Hostname: page-component-7c8c6479df-fqc5m Total loading time: 0 Render date: 2024-03-29T15:56:54.274Z Has data issue: false hasContentIssue false

Validation of a BMI cut-off point to predict an adverse cardiometabolic profile with adiposity measurements by dual-energy X-ray absorptiometry in Guatemalan children

Published online by Cambridge University Press:  17 June 2014

Olga Redondo
Affiliation:
Social and Cardiovascular Epidemiology Research Group, School of Medicine, University of Alcalá, Alcalá de Henares, Madrid 28871, Spain Research Support Unit, La Mancha Centro General Hospital, Alcázar de San Juan, 13600 Ciudad Real, Spain
Eduardo Villamor
Affiliation:
Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, MI, USA
Javiera Valdés
Affiliation:
Department of Public Health, Medical Faculty, Catholic University of the North, Santiago de Chile, Chile
Usama Bilal
Affiliation:
Social and Cardiovascular Epidemiology Research Group, School of Medicine, University of Alcalá, Alcalá de Henares, Madrid 28871, Spain Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
Benjamín Caballero
Affiliation:
Center for Human Nutrition, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
Dina Roche
Affiliation:
INCAP Comprehensive Center for the Prevention of Chronic Diseases (CIIPEC), Institute of Nutrition of Central America and Panama (INCAP), Guatemala 1188, Central America
Fernanda Kroker
Affiliation:
INCAP Comprehensive Center for the Prevention of Chronic Diseases (CIIPEC), Institute of Nutrition of Central America and Panama (INCAP), Guatemala 1188, Central America
Manuel Ramírez-Zea
Affiliation:
INCAP Comprehensive Center for the Prevention of Chronic Diseases (CIIPEC), Institute of Nutrition of Central America and Panama (INCAP), Guatemala 1188, Central America
Manuel Franco*
Affiliation:
Social and Cardiovascular Epidemiology Research Group, School of Medicine, University of Alcalá, Alcalá de Henares, Madrid 28871, Spain Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
*
*Corresponding author: Email manuel.franco@uah.es
Rights & Permissions [Opens in a new window]

Abstract

Objective

To identify a body fat percentage (%BF) threshold related to an adverse cardiometabolic profile and its surrogate BMI cut-off point.

Design

Cross-sectional study.

Setting

Two public schools in poor urban areas on the outskirts of Guatemala City.

Subjects

A convenience sample of ninety-three healthy, prepubertal, Ladino children (aged 7–12 years).

Results

Spearman correlations of cardiometabolic parameters were higher with %BF than with BMI-for-age Z-score. BMI-for-age Z-score and %BF were highly correlated (r=0·84). The %BF threshold that maximized sensitivity and specificity for predicting an adverse cardiometabolic profile (elevated homeostasis model assessment–insulin resistance index and/or total cholesterol:HDL-cholesterol ratio) according to receiver operating characteristic curve analysis was 36 %. The BMI-for-age Z-score cut-off point that maximized the prediction of BF ≥ 36 % by the same procedure was 1·5. The area under the curve (AUC) for %BF and for BMI data showed excellent accuracy to predict an adverse cardiometabolic profile (AUC 0·93 (sd 0·04)) and excess adiposity (AUC 0·95 (sd 0·02)).

Conclusions

Since BMI standards have limitations in screening for adiposity, specific cut-off points based on ethnic-/sex- and age-specific %BF thresholds are needed to better predict an adverse cardiometabolic profile.

Type
Research Papers
Copyright
Copyright © The Authors 2014 

The International Obesity Task Force estimated in 2010 that up to 200 million schoolchildren were overweight; of those, 40–50 million were obese( 1 ). The great increase of childhood overweight/obesity over the last three decades seen in developing countries is now being observed in low- and middle-income countries elsewhere in the world, particularly in urban settings, where racial and ethnic disparities in obesity prevalence exist( 2 ). As an example, Guatemala is experiencing ‘the double burden of disease’, with the coexistence of chronic malnutrition and an obesity prevalence of 12·3 % in pre-adolescent children( Reference Mbowe, Díaz and Wallace 3 ). Half of Guatemalan children aged 8–13 years have at least one component of the metabolic syndrome (MetS). However, it is uncertain whether conventional BMI-based definitions of overweight/obesity appropriately identify children with excess adiposity who are then at cardiometabolic risk( Reference Freedman, Ogden and Berenson 4 ).

Obesity is a medical condition in which excessive body fat (BF) may impair health and represents the fifth leading risk for death globally( 1 ). Although excess fat is the indicator of obesity, BMI is considered a surrogate measure of BF to classify individuals as obese( Reference Lazarus, Baur and Webb 5 ). Nevertheless, the accuracy of BMI to estimate BF is debatable in childhood because adiposity is affected by growth and stages of maturation( Reference Taylor, Jones and Williams 6 Reference Brann 8 ). Moreover, variations in adiposity and its relationship with BMI by sex and race–ethnicity can lead to misclassification of obesity( Reference Flegal, Ogden and Yanovski 9 ).

The WHO cut-off points to define overweight inadequately estimate percentage body fat (%BF) in Caucasian populations and different racial groups in Asia( Reference Deurenberg-Yap, Schmidt and Van Staveren 10 ). Also, it is known that common BMI cut-offs to diagnose overweight have high specificity, but low sensitivity in identifying adiposity, failing to identify half of the people with excess %BF( Reference Veiga, Dias and Anjos 7 , Reference Okorodudu, Jumean and Montori 11 ). Despite the significant correlation between %BF and BMI, children grouped by their BMI do not group similarly based on their %BF( Reference Brann 8 ). As BMI does not account for wide variations in BF distribution, it may not correspond to the same degree of adiposity and associated health risks in different individuals and populations( Reference Veiga, Dias and Anjos 7 , Reference Kurtoglu, Mazicioglu and Ozturk 12 ).

Then, it is of relevance to continue studying the relationship between BMI and %BF to properly identify the possible components of MetS( Reference Biro and Wien 13 ). Several studies have assessed BF cut-offs correlating with biological risk( Reference Higgins, Gower and Hunter 14 , Reference Williams, Going and Lohman 15 ). Others have evaluated the ability of different BMI cut-off points in childhood to predict cardiovascular risk in mid-adulthood( Reference Li, Pinot de Moira and Power 16 ). However, to our knowledge, no study has defined BMI cut-offs based on the metabolic disorders related to adiposity by means of a two-step procedure. Health-care professionals need a more sensitive BMI reference tool for children to enable early detection and to quantify the degree of obesity in the prevention of future cardiovascular health hazards( Reference Monasta, Lobstein and Cole 17 Reference Shields and Tremblay 19 ).

Our objective was to identify a %BF threshold associated with an adverse cardiometabolic profile and its surrogate BMI cut-off point in a population of Guatemalan children.

Methods

Participants and setting

The present study is part of a cross-sectional study carried out by the Comprehensive Center for the Prevention of Chronic Diseases (CIIPEC), at the Institute of Nutrition of Central America and Panama (INCAP). The main objective was to measure the prevalence of cardiovascular risk factors in normal-weight and overweight children in order to understand cardiovascular health in poor Guatemalan children.

Ninety-three healthy, prepubertal, Ladino children (a distinct ethnic group composed mostly of Spanish/American Indian mestizos), aged between 7 and 12 years, were randomly recruited within two elementary public schools of Mixco in the peri-urban area of Guatemala City (forty-six boys and forty-seven girls). Schools were selected by convenience, for being located in a poor area (children from low- and medium- to low-income families), with the agreement of the principals and teachers. They were similar to the applicant schools that later on would be included in an intervention programme.

Inclusion criteria were: apparently healthy children with BMI within the normal range (n 53) or overweight (n 40) according to WHO classification (BMI Z-score of 0 to 1 for normal weight and >1 for overweight) and living in the peri-urban area. Exclusion criteria were: chronic disease, menarche, having a brother/sister already included in the study and undernutrition (WHO BMI Z-score <−2).

INCAP’s Institutional Ethics Committee approved the study protocol. Informed written consent was obtained from parents and verbal assent from each child before the tests.

Measurements

Anthropometry, body composition and cardiometabolic measurements were performed early in the morning on each child by three trained members of the research team at the Physiology and Body Composition Laboratory of CIIPEC, using standard protocols, from August to December 2010. Children, who were accompanied by one parent or tutor, had fasted overnight and had avoided any kind of physical exertion for at least 12 h before the test. All measures are summarized in Table 1.

Table 1 Descriptive characteristics by sex in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

CDC, Centers for Disease Control and Prevention; IOTF, International Obesity Task Force; DXA, dual-energy X-ray absorptimetry; %BF, percentage body fat; %TF, percentage trunk fat; TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol; FPG, fasting plasma glucose; HOMA-IR, homeostasis model assessment–insulin resistance index; SBP, systolic blood pressure; DBP, diastolic blood pressure.

* According to the WHO reference( 20 ).

Values reported are percentage (%) and number of subjects (n) for total, boys and girls.

Significantly different by sex (P<0·05).

Anthropometry

Weight (kg) was obtained to the nearest 0·01 kg using a calibrated digital scale (Mettler Toledo IND 221; Mettler Toledo Inc., Columbus, OH, USA), with the child wearing a swimsuit. Height (cm) was obtained using a stadiometer (Perspective Enterprises, Portage, MI, USA) calibrated to the nearest 1·0 mm, with the child standing in bare feet with his/her head, shoulders, buttocks and heels leaning against a surface that was at a 90° angle to the floor. BMI was calculated as weight/height2 (kg/m2). BMI Z-scores were evaluated according to the WHO standards( 20 ). Precision was estimated with the mean of two measurements. If the difference between both measurements was 0·5 kg for weight and 0·5 cm for height, a third was obtained, and the mean of the two closest was used.

Body composition

Percentage of trunk fat (%TF), %BF and fat mass (kg) were measured by dual-energy X-ray absorptiometry using a Lunar iDXA instrument (GE Healthcare Worldwide, General Electric Company, Bucks, UK) and enCORE 2008 (version 12·30·008) software. This is the most used reference technique to measure fatness owing to its high accuracy( Reference Hangartner, Warner and Braillon 21 , Reference Fields and Goran 22 ). Fat mass is adjusted to be more accurate: fat mass/(fat mass + lean mass + bone mineral content) × 100( Reference Goran, Driscoll and Johnson 23 ). On the day of each test, the equipment was calibrated following the manufacturer’s guidelines.

Cardiometabolic measurements

The study physician drew a blood sample (8 ml) by venepuncture. Plasma or serum was separated by centrifugation within 3 h after phlebotomy and immediately frozen until analysis to obtain biochemical variables. Colorimetric methods (Cobas c111 analyser; Roche Diagnostics GmbH, Indianapolis, IN, USA) were used for total cholesterol (TC), LDL-cholesterol (LDL-C), HDL-cholesterol (HDL-C), TAG, fasting plasma glucose (FPG) and serum homocysteine analyses. Serum insulin was obtained by chemiluminescent enzyme immunoassay (Immulyte 2000; Siemens Healthcare Global, Deerfield, IL, USA). Insulin resistance (homeostasis model assessment–insulin resistance index) was estimated as: HOMA-IR=(glucose × insulin)/405 (for glucose in mg/dl and insulin in μUI/ml)( Reference Keskin, Kurtoglu and Kendirci 24 ). The derived TC:HDL-C ratio was determined. Blood pressure (BP) was measured in triplicate, at least 1 min apart, using a standardized technique and an appropriate size for the arm cuff. The mean of the second and third measurements was reported. If the difference between them was 10 mmHg for systolic or diastolic BP, a fourth measurement was obtained, and the mean of the two closest was used.

Definition of adverse cardiometabolic profile

We considered adverse lipid levels as those referred to in the American Academy of Pediatrics’ 2008 report( Reference Daniels and Greer 25 ), corresponding to 95th percentile values (5th for HDL-C) for children <10 and ≥10 years old. We used the cut-off of ≥100 mg/dl for FPG from the American Diabetes Association( 26 ). HOMA-IR higher than 3·2 was considered abnormal( Reference Keskin, Kurtoglu and Kendirci 24 ). Based on adult observations, TC:HDL-C greater than 4·5 was considered adverse( Reference Ingelsson, Schaefer and Contois 27 , Reference Barbosa, Volp and Hermsdorff 28 ).

From 6 to 9 years old and from 10 to 12 years old, hypertension was considered as systolic and diastolic BP levels ≥122/78 mmHg and ≥126/82 mmHg, respectively( Reference Horan and Sinaiko 29 ).

The term ‘adverse cardiometabolic profile’ (ACMP) has been thoroughly studied in relation to the MetS( Reference Rodrigues, Abreu and Resende 30 ). A definition of MetS is established for adults( Reference Alberti, Zimmet and Shaw 31 ), but remains controversial for children( Reference Gustafson, Yanoff and Easter 32 ). Therefore, an ad hoc definition of risk profile was used. Considering the definitions above, we defined an ACMP from those parameters that proved to have a moderate to good significant correlation with %BF (r=0·50 to 0·75) in both sexes, which resulted as HOMA-IR higher than 3·2 and/or TC:HDL-C greater than 4·5 (Table 2).

Table 2 Cross-correlation between adiposity and BMI with cardiometabolic factors: partial Spearman correlation coefficients adjusted for age and/or sex* in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

%BF, percentage body fat; DXA, dual-energy X-ray absorptimetry; TC, total cholesterol; HDL-C, HDL-cholesterol; LDL-C, LDL-cholesterol; FPG, fasting plasma glucose; HOMA-IR, homeostasis model assessment–insulin resistance index; SBP, systolic blood pressure; DBP, diastolic blood pressure.

* ρ p, Spearman correlation coefficients adjusted for age and/or sex.

No significant correlation (P≥0·05).

Statistical analyses

The sex-stratified profile of the sample was described quantitatively. Significant differences across anthropometric and metabolic parameters between sexes were investigated using Student’s t test or the Mann–Whitney U test.

The prevalence of adverse lipid profile, TC:HDL-C, FPG, HOMA-IR and BP was estimated. Afterwards, the prevalence of obesity with the WHO (BMI-for-age Z-score >2), International Obesity Task Force (BMI-for-age equivalent to ≥30 kg/m2 in adults) and Centers for Disease Control and Prevention standards (BMI-for-age ≥95th percentile) was determined( 20 , Reference Cole, Bellizzi and Flegal 33 , 34 ). The differences between the three criteria were tested using the χ 2 test.

Age- and sex-adjusted Spearman analyses were used to obtain cross-correlations between adiposity (%BF) and BMI-for-age Z-scores with cardiometabolic risk parameters (lipid profile, insulin resistance and BP).

A threshold of %BF that predicted an ACMP with a receiver operating characteristic (ROC) curve was determined( Reference Søreide 35 ). The sample distribution of %BF was divided into 2 % thresholds. For each threshold, we calculated sensitivity (percentage of all children with an ACMP defined as obese by the threshold), specificity (percentage of all children without an ACMP defined as non-obese by the threshold), positive likelihood ratio (LHR; ${\rm sensitivity}/(1{\minus}{\rm specificity})$ , the odds of having an ACMP above the threshold) and negative LHR ( $(1{\minus}{\rm sensitivity})/{\rm specificity}$ , the odds of not having an ACMP below the threshold)( Reference Zhou X-H and McClish 36 ). According to the ROC curve analysis, a threshold based on the maximization of sensitivity and specificity was selected. The area under the curve (AUC) was calculated with the parametric ROC curve analysis method, suggesting the accuracy of %BF to discriminate between children with and without the risk profile.

The explained ROC curve analyses for BMI-for-age Z-scores and %BF were repeated to obtain the surrogate threshold of BMI Z-score that divided our sample into children above and below the previously defined %BF threshold. In this case, BMI-for-age Z-scores were divided into thresholds of 0·5.

Statistical analyses were performed using the statistical software package STATA SE version 12·1. Statistical significance was defined as P<0·05.

Results

Mean age of the participants was 9·7 (sd 1·4) years. The prevalence of elevated TC, TC:HDL-C, TAG, LDL-C, FPG and HOMA-IR levels was 7·5 %, 10·8 %, 54·8 %, 3·2 %, 1·1 % and 5·4 %, respectively. No child was hypertensive, 43 % had low HDL-C and 43 % were overweight (BMI-for-age Z-score >1). The prevalence of obesity varied from 20 % under the WHO reference to 19 % by the Centers for Disease Control and Prevention and 13 % by the International Obesity Task Force standards (P<0·001). %BF and %TF were both significantly higher in girls (Table 1).

HOMA-IR and TC:HDL-C had the highest association with %BF (r=0·64 and 0·60, respectively; Table 2). Correlations of these two and most cardiometabolic risk factors were somewhat lower with BMI-for-age Z-scores than with %BF. Correlation of BMI-for-age Z-score and %BF was high (r=0·84, P<0·001). The prevalence of the ACMP profile was 16 %.

Table 3 shows the specificity, sensitivity, positive and negative LHR for each %BF threshold in the prediction of an ACMP. The %BF threshold that maximized sensitivity and specificity was 36 %. The prevalence of children having BF ≥ 36 % was 27 %, who were five times more likely to have an ACMP compared with those having BF < 36 %. The AUC for the %BF data was 0·93 (sd 0·04; Fig. 1), indicating excellent accuracy of %BF predicting an ACMP.

Fig. 1 Empirical receiver operating characteristic curve predicting children with abnormal homeostasis model assessment–insulin resistance index and/or total cholesterol:HDL-cholesterol levels using specific percentage body fat thresholds in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Table 3 Specificity, sensitivity and likelihood ratios for each %BF threshold to predict an adverse cardiometabolic risk profile in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

%BF, body fat percentage; LHR, likelihood ratio; pos, positive; neg, negative; HOMA-IR, homeostasis model assessment–insulin resistance index; TC, total cholesterol; HDL-C, HDL-cholesterol.

* For each %BF demarcation range, adverse CVD risk profile is defined as HOMA-IR ≥ 3·16 and/or TC:HDL-C ≥ 4·5.

Sensitivity/(1 – specificity).

(1 – Sensitivity)/specificity.

Table 4 shows the specificity, sensitivity, positive and negative LHR for each BMI-for-age Z-score threshold in the prediction of excess adiposity, defined by the selected threshold of BF ≥ 36 %. A BMI-for-age Z-score threshold of 1·5 was selected as it maximized sensitivity and specificity. The prevalence of children having BMI-for-age Z-score ≥ 1·5 was 32 %, who were eight times more likely to have excess adiposity compared with those having BMI-for-age Z-score < 1·5. Children with a BMI-for-age Z-score < 1·5 were 86 % less likely to have excess adiposity. The AUC for the BMI data was 0·95 (sd 0·02; Fig. 2), indicating excellent accuracy of BMI-for-age Z-score in predicting the critical %BF that implies an ACMP.

Fig. 2 Empirical receiver operating characteristic curve predicting children with excess adiposity levels implying an adverse cardiometabolic profile using specific BMI-for-age Z-score thresholds in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Table 4 Specificity, sensitivity and likelihood ratios for each BMI-for-age Z-score threshold to predict adiposity in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

LHR, likelihood ratio; pos, positive; neg, negative; BF, body fat.

* BMI-for-age Z-score demarcation ranges according to the WHO reference( 20 ). For each range, 36 % BF is the cut-off point defining excess of adiposity implying risk profile, as revealed in Table 3.

Sensitivity/(1 – specificity).

(1 – Sensitivity)/specificity.

Discussion

The BMI Z-score cut-off point of 1·5 was estimated from a 36 % BF threshold built from adverse cardiometabolic parameters with which the correlation was moderate to good. That resulted in the best trade-off between sensitivity and specificity for Guatemalan Ladino children to predict excess adiposity compromising health. Following the WHO classification, eleven out of ninety-three children would have been misclassified downwards with the BMI Z-score cut-off of >2 (sensitivity 64 %); and nine out of ninety-three would have been misclassified upwards with the BMI Z-score cut-off of >1 (sensitivity 100 %).

Most studies have evaluated excess adiposity according to a sex- and age-specific %BF percentile( Reference Li, Pinot de Moira and Power 16 , Reference Zimmermann, Gubeli and Puntener 37 , Reference Mei, Grummer-Strawn and Pietrobelli 38 ). Three heterogeneous studies proposed %BF cut-offs in childhood based on cross-sectional associations with CVD risk factors( Reference Higgins, Gower and Hunter 14 , Reference Williams, Going and Lohman 15 , Reference Dwyer and Blizzard 39 ). They differed in the selection of the risk factors’ threshold levels.

The MetS is a cluster of risk factors for CVD and type 2 diabetes( Reference Müller, Lagerpusch and Enderle 40 ). Obesity plays a primary role and leads to insulin resistance, increased BP and dyslipidaemia( Reference Alberti, Zimmet and Shaw 31 ). Its definition for children is problematic because of changes in lipid profile, BP, insulin sensitivity and anthropometry with age and pubertal development. Thus, several cut-off values for the risk factors have been used( Reference Reinehr, de Sousa and Toschke 41 ). The International Diabetes Federation made a proposal, but also suggested that MetS should not be diagnosed as an entity in children younger than 10 years( Reference Zimmet, Alberti and George 42 ). Therefore, we defined an ad hoc ACMP based on the clustering of risk factors and their correlation with %BF. Insulin resistance is a major precursor of CVD and type 2 diabetes( Reference Rodrigues, Abreu and Resende 30 ) and among children, the reliability of HOMA-IR for its measurement is well established( Reference Keskin, Kurtoglu and Kendirci 24 ). Moreover, the ratio TC:HDL-C is a positive predictor of oxidized LDL concentrations, playing an important role in the atherosclerotic process( Reference Barbosa, Volp and Hermsdorff 28 ).

Comparison of studies that explore the feasibility of overweight to identify fat children is challenging, as different BMI and adiposity cut-offs are proposed( Reference Monasta, Lobstein and Cole 17 , Reference Reilly, Dorosty and Emmett 43 , Reference Neovius, Rössner and Vågstrand 44 ). Moreover, a high BMI level is a specific, but not a sensitive, predictor of %BF( Reference Zimmermann, Gubeli and Puntener 37 , Reference Reilly, Dorosty and Emmett 43 , Reference Wickramasinghe, Cleghorn and Edmiston 45 ). Among children, BMI has inconsistently been correlated with adiposity( 2 , Reference Wells 46 ) and its accuracy increases with the level of BF( 2 , Reference Widhalm, Schonegger and Huemer 47 , Reference Bray, DeLany and Harsha 48 ). Then, the high correlation found between BMI and %BF in the present study might be due to our oversampling of overweight children. However, some ethnic groups, across all levels of BMI, have higher %BF( Reference Flegal, Ogden and Yanovski 9 , Reference Deurenberg-Yap, Schmidt and Van Staveren 10 ). Our 36 % BF threshold, higher than those of Higgins, Williams or Dwyer( Reference Higgins, Gower and Hunter 14 , Reference Williams, Going and Lohman 15 , Reference Dwyer and Blizzard 39 ), suggests that Guatemalan Ladino children may have higher %BF basal levels. Although there is a significant association between overweight classification by WHO and cardiometabolic risk( Reference de Onis, Martínez-Costa and Núñez 49 ), BMI misclassifies individuals with increased risk related to elevated adiposity( Reference Gómez-Ambrosi, Silva and Galofré 50 ).

BMI cut-offs detecting excess adiposity and %BF thresholds detecting risk profile with excellent discrimination were assessed( Reference Simundic 51 ). In the whole spectrum of possible decision thresholds, we selected those optimizing both sensitivity and specificity. The positive LHR for the BMI-for-age Z-score of 2·0 doubles that of the 1·5 cut-off, and would therefore be more accurate in showing the likelihood of having ACMP in those with elevated adiposity, but at the cost of worsening the sensitivity (Table 4). However, the health care of children puts a greater emphasis on more sensitive references for prevention tasks. Instead, the WHO BMI Z-score of >1 would increase the false-positive rate.

We acknowledge several important limitations of the present study. About half of the children were preselected to be overweight, so we cannot know the real prevalence of overweight/obesity. Moreover, Tanner stage was not assessed, so we could expect an influence of sex on biochemistry and %BF. To minimize this, we chose BMI-for-age Z-scores, which are sex- and age-adjusted, and lipid profile cut-offs above and below the age of 10 years, because most children below 10 years old are in Tanner stage 1 (prepubertal)( Reference Horan and Sinaiko 29 ). The proposed risk profile definition is a modest approach to the multifaceted obese phenotype, lacking the dynamic view of muscle and fat( Reference Müller, Lagerpusch and Enderle 40 ). Nevertheless, empirical categories are required for clinical practice. Estimating cut-off points for boys and girls separately would be the preferred choice but our limited sample size precluded it. Finally, parametric ROC curve analyses have the drawback of discarding data when they are grouped( Reference Søreide 35 ). Therefore, ROC curve points and the AUC may be biased. However, as ROC curve analysis is independent of prevalence( Reference Søreide 35 ), we had the advantage of equal numbers of participants with both conditions to evaluate without the need for representative samples.

Another limitation comes from the cross-sectional design of the study and the inherent lack of temporality in our causal hypothesis, but several colleagues( Reference Higgins, Gower and Hunter 14 , Reference Williams, Going and Lohman 15 , Reference Dwyer and Blizzard 39 ) have previously used cross-sectional data to develop %BF cut-off points for defining obesity implying cardiometabolic risk.

Internal validity may somehow be affected in our study by the fact that the BMI Z-score threshold of 1·5 is based on a 36 % BF threshold, implying a false-negative rate of 8 % (92 % sensitivity) in the first step, and in our final step we found a false-negative rate of 12 % (88 % sensitivity). Like Higgins et al.( Reference Higgins, Gower and Hunter 14 ), we used dual-energy X-ray absorptiometry to obtain a robust estimation of %BF, as this technique is not affected by intra- and inter-observer variability.

According to our criteria of ACMP, ten children (10·8 % of the sample) had TC:HDL-C>4·5 and five children (5·4 %) presented a HOMA-IR>3·2; altogether, fifteen children. It is difficult to quantify the validity and reliability of these thresholds because there is no clear standard definition of ACMP( Reference Reinehr, de Sousa and Toschke 41 ).

Nevertheless, to our knowledge, this is the first attempt to assess the accuracy of BMI as an indicator of the degree of harmful adiposity using a two-step method. It represents an approach to obesity classification and a novel contribution to the literature concerning MetS.

Conclusion

In summary, as current BMI definitions have limitations in screening for adiposity, further efforts to build specific BMI cut-off points derived from ethnic-/sex- and age-specific %BF thresholds remain necessary. Such tailored tools may help developing programmes aimed to identify early metabolic risk in other child populations.

Acknowledgements

Acknowledgements: The study was accomplished thanks to four collaborating institutions: the School of Medicine at the University of Alcalá in Spain; INCAP in Guatemala; the Department of Epidemiology at Michigan University; and the Department of International Health at the Johns Hopkins Bloomberg School of Public Health. The authors thank all the study participants, parents and children in the two schools in Guatemala. They also thank all the technicians who performed the fieldwork measurements at the Physiology and Body Composition Laboratory of CIIPEC. Financial support: This project was funded in part with Federal funds from the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health, Department of Health and Human Services (under contract no. HHSN26820900032C). The NHLBI had no role in the design, analysis or writing of this article. Conflicts of interest: None. Authorship: O.R., M.F. and E.V. contributed to the original design. M.R.-Z., D.R. and F.K. organized and conducted data collection. O.R. reorganized, reordered and recoded variables of the database. O.R., E.V., U.B. and M.F. conducted the statistical analyses. O.R., E.V., U.B., B.C., M.R.-Z. and M.F. were active in the interpretation of results. The manuscript was drafted by O.R., M.F., U.B. and E.V. and reviewed by all authors. All authors have approved the final report. All authors had full access to the data in the study and take responsibility for its integrity and the accuracy of the data analysis. M.F. is the guarantor and corresponding author for this study. Ethics of human subject participation: This study was approved by the INCAP’s Institutional Ethics Committee.

References

1. International Obesity Taskforce (2012) Obesity & Research/Obesity the Global Epidemic. http://www.iaso.org/iotf/obesity/obesitytheglobalepidemic/ (accessed October 2012).Google Scholar
2. Centers for Disease Control and Prevention (2012) Childhood Overweight and Obesity. http://www.cdc.gov/obesity/childhood/index.html (accessed October 2012).Google Scholar
3. Mbowe, O, Díaz, A, Wallace, J et al. (2013) Prevalence of metabolic syndrome and associated cardiovascular risk factors in Guatemalan school children. Matern Child Health J (Epublication ahead of print version).Google Scholar
4. Freedman, DS, Ogden, CL, Berenson, GS et al. (2005) Body mass index and body fatness in childhood. Curr Opin Clin Nutr Metab Care 8, 618623.CrossRefGoogle ScholarPubMed
5. Lazarus, R, Baur, L, Webb, K et al. (1996) Body mass index in screening for adiposity in children and adolescents: systematic evaluation using receiver operating characteristic curves. Am J Clin Nutr 63, 500506.CrossRefGoogle ScholarPubMed
6. Taylor, RW, Jones, IE, Williams, SM et al. (2002) Body fat percentages measured by dual-energy X-ray absorptiometry corresponding to recently recommended body mass index cutoffs for overweight and obesity in children and adolescents aged 3–18 y. Am J Clin Nutr 76, 14161421.CrossRefGoogle ScholarPubMed
7. Veiga, GV, Dias, PC & Anjos, LA (2001) A comparison of distribution curves of body mass index from Brazil and the United States for assessing overweight and obesity in Brazilian adolescents. Pan Am J Public Health 10, 7985.CrossRefGoogle ScholarPubMed
8. Brann, LS (2008) Classifying preadolescent boys based on their weight status and percent body fat produces different groups. J Am Diet Assoc 108, 10181022.CrossRefGoogle ScholarPubMed
9. Flegal, KM, Ogden, CL, Yanovski, JA et al. (2010) High adiposity and high body mass index-for-age in US children and adolescents overall and by race–ethnic group. Am J Clin Nutr 91, 10201026.CrossRefGoogle ScholarPubMed
10. Deurenberg-Yap, M, Schmidt, G, Van Staveren, WA et al. (2000) The paradox of low body mass index and high body fat percentage among Chinese, Malays and Indians in Singapore. Int J Obes Relat Metab Disord 24, 10111017.CrossRefGoogle ScholarPubMed
11. Okorodudu, DO, Jumean, MF, Montori, VM et al. (2010) Diagnostic performance of body mass index to identify obesity as defined by body adiposity: a systematic review and meta-analysis. Int J Obes (Lond) 34, 791799.CrossRefGoogle ScholarPubMed
12. Kurtoglu, S, Mazicioglu, MM, Ozturk, A et al. (2010) Body fat reference curves for healthy Turkish children and adolescents. Eur J Pediatr 169, 13291335.CrossRefGoogle ScholarPubMed
13. Biro, FM & Wien, M (2010) Childhood obesity and adult morbidities. Am J Clin Nutr 91, issue 5, 1499S1505S.CrossRefGoogle ScholarPubMed
14. Higgins, PB, Gower, BA, Hunter, GR et al. (2001) Defining health-related obesity in prepubertal children. Obes Res 9, 233240.CrossRefGoogle ScholarPubMed
15. Williams, DP, Going, SB, Lohman, TG et al. (1992) Body fatness and risk for elevated blood pressure, total cholesterol, and serum lipoprotein ratios in children and adolescents. Am J Public Health 82, 358363.CrossRefGoogle ScholarPubMed
16. Li, L, Pinot de Moira, A & Power, C (2011) Predicting cardiovascular disease risk factors in midadulthood from childhood body mass index: utility of different cutoffs for childhood body mass index. Am J Clin Nutr 93, 12041211.CrossRefGoogle ScholarPubMed
17. Monasta, L, Lobstein, T, Cole, TJ et al. (2011) Defining overweight and obesity in pre-school children: IOTF reference or WHO standard? Obes Rev 12, 295300.CrossRefGoogle ScholarPubMed
18. Twells, LK & Newhook, LA (2011) Obesity prevalence estimates in a Canadian regional population of preschool children using variant growth references. BMC Pediatr 11, 21.CrossRefGoogle Scholar
19. Shields, M & Tremblay, MS (2010) Canadian childhood obesity estimates based on WHO, IOTF and CDC cut-points. Int J Pediatr Obes 5, 265273.CrossRefGoogle ScholarPubMed
20. World Heath Organization (2012) Growth Reference 5–19 years. BMI-for-age. http://www.who.int/growthref/en/ (accessed October 2012).Google Scholar
21. Hangartner, TN, Warner, S, Braillon, P et al. (2013) The Official Positions of the International Society for Clinical Densitometry: acquisition of dual-energy X-ray absorptiometry body composition and considerations regarding analysis and repeatability of measures. J Clin Densitom 16, 520536.CrossRefGoogle ScholarPubMed
22. Fields, DA & Goran, MI (2000) Body composition techniques and the four-compartment model in children. J Appl Physiol 89, 613620.CrossRefGoogle ScholarPubMed
23. Goran, MI, Driscoll, P, Johnson, R et al. (1996) Cross-calibration of body-composition techniques against dual-energy X-ray absorptiometry in young children. Am J Clin Nutr 63, 299305.CrossRefGoogle ScholarPubMed
24. Keskin, M, Kurtoglu, S, Kendirci, M et al. (2005) Homeostasis model assessment is more reliable than the fasting glucose/insulin ratio and quantitative insulin sensitivity check index for assessing insulin resistance among obese children and adolescents. Pediatrics 115, e500e503.CrossRefGoogle Scholar
25. Daniels, SR & Greer, FR (2008) Lipid screening and cardiovascular health in childhood. Pediatrics 122, 198208.CrossRefGoogle ScholarPubMed
26. American Diabetes Association (2006) Diagnosis and classification of diabetes mellitus. Diab Care 29, Suppl. 1, S43S48.CrossRefGoogle Scholar
27. Ingelsson, E, Schaefer, EJ, Contois, JH et al. (2007) Clinical utility of different lipid measures for prediction of coronary heart disease in men and women. JAMA 298, 776785.CrossRefGoogle ScholarPubMed
28. Barbosa, KB, Volp, AC, Hermsdorff, HH et al. (2011) Relationship of oxidized low density lipoprotein with lipid profile and oxidative stress markers in healthy young adults: a translational study. Lipids Health Dis 10, 61.CrossRefGoogle ScholarPubMed
29. Horan, MJ & Sinaiko, AR (1987) Synopsis of the report of the second task force in blood pressure control in children. Hypertension 10, 115121.CrossRefGoogle ScholarPubMed
30. Rodrigues, AN, Abreu, GR, Resende, RS et al. (2013) Cardiovascular risk factor investigation: a pediatric issue. Int J Gen Med 6, 5766.CrossRefGoogle ScholarPubMed
31. Alberti, KG, Zimmet, P & Shaw, J (2006) Metabolic syndrome – a new world-wide definition. A Consensus Statement from the International Diabetes Federation. Diabet Med 23, 469480.CrossRefGoogle ScholarPubMed
32. Gustafson, JK, Yanoff, LB, Easter, BD et al. (2009) The stability of metabolic syndrome in children and adolescents. J Clin Endocrinol Metab 94, 48284834.CrossRefGoogle ScholarPubMed
33. Cole, TJ, Bellizzi, MC, Flegal, KM et al. (2000) Establishing a standard definition for child overweight and obesity worldwide: international survey. Br Med J 320, 12401243.CrossRefGoogle ScholarPubMed
34. Centers for Disease Control and Prevention (2012) Children’s BMI Group calculator (metric version). http://www.cdc.gov/healthyweight/assessing/bmi/childrens_bmi/tool_for_schools.html (accessed November 2012).Google Scholar
35. Søreide, K (2009) Receiver-operating characteristic curve analysis in diagnostic, prognostic and predictive biomarker research. J Clin Pathol 62, 15.CrossRefGoogle ScholarPubMed
36. Zhou X-H, Obuchowski NA & McClish, DK (2002) Statistical Methods in Diagnostic Medicine. New York: Wiley-Interscience.CrossRefGoogle Scholar
37. Zimmermann, MB, Gubeli, C, Puntener, C et al. (2004) Detection of overweight and obesity in a national sample of 6–12-y-old Swiss children: accuracy and validity of reference values for body mass index from the US Centers for Disease Control and Prevention and the International Obesity Task Force. Am J Clin Nutr 79, 838843.CrossRefGoogle Scholar
38. Mei, Z, Grummer-Strawn, LM, Pietrobelli, A et al. (2002) Validity of body mass index compared with other body-composition screening indexes for the assessment of body fatness in children and adolescents. Am J Clin Nutr 75, 978985.CrossRefGoogle ScholarPubMed
39. Dwyer, T & Blizzard, CL (1996) Defining obesity in children by biological endpoint rather than population distribution. Int J Obes Relat Metab Disord 20, 472480.Google ScholarPubMed
40. Müller, MJ, Lagerpusch, M, Enderle, J et al. (2012) Beyond the body mass index: tracking body composition in the pathogenesis of obesity and the metabolic syndrome. Obes Rev 13, 613.CrossRefGoogle ScholarPubMed
41. Reinehr, T, de Sousa, G, Toschke, AM et al. (2007) Comparison of metabolic syndrome prevalence using eight different definitions: a critical approach. Arch Dis Child 92, 10671072.CrossRefGoogle ScholarPubMed
42. Zimmet, P, Alberti, K, George, MM et al.; IDF Consensus Group (2007) The metabolic syndrome in children and adolescents – an IDF Consensus Report. Pediatr Diabetes 8, 299306.CrossRefGoogle ScholarPubMed
43. Reilly, JJ, Dorosty, AR, Emmett, PM; ALSPAC Study Team (2000) Identification of the obese child: adequacy of the body mass index for clinical practice and epidemiology. Int J Obes Relat Metab Disord 24, 16231627.CrossRefGoogle ScholarPubMed
44. Neovius, M, Rössner, SM, Vågstrand, K et al. (2009) Adiposity measures as indicators of metabolic risk factors in adolescents. Obes Facts 2, 294301.CrossRefGoogle ScholarPubMed
45. Wickramasinghe, VP, Cleghorn, GJ, Edmiston, KA et al. (2005) Validity of BMI as a measure of obesity in Australian white Caucasian and Australian Sri Lankan children. Ann Hum Biol 32, 6071.CrossRefGoogle ScholarPubMed
46. Wells, JC (2001) A critique of the expression of paediatric body composition data. Arch Dis Child 85, 6772.CrossRefGoogle ScholarPubMed
47. Widhalm, K, Schonegger, K, Huemer, C et al. (2001) Does the BMI reflect body fat in obese children and adolescents? A study using the TOBEC method. Int J Obes Relat Metab Disord 25, 279285.CrossRefGoogle ScholarPubMed
48. Bray, GA, DeLany, JP, Harsha, DW et al. (2001) Evaluation of body fat in fatter and leaner 10-y-old African American and white children: the Baton Rouge Children’s Study. Am J Clin Nutr 73, 687702.CrossRefGoogle ScholarPubMed
49. de Onis, M, Martínez-Costa, C, Núñez, F et al. (2013) Association between WHO cut-offs for childhood overweight and obesity and cardiometabolic risk. Public Health Nutr 16, 625630.CrossRefGoogle ScholarPubMed
50. Gómez-Ambrosi, J, Silva, C, Galofré, JC et al. (2012) Body mass index classification misses subjects with increased cardiometabolic risk factors related to elevated adiposity. Int J Obes (Lond) 36, 286294.CrossRefGoogle ScholarPubMed
51. Simundic, AM (2008) Measures of diagnostic accuracy: basic definitions. Med Biol Sci 22, 6165.Google Scholar
Figure 0

Table 1 Descriptive characteristics by sex in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Figure 1

Table 2 Cross-correlation between adiposity and BMI with cardiometabolic factors: partial Spearman correlation coefficients adjusted for age and/or sex* in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Figure 2

Fig. 1 Empirical receiver operating characteristic curve predicting children with abnormal homeostasis model assessment–insulin resistance index and/or total cholesterol:HDL-cholesterol levels using specific percentage body fat thresholds in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Figure 3

Table 3 Specificity, sensitivity and likelihood ratios for each %BF threshold to predict an adverse cardiometabolic risk profile in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Figure 4

Fig. 2 Empirical receiver operating characteristic curve predicting children with excess adiposity levels implying an adverse cardiometabolic profile using specific BMI-for-age Z-score thresholds in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala

Figure 5

Table 4 Specificity, sensitivity and likelihood ratios for each BMI-for-age Z-score threshold to predict adiposity in ninety-three healthy, prepubertal, Ladino children (aged 7–12 years), Guatemala