lack of CVD cases, and lack of clinical trials [18]. Consequently, the discrepancies in the prevalence of MetS in children and adolescents is remarkably large as there is no uni ed diagnostic criteria. Many studies have modi ed the NCEP ATP III's de nition of MetS in adults [19], while recent studies commonly adopt the modi ed NCEP ATP III criteria and the 2007 IDF criteria [20].
This study aimed to elucidate the in uence that different de nitions of MetS used to categorize rst year senior high school students in Taipei City had on the prevalence of MetS reported, as well as to analyze the differences in the discriminatory power and the optimal cutoff points for relevant risk factors. The results can serve as a reference for the future de nition of MetS in adolescents and to enable early intervention measures.

Study population and data collection
This study was approved by the institutional review board of the Taipei City Hospital (TCHIRB-10811003-E). The personal health data of high school students aged between 15 to 17 years who underwent health checkups at a district hospital from 2011 to 2014 was sourced from the hospital's database. There were 50,280 sets of data. After omitting those with missing information, a total of 45,756 sets of data were included in this study.
The database of this study included the subjects' gender, age, height, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), and waist circumference (WC), as well as biochemical markers such as triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and fasting glucose (FG) levels.
Height was measured using a standard stadiometer and measured height (in cm) was rounded to the rst decimal place. Weight was measured with a standard weighing scale and measured weight (in kg) was rounded to the rst decimal place. To measure blood pressure (BP), a subject was told to sit in a relaxed manner, and their BP was measured by placing their left arm into a tunnel-type electronic sphygmomanometer. The recorded BP was taken as the mean of two measurements. To measure WC, a measuring tape was kept parallel to the ground and was wrapped around a subject's waist, starting from a point midway between the upper iliac crest margin and the lower rib margin. The recorded WC was taken after the subject had breathed out. A subject's body mass index (BMI) was taken as their weight in kilograms divided by the square of their height in meters.

De nition of MetS
In this study, three de nitions for the prevalence of MetS were compared: the modi ed NCEP ATP III proposed by Ford et al. [26]; the 2007 IDF [20]; and the new criteria based on the cutoff points of this study's receiver operating characteristic (ROC) curves. All of these diagnostic criteria include the ve components of WC, TG, HDL-C, BP, and FG. The modi ed NCEP ATP III served as a baseline for the criteria of this study, in which the optimal cutoff point of each MetS component with an area under the ROC curve (AUC) greater than 0.8 were taken as new components (modi ed WC, TG, and HDL-C). The components of BP and FG remained unchanged (see Table 1).

Statistics
The statistical approaches used to analyze the discriminatory power and optimal cutoff points of relevant MetS risk factors in this study included descriptive statistics, chi-squared tests, independent sample t-tests, and ROC curves.

Results
The results of this study are shown in Table 2. The male subjects had a mean age of 15.18 ± 0.41 years; a mean height of 170.63 ± 6.14 cm; a mean weight of 63.74 ± 13.54 kg; a mean BMI of 21.81 ± 4.33 kg/m 2 ; a mean WC of 72.29 ± 10.66 cm; a mean SBP of 117.23 ± 15.18 mmHg; a mean DBP of 61.75 ± 11.23 mmHg; a mean TG level of 72.87 ± 36.20 mg/dL; a mean HDL-C level of 56.66 ± 11.37 mg/dL; and a mean FG level of 84.76 ± 9.55 mg/dL.
The female subjects had a mean age of 15.19 ± 0.41 years; a mean height of 159.28 ± 5.53 cm; a mean weight of 52.60 ± 9.35 kg; a mean BMI of 20.67 ± 3.43 kg/m 2 ; a mean WC of 67.02 ± 7.80 cm; a mean SBP of 106.13 ± 13.33 mmHg; a mean DBP of 62.82 ± 9.81 mmHg; a mean TG level of 68.89 ± 27.16 mg/dL; a mean HDL-C level of 63.38 ± 12.74 mg/dL; and a mean FG level of 83.73 ± 9.09 mg/dL.
The relevant MetS risk factors of the rst year senior high school students have different discriminatory powers toward the diagnosis of MetS ( Fig. 1 and Table 3). Table 3 Page 7/18 0.902); and HDL-C (AUC = 0.827, sensitivity = 0.763, speci city = 0.752). Generally speaking, the subjects' WC, BMI, and TG had the highest discriminatory power, sensitivity, and speci city.
For the male subjects, the cutoff points for the main components of MetS were WC, TG, HDL-C, SBP, DBP,

MetS in adolescents and the prevalence of its components
At present, there are no uni ed criteria for diagnosing MetS in adolescents, which makes it di cult to compare between studies in the literature. The prevalence of MetS differs across studies, which could be associated with differences in diagnostic criteria, age (especially IR in adolescents), gender, regions, and races [7]. In this study, the prevalence of MetS as measured using the modi ed NCEP ATP III and IDF criteria was 2.3% and 1.2%, respectively. These values were lower than those summarized in literature [21][22][23][24][25]. The prevalence of MetS measured using this study's criteria was 4.3% and there were signi cant differences observed between males and females. In particular, MetS was more prevalent in males than females, which is in line with an American study [7].
Differences in the percentage of abnormal measures of MetS components are a result of different diagnostic criteria. According to this study's criteria, 10.7% of high school students had a slightly large WC, 10.8% had high TG, and 21.2% had low HDL-C. The study population had a higher percentage of individuals with abnormal measures for these three components than for the other two components (elevated BP and elevated FG).
The criteria developed in this study identi ed a signi cantly higher percentage of subjects with low HDL-C (21.2%), compared to that seen using the modi ed NCEP ATP III (4.2%) and the IDF (5.1%). A possible explanation could be that a uni ed criterion (< 40 mg/dL) was used in the modi ed NCEP ATP III (in which subjects were 12 to 19 years old) and the IDF (in which subjects were below 16 years old). Moreover, gender differences were not adjusted for in the cutoff values, which resulted in a low prevalence among females. This has also been observed in another study [24].
Differences were observed between the results from the modi ed NCEP ATP III and IDF in the percentage of individuals with abnormal measures for TG (9.7% vs. 2.9%) and BP (18.6% vs. 15.7%). This is because the modi ed NCEP ATP III criteria de ne elevated BP as ≧ 90th percentile after adjusting for age, gender, and height, while high TG are de ned as ≧ 110 mg/dL after adjusting for age [3], whereas the IDF criteria de ne elevated BP in adults as ≧ 130/85 mmHg and high TG as ≧ 150 mg/dL.
Compared with the modi ed NCEP ATP III criteria adjusted by Ford et al. [26], the percentage of subjects identi ed as having a large WC in this study(10.6%) was similar to that in a Korean population (9.7%) [22], but lower than that in an American population (19.1%) [21]; the percentage of subjects with elevated BP in this study (18.6%) was also similar to that in a Korean population (20.4%), but higher than that in an American population (6.9%). These differences could due to differences in the populations, which indicates the importance of establishing a large database on the WC and BP of people in different regions. The percentage of subjects with high TG in this study(9.7%) was lower than that in studies conducted in Korea (21.2%) and the USA (25.6%); the percentage of subjects with low HDL-C in this study(4.2%) was lower than that in Korea (11.6%) and the USA (19.3%); and the percentage of subjects with elevated FG was 2.9%, which is signi cantly lower than that in the USA (14%) and Korea (11.4%) [21,22]. The components of MetS in adolescents must be adjusted for regional differences. This highlights the importance of the results of this study.
Moreover, according to the criteria of the modi ed NCEP ATP III, the IDF, and this study, about 34.4%, 28.3%, and 44.6% of adolescents, respectively, had at least one MetS component. The fewer of these components that are present during childhood, the lower the cardiovascular risk in the future [27]. Some researchers [28] have emphasized that the effects of metabolic risk factor clustering are more important than diagnosing MetS in children. Based on these arguments, it is not only crucial to detect MetS in adolescents, but those present with MetS components despite not yet reaching the diagnostic criteria should receive attention as well, to provide prompt intervention and prevention [7,28].

The predictive power of MetS components
According to the results of this study, WC has the highest predictive power, sensitivity, and speci city, regardless of gender. This is in agreement with other studies which have suggested that WC is a good indicator for predicting MetS during adolescence [25, 29,30] and adulthood [31]. A study on American adolescents between 12 and 19 years of age revealed that abdominal obesity was closely associated with MetS and other MetS components [25]. Another study on 15-year old Greek teenagers showed that a WC at the 75th percentile or higher is closely related to the phenotypes of MetS [29]; while a study on Chinese adolescents between 11 and 16 years old indicated that WC has the best predictive power toward MetS [30].
To determine the optimal cutoff point for WC, Cook et al. (2003) [3] took into account the differences between adolescents and adults, and de ned abdominal obesity as at the 90th percentile or higher; in 2004, de Ferranti et al. [32] adopted a value at the 75th percentile or higher as their standard; afterward, in 2007, the IDF study (in which subjects were below 16 years of age) [20] and numerous studies [21,22,26] adopted a WC at the 90th percentile or higher as their standards. A Chinese study on children and adolescents between 7 and 18 years of age found that a WC at the 75th percentile and the 90th percentile was the optimal cutoff point for predicting the risk of cardiovascular risk [33]. The optimal cutoff point for WC speci ed in this study was 86.8 cm for males and 76.25 cm for females, which was around the 90th percentile and similar to that of the previous studies.
In this study, TG level also had adequate predictive power, with an optimal cutoff point of 108 mg/dL for males and 104.05 mg/dL for females, similar to that of the modi ed NCEP ATP III criteria (< 110 mg/dL). Even though the sensitivity and speci city of HDL-C were slightly inadequate, gender differences were also observed for HDL-C in this study. These results highlight that de nitions should be speci c for gender. According to the criteria of the modi ed NCEP ATP III and the IDF, male subjects in this study had a higher prevalence of high TG, low HDL-C, elevated BP, and high FG than their female counterparts. After adjusting the criteria of this study, the prevalence of low HDL-C signi cantly increased for both genders, in which females (25.5%) had a higher prevalence than males (18.3%).
SBP, DBP, and FG had weaker predictive powers. In particular, FG had the weakest predictive power, which was also observed in other studies with adolescents [30] and adult [31] subjects. Furthermore, according to the IDF criteria, the prevalence of elevated BP was signi cantly higher in males (22.3%) than females (6.1%). This could be caused by the use of uni ed criteria for adults ( ≧ 130/85 mmHg) that neglects gender differences in BP.
In this study, the cutoff points for MetS components were rede ned based on the results of rst year high school students in Taipei. This indicates that it is necessary to take into account regional differences when determining de nition criteria.
The predictive power of BMI Based on the results of this study, BMI also had good predictive power on MetS in adolescents, after that of WC and TG. The optimal cutoff point for males was between the 80th to 85th percentile (25.6 kg/m 2 ) for males and approximately the 90th percentile (24.65 kg/m 2 ) for females.
The WC and BMI of adolescents are good predictive indicators of cardiovascular risk factors [34]. A longitudinal study highlighted the close association between BMI and many other cardiometabolic risk factors, while changes in WC mainly have a stronger correlation with FG [35]. A study on children and adolescents between 8 and 19 years of age [36] revealed that a high BMI has strong predictive power for cardiometabolic risk factors. In addition, the sensitivity of BMI is higher among obese adolescents while its speci city is higher among overweight adolescents [37]. The de nitions of overweight and obesity are currently based on a person's BMI, and their criteria differ for adolescents [37][38][39]. In 2007, the WHO de ned overweight as having a BMI between the 85th and 95th percentiles, while obesity is de ned as having a BMI greater than the 95th percentile [37]; in 2012, the International Obesity Task Force (IOTF) deduced the cutoff points for BMI in adolescents and adults after mathematical adjustments based on the de nition of overweight and obesity in adults I [38,39]. One study [40] used the three aforementioned criteria (WHO; Conde and Monteiro; IOTF) to analyze Brazilian adolescents between 12 and 20 years old; it revealed that the IOTF criteria had the best predictive power for MetS (AUC = 0.75-0.89), with a sensitivity ranging from 59.4-84.2% and a speci city ranging from 88.2% to 93.6%. In contrast, according to this study's adjusted de nition of MetS, BMI had a better predictive power (AUC = 0.915-0.926) and sensitivity (0.817-0.9) for the high school students. The differences between these results and those of the aforementioned studies could be due to the differences in the diagnostic criteria, age, region, and race in adolescents [7]. Limitations The cross-sectional research design of this study hindered observations of the causal relationships in the data. Moreover, the subjects were adolescents from an urban region of northern Taiwan, which limited the extrapolation of results to adolescents in rural locations as well as those with special circumstances.
However, the results of this study are still valuable and can serve as a reference for de ning MetS in Taipei adolescents. This study analyzed the differences in the discriminatory power of relevant risk factors as well as their optimal cutoff points, which could provide markers for early interventions in the future. Subsequent research could include the potential confounders of MetS, such as the in uence of puberty and temporary IR during adolescence [7]; as well as taking into account more biochemical markers, and participating in cross-regional and cross-cultural cohort studies. A combination of these approaches would make the understanding of MetS in Taiwanese adolescents more comprehensive.

Conclusion
The prevalence of MetS in adolescents differs when measured using different criteria. In comparison to the results with the criteria of the modi ed NCEP ATP III and the IDF, the criteria de ned in this study found that the prevalence of MetS among adolescents in Taipei City was higher (4.3%). Of the MetS components, WC and TG had the strongest discriminatory power, sensitivity, and speci city, while FG had the weakest discriminatory power. The optimal cutoff point for WC was approximately at the 90th percentile; while the optimal cutoff point for TG was similar to the criteria of the modi ed NCEP ATP III.
Moreover, about 44.6% of adolescents had at least one MetS component. The early detection of relevant risk factors during adolescence is a crucial issue as it provides vital information for prevention and intervention, to reducing the risk of CVD and T2DM in the future. For accurate detection, it is necessary to rede ne the cutoff points for MetS components speci c to adolescents in different regions, to establish suitable criteria for adolescents in Taiwan as a whole. Declarations Figure 1 ROC curves of relevant MetS risk factors for boys (left) and girls (right).