Stability and transformation of metabolic syndrome in adolescents A prospective assessment in relation to the change of metabolic risk factors

Pei-Wen Wu Kaohsiung Medical University Yi-Wen Lai Kaohsiung Medical University Yu-Ting Chin Kaohsiung Medical University Sharon Tsai Kaohsiung Municipal Hsiao-kang Hospital: Kaohsiung Municipal Siaogang Hospital Tun-Min Yang Kaohsiung Medical University Wei-Ting Lin Tulane University School of Public Health and Tropical Medicine Chun-Ying Lee Kaohsiung Medical University Hospital: Kaohsiung Medical University Chung Ho Memorial Hospital Wei-Chung Tsai Kaohsiung Medical University Hsiao-Ling Huang Kaohsiung Medical University David W. Seal Tulane University School of Public Health and Tropical Medicine Chien-Hung Lee (  cnhung@kmu.edu.tw ) Kaohsiung Medical University https://orcid.org/0000-0002-0988-264X


Introduction
The pathogenesis of cardiovascular disease starts in childhood, when cardiometabolic risk factors are rst observed [1]. In adolescents, metabolic syndrome (MetS)-a syndrome involving the clustering of abdominal obesity, hypertriglyceridemia, high fasting plasma glucose (FPG), low high-density lipoprotein cholesterol (HDL-C), and elevated blood pressure (BP)-is a vital risk marker for future cardiometabolic disease [2]. Longitudinal studies have demonstrated that childhood MetS is associated with a 2.0-2.9-fold risk for subclinical atherosclerosis, 2.3-11.5-fold risk for type 2 diabetes mellitus (T2DM), and 14.6-fold risk for cardiovascular disease after 14-30 years [3][4][5].
The clustering of cardiometabolic risk factors exceeds what should be observed by chance aggregation, indicating the presence of common underlying pathophysiological mechanisms [2,6,7]. Accordingly, the questions of what cardiometabolic factor clustering structure re ect the latent pathogenic mechanisms and whether the clusters retain stable through growth stages in adolescents remain to be answered. An individual's MetS status may change from adolescence to young adulthood [8,9]. One longitudinal study following adolescents up to adulthood revealed that 76.9% of the adolescents never had MetS, 16.4% had incident MetS, 5.7% experienced unstable/remitted MetS, and 1.1% had persistent MetS [9]. However, the typological changes in MetS status and their stability during adolescence remains unclear. A comparison of MetS transformation using diverse MetS diagnostic criteria may help clarify this.
Prospective investigations have observed that 3.8%-5.2% of MetS-negative adolescents (i.e., individuals aged 12-19 years) developed new MetS after 3 years; by contrast, 48.6%-56.1% of MetS-positive adolescents achieved remission 3 years later, and in 43.9%-51.4% of MetS-positive adolescents, MetS persisted [8]. If the structure of cardiometabolic parameter clustering for MetS is stable over adolescent growth, an investigation into the effect of metabolic risk determinants on typological transformation of adolescent MetS is warranted. In cross-sectional studies involving adolescents, metabolic risk factors have been observed to occur in childhood, but whether and to what extent they persist is unknown [10][11][12][13]. Insights into the persistence of metabolic risk factors and how they contribute to typological change of MetS can help create effective strategies for preventing MetS, increasing remission, and managing persistent MetS.
In a nationwide survey conducted in 2010-2011 in Taiwan, the prevalence of adolescent MetS, as de ned by the Taiwan Pediatric Association (TPA) and International Diabetes Federation (IDF) diagnostic criteria, was 4.1% and 3.0%, respectively, with 22.1%, 12.3%-19.3%, and 17.7%-18.1% of adolescents having increased FPG, low HDL-C, and abdominal obesity, respectively [14]. Continuously monitoring and assessing the clustering of MetS risk factors and their impact on adolescent cardiometabolic health are warranted. This community-based longitudinal study evaluated the latent clustering structure and its stability for MetS during adolescence and investigated the relationship between changes in metabolic risk factors and transformation of MetS over 2 years of follow-up.

Participants
The adiposity-cardiovascular disease axis (adi-Cars) investigation was a large representative cohort study conducted to investigate multilevel determinants and risk pro les of cardiometabolic disease, prediabetes, and hyperuricemia among adolescents aged 12-14 years from southern Taiwan. The adolescents lived in three regions with varying levels of urbanization: Kaohsiung City, Pingtung County, and Taitung County. At the baseline survey, a three-stage procedure of random sampling of geographically strati ed clusters was introduced to recruit the study participants. In stage one, Kaohsiung City, Pingtung County, and Taitung County were geographically strati ed into nine, six, and four divisions, respectively. In stage two, all junior high schools within each division were compiled and listed, and 31 schools were randomly selected through computer-generated random numbers. In stage three, three classes (20-25 students per class) were randomly selected from each chosen school. One class was considered a cluster, and all students in the chosen classes were invited to participate in this study.
Baseline data were collected between September 2014 and June 2018 and follow-up data were measured between September 2017 and June 2021. A total of 2,046 adolescents agreed to participate in the baseline anthropometric and questionnaire surveys (response rate: 94.9%). Of them, 1,516 adolescents (74.1%) participated in clinical biochemical examinations. In May 2021, a Covid-19 outbreak occurred in Taiwan and prevented follow-up by participants from three schools (these participants will be revisited after the outbreak). We excluded the students in these three schools in data analyses. Eventually, 1,246 adolescents from 28 schools with complete anthropometric and clinical blood data were followed. Of them, 1,155 participated in the anthropometric and questionnaire survey at follow-up (retention rate: 92.7%; mean length of follow-up: 2.2 years), but only 896 participated in clinical blood examinations. This study adhered to the principles expressed in the Declaration of Helsinki. The Institutional Review Board of Kaohsiung Medical University Hospital reviewed and approved this research project. Written assents from the adolescents and consents from their parents/guardians were collected for both the baseline and follow-up surveys.

Demographic and metabolic risk factors
Multilevel-structured questionnaires were developed to obtain information on sociodemographic factors and lifestyle behaviors from adolescent participants and their parents. The urbanization of the township where each school is located was categorized into seven levels according to a socioeconomic cluster analysis of 359 Taiwan townships, with level 1 denoting the most urbanized [15]. Anthropometric parameters, including height, weight, hip circumference, waist circumference (WC), systolic BP (SBP), and diastolic BP (DBP), were measured at baseline and follow-up by a research team trained according to the World Health Organization guide to physical measurements [16]. Details of anthropometric measurements have been described previously [10,17,18]. Body mass index (BMI) was calculated as weight divided by height squared (kg/m 2 ). Moreover, venous blood samples were obtained in school health centers in the morning after a >10-h overnight fast. Triglycerides (TG) and HDL-C concentrations were enzymatically determined using a chemistry autoanalyzer and commercially available reagents, and FPG levels were assessed using a glucose oxidase method (TBA-c16000, Toshiba, Tokyo, Japan) [19]. High-performance liquid chromatography (Bio-Rad Variant Turbo II HbA1c analyzer, Hercules, CA, USA) was used to measure glycosylated hemoglobin (HbA1c) values.

MetS diagnosis
MetS and MetS abnormal components were diagnosed using the IDF criteria for adolescents aged 10-18 years, TPA criteria for adolescents aged 8-18 years, and the Joint Interim Statement for adult MetS (JIS-Ad) [6,20,21]. Abdominal obesity is de ned as WC ≥90 percentile (or adult cutoff if lower) by the IDF, as BMI >95th percentile of age-sex-speci c groups by the TPA, and as WC ≥90 cm in boys and WC ≥80 cm in girls by the JIS-Ad. Low HDL-C is de ned as HDL-C <40 mg/dL for adolescents aged 10-15 years and HDL-C <40 mg/dL in adolescent boys and <50 mg/dL in adolescent girls aged 16-18 years by the IDF; however, the TPA and JIS-Ad de ne low HDL-C as HDL-C <40 mg/dL in boys and <50 mg/dL in girls. The IDF, TPA, and JIS-Ad have the same criteria for increased TG (≥150 mg/dL), high FPG (≥100 mg/dL or previously diagnosed T2DM), and elevated BP (SBP ≥130 mmHg, DBP ≥85 mmHg, or antihypertensive drug treatment). Supplementary Table S1 presents the complete de nitions for the ve abnormal components given by the IDF, TPA, and JIS-Ad. Because the IDF and TPA have similar de nitions for MetS abnormal components, we combined them as IDF-TPA criteria to accommodate any outlier de nitions of the IDF and TPA and include all potential abnormal components. IDF-, TPA-, and IDF-TPA-de ned MetS were the presence of abdominal obesity and any two other abnormal components, and JIS-Ad-de ned MetS was the presence of any three abnormal components.

Transformation of MetS status
The IDF-TPA criteria for MetS and its abnormal components were used to investigate the transformation of MetS status over 2 years of follow-up in the adi-Cars cohort. Adolescents who were MetS-negative at both baseline and follow-up were de ned as the never MetS group. Those who were MetS-negative at baseline but MetS-positive at follow-up were de ned as the incident MetS group. Those who were MetS-positive at baseline and MetS-negative at follow-up were de ned as the remitted MetS group. Those who were MetS-positive at both baseline and follow-up were de ned as the persistent MetS group. The four groups were used as the main outcome of this investigation.

Statistical analysis
We applied 6 statistical procedures for data analysis. First, the demographic and metabolic risk factors measured are presented as mean ± standard deviation for continuous variables and percentages for categorical variables. Second, exploratory factor analysis (EFA) was employed to investigate the latent factor clustering structure across metabolic parameters for baseline and follow-up surveys, similar to a prior study [8]. Before performing EFA, all variables were assessed for Gaussian normality, and nonnormally distributed variables were converted using the logarithm function. Next, we performed Bartlett's test of sphericity to examine whether metabolic risk factors have a signi cant correlation structure. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was used to evaluate the suitability of study data for structure detection. A KMO value of >0.50 was considered appropriate for factor analysis. In EFA, we used principal component analysis as the factor extraction approach and applied the eigenvalue >1 rule and surpassing the break in scree plots to extract factors [22,23]. Varimax rotation was applied to obtain more interpretable factor loadings, and the parameters with loadings > 0.4 were used in interpreting factors.
Third, Cohen's Kappa coe cient (k) was calculated to evaluate the agreement of MetS (i.e., prevalence) de ned by four MetS criteria between baseline and follow-up surveys [24]. The k values of 0.21-0.40, 0.41-0.60, 0.61-0.80, and 0.81-1.00 were interpreted as fair, moderate, substantial, and almost perfect agreement, respectively [25]. Fourth, strati ed by the typological group of MetS transformation, we applied a mixed model for repeated measures to assess within-person changes in the levels of metabolic risk factors measured at baseline and follow-up. For each participant, change in status between baseline and follow-up surveys (no change, turning negative, or turning positive) for each MetS component was also appraised. Fifth, because quaternary outcomes (i.e., never, incident, remitted, and persistent MetS) were investigated, we used a multinomial logistic regression model to assess the association of changes in metabolic risk factors with changes in MetS status over 2 years of follow-up. This modeling technique enables the simultaneous comparison of an outcome variable with >2 categories and has been veri ed to have a higher precision and statistical power compared with simple binary outcome analysis [26,27]. Finally, incidence density was used to assess the status of new-onset MetS components in follow-up survey. Multivariable Cox proportional hazards models and adjusted hazard ratios (aHRs) were applied to evaluate the association between initial risk status and subsequent risk occurrence for each MetS component. All multivariable models were adjusted for sex, age, urbanization level, and metabolic risk factors, as appropriate. Table 1 presents the distributions of demographic and metabolic risk factors measured at baseline and follow-up for the adolescent cohort. The distributions of sex and urbanization level for the participants were similar between the two time-points; age was obviously increased by 2 years. Compared with baseline, the adolescents had higher SBP, DBP, and weight-related variables and lower HDL-C and FPG at follow-up.  Fig. S1). The factor structures were similar at baseline and follow-up; both comprised a fat factor (BMI, WC, hip circumference, HDL-C, and TG), a BP factor (SBP and DBP), and a glucose factor (FPG and HbA1c). Factor loadings for each factor and the proportion of variance explained by the three factors (68.8% and 69.7%, respectively) were very comparable between the data for the two time-points. All metabolic risk factors used to interpret each factor structure had a factor loading of ≥0.508.   a Kappa coe cient was used to examine the agreement of MetS prevalence between baseline and follow-up. Table 4 presents within-person changes in metabolic risk factors between baseline and follow-up, strati ed by MetS typology per the IDF-TPA criteria. In the never group, BMI, WC, SBP, and DBP at follow-up were higher and HDL-C and FPG were lower than the corresponding values at baseline. In the incident group, the within-person changes in BMI, WC, SBP, DBP, and HDL-C were signi cant and greater than those in the never group. In the remission group, TG and FPG levels were noticeably reduced at follow-up. In the persistent group, the values of metabolic risk factors were relatively high at baseline, and SBP also increased at the 2-year follow-up. a IDF-TPA criteria were used to determine adolescent metabolic syndrome. b The mean level for within-person change in the metabolic risk factors between baseline and follow-up. Here, * denoting P <0.05 for the pair difference betwee c P value for WP change was obtained from the mixed model for repeated measures adjusted for sex, age, and urbanization level. Table 5 displays the association of changes in metabolic risk factors with changes in MetS status at follow-up. After adjustment for covariates, adolescents with a 1 mmHg increase in ΔSBP had a 1.07-fold risk of incident MetS at follow-up. Compared with the never group, the remission group had a greater elevation in HDL-C (ΔHDL-C, 0.61 vs -3.67 mg/dL) and a greater decrease in TG and FPG (ΔTG, -28.12 vs -0.35 mg/dL and ΔFPG, -8.94 vs -2.29 mg/dL).

Results
Compared with those with persistent MetS, participants having a 1 unit increase in ΔSBP, ΔTG, and ΔFPG had a 0.95, 0.98, and 0.94-fold, respectively, lower likelihood of MetS remission after 2 years.  Abbreviations: aOR, adjusted odds ratio; WC, waist circumference; HDL-C, high-density lipoprotein cholesterol; TG, triglyceride; BP, blood pressure; FPG, fasting plasma glucose; na, non-appreciable due to no subject in that group. a IDF-TPA criteria were used to determine adolescent MetS. Never presence of MetS at baseline and follow-up was used as the reference group for outcome comparisons. b aOR were obtained from polytomous logistic regression models adjusted for sex, age, urbanization level, and covariates in the Table.   Table 7 displays the baseline prevalence, follow-up incidence density, and aHR of MetS abnormal components associated with their original status. The baseline prevalence was 24.9%, 10.6%, 21.1%, 5.5%, and 11.4% for abdominal obesity, elevated BP, low HDL-C, increased TG, and high FPG, respectively. Among the ve abnormal components, low HDL-C had the highest incidence density (9.6% per year) in adolescents with an initial normal status. Abdominal obesity and low HDL-C had a greater persistent incidence density (34.3% and 36.5% per year, respectively) in participants with an original positive status.
Compared with a normal status at baseline in the ve MetS components, an abnormal status at baseline was associated with a higher risk of the abnormal status persisting at follow-up, with abdominal obesity and increased TG rendering a >5.0-fold risk each (aHR, 15.0 and 5.7, respectively). Table 7 Baseline prevalences, follow-up incidence densities, and adjusted hazard ratios of abnormal components of metabolic syndrome associated with initial status over 2 years of follow-up in adolescents.

Discussion
This study presents ndings that demonstrate that the structure of parameter clustering for adolescent MetS at baseline and follow-up were comparable.
Elevated SBP was associated with MetS incidence and persistence, and decreased SBP, TG, and FPG were associated with MetS remission after 2 years. Adolescents who had an abnormal MetS component at baseline were more likely to have the component be abnormal at follow-up than those who were normal for each MetS component.
MetS re ects a clustering of metabolic risk parameters, which are believed to originate from the common pathophysiological mechanisms of insulin resistance [28]. In this study, a fat-BP-glucose three-factor structure for MetS was observed both in adolescents aged 12-14 years (baseline survey) and 15-17 years (follow-up survey), with analogous factor loadings and proportions of total variance explained (68.8% vs 69.7%). In one school-based longitudinal investigation of adolescents aged 12-19 years, the overall parameter clustering structure of metabolic risks was identi ed to be no change after 3-years of follow-up [8]. These ndings strengthen the argument that the mechanistic underpinning for MetS is stable during adolescence. Alternatively, the veri cation of a multifactor structure highlights the necessity of evaluating multisystem dysregulation in MetS using factor analysis. A risk score for MetS derived from con rmatory factor analysis has been applied to measure the effect of the spectrum of MetS severity on T2DM and cardiometabolic disease [29][30][31].
Using four criteria to diagnose MetS in adolescents, we identi ed that 3.  [32,33]. Determining transformations in MetS status over time is thus a promising approach for estimating its in uence on cardiometabolic disorders.
Our data revealed that intraindividual changes between baseline and follow-up in weight-and BP-related variables, HDL-C, and FPG were signi cant in the never MetS group, thus demonstrating the variability of metabolic risk parameters in childhood development. After adjustment for all covariates, a high SBP increase (ΔSBP) was associated with a high risk of new-onset MetS (aOR, 1.07 for 1 mmHg increase, Table 5), implying that SBP elevation is critical for MetS occurrence in this population. Compared with adolescents having persistent MetS, decreases in the SBP, TG, and FPG levels were associated with an increased likelihood of MetS remission (aOR, 1.05, 1.02, and 1.06, respectively, for 1 unit decrease, Table 5). Clinical studies have revealed that patients who received a short-term intensive drug treatment to lower blood glucose, BP, and cholesterol levels had a long-term reduction in the risk of T2DM and cardiovascular disorders, even after treatment cessation (known as the cardiometabolic memory phenomenon) [34]. Our ndings underline the need for interventions among adolescents with MetS such as dietary improvement and exercise promotion with the aim to reducing SBP, TG, and FPG levels. Surveillance and monitoring of the incidence and persistence of ve MetS components are vital tasks in adolescent cardiometabolic health [2,14]. In our cohort, among the ve MetS components, low HDL-C had the highest incidence rate (9.6% per year), whereas abdominal obesity and low LDL-C had the greatest persistence rate (34.3% and 36.5% per year, respectively). These data indicate speci c risk factors that need enhanced monitoring. A combined assessment involving two longitudinal investigations demonstrated that incident MetS and persistent MetS during the transition from adolescence to young adulthood are associated with a 1.7-and 3.4-fold risk, respectively, of high carotid artery intima-media thickness and a 4.4-and 12.2-fold of T2DM in adulthood [32]. Because the likelihood of persistence of each MetS component was higher than that for new onset (aHR, 3.4-15.0, Table 7), screening for existing abnormal MetS components can be more bene cial than preventing new-onset abnormal components.
This study had several strengths. First, a large-scale representative community-based cohort was used to prospectively assess the stability and typological transformation of adolescent MetS and their relation to the change of metabolic risk factors. Second, our investigative framework and methodology can be adopted to other countries that wish to evaluate their own typological transformations in MetS status and attendant in uences on cardiometabolic disorders among adolescents and adults. Third, several criteria with speci c cutoff points for MetS diagnosis were simultaneously used to determine the agreement and stability for changes in MetS status over 2 years.
This study also had a few limitations. First, adolescents from three schools could not be followed due to the COVID-19 outbreak in Taiwan, and their data were excluded from this evaluation. However, the distributions of sex, age, urbanization level, and weight variables were comparable between the participants in the excluded and remaining schools. Second, our cohort included only Taiwanese adolescents, and thus, our ndings may not be generalizable to other adolescent populations.

Declarations
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate This study was approved by the Institutional Review Board of Kaohsiung Medical University. All participants and their parents/guardians provided signed informed assent and consent for the investigation.

Consent for publication
Not applicable.