A confirmatory factor analysis of the metabolic syndrome in adolescents: an examination of sex and racial/ethnic differences
- First Online:
- Cite this article as:
- Gurka, M.J., Ice, C.L., Sun, S.S. et al. Cardiovasc Diabetol (2012) 11: 128. doi:10.1186/1475-2840-11-128
The metabolic syndrome (MetS) is a cluster of clinical indices that signals increased risk for cardiovascular disease and Type 2 diabetes. The diagnosis of MetS is typically based on cut-off points for various components, e.g. waist circumference and blood pressure. Because current MetS criteria result in racial/ethnic discrepancies, our goal was to use confirmatory factor analysis to delineate differential contributions to MetS by sub-group.
Research Design and Methods
Using 1999–2010 data from the National Health and Nutrition Examination Survey (NHANES), we performed a confirmatory factor analysis of a single MetS factor that allowed differential loadings across sex and race/ethnicity, resulting in a continuous MetS risk score that is sex and race/ethnicity-specific.
Loadings to the MetS score differed by racial/ethnic and gender subgroup with respect to triglycerides and HDL-cholesterol. ROC-curve analysis revealed high area-under-the-curve concordance with MetS by traditional criteria (0.96), and with elevations in MetS-associated risk markers, including high-sensitivity C-reactive protein (0.71), uric acid (0.75) and fasting insulin (0.82). Using a cut off for this score derived from ROC-curve analysis, the MetS risk score exhibited increased sensitivity for predicting elevations in ≥2 of these risk markers as compared with traditional pediatric MetS criteria.
The equations from this sex- and race/ethnicity-specific analysis provide a clinically-accessible and interpretable continuous measure of MetS that can be used to identify children at higher risk for developing adult diseases related to MetS, who could then be targeted for intervention. These equations also provide a powerful new outcome for use in childhood obesity and MetS research.
KeywordsMetabolic syndromeFactor analysis, StatisticalInsulin resistancePediatricsAdolescentsEpidemiologyClinical studiesObesityRisk factors
Akaike’s information criteria
Area under the curve
Adult treatment panel III
Body mass index
Confirmatory factor analysis
Diastolic blood pressure
Goodness of fit index
High-density lipoprotein cholesterol
High-sensitivity C-reactive protein
Bentler-bonett normed fit index
National health and nutrition examination survey
Root mean square error of approximation
Systolic blood pressure
Standardized root mean square residual
Type 2 diabetes mellitus
The metabolic syndrome (MetS) is a cluster of interrelated individual factors that increase risk for future Type 2 diabetes mellitus (T2DM) and cardiovascular disease (CVD)[1, 2]. These individual components of MetS include elevations in adiposity, triglycerides, blood pressure (BP) and fasting glucose, and low levels of high-density lipoprotein (HDL) particles (a surrogate for which is HDL cholesterol). While the pathophysiologic processes that drive abnormalities in these individual components are not fully understood, these underlying processes appear to be related to systemic insulin resistance. In an attempt to understand the existence of MetS and the contributions of clinical measures of its components more fully, numerous researchers have used factor analysis, a model that explains the correlation among a set of variables in terms of a smaller set of unobserved “factors”[4–6]. A previous review highlighted the motivations and the pitfalls of the use of various forms of factor analysis for this purpose. Namely, the majority of studies have been exploratory, have not fully evaluated the appropriateness of a linear factor model in exploring MetS, and have not taken great care in the factor analysis itself, for example including in the model(s) highly correlated variables such as systolic blood pressure (SBP) and diastolic blood pressure simultaneously.
For most sets of MetS criteria—including the Adult Treatment Panel III (ATP-III)—MetS is classified based on cut-off values for each of the individual components[3, 7–9]. A given person is classified as having MetS in the presence of three or more abnormalities in these components. When applied to populations, the ATP-III MetS criteria for adults—or an adolescent adaptation of these criteria—predict elevations in measures of obesity-related inflammation (such as serum levels of high-sensitivity C-reactive protein, hsCRP), oxidative stress (such as uric acid[11, 12]) and insulin resistance (often assessed in epidemiological studies as fasting insulin[13–15]). In a prospective manner, a classification of MetS can be tracked over time and predicts future T2DM in adolescents and future CVD[2, 17] and T2DM in adults. As such, the presence of MetS in this era of pediatric obesity has been proposed as a trigger for increased intervention[9, 20].
Nevertheless, controversy exists over which of a variety of sets of MetS criteria to use among adolescents (Additional file1: Table S1)[7, 9, 21–24], and evidence suggests that current criteria exhibit racial/ethnic and gender differences in the ability of MetS criteria to identify increased risk. Non-Hispanic-black individuals have lower rates of MetS[26, 27] despite having higher rates T2DM and death from CVD. Currently-used criteria for MetS exhibit a lower sensitivity to detect insulin resistance, underlying inflammation and oxidative stress (as assessed by levels of uric acid) among non-Hispanic-black adolescents compared to non-Hispanic whites or Hispanics. A major reason for this is that in non-Hispanic-black adolescents triglyceride levels are lower and HDL levels are higher (i.e., more favorable) at baseline, and although these levels continue to worsen with progressive insulin resistance, they are overall less likely to exceed population cut-offs required for MetS diagnosis[27, 31, 32].
Because of these drawbacks of current MetS criteria, many have advocated for using a continuous scale for MetS diagnosis[33, 34] or for criteria that are race/ethnicity-specific[27, 31, 32] Eisenmann provides an overview of multiple proposed continuous pediatric MetS scores, the majority of which use a sum of z-scores of individual MetS components. While these z-scores can be calculated to account for age, sex and race/ethnicity, the method of standardizing components (e.g., waist circumference (WC), blood pressure) and summing the resulting z-scores to formulate a MetS score does not account for the strong correlations that occur across the components themselves and does not account for possible differential influences of individual components on the overall score. To better account for these drawbacks, Li and Ford performed a confirmatory factor analysis on adolescents to assess the validity of one factor in explaining the covariance across the traditional MetS components; however, they did not detect hypothesized differences by sex and race/ethnicity, potentially due to being inadequately powered. Others have performed such an analysis in ethnically homogeneous populations[6, 35].
Our study had two primary goals. Our first goal was to perform a multi-group confirmatory factor analysis that assumed a one-factor model of the traditional MetS components while allowing for differences across sex and race/ethnicity. Thus, we are testing our hypothesis that the factor structure is the same across all sex and race/ethnicity groups, but that the manner in which these MetS components correlate with this factor differs across the groups in a meaningful way. Second, if we found differences in these correlations utilizing this one-factor model, we would utilize the factor score from this model among adolescents as a continuous MetS “risk score” that was sex- and race/ethnicity specific. The ability of this risk score to detect elevations in surrogate factors related to processes underlying MetS, including hsCRP, uric acid and fasting insulin, could then be assessed on a race/ethnicity-specific basis. The data for this study were from the past twelve complete years released from NHANES (1999–2010), and thus would provide substantially more power than prior attempts to detect sex and racial/ethnic differences if they exist and would provide a subsequent reliable set of equations from which to base a continuous score. Our hypothesis was that such a race/ethnicity and sex-specific score, if determined necessary by the confirmatory factor analysis, would perform better than traditional MetS criteria at predicting surrogate factors related to MetS.
Methods and procedures
Data were obtained from NHANES (1999–2010), a complex, multistage probability sample of the US population. These annual cross-sectional surveys are conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control (CDC), with randomly-selected subjects undergoing anthropometric and blood pressure measurements, answering questionnaires and undergoing phlebotomy. The NCHS ethics review board reviewed and approved the survey and participants gave informed consent prior to participation. Body mass index (BMI), SBP, and laboratory measures of triglycerides, HDL-C, and fasting glucose were obtained using standardized protocols and calibrated equipment. For SBP, the mean of up to four readings taken on each individual was used. All blood samples used for analyses were obtained following a fast ≥8 hours prior to the blood draw.
Data from non-Hispanic-white, non-Hispanic-black, or Hispanic (Mexican-American/other Hispanic) adolescents 12–19 years old were analyzed. Children <12 years old were excluded since fasting values for triglycerides and glucose were only obtained in participants ≥12 years old. Subjects were excluded if they had known diabetes or unknown diabetes (fasting plasma glucose >125 mg/dL), as each of these limit insulin release. Pregnant women were also excluded, as well as individuals taking antihyperlipidemic or anti-diabetic medications as these are all likely to alter lipid and insulin levels.
We combined all data sets from the 6 two-year cycles (1999–2010) for statistical analyses to increase our total sample size. Prevalence rates of MetS were calculated by sex and race/ethnicity, according to Ford’s pediatric adaptation of the ATP III adult criteria, and mean levels (95% CI’s) of the MetS components of interest (BMI z-score, SBP, HDL, triglycerides, and glucose) as well as the surrogate outcomes (hsCRP, uric acid, fasting insulin) were calculated by sex and race/ethnicity.
Model 1: constrained the factor loadings to be equal across the six combinations of sex and race/ethnicity;
Model 2: allowed the factor loadings to vary across the six groups.
Chi-square tests of the equality of the factor loadings across the six groups in Model 2 were performed. The two models were compared using various fit statistics, both overall and by group. Chi-square and Akaike’s Information Criteria (AIC) were used for model comparisons; smaller chi-square and AIC values indicated a better fit. A chi-square difference test was calculated; a significant difference between two nested models implies that the model with more paths explains the data better. Other goodness of fit indices included the Root Mean Square Error of Approximation (RMSEA; >0.06 poor fit), the Standardized Root Mean Square Residual (SRMR; >0.08 poor fit), the Goodness of Fit Index (GFI; <0.90 poor fit), and the Bentler-Bonett Normed Fit Index (NFI; <0.90 poor fit).
The standardized factor coefficients from the better-fitting model were used to calculate the MetS factor score on each individual. This score can be interpreted as a Z-score (mean 0, SD=1), with higher scores representing an increased risk of MetS. Receiver operating characteristic (ROC) analysis was used to assess the ability of this new score to discriminate against the traditional MetS criteria as well as elevated levels of the three identified CVD/T2DM surrogates. Specifically, we were interested in the ability to predict elevated fasting insulin (>16 IU/mL, the 95th percentile among normal-weight adolescents), elevated hsCRP (>4.5 mg/L), and elevated uric acid (approximately the 95th percentile among lean individuals: 7.0 mg/dL for males, 5.5 mg/dL for females). In this analysis, individuals with CRP >10.0 mg/L were excluded. In addition, the ability to predict 1 and ≥2 elevations amongst these surrogates was of interest, as this indicates the more at-risk adolescents. Overall predictive performance was measured by the area under the curve (AUC) of the ROC curve, with 0.5 and 1.0 indicating no and perfect predictive ability, respectively. We considered AUC values >0.70 to be reasonably accurate and AUC >0.90 to be very accurate. Sensitivities and specificities to predict ≥2 elevations among the three surrogates were compared between the traditional MetS classification and a definition of MetS using a cutoff identified by the ROC analysis; these statistics were done on a sex and race/ethnicity-specific basis.
Statistical significance was defined as a p-value<0.05. Statistical analysis was performed using SAS (version 9.3, Cary, NC). Descriptive statistics as well as sensitivity estimates used SAS survey procedures (SURVEYMEANS, SURVEYFREQ), which accounts for the survey design when estimating standard errors to obtain population-based estimates. The CFA itself did not account for the survey design due to the inability of standard software to perform multigroup CFA’s within subpopulations while accounting for the survey design.
NHANES 1999–2010 Characteristics: Children 12–19 Years Old with Data on all Metabolic Syndrome Components (n = 4,174)
Mean (95% Confidence Interval)
% with Mets (95% CI)*
By Gender and Race/Ethnicity
Confirmatory Factor Analysis Results*
Model Fit Indices
Akaike’s Information Criteria (AIC)
Root Mean Square Error of Approximation (RMSEA)
Standardized Root Mean Square Residual (SRMR)
Goodness of Fit Index (GFI)
Bentler-Bonett Normed Fit Index (NFI)
Equations for New Sex and Race/Ethnic-Specific Childhood Metabolic Syndrome Risk Z-Score
* BMI Z-score
* BMI Z-score
* BMI Z-score
* BMI Z-score
* BMI Z-score
* BMI Z-score
The present analysis reveals racial/ethnic and sex differences among adolescents in loading weights to MetS that result in improvements in the ability of MetS to predict elevations in MetS-associated risk markers. While the value of a diagnosis of MetS has been questioned as being more valuable than the sum of its parts, the concept of MetS remains a frequently-used research tool that has validity in being able to predict future occurrence of T2DM in children and CVD in adults. Most currently-employed criteria for diagnosing MetS in children and adolescents utilize somewhat arbitrarily-determined cut-off values for individual components (Table1) that appear to have racial/ethnic biases[25, 31] while other sets of criteria have used a sum of z-scores, which does not account for the strong correlations across the components and does not account for possible differential influences of individual components on the overall score. Our current approach offers multiple improvements on these prior approaches by using a confirmatory factor analysis, allowing for sex and racial/ethnic differences in the weighting of individual components to the risk score. Restated, this allows for the possibility that MetS manifests itself differently between sex and racial/ethnic groups in a way that may affect our ability to identify MetS-associated risks. The sex- and race/ethnicity-specific differences result in unique equations to calculate the risk score based on sex and racial/ethnic group. Once this score has been more fully validated it could be placed on an internet web site or smart phone application to assist in clinical use, alerting clinicians and patients regarding an adolescent’s risk score—along with any risk implications specific to that individual’s sex and racial/ethnic group.
In evaluating race/ethnicity-specific differences in this score, we noted in particular variations in contributions of lipid measures by sex and race/ethnicity to the single MetS factor. This is perhaps not surprising, given that non-Hispanic-black adolescents have lower levels of triglycerides and higher levels of HDL at baseline and although they exhibit worsening dyslipidemia with insulin resistance, they are less likely to exhibit gross abnormalities in lipids when using population-based cut-off values[25, 27]. This results in a lower prevalence of MetS among non-Hispanic-black males using traditional criteria and a poorer sensitivity for MetS to detect elevations in surrogates of MetS-related processes[14, 29, 30]. In the current analysis (Table3), we noted that as compared to the factor loadings for non-Hispanic-white males, non-Hispanic-blacks had a higher loading of HDL (0.51 non-Hispanic whites vs. 0.66 non-Hispanic blacks) but a lower loading of triglycerides (0.62 vs. 0.50). This would suggest that lower levels of HDL may be even more indicative of worsening MetS severity in non-Hispanic-black males compared to non-Hispanic whites while higher levels of triglycerides may not be as strong an indicator in non-Hispanic-black males. Hispanic males exhibited high loading factors of both HDL and triglycerides, potentially suggesting that worsening levels of both components are important indicators of increasing degree of MetS severity. Non-Hispanic-white males had high loadings for SBP (0.50) compared to non-Hispanic blacks (0.33) and Hispanics (0.30). This is interesting given that while elevated SBP is more common in non-Hispanic blacks and less common in Hispanics compared to non-Hispanic whites, SBP appears to have a greater relative importance to MetS in non-Hispanic whites compared to the other groups.
Non-Hispanic-white females exhibited lower factor loadings of HDL and triglycerides relative to either of the other racial/ethnic groups, while Hispanic females exhibited the highest loadings for HDL and triglycerides. These data again suggest that changes in these lipid values are more likely to indicate worsening of underlying MetS among Hispanics. Overall, the poor model fit indices for non-Hispanic-white females in particular, may indicate that a one-factor model of MetS may not be appropriate for this group.
It was notable that fasting glucose had a low factor loading for all sex and racial/ethnic groups. While others have included insulin in some manner as a measure in factor analyses[5, 33], we chose to use glucose because the lack of standardized assays for insulin would impede use of the risk score for clinical purposes. It has been noted previously that glucose is maintained in a relatively narrow range among obese children, and this range was narrowed further in our analysis by excluding diabetic individuals (glucose>125 mg/dL). We ultimately elected to retain glucose in the score because of the near universality of its inclusion in prior MetS criteria[3, 7–9, 21–24] and its common use in screening for undiagnosed diabetes.
Most telling of the accuracy of this new score would be in its ability to predict future disease risk. The optimal testing for such risk would require long-term data including childhood factors and adult disease outcomes. Lacking these, such a score could also be used to assess the quantity of markers associated with processes associated with MetS-related risk, including adiponectin (which appears to be in the causative pathway of insulin resistance[20, 48]) or markers of atherogenic dyslipidemia, including ApoB, small, dense LDL particles. In the present analysis, we instead assessed the accuracy of the new score for its ability to identify individuals with elevations in clinical measures that are part of processes related to MetS. These measures were serum levels of fasting insulin (as an assessment of insulin resistance), hsCRP (as an assessment of underlying inflammation) and uric acid (as an assessment of oxidative stress[11, 12]). Comparing the score’s performance in predicting these elevations to the more traditional MetS diagnosis required using a cutoff value for the MetS risk score itself—even though one of the benefits of such a score is its lack of binary nature. We chose a cutoff of a z-score of 0.75 based on the ROC curve for the score to identify individuals with traditional MetS. Using this cutoff, there was a higher prevalence of adolescents with MetS in each sex/racial/ethnic group and in particular among non-Hispanic-black males and females (Figure2A). While traditional MetS criteria (ATP-III based) performed poorly in predicting elevations in these measures (sensitivity 21-65%), the new risk score (cut-off of 0.75) performed significantly better (sensitivity 43-81%) without clinically meaningful differences in specificity (90-98% for traditional score vs. 78-92%). This cut-off, if used, could help to identify a larger number of at-risk children and adolescents that currently-used MetS criteria—particularly among non-Hispanic-black individuals.
It is important to note that our aim was not to question the existence of the metabolic syndrome in general, nor study in an exploratory fashion the precise number of factors. We operated under the assumption that one “MetS” factor exists in the pediatric population, and under the assumption, we assessed whether the components contribute to that factor differentially by sex and race/ethnicity. Along those lines, we focused only on traditional MetS components that are common to almost all existing MetS criteria based on cutoffs of these components in order to ensure a clinically accessible risk score that results from the analysis. The limited number of components that comprise traditional MetS criteria did not allow for examination of more than one factor. Thus, our examination here was to examine the one-factor model of MetS in adolescents that would thus allow for a continuous representation of traditionally-defined MetS while simultaneously allowing for sex and race/ethnic differences within this one-factor model. Our comparison of the predictive ability of this new score to the traditional MetS diagnosis required the use of a cutoff value. However, a clear benefit of this score is its potential to identify elevated risk in an individual who has a high score but would not be classified as having MetS using traditional criteria based on having elevations in components of MetS that did not exceed population-based or perhaps even arbitrary thresholds (see Additional file1: Table S1 for examples). Importantly, this score could also be used to assess degree of improvement during lifestyle modification treatment for weight loss.
This study had several weaknesses. We utilized NHANES data which, although powerful, are cross-sectional. Future use of this score will need to utilize longitudinal databases that include information regarding long-term disease outcomes. We used BMI z-score as an assessment of obesity. While this has been done for prior MetS criteria[8, 39], it is generally recognized that markers of visceral obesity (such as WC) are more strongly associated with MetS risk than BMI. Nevertheless, BMI is known to be highly associated with MetS risk[8, 42] and has been noted in prior factor analyses of MetS[1, 33]. In addition, our MetS risk score that was developed using BMI z-score had near-perfect ability (ROC AUC=0.96) to discriminate against an ATP-III-based classification that utilized WC percentile cut-offs, indicating use of BMI in this instance is sufficient in the creation of a continuous representation of the traditional MetS diagnosis.
In summary, using confirmatory factor analysis, we have demonstrated significant sex- and racial/ethnic differences in factor loading of MetS components that has resulted in a novel sex- and race/ethnicity-specific MetS risk score. This continuous score demonstrates strong predictive ability to detect MetS-associated processes while being less prone to racial/ethnic differences than traditional pediatric MetS criteria. Future research is needed to ascertain the ability of this score to identify individuals at risk for long-term CVD and T2DM, as well as its ability to monitor MetS in the setting of lifestyle modification for obesity treatment.
This work was supported by NIH grants 5K08HD060739-03 (MDD), U54GM104942 (MJG), and 1R21DK085363 (MDD and MJG). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.