Body composition indicators in the metabolic syndrome risk prediction in 6-10-year-old children

ABSTRACT The aim of this study was to develop percentiles of body composition indicators and determine cutoff points to predict metabolic syndrome (MS) risk in 6-10-year-old children. This is a cross-sectional, population-based epidemiological study with the participation of 1480 schoolchildren aged 6-10-year. Anthropometric assessment (body mass, height, and skinfolds) and blood pressure measurement were performed in schools. The body mass index (BMI), as well as the body fat percentage (%BF), lean body mass (LBM), fat body mass (FBM), were calculated according to standardized formulas for children. Blood collection to assess the lipid and glycemic profile was also performed at school, on pre-established days and times. The MS diagnosis was determined based on changes in triglycerides, HDL-c, blood glucose, waist circumference, and blood pressure. The LMS method was used to develop the percentiles, the area under the ROC curve (AUC) to identify the accuracy of the indicators, and the sensitivity and specificity to determine the cutoff points. FBM and %BF had significantly higher values in girls, who also had lower values for %LM compared to boys (p<0.05). The indicators of body composition, BMI, FBM, and %BF were accurate in predicting the MS risk for both sex at all ages. The main indicators of body composition to predict the MS risk, in both sex, were BMI, FBM, and %BF. These findings suggest that simple anthropometric measurements, which can be performed in clinical practice, have the potential to direct non-pharmacological actions.


INTRODUCTION
Chronic non-communicable diseases are among the main passive harms of prevention that burden the health system and are responsible for a high burden of morbidity and mortality 1 .Among these diseases, we highlight metabolic syndrome (MS), which can be characterized by the simultaneity of risk factors such as central obesity, arterial hypertension, low levels of HDL-c, hypertriglyceridemia, and hyperglycemia 2 .The presence of MS increases the risk for diabetes by five times and doubles the chances of developing cardiovascular diseases. 3Furthermore, longitudinal data demonstrate that children diagnosed with MS are more likely to have MS in adulthood. 4Therefore, two points seem to be crucial to reduce the prevalence of MS, I) to establish environments that control risk factors from childhood; and II) to validate simple, non-invasive, and low-cost methods for early detection of MS risk 5,6 .
The worldwide MS prevalence among children and adolescents varies between 1.2% and 26%, reaching up to 60% among overweight/obese young people 7 .In Brazil, a systematic review of the literature showed a prevalence of MS in the pediatric population ranging from 0% to 40% 8 .The marked differences in the prevalence of MS between studies may be associated, especially, with differences in the sample profile and the criteria used for the MS diagnosis.According to Haffner 9 , the difficulty in defining accurate cutoff points that are adjusted to the growth curves also represents a challenge for epidemiologists in the MS diagnosis in children and adolescents.Nevertheless, assessments for the MS diagnosis involve laboratory collections to obtain indicators of lipid, glycemic, and insulinemic profiles.These techniques are invasive, expensive, and difficult to access, highlighting the need to identify practical and low-cost methods to enable the assessment of this syndrome at a population level.
Anthropometric measurements have received special attention due to their consistent association with this outcome and their possibility of use in many sectors of public health, such as schools, family health units, and hospitals.Considering the consensus in the literature that obese and overweight young people are more likely to have MS 7,8 ; anthropometric indicators of body composition could be used as important tools for MS screening in pediatric population.Thus, this study aimed: I) to verify the accuracy of anthropometric indicators of body composition to predict MS risk in children, and II) to develop percentiles and determine cutoff points for body composition indicators to predict the MS risk in children.

Study design and population
This is a cross-sectional, population-based epidemiological study with a probabilistic sample of 6-10-year-old children, enrolled in public and private schools in urban and rural areas of the city of Uberaba/MG, Brazil.
For the sample calculation, we considered the number of children enrolled in elementary school (1 st to 9 th grade), the MS prevalence of 50% (prevalence unknown in the municipality), tolerable error of 3.5%, and confidence level of 95%.The minimum sample number was 768 children and after adding 10% to compensate for losses and refusals and 20% to minimize confounding factors, the sample was estimated at 1014 children.To select the sample, schools were stratified according to educational segment (municipal, state, and private).We randomly selected 15 schools in the city.The number of children in each stratum was determined in proportion to the number of enrollments, according to data provided by the State Department of Education.
After approval by the Ethics Committee for Research with Human Beings (Protocol CEP/UFTM: 1710), the directors of the selected schools were contacted to obtain authorization and schedule for data collection.The selected students who met the study's inclusion criteria and were interested in participating in the research received the Informed Consent Form for their parents to know and sign.Data collection was carried out at the school from August 2011 to August 2012.

Evaluations
Body mass (BM) was obtained using a digital electronic scale (Plenna, Ice, São Paulo, Brazil).Height was measured using a portable anthropometer (Welmmy, Santa Bárbara d'Oeste/SP, Brazil).Body mass index (BMI) was calculated with measurements of body mass and height.The classification of overweight and obese children was performed based on the criteria proposed by Cole et al. 10 Waist circumference (WC) was obtained at the end of normal expiration using a flexible and inelastic measuring tape with a length of 2m (TBW, São Paulo).All measurements were taken in triplicate and the mean value of the three measurements was considered.Measurements were taken at the midpoint between the iliac crest and the last floating rib, as recommended by the World Health Organization.
Triceps (Tri-SK) and subscapular (Sub-SK) skinfolds were obtained using an adipometer (Lange Skinfold Caliper, Cambridge, MA, USA) which exerts a constant pressure of 10g/mm2, on the right side of the body and with three non-consecutive repetitions for each measurement.The final measure was obtained by the average of the three values.To calculate the body fat percentage (%BF) we used the equations proposed by Slaughter et al. 11 based on ethnicity and Tri-SK and Sub-SK thickness summation.The lean body mass (LBM) was obtained by subtracting the fat body mass (FBM) from the body mass total.
Blood pressure was measured with a mercury column sphygmomanometer (Unitec, São Paulo, Brazil), following standards protocols for children 12 .Two other measurement moments were set aside for children with high blood pressure levels.Children who had systolic or diastolic blood pressure above the 90 th percentile were diagnosed with high blood pressure National High Blood Pressure Education Program Working Group on High Blood Pressure in Children and Adolescents 13 on three different occasions.
Blood samples (8mL) were collected in BD Vacutainer® vacuum tubes and centrifuged at 3400rpm for 8-minute.Serum samples were analyzed for the determination of HDL-c and triglycerides and plasma for blood glucose.We used the semi-automated Bio 200F analyzer (Bioplus, São Paulo, Brazil).
The MS diagnosis was determined by the presence of, at least, three of the following alterations triglycerides ≥ 100mg/dL; HDL-c < 50mg/dL; blood glucose ≥110mg/dL; waist circumference ≥ 75 th percentile of the sample for Quadros et al.Body composition indicators and metabolic syndrome age and sex, blood pressure (diastolic or systolic) > 90 th percentile adjusted for age, height and sex 13 .

Statistical analysis
The SPSS statistical program was used to perform the Kolmogorov-Smirnov, Mann-Whitney, and Chi-square tests.Outliers were identified and removed using the quartile interval method.The ROC curve assessed the ability of anthropometric indicators of body composition to predict MS.The sensitivity (SE) and specificity (SP) values of the anthropometric indicators of body composition were calculated for each cutoff point present in the sample.The statistical significance of each analysis was verified by the area under the ROC curve (AUC) and by the lower limit of the 95% confidence interval > 0.5.For comparability with other studies, the 5th, 10th, 25th, 50th, 75th, 90th, and 95th percentiles were chosen as reference values.The percentiles, Box Cox (L), adjusted median (M), and coefficient of variation (S) values were performed using the LMS method proposed by Cole, using the LMS Chartmaker program 14 .

RESULTS
The participants of this investigation were 1.480 children of both sex (52.2% girls), with a mean age of 8.55 (SD=1.53),students from the urban and rural areas in the municipality of Uberaba/MG.The BM, height, and BMI did not differ significantly between boys and girls (p>0.05).The FBM and %BF was significantly higher in girls, who also had lower values for the LBM compared to boys (p<0.05)(Table 1).The MS prevalence was 12.6% and 8.5% for girls and boys, respectively, with a statistically significant difference between sex (p<0.05).
The percentiles (5th, 10th, 25th, 50th, 75th, 90th, and 95th) of body composition indicators (BMI, BFM, LBM, and %BF), Box Cox (L), adjusted median (M), and coefficient of variation (S) values for age and sex for all variables analyzed were presented in Tables 2 and 3. BMI, FBM, and %BF showed statistically significant accuracy for predicting the MS risk for both sex at all ages evaluated.On the other hand, the LBM showed significant accuracy values only for girls aged 7-10-year and boys aged 6, 9, and 10 years (Table 4).

DISCUSSION
This study investigated different body composition indicators as predictors of MS risk in Brazilian children, including the proposition of percentiles and cutoff points.Among the investigated body composition indicators, those that estimate the amount of body fat were the most accurate for predicting MS in 6-10-year-old children.We believe that the percentiles and cutoff points developed in this study may be useful in screening for MS risk at an early age, aiming at the development of prevention strategies by health-promoting agents.
The prevalence of MS observed in this study, compared with Brazilian 8 and international studies 7 , suggests that actions to deal with this health problem are necessary.BMI cutoff points have been used to determine overweight and obesity in population studies involving children 10,15 .On the other hand, BMI percentiles were also used as one of the diagnostic MS factors, for example, BMI 97th percentile for age and sex 16,17 ; > 95th percentile 18,19 ; ≥ 85th percentile of the sample 20 .
The suggested BMI cutoff point to predict MS risk among girls in this study was 18.5 kg/m 2 (S=80.33%,E=76.33%).BMI had the highest AUC (0.867) among girls.For boys, BMI was the fourth most accurate predictor (AUC=0.872),behind FBM (AUC=0.880).The suggested BMI cutoff point to predict MS among boys was 18.6 kg/m 2 (S=80%, E=80%).A previous study with Spanish children 21 suggested the 65th percentile of BMI, regardless of sex and age, in predicting the set of cardiovascular risk factors that compose MS (AUC=0.868,S=80%, and E=75%).
The BMI cutoff points (≥ 75th percentile of the sample) suggested to predict MS, by age and sex, in the present study, approached the overweight diagnosis proposed by the International Obesity Task Force/IOTF (criterion internationally used with a representative sample of six countries, including Brazil) 10 and Conde and Monteiro 15 , the national reference.
In another study, Ferreira et al. 22 reported that BMI was the most accurate predictor of MS risk in a study with 109 Brazilian children aged 7-11-year-old.Among the different anthropometric measures tested, the BMI > 24.5 kg/m 2 , WC > 78cm, and %BF > 41% (measured by DEXA) were considered to be MS predictors.The lack of description of the anatomical point of PC measurement and generalization of data (without distinction of age and sex) limits the possibility of comparisons between studies.
The BMI values in this study, which are close to those reported for the diagnosis of obesity 10,15 , are found in the reference curve ≥ 90th percentile, with high specificity values in predicting the risk of MS.Halley Castillo et al. 23 reported that high BMI values have an impact on the MS prevalence, as eutrophic and overweight boys had 3% and 66% of MS prevalence, respectively.However, the MS prevalence in 6-12-year-old Mexican children was 23.3%, while in the overweight group, the prevalence was 28.5% 24 .
Sellers et al. 25 found unexpected results in a cohort of Australian Aboriginal children, as BMI had low values in the sample, with a prevalence of overweight of 6.4% and obesity of 4.9%, while the MS prevalence was 14%, a result contradictory to that found in an American study with the same parameters, prevalence of obesity of 15% and MS of 4% 26 .The use of BMI as a diagnostic criterion for overweight and obesity in children is consolidated in the scientific literature; however, its use, as a predictive criterion for the MS risk, should be Quadros et al.Body composition indicators and metabolic syndrome carried out with caution, as children with a classification of "normal weight" by BMI, they can be considered falsely thin, and, as a result of excess body fat, especially in the central region of the body, present metabolic changes in MS components.
The %BF was an accurate predictor of MS for girls (AUC = 0.867) and boys (AUC = 0.864) with cutoff points between the 50th and 75th percentile.The values suggested for predicting the MS risk in 6-8-year-old girls and 7-8-year-old boys approached the cutoff points they classify as adequate 27 or excellent 28 (SE≥93.3%and SP=72%) the amount of %BF.Cutoff points for 7, 9, and 10-year-old girls were included in the classification of moderately high and high body fat 27,28 , which was repeated among 6,9, and 10-year-old boys.A previous study carried out with Brazilians suggested a cutoff point of 25.7% of %BF (SE=75% and SP=71.25%) in the MS prediction in girls with a mean age of 9.9 years, classification of the moderately high amount of %BF 28 , and among boys, there was no significant accuracy for this predictor.The recommendations for the amount of %BF in children were proposed only by sex, without age stratification 27,28 .Thus, divergences between cutoff points for %BF presented in this study (by age and sex) and the recommendations may have occurred due to the lack of specificity attributed to different age groups in previous recommendations.
Sun et al. 29 warn of the secular trend of increasing %BF.The average %BF of North American children born in the 1990s was significantly higher than the average of children born three decades earlier.The estimate of the growth trajectories of the %BF differed between sex, with an increase in the %BF among boys up to 12 years old and a subsequent decrease, and a linear increase in the %BF with age among girls.Girls had significantly higher mean %BF than boys at all ages (8-18-year-old) and in all four decades of birth.In the present study, girls, in addition to having higher %BF, following the secular trend, had a higher prevalence of MS compared to boys (p<0.05),strengthening the consistent relationship between %BF and risk of MS in the pediatric population.
We should highlight the FBM, which presented AUC>0.737 for boys and girls.Suggested cutoff points were ≥ 75th percentile for girls and > 50th percentile for boys.The girls in this study had significantly higher FBM values compared to boys (p<0.05).It is difficult to compare our findings with other studies, as values are expressed, most of the time, in percentages and not in kilograms.
The LBM was not accurate in predicting the risk of MS in different age groups in both sex, and it had the worst performance when compared to the other indicators analyzed in this study for the MS prediction.These findings are in agreement with the body of evidence that suggests a strong relationship between body fat indicators and MS during childhood and adolescence. 7,8ne of the limitations of the present investigation is the lack of comparability with other studies that have been conducted on this topic, as there is no standardization of the measurements and cutoff points used.Another important aspect is the standardization of data collection on anthropometric measurements, which must be considered when using this technique.In our study, all anthropometric measurements were performed by two trained evaluators, ensuring greater reliability of the results.Although we have a representative sample of the population with children of different ethnicities, the results were not stratified in this way in our study.

Table 1 .
Anthropometric characteristics and body composition of 6-10-year-old children according to sex.
Legend: n = sample; SD = standard deviation; min = minimum value; max = maximum value; BMI = body mass index; LBM = lean body mass; FBM = fat body mass; % BF = body fat percentage.Note: Significant difference between sex p ≤ 0.05 from Mann Whitney test.

Table 2 .
Coefficient Box Cox (L), Median (M) and coefficient of variation (S), followed by percentile values for body mass index, lean body mass, fat body mass, and body fat percentage by age for girls.
BMI = body mass index; LBM = lean body mass; FBM = fat body mass; % BF = body fat percentage

Table 3 .
Coefficient Box Cox (L), Median (M) and coefficient of variation (S), followed by percentile values for body mass index (BMI), lean body mass, fat body mass, and body fat percentage by age for boys.

Table 4 .
Area under the curve, confidence interval (95%), cutoff points, sensitivity, and specificity of body composition indicators as predictors of metabolic syndrome in schoolchildren according to sex.indicates that LBM did not present significant AUC to predict metabolic syndrome.