Journal of Diabetes Research

Volume 2015 (2015), Article ID 539835, 10 pages

http://dx.doi.org/10.1155/2015/539835

## A New Approach to Define and Diagnose Cardiometabolic Disorder in Children

^{1}Center of Research in Childhood Health, Department of Sport Sciences and Clinical Biomechanics, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark^{2}Department of Sports Medicine, Norwegian School of Sport Sciences, Sognsveien 220, 0806 Oslo, Norway^{3}Exercise and Health Laboratory, CIPER, Fac Motricidade Humana, Universidade de Lisboa, Estrada Dacosth, Cruz-Quebrada, 1499 Lisbon, Portugal^{4}Department of Exercise and Sport Science, University of North Carolina, 025 Fetzer Gym, CB No. 8700, Chapel Hill, NC 27599-8700, USA^{5}School of Physical Education, University of Pernambuco, Campus Universitario HUOC-ESEF, Arnobio Marques 310, Santo Amaro, 50.100-130 Recife, PE, Brazil^{6}Epidemiology, Biostatistics and Prevention Institute, University of Zürich, Hirschengraben 84, 8001 Zürich, Switzerland^{7}The Centre of Inflammation and Metabolism and Trygfondens Center for Aktiv Sundhed, Department of Infectious Diseases and CMRC, Rigshospitalet, Faculty of Health Sciences, University of Copenhagen, Tagensvej 20, 2100 Copenhagen, Denmark^{8}MRC Epidemiology Unit, Institute of Metabolic Science, Addenbrooke’s Hospital, University of Cambridge, Hills Road, Cambridge CB2 0QQ, UK

Received 26 November 2014; Accepted 17 March 2015

Academic Editor: Francesco Chiarelli

Copyright © 2015 Lars Bo Andersen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

The aim of the study was to test the performance of a new definition of metabolic syndrome (MetS), which better describes metabolic dysfunction in children. *Methods*. 15,794 youths aged 6–18 years participated. Mean *z*-score for CVD risk factors was calculated. Sensitivity analyses were performed to evaluate which parameters best described the metabolic dysfunction by analysing the score against independent variables not included in the score. *Results*. More youth had clustering of CVD risk factors (>6.2%) compared to the number selected by existing MetS definitions (International Diabetes Federation (IDF) < 1%). Waist circumference and BMI were interchangeable, but using insulin resistance homeostasis model assessment (HOMA) instead of fasting glucose increased the score. The continuous MetS score was increased when cardiorespiratory fitness (CRF) and leptin were included. A mean *z*-score of 0.40–0.85 indicated borderline and above 0.85 indicated clustering of risk factors. A noninvasive risk score based on adiposity and CRF showed sensitivity and specificity of 0.85 and an area under the curve of 0.92 against IDF definition of MetS. *Conclusions*. Diagnosis for MetS in youth can be improved by using continuous variables for risk factors and by including CRF and leptin.

#### 1. Introduction

Metabolic syndrome (MetS) was first described by Reaven in the mid 1980s [1]. MetS is a conceptual framework, which links several apparently unrelated biological events into a single pathophysiological assemble; that is, several cardiovascular disease (CVD) risk factors seemed to be increased simultaneously in some individuals. Despite different definitions most agree on the individual components constituting the MetS, which include dyslipidemia (triglycerides and cholesterol), hypertension, glucose intolerance, and adiposity [2–5]. Originally, the different CVD risk factors were mainly treated with drugs and it therefore seemed logical to use the cutoff point for each risk factor in the definition of the criteria for MetS. Criteria have been suggested by several organizations and researchers and have changed over time even if they all build on the same concept. Insulin resistance triggered a common mechanism affecting blood pressure (BP), high density lipoprotein cholesterol (HDL), triglycerides (TG), and glucose tolerance, and central obesity may cause the insulin resistance and therefore be a central part of MetS. This concept subsequently evolved to encompass a number of multiple definitions [2] established by the WHO [3, 5], the National Cholesterol Education Program Adult Treatment Panel III (ATP III) [4], and the International Diabetes Federation (IDF) [5]. Modified criteria based on the same concept as in adults have been suggested in children [6–9].

Existing MetS definitions have shortcomings, especially for children, and because studies use different definitions comparison between studies is difficult. All definitions are based on dichotomisation of the CVD risk factors and to be clinically diagnosed with the MetS the thresholds for at least three risk factors including obesity must be attained. Limitations include (1) reduction of available information of risk by dichotomizing variables; (2) different risk factors that are given different weight (i.e., prevalence of the risk factors differs, which means that few are selected based on the rare risk factors); (3) thresholds for the individual risk factors that are arbitrarily chosen in children, where no hard endpoints exist; (4) selection of risk factors that exclude potentially important variables; for example, the use of fasting glucose in children rather than fasting insulin or HOMA score as measure of impaired glucose regulation may conceal important information; many children with severe insulin resistance are still able to regulate their fasting blood glucose well [10]; (5) different definitions that use different blood variables and fatness variables. This problem makes it difficult to compare prevalence between populations. These shortcomings result in substantial misclassification, and comparison between studies using different criteria is difficult. When children are analysed, there is also a major difference between the number of children diagnosed with MetS compared to the number where a clustering of risk factors occurs [11, 12].

The current definitions in use preclude accurate estimates of the MetS prevalence between populations and the dichotomisation of variables may attenuate or preclude real associations between lifestyle behaviours and metabolic risk.

The aim of this study was to develop a novel method for identifying young people with increased metabolic risk and solve weaknesses of former definitions. Furthermore, we tested a simple noninvasive screening tool to identify children where further investigation is indicated. The steps in the analysis were (a) to calculate the number of children where risk factors were not independently distributed; (b) to construct age adjusted* z*-scores for each risk factor based on common means and SDs for the whole database for genders separately; (c) to define a cutoff point in mean of summed* z*-scores, which selected the same number of children as calculated above, where risk factors were not independently distributed; (d) to construct a program where absolute values of the risk factors can be entered, and the program calculates the mean* z*-score of the included variables (age and sex adjusted) to evaluate if the child has a disorder; (e) to do sensitivity analyses to evaluate if other risk factors may improve diagnostic criteria; and (f) if it matters to substitute a measure of a certain trait with another such as including* z-*score of body mass index (BMI) instead of waist circumference. The program makes it easy for the general practitioner to use the diagnostic tool. Last, we tested a noninvasive measure of metabolic risk based on physical fitness and fatness.

This approach will make it possible to use the available information and get a better measure of the risk of the child. Information is not reduced when we use continuous variables and the composite* z-*score can be used as a continuous or dichotomous variable. It is hypothesized that the* z-*score of the different variables covering the same trait will only differ slightly; that is, the mean* z-*score only changes marginally if BMI is entered instead of waist, which makes the program flexible and enables comparison between studies using different measures. Further, this approach makes it possible to include other risk factors in the calculation of mean* z*-score or exclude them if data is not available. The constructed program is available on the Internet or as an APP for smartphone, and the general practitioner can enter the available information, which should include as many of the risk factors as possible.

#### 2. Methods

##### 2.1. Participants

We pooled cross-sectional data from 23 population based cohorts in children and adolescents aged 6–18 years. The 23 cohorts were merged into 18 cohorts before analysis. Ten cohorts () of 9- and 15-year-old boys and girls from the European Youth Heart Study (EYHS) (1997 through 2007) were included. Nine-year-olds and 15-year-olds, respectively, assessed at two different time points from the same country were pooled before analysis. Additionally, four cohorts of 9- and 15-year-olds from EYHS Norway () were included but analysed as four separate cohorts, as two cohorts lacked important blood variables. The Copenhagen School Child Intervention Study (CoSCIS) () included three measurements collected in 2001, 2004, and 2008. Data from the US NHANES () collected in 2003 and 2005 were merged and recoded into three age groups (<10 years, 10–14 years, and >14 years). Finally, children from Switzerland (KISS) were sampled from 1st (mean age 8 years) and 5th grade school children (mean age 12 years) and were measured three times over two years (). Successive measurements of a child were regarded as multiple individuals. In total 7902 girls and 7892 boys were included in the analysis. Data collection procedures and analytical methods have previously been described in detail (EYHS [13], CoSCIS [14, 15], KISS [16], NHANES [17], Norwegian EYHS 2000 [18], and Norwegian EYHS 2005 [19]). Ethical approval has been obtained for all included cohort studies in their respective countries.

##### 2.2. Analytical Process and Statistics

Statistical analysis was performed in IBM SPSS Statistics 22 and Stata 12 (meta-analysis).

Common estimates were calculated using random effect meta-analysis [13].

Analysis was progressed by the following steps.(1)First step is calculation of the number of children where risk factors were not independently distributed; that is, risk factors clustered. MetS risk factors follow a binomial distribution if they are independently distributed. If the observed number of children with a given number of risk factors exceeds the expected, risk factors exhibit clustering. For this analysis we defined the upper quartile of waist circumference, systolic blood pressure (BP), triglyceride (TG), and HOMA together with the lower quartile of high density lipoprotein cholesterol (HDL) to be at risk for each cohort separately. Proportion of expected subjects with a specific number of risk factors according to the binomial distribution was , where is the possible number of risk factors (5), is the proportion having the risk factor (25%), and is the number of risk factors the probability is calculated for. The expected proportions of children having 0 to 5 risk factors were 0.178, 0.356, 0.297, 0.132, 0.057, 0.0044, and 0.0002, respectively. We then divided the observed number of children with a specific number of risk factors with the expected and calculated 95% confidence intervals as exp(Ln)), where SE(Ln(OR) = SQRT(1/) + (1/)). is the total number of children in the cohort and is the number of children with the specific number of risk factors.(2)The association between age and each risk factor was calculated using linear regression in order to compute age adjusted risk factor levels for each child. All risk factors were adjusted to the age of 12 years for each gender to enable comparison of levels between cohorts of different age by the following formula: risk . BMI, waist circumference, sum of 4 skinfolds, TG, glucose, HOMA, C-reactive protein (CRP), APOA1, APOB, and leptin were skewed and therefore log-transformed before* z*-scores were constructed.* z*-scores were constructed using mean and SD of the age adjusted risk factors for the total sample. The mean of the 5 risk factors included in the IDF definition of MetS [6] was calculated using HOMA rather than fasting glucose as an indicator of glucose metabolism.(3)Third step is definition of a cut-point in mean of summed* z*-scores, which selected the number of children with clustered CVD risk as calculated in step (1).(4)Fourth step is determination of the significance of substituting a measure of a certain trait with another, for example, including* z*-score of body mass index (BMI) or skinfold instead of waist circumference or HOMA instead of fasting glucose.(5)Fifth step is sensitivity analyses to evaluate if addition of other risk factors improved diagnostic criteria. For this purpose we examined the associations between CRF (independent variable) by logistic regression and calculated odds ratio (OR) for the three lower quartiles of CRF against the MetS variables (dependent variable). Additional risk factors (APOA1, APOB, leptin, and CRP) beyond those included in the IDF criteria were tested to examine if their inclusion strengthened the magnitude of association between CRF and the mean* z*-score of the traditional MetS risk factors. In all analyses a threshold in the mean* z*-score, identifying the same proportion of children, was identified. Inclusion of CRF in the mean* z*-score was tested by removing waist circumference from the MetS and using it as independent variable to see if the association became stronger when fitness was included in the MetS outcome.(6)Prognostic value of a MetS outcome including only noninvasive variables (waist/height and inverse CRF) was also tested. We used waist circumference/height as a proxy for fatness, because this variable is independent of age [20]. The mean* z*-score of ((waist/height) + (1/CRF))/2 was examined in a receiver operating characteristic (ROC) analysis against the IDF definition of the MetS and the clustered MetS score variable constructed from* z-*scores of log TG, log waist, systolic BP, log HOMA, inverse HDL, and inverse fitness.

All analytical steps were built into an Internet-based program which calculates MetS score in order to facilitate the use in general practice (http://www.video4coach.com/zscore/index.html).

The proportion of children categorised as having the metabolic syndrome (IDF definition) was calculated according to Zimmet et al. [6], where age- and sex-specific 90th percentiles for waist circumference were defined according to Fernández et al. [21].

#### 3. Results

Age, anthropometrics, and CVD risk factors are described for each cohort in web-appendix (see web-only Table A in Supplementary Material available online at http://dx.doi.org/10.1155/2015/539835). Characteristics of CVD risk factors are also described after adjustment for age to enable comparison of levels between cohorts of different age (Table 1). The association between age and each risk factor is shown in Table 2.