## Services on Demand

## Article

## Indicators

## Related links

## Share

## Jornal de Pediatria

##
*On-line version* ISSN 1678-4782

### J. Pediatr. (Rio J.) vol.84 no.1 Porto Alegre Jan./Feb. 2008

#### http://dx.doi.org/10.1590/S0021-75572008000100009

**Predicting insulin
resistance in children: anthropometric and metabolic indicators**

**Sérgio
R. Moreira ^{I}; Aparecido P. Ferreira^{I}; Ricardo M. Lima^{I};
Gisela Arsa^{I}; Carmen S. G. Campbell^{II}; Herbert G. Simões^{II};
Francisco J. G. Pitanga^{III}; Nanci M. França^{II} **

^{I}Programa
de Mestrado e Doutorado em Educação Física, Universidade
Católica de Brasília (UCB), Brasília, DF, Brazil. Bolsista,
Coordenação de Aperfeiçoamento de Pessoal de Nível
Superior (CAPES)

^{II}Programa de Mestrado e Doutorado em Educação Física,
UCB, Brasília, DF, Brazil

^{III}Programa de Mestrado e Doutorado em Educação Física,
UCB, Brasília, DF, Brazil. Departamento de Educação Física,
Universidade Federal da Bahia (UFBA), Salvador, BA, Brazil

**ABSTRACT**

**OBJECTIVE:**
To predict insulin resistance in children based on anthropometric and metabolic
indicators by analyzing the sensitivity and specificity of different cutoff
points.

**METHODS:**
A cross-sectional study was carried out of 109 children aged 7 to 11 years,
55 of whom were obese, 23 overweight and 31 well-nourished, classified by body
mass index (BMI) for age. Measurements were taken to determine BMI, waist and
hips circumferences, waist circumference/hip circumference ratio, conicity index
and body fat percentage (dual emission X-ray absorptiometry). Fasting blood
samples were taken to measure triglyceridemia, glycemia and insulinemia. Insulin
resistance was evaluated by the glycemic homeostasis method, taking the 90th
percentile as the cutoff point. Receiver operating characteristic curves were
analyzed to a 95% confidence interval in order to identify predictors of glycemic
homeostasis, and sensitivity and specificity were then calculated.

**RESULTS:**
After analysis of the area under the receiver operating characteristic curve
(confidence interval), indicators that demonstrated the power to predict insulin
resistance were, in the following order: insulinemia = 0.99 (0.99-1.00), 18.7
µU×mL^{-1}; body fat percentage = 0.88 (0.81-0.95), 41.3%;
BMI = 0.90 (0.83-0.97), 23.69 kg×m^{2-¹}; waist circumference=
0.88 (0.79-0.96), 78.0 cm; glycemia = 0.71 (0.54-0.88), 88.0 mg×dL^{-1};
triglyceridemia = 0.78 (0.66-0.90), 116.0 mg×dL^{-1} and conicity
index = 0.69 (0.50-0.87), 1.23 for the whole sample; and were: insulinemia =
0.99 (0.98-1.00), 19.54 µU×mL^{-1}; body fat percentage =
0.76 (0.64-0.89), 42.2%; BMI = 0.78 (0.64-0.92), 24.53 kg×m^{2-¹};
waist circumference = 0.77 (0.61-0.92), 79.0 cm and triglyceridemia = 0.72 (0.56-0.87),
127.0 mg×dL^{-1}, for the obese subgroup.

**CONCLUSIONS: **Anthropometric
and metabolic indicators appear to offer good predictive power for insulin resistance
in children between 7 and 11 years old, employing the cutoff points with the
best balance between sensitivity and specificity of the predictive technique.

**Keywords:**
Prediction, insulin resistance, cutoff points, children.

**Introduction**

Insulin
resistance is a clinical condition that is characterized by reduced cellular
glucose uptake in response to a given concentration of insulin and which has
been identified as a public health problem,^{1} while attention has
also been called to the condition in populations such as children and adolescents.^{2-4}
The disorder is associated with a defect in post-receptors of the insulin signaling
pathway,^{5} which interferes with the translocation process of the
muscular glucose transporter (GLUT-4), which itself performs an important role
in glucose uptake. Recently, some authors^{6,7} have extrapolated this
initial theory and proposed an explanation for insulin resistance based on a
lipocentric perspective, by which an accumulation of intramuscular lipids originating
from long chain fatty acids penetrating cells would inhibit translocation of
GLUT-4 to the plasmatic membrane, thereby also suggesting a possible alternative
method of identifying insulin resistance based on indicators associated with
body fat content.

The techniques
for diagnosing insulin resistance based on biomolecular evaluation of insulin
receptors and post-receptors^{5} and on the euglycemic-hyperinsulinemic
clamp test (which analyzes glucose uptake during induced hyperinsulinemia),^{2,8}
are expensive and, for many health professionals, difficult to access. Huang
et al.^{9} validated the glucose homeostasis index (HOMA) for the identification
of insulin resistance in children, demonstrating it to be an interesting proposal
when compared with the gold standard. Nevertheless, the HOMA calculations require
values for fasting insulinemia and glycemia, which in turn demand invasive sample
collections. These procedures make the use of this index problematic, especially
for the diagnostic evaluation of large population samples.

It is clear that
there is a need for diagnostic tests to be developed that are easy to apply,
offer good precision and are of low cost, with the objective of predicting insulin
resistance based on risk factors.^{10} It is a fact that childhood obesity
is associated with negative consequences for children's health, and its prevalence
has been increasing progressively over recent years.^{11-13} In this
context, excess body fat is a variable which may have the potential to predict
insulin resistance in children.^{9,14} To give one example, waist circumference
(WC) has been highlighted as an independent predictor of metabolic and hemodynamic
disorders.^{15,16} However, these studies identified cutoff points for
the variable based on the 90th percentile for a given population and further
research is required to suggest diagnostic tests and their respective advantages,
and with the inclusion of data on the degree of sensitivity and specificity
of the methods being proposed.

Although the health
sciences have identified indicators based on body composition to predict insulin
resistance, diabetes type 2 and other diseases of a cardiovascular character,^{16-18}
no studies have been carried out with children to investigate indicators with
cutoff points determined based on an analysis of the balance between the sensitivity
and specificity of the prediction technique. This being so, the objective of
this study was to test insulin resistance prediction in children based on anthropometric
and metabolic indicators while simultaneously calculating the sensitivity and
specificity of cutoff points.

**Methods**

This was a population-based
cross-sectional study of an initial randomized sample selected from the public
schools of Taguatinga, a satellite city of Brasília, DF, Brazil, in accordance
with a sample size calculation with a confidence interval (CI) of 97%. Schools,
grades and classes were chosen at random, preserving the proportionality of
the children enrolled in the educational sector chosen for study. The sample
analysis had demonstrated that 394 children were needed to achieve a number
of participants (p = 0.05) representative of the population of schoolchildren
enrolled in the public school system. However, with the intention of guaranteeing
a more expressive number, the initial analysis included 958 children from 10
public schools (p = 0.03), observing a prevalence of overweight of 10.6% (n
= 102) and of 7.7% (n = 74) of obesity, meaning that 18.3% of the total number
of children were overweight. After screening the 958 initial subjects, 109 children
of both sexes were chosen with a variety of nutritional status classifications
and aged from 7 to 11 years. The sample studied was classified according to
body mass index/age (BMI/age),^{19} defining 55 children as obese (over
the 95th percentile), 23 children as overweight (between percentiles 85 and
95) and 31 children as well-nourished (between percentiles 5 and 75). The number
of overweight and obese participants was defined based on the prevalence of
overweight and obesity observed previously (18.3%) in this population, and resulting
in an estimate that 71 individuals (p = 0.05) would be sufficient to represent
the population of overweight and obese children enrolled in the public education
system. A further subgroup of 31 children classified as well-nourished comprised
the control group, completing the breakdown of the whole sample studied.

The protocol for this research was approved by the Ethics Committee at the Universidade Católica de Brasília (UCB) and by the Taguatinga Regional Education Department (Secretaria Regional de Ensino de Taguatinga). Those responsible for the study participants signed a free and informed consent form giving authorization for the children selected to participate in the study.

The weight, height
and BMI^{20} of each child were measured using a Plena brand balance
with a digital readout and a stadiometer by Seca. Waist circumference and hip
circumference (HC) were measured,^{20} using a tape measure, Seca brand,
and then calculations were performed to obtain the waist-to-hip ratio (WHR)^{21}
and conicity index (C index),^{18} as follows:

Body fat was evaluated
using dual emission X-ray absorptiometry (DEXA), with Lunar DPX-IQ apparatus
(Lunar Corporation, Madison, WI, United States) with software version 4.6A.
The volunteers were requested to remove all metal objects that they might be
wearing or carrying. Each subject was then positioned in decubitus dorsal on
the DEXA machine for a whole body scan with the pediatric analysis option selected,
in accordance with the manufacturer's recommendations. The equipment had been
duly calibrated before use. Fat mass was calculated for each participant in
relative terms (%F), and all analyses were carried out by the same researcher.^{13}

After an overnight fast of 12 hours, venous blood was taken at the UCB Hospital between 7:45 and 9:00 am for biochemical analysis. The samples were conditioned in vacuum tubes with separator gel and without anticoagulant. After collection, the blood was centrifuged for 10 minutes at 3,000 rpm to separate the serum from the remaining components, and the serum was used for analysis. Triglycerides and blood glucose were assayed using an enzymatic colorimetric kit, processed in an Autohumalyzer A5 (Human-2004). Insulin was assayed using the Automated Chemiluminescence System ACS-180 (Ciba-Corning Diagnostic Corp., 1995, United States).

Insulin resistance
was calculated using the HOMA method,^{9} as illustrated by the following
equation:

The HOMA index
has been validated for children by Huang et al.^{9} against the euglycemic-hyperinsulinemic
clamp technique. The criterion adopted here for a diagnosis of insulin resistance
was a HOMA index over the 90th percentile (p. 90), as has been proposed in the
past.^{22-24}

Receiver operating
characteristic (ROC) analysis was adopted to select the cutoff points that identified
insulin resistance for each of the indicators studied.^{25} For this
procedure the sample was divided into a total group (n = 109; 9.24±1.38
years) and a subgroup comprising just the obese children (n = 55; 9.20±1.16
years). Briefly, a ROC curve is generated by plotting sensitivity on the y-axis
as a function of [1 - specificity] on the x-axis. Sensitivity is the
percentage of individuals who exhibited the outcome (in the case studied here,
insulin resistance) and who have been correctly diagnosed by the indicator in
question (i.e. true positives), while specificity describes the percentage of
individuals who did not exhibit the outcome and were correctly diagnosed by
the indicator (i.e. true-negatives). The criterion utilized to choose the cutoff
points was to select the values at which sensitivity and specificity were most
similar and were not less than 60%. The statistical significance of each analysis
was verified by the area under the ROC curve and by the 95% confidence interval
(95%CI). Thus, a perfect indicator would offer an area under the ROC curve of
1.00, while a diagonal line would represent an area of 0.50. For an indicator
to be exhibiting any discriminative power its area under the ROC curve must
be between 0.50 and 1.00, and the greater the area the greater the indicator's
discriminative power. Another way to determine predictive capacity is using
the 95%CI, where, for an anthropometric or metabolic indicator to be considered
a significant predictor of insulin resistance, the lower limit of the CI (LL-CI)
must not be less than < 0.50.^{26} Additionally, Pearson's linear
correlation test was applied to the relationships between each of the indicators
being tested and insulin resistance, to a significance level of p < 0.05.
Statistical analysis of the data was carried out using the software programs
Stata^{tm} version 9.1 and Statistica^{®} version 5.1.

**Results**

Table 1 lists the areas under the ROC curves for the anthropometric and metabolic insulin resistance predictors together with their respective CIs. Neither the WHR for the whole group or for the obese subgroup, nor the C index or glycemia for the obese subgroup demonstrated significant discriminatory power for insulin resistance (LL-CI < 0.50). In contrast, after analysis of the areas under the ROC curves, the anthropometric indicators C index, BMI, WC and %F for the whole group and BMI, WC and %F for the obese subgroup did prove to be significant predictors of insulin resistance (LL-CI ³ 0.50). Furthermore, the metabolic indicators glycemia, insulinemia and triglyceridemia, for the whole group, and insulinemia and triglyceridemia, for the group made up of obese children, all demonstrated significant discriminatory power for insulin resistance prediction (LL-CI ³ 0.50).

With relation to the ROC curves, it is worth drawing attention to the fact that the x-axis represents [1 - specificity] and the y-axis the sensitivity of possible indicators for predicting insulin resistance (reference). Therefore, the points at which the indicators proposed in this study as having predictive power for insulin resistance exhibit the greatest similarity between the two axes (x and y) were defined as cutoff points and these are listed in Table 2. Furthermore, the correlations between these predictors and insulin resistance are also given in Table 2.

**Discussion**

The principle findings of this study demonstrate the possibility of predicting insulin resistance in children based on anthropometric and metabolic indicators. Analysis of the ROC curves (Table 1), a method not previously used for this purpose, suggests the cutoff points with greatest similarity between sensitivity and specificity, offering information related to the degree of validity of the indicator used in the prediction. The predictors of insulin resistance thus proposed, in decreasing order of sensitivity and specificity, were: insulinemia, %F, BMI, WC, glycemia, triglyceridemia and the C index for the whole sample; and insulinemia, %F, BMI, WC and triglyceridemia for the subgroup composed of obese children (Table 2).

The euglycemic-hyperinsulinemic
clamp test has been described as being the gold standard for the identification
of insulin resistance in children and adolescents.^{2,9,17} It was not
possible to use the euglycemic-hyperinsulinemic clamp technique to test for
insulin resistance in this study, and the HOMA index was used instead, which
could be characterized as a limitation. However, Huang et al.^{9} have
validated the HOMA technique for identifying insulin resistance in children,
and several authors^{22-24} have used the index successfully. Despite
the practicality of using HOMA when compared with the gold standard, it is still
necessary to measure two variables in order to calculate it (glycemia and insulinemia),
and these are obtained invasively. Furthermore, measuring insulinemia is considered
difficult to apply within the daily practice of many different health professionals,
since biochemical assays are needed that must be carried out in a laboratory
environment by a fully trained technician.

Many studies^{15,18,23,27}
have attempted to identify practical and precise indices for predicting diseases,
including insulin resistance,^{9,16,17} which may later trigger diabetes
type 2 early in life.^{28} Information related to detection of insulin
resistance during childhood, acquired in a simple and inexpensive manner, can
be of benefit to a variety of professionals working with child health during
their prophylactic and therapeutic practice, in addition to reducing healthcare
costs.

In this study it
was possible to identify predictors of insulin resistance based on a single
metabolic measurement, such as glycemia, triglyceridemia or insulinemia itself.
As would be expected, insulinemia demonstrated the greatest predictive power
when its area under the ROC curve was analyzed^{25,26} (Table
1), in addition to a high correlation and better sensitivity and specificity
when compared with the other indicators (Table
2). On the other hand, triglyceridemia and glycemia, although having lower
percentages for sensitivity and specificity when compared with insulinemia (Table
2), proved to be good predictors of insulin resistance. When the area under
the ROC curve^{25} and the CI were analyzed, in particular the CI lower
limit greater than 0.50, it was confirmed that there was a significant predictive
ability^{26} for glycemia for the whole sample and for triglyceridemia
for both the whole sample and for the obese subgroup (Table
1). Nowadays, triglyceridemia and glycemia can be tested using low cost
portable analyzers, making it easily possible to use these measurements for
the prediction of insulin resistance in children.

Furthermore, anthropometric
indicators such as %F, the C index, BMI and WC, also demonstrated significant
predictive power^{25,26} for insulin resistance (Table
1). Notwithstanding, %F was measured using DEXA, an expensive method that
is highly complex to apply clinically. However, similar results are observed
when the areas under the ROC curves for the indicators BMI and WC are analyzed
with relation to the area under the ROC curve for %F as measured by DEXA, for
the whole group and the obese subgroup (Table
1). Furthermore, there were significant moderate to high correlations between
%F measured by DEXA and BMI (r = 0.89), WC (r = 0.84) and the C index (r = 0.53)
in this study, and with BMI (r = 0.73) and WC (r = 0.61) in a study by Gomes
et al.^{27} The power of the variable WC to predict insulin resistance
that was detected for both groups in this study (Tables
1 and 2), is in
keeping with other studies^{10,15,16} that have demonstrated that this
variable is an independent predictor, in a variety of populations, for insulin
resistance, lipid content and arterial blood pressure. It is therefore suggested
that the anthropometric indicators studied here be used for predicting insulin
resistance in children, since they offer the advantages of ease of measurement,
low cost, a noninvasive nature and values referring to the degree of sensitivity
and specificity of the cutoff point proposed.

Currently, in places
where morphophysiological, postural and nutritional characteristics are assessed,
such as at sports clubs, gymnasiums and physiotherapy, nutrition and pediatrics
consulting rooms, there is an ever rising prevalence of patients with a variety
of risk factors, of which obesity is of greatest prominence,^{11,12,29}
which is itself associated with insulin resistance at early ages.^{13,14,16}
This being so, the practicality of using the indicators proposed here represents
easily applied procedures and great clinical importance for future therapeutic
and preventative interventions. These practices are even more relevant to the
assessment of children, since they make it possible to prevent the complications
associated with insulin resistance and diabetes type 2 in later life.

Based on the results observed, we conclude that it has been possible to identify anthropometric and metabolic indicators with discriminatory power for the prediction of insulin resistance in children aged 7 to 11 years, based on the cutoff points with the best balance between sensitivity and specificity. The predictors of insulin resistance proposed were insulinemia, %F, BMI, WC, glycemia, triglyceridemia and the C index for the whole sample, and insulinemia, %F, BMI, WC and triglyceridemia for the subgroup of obese children. The ease with which the indicators proposed can be measured makes them important tools to be used in the routines of health professionals. Further studies with similar methodologies are needed to examine the application of these indicators to different populations and to stratify them by characteristics such as ethnicity and family history of diabetes type 2.

**References **

1. Stephens JW, Ambler G, Vallance P, Betteridge DJ, Humphries SE, Hurel SJ. Cardiovascular risk and diabetes. Are the methods of risk prediction satisfactory? Eur J Cardiovasc Prev Rehabil. 2004;11:521-8. [ Links ]

2. Moran A, Jacobs DR Jr., Steinberger J, Hong CP, Prineas R, Luepker R, et al. Insulin resistance during puberty: results from clamp studies in 357 children. Diabetes. 1999;48:2039-44. [ Links ]

3. Sinaiko AR, Jacobs DR Jr., Steinberger J, Moran A, Luepker R, Rocchini AP, et al. Insulin resistance syndrome in childhood: associations of the euglycemic insulin clamp and fasting insulin with fatness and other risk factors. J Pediatr. 2001;139:700-7. [ Links ]

4. Valente AM, Strong W, Sinaiko AR. Obesity and insulin resistance in young people. Am Heart J. 2001;142:440-4. [ Links ]

5. Klip A. Regulation of glucose transport by insulin and non hormonal factors. Life Sci. 1982;31:2537-48. [ Links ]

6. Berggren JR, Hulver MW, Dohm GL, Houmard JA. Weight loss and exercise: implications for muscle lipid metabolism and insulin action. Med Sci Sports Exerc. 2004;36:1191-5. [ Links ]

7. Bruce CR, Hawley JA. Improvements in insulin resistance with aerobic exercise training: a lipocentric approach. Med Sci Sports Exerc. 2004;36:1196-201. [ Links ]

8. DeFronzo RA, Tobin JD, Andres R. Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol. 1979;237:E214-23. [ Links ]

9. Huang TT, Johnson MS, Goran MI. Development of a prediction equation for insulin sensitivity from anthropometry and fasting insulin in prepubertal and early pubertal children. Diabetes Care. 2002;25:1203-10. [ Links ]

10. Zimmet P, Alberti G, Kaufman F, Tajima N, Silink M, Arslanian S, et al. The metabolic syndrome in children and adolescents. Lancet. 2007;369:2059-61. [ Links ]

11. Mello ED, Luft VC, Meyer F. Obesidade infantil: como podemos ser eficazes? J Pediatr (Rio J). 2004;80:173-82. [ Links ]

12. Lobstein T, Baur L, Uauy R; IASO International Obesity TaskForce. Obesity in children and young people: a crisis in public health. Obes Rev. 2004;5 Suppl 1:4-104. [ Links ]

13. Ferreira AP, Oliveira CE, França NM. Metabolic syndrome and risk factors for cardiovascular disease in obese children: the relationship with insulin resistance (HOMA-IR). J Pediatr (Rio J). 2007;83:21-6. [ Links ]

14. Bacha F, Saad R, Gungor N, Arslanian SA. Are obesity-related metabolic risk factors modulated by the degree of insulin resistance in adolescents? Diabetes Care. 2006;29:1599-604. [ Links ]

15. Fernandez JR, Redden DT, Pietrobelli A, Allison DB. Waist circumference percentiles in nationally representative samples of African-American, European-American, and Mexican-American children and adolescents. J Pediatr. 2004;145:439-44. [ Links ]

16. Lee S, Bacha F, Gungor N, Arslanian SA. Waist circumference is an independent predictor of insulin resistance in black and white youths. J Pediatr. 2006;148:188-94. [ Links ]

17. Yeckel CW, Weiss R, Dziura J, Taksali SE, Dufour S, Burgert TS, et al. Validation of insulin sensitivity indices from oral glucose tolerance test parameters in obese children and adolescents. J Clin Endocrinol Metab. 2004;89:1096-101. [ Links ]

18. Pitanga FJ, Lessa I. Sensibilidade e especificidade do índice de conicidade como discriminador do risco coronariano de adultos em Salvador, Brasil. Rev Bras Epidemiol. 2004;7:259-69. [ Links ]

19. National Center for Chronic Disease Prevention and Health Promotion, National Center for Health Statistics. CDC table for calculated body mass index values for selected heights and weights for ages 2 to 20 years. http://www.cdc.gov/growthcharts. Acesso: 10/02/2006. [ Links ]

20. Marins JC, Giannichi RS. Avaliação e prescrição de atividade física: guia prático. 2ª ed. Rio de Janeiro: Shape; 1998. [ Links ]

21. Lohman TG, Roche AE, Martorell R. Anthropometric standardization reference manual. Illinois: Human Kinetics; 1988. [ Links ]

22. Davis CL, Flickinger B, Moore D, Bassali R, Domel Baxter S, Yin Z. Prevalence of cardiovascular risk factors in schoolchildren in a rural Georgia community. Am J Med Sci. 2005;330:53-9. [ Links ]

23. Hirschler V, Aranda C, Calcagno Mde L, Maccalini G, Jadzinsky M. Can waist circumference identify children with the metabolic syndrome? Arch Pediatr Adolesc Med. 2005;159:740-4. [ Links ]

24. Srinivasan SR, Myers L, Berenson GS. Changes in metabolic syndrome variables since childhood in prehypertensive and hypertensive subjects: The Bogalusa Heart Study. Hypertension. 2006;48:33-9. [ Links ]

25. Erdreich LS, Lee ET. Use of relative operating characteristic analysis in epidemiology. A method for dealing with subjective judgement. Am J Epidemiol. 1981;114:649-62. [ Links ]

26. Schisterman EF, Faraggi D, Reiser B, Trevisan M. Statistical inference for the area under the receiver operating characteristic curve in the presence of random measurement error. Am J Epidemiol. 2001;154:174-9. [ Links ]

27. Gomes MA, Rech CR, Gomes MBA, Santos DL. Correlação entre índices antropométricos e distribuição de gordura corporal em mulheres idosas. Rev Bras Cineantropom Desempenho Hum. 2006;8(3):16-22. [ Links ]

28. Rotteveel J, Belksma EJ, Renders CM, Hirasing RA, Delemarre-Van de Waal HA. Type 2 diabetes in children in the Netherlands: the need for diagnostic protocols.Eur. J Endocrinol. 2007;157:175-80. [ Links ]

29. Seidell JC. Environmental influences on regional fat distribution. Int J Obes. 1991;15 Suppl 2:31-5. [ Links ]

**
Correspondence:
**Sérgio
Rodrigues Moreira

SCLN 106, Bloco A/211

CEP 70742-510 - Brasília, DF - Brazil

Tel.: +55 (61) 3036.9147, +55 (61) 8128.7745

Email: sergior@pos.ucb.br

Manuscript received Jun 26 2007, accepted for publication Oct 16 2007.

No conflicts of interest declared concerning the publication of this article.