Excesso de peso em crianças: comparação entre o critério internacional e nacional de classificação do índice de massa corpórea

OBJETIVO: Avaliar o desempenho dos pontos de corte para o indice de massa corporea recomendados por Cole et al e Conde e Monteiro para diagnosticar o excesso de peso em criancas. METODOS: A amostra foi composta por 585 escolares com idades entre seis e nove anos, do municipio de Ponta Grossa, no estado do Parana. Foram realizadas medidas de peso, estatura e da dobra cutânea subescapular, a qual foi considerada como padrao-ouro para a caracterizacao do excesso de adiposidade, utilizando como ponto de corte o percentil 85 da curva de referencia do National Center for Health Statistics. Foram calculados a sensibilidade, a especificidade e os valores preditivos positivo e negativo dos criterios de classificacao do indice de massa corporea desenvolvidos por Cole et al e Conde e Monteiro. RESULTADOS: A prevalencia do excesso de peso foi de 20,7 e 28,9%, de acordo com os criterios de Cole et al e de Conde e Monteiro, respectivamente. Independentemente da idade analisada, a sensibilidade do criterio nacional foi superior ao internacional para o sexo masculino, com extensao de 53,3 a 100,0% versus 33,3 a 92,3%, e semelhante para o feminino, variando de 68,4 a 100,0% para ambos os criterios. Em contrapartida, os pontos de corte de Cole et al apresentaram especificidade superior aos de Conde e Monteiro em ambos os sexos, variando de 87,3 a 98,4% versus 74,6 a 93,5%. CONCLUSOES: Tanto o criterio internacional quanto o nacional apresentaram resultados satisfatorios para o diagnostico do excesso de peso em criancas brasileiras. Entretanto, o criterio de Conde e Monteiro mostrou-se mais sensivel, acarretando menor numero de falsos-negativos.

do índice de massa corpórea desenvolvidos por Cole et al e Conde e Monteiro.

RESUMEN
Objetivo: Evaluar el desempeño de los puntos de corte para el Índice de Masa Corporal (IMC), recomendados por Cole et al y Conde y Monteiro para diagnosticar el exceso de peso en niños.

Introduction
The increase in the prevalence of overweight and obesity in children in the last decades (1) has become a public health problem.Overweight children are more susceptible to cardiovascular risk factors, such as dyslipidemia (2) , hypertension (3,4) , hyperglycemia (3) and metabolic syndrome (5) .Moreover, obese children have greater chances of becoming obese adults (6) , which accelerates the early outcomes of morbidity and mortality due to chronic noncommunicable diseases.
The body mass index (BMI) is frequently used in clinical practice and epidemiological studies to evaluate the nutritional status of populations.The World Health Organization (WHO) (7) recommends the use of the BMI in screening overweight and obesity in children and adolescents because measurements are easy to make, the cost of the equipment for assessments is low, and the index is strongly correlated with body fat.
Cole et al (8) developed an international criterion to define underweight, overweight and obesity based on BMI cut-off points in individuals 2 to 18 years of age using data reported by studies conducted in six countries.As there was no Brazilian criterion to classify BMI, Conde & Monteiro (9) established cut-off points for underweight, overweight and obesity for Brazilian children and adolescents based on the Health and Nutrition National Survey, conducted by the Brazilian Institute of Geography and Statistics in 1989 (10) .The cut-off points for overweight and obesity were estimated based on the WHO (7) criteria for adults, and the method used to build the Brazilian curve was basically the same as the one used for the international criterion.
Several studies evaluated the performance of different BMI criteria to diagnose overweight and obesity in children and adolescents by calculating specificity and sensitivity (11)(12)(13)(14)(15)(16) .However, few studies with Brazilian children have been conducted, particularly to compare the international and the national criteria.Therefore, this study evaluated the performance of BMI cut-off points recommended by Cole et al (8) and Conde & Monteiro (9) to diagnose overweight and obesity in children using the subscapular skinfold (SSF) as the criterion standard.

Method
This study used the database of the study Prevalence of overweight and obesity in schoolchildren in the public and private healthcare system, conducted in Ponta Grossa, Brazil, after approval by the Ethics in Research Committee of Universidade Estadual de Ponta Grossa.
The study population comprised 23,931 first to fourth graders in the city of Ponta Grossa, Brazil, who studied in 22 private (3,249) and 77 public (20,682) schools.A representative sample of the population was calculated using the method described by Silva (17) for an estimated maximum obesity prevalence of 30%, considering the highest values found in the literature (18) ; the confidence interval was set at 95%, and precision, at 4% around the prevalence adopted.After the addition of a 5% safety margin, the final sample size was 557 children, and 585 children were evaluated after they submitted an informed consent form signed by their parents or guardians.
Data were collected from 2004 to 2005 by a team of six researchers previously trained to ensure to precision of measurements and the reliability of the study.Anthropometric data were obtained for each patient on the same day: body mass (kg), height (cm) and SSF (mm).During data collection, each measurement was made by the same observer.
Body mass was measured using a Filizola scale and recorded to the nearest 100g.Height was measured at a precision of 0.1cm using a stadiometer.Both measurements were made according to standardized procedures and techniques (19) .
BMI was calculated as a weight-to-height ratio (BMI=weight(kg)/height(m 2 )); overweight and obesity according to BMI were defined based on the criteria developed by Cole et al (8) , as an international reference, and Conde & Monteiro (9) , as a Brazilian reference.In both criteria, overweight was defined based on cut-off points that were equivalent to a BMI of 15kg.m -2 for adults.The cut-off points selected for overweight corresponded to the mean value of the child's age in years (for example, 6.5 years).
SSF was measured using a skinfold caliper (Cescorf, Porto Alegre, Brazil) at 1/10mm, and measurements were made according to standardized procedures and techniques (20) .
Measurements were made on the right side of the child and repeated three successive times at each site.The mean value of the three measurements was recorded.Overweight and obesity were defined based on the SSF measurement, used as the criterion standard.This skinfold value is an important marker of cardiovascular risk factors in children and adolescents (4) .The cut-off point used to diagnose overweight by means of SSF measurement was the 85th percentile of the reference distribution for sex and age according to the reference curve of the National Center for Health Statistics (NCHS) (21) .The 85th percentile has been used to describe overweight in studies conducted with children (22) .
The variables body mass, height and BMI were not normally distributed according to the Kolmogorov-Smirnov test and were, therefore, log-transformed.Two-way ANOVA was used to evaluate the interaction between sex and age for body mass, height and BMI.Specific differences were defined using confidence intervals (interaction) and the Tukey post hoc test, and the significance level was set at p<0.05.The SSF measurements of the sample under study and the NCHS references (21) were described as mean and standard deviations, and the values corresponding to the 85th percentile of the NCHS reference (21) were also presented.
Sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV) and their respective 95% confidence intervals (95%CI) were calculated for the Cole et al (8) and Conde & Monteiro (9) criteria based on the criterion standard.SEN was the percentage of schoolchildren diagnosed with overweight by the BMI classification criteria and the criterion standard.SPE was the percentage of schoolchildren classified as not overweight by the BMI classification criteria and the criterion standard.PPV was the probability of schoolchildren being overweight according to the BMI classification criteria and the criterion standard, whereas NPV was the probability of not being overweight according to the two criteria and the criterion standard.

Results
A total of 585 schoolchildren were enrolled in the study; their mean age was 7.5 (standard deviation=1.0)years, and 50.4% were boys.The prevalence of overweight in this study was 28.9% (31.2% in boys and 26.6% in girls) according to the Conde & Monteiro (9) criterion and 20.7% (20.7% in boys and 20.7% in girls) according to the cut-off points described by Cole et al (8) .The prevalence of overweight based on SSF was 19.7% (21.7% in boys and 17.6% in girls) considering the 85th percentile of the NCHS reference (21) .
The descriptive characteristics and two-way ANOIVA for body mass, height and BMI are presented in Table 1.There were no differences in any of the variables under analysis between sexes.However, there was an interaction between sex and age for the variable height, and there was a difference between all ages for both boys and girls.There was a difference in body mass between ages, and the older children had a higher body mass value than the younger children.Six-year-old children had a BMI below that of eight-and nine-year-old children, and seven-year-old children had a BMI below that of the nine-year-old children.
Table 2 shows SSF means and standard deviations for the sample under analysis and for the NCHS reference (21) , as well as the values corresponding to the 85th percentile of that reference.The SSF values were very similar from six to eight years of age for both boys and girls, and about 2mm below the reference at nine years when compared with the NCHS reference (21) .
Table 3 shows the results of sensitivity, specificity, PPV and NPV of the criteria under evaluation according to age and sex.Findings show that sensitivity of the Brazilian criterion was superior to that of the international criterion, particularly for boys.In contrast, the Cole et al (8) cut-off point had a greater specificity than those described by Conde & Monteiro (9) .The results using the Conde & Monteiro (9) criterion showed higher NPV than the Cole et al (8) criterion for boys in all age groups under analysis.However, except for six years of age, PPV for the international criterion was higher for both boys and girls.The same NPV was found for both criteria at eight and nine years for girls.At seven years of age, the Brazilian criterion yielded higher values.
The Cole et al (8) criterion was less efficient according to the number of false negative cases.The Conde & Monteiro (9) criterion did not diagnose 13 boys (4.4%) as having overweight, and the Cole et al (8) criterion, about twice that number (7.5%).The Brazilian criterion had fewer false negatives than the international criterion for girls (n=7 vs. 10)., and the values corresponding to the 85th percentile of the NCHS reference (21)

Discussion
According to the study results, there were no differences in body mass, height and BMI between sexes.However, younger schoolchildren had lower BMI than the older children.These findings are in agreement with the literature, in relation to both sex (23) and age (24) , which suggests that age may be more relevant to BMI than sex during childhood.However, several studies to evaluate the performance of different criteria to classify BMI in young individuals have conducted their analyses by grouping ages (12)(13)(14)(15)25) . Accoding to this study, the analysis of both sexes together would result in a lower result interpretation error than if ages were grouped.
The criterion developed by Conde & Monteiro (9) for the nutritional evaluation of the Brazilian population yielded a lower number of false negative results and higher sensitivity values than those found when using the Cole et al (8) criterion.The lower the number of false negative results, the better the test.To screen changes in the nutritional status of the population, BMI cut-off points should be defined to identify the highest numbers of overweight individuals, because the adoption of sensitive criteria may contribute to the early establishment of measures to reduce overweight.An overweight person has greater chances of regaining normal weight than a person that is already obese (26) .
Considering all the characteristics analyzed in this study, the Brazilian criterion had more advantages in the evaluation of overweight in children, particularly due to its sensitivity.The lower cut-off points (except for boys at six years) result in a smaller number of false negative results and indicate that this criterion is the most adequate for Brazilian children.Therefore, the main advantage of the Brazilian criterion was its greater capacity to accurately classify overweight children according to the criterion standard.However, the specificity of the international criterion was higher than that of the Brazilian criterion, and its sensitivity, although lower in many cases, was also satisfactory.
The results of comparisons between the two criteria to classify BMI for Brazilian children are controversial.Two recent studies (16,25) compared the BMI criteria developed by Conde & Monteiro (9) and by Cole et al (8) .Dumith and Farias Júnior (25) evaluated the agreement between the Brazilian and the international criteria using kappa statistics; they enrolled a group of 525 schoolchildren aged seven to 15 years in the city of Rio Grande, Brazil, and found that the agreement between criteria was relatively strong for children seven to nine years.However, the Cole et al (8) criterion resulted in a lower prevalence of overweight than the Conde & Monteiro (9) criterion for both sexes, and the difference was statistically significant for girls.The analyses made by Dumith and Farias Júnior (25) did not separate age groups, which may result in an imprecise evaluation of BMI as a marker of overweight because both height and body mass change with growth and development as in children and adolescents grow older (27) .Moreover, sensitivity and specificity were not analyzed, and Table 3 -Sensitivity (SEN), specificity (SPE), positive predictive value (PPV) and negative predictive value (NPV) for the Cole et al (8) and Conde & Monteiro (9) criteria according to age and sex

BMI classification criteria
Cole et al (8) Conde & Monteiro (9) SEN the numbers of false negative and false positive results were not reported.These differences in methods make it impossible to compare their findings and the results of our study.
The second study was conducted by Silva et al (16) and enrolled 1570 schoolchildren aged 7 to 12 years in João Pessoa, Brazil.The authors evaluated sensitivity and specificity of the Brazilian and international criteria using the percentage of body fat using skinfold measurement as the criterion standard.Their findings for children (seven to nine years) revealed elevated and very similar values for both criteria: sensitivity ranged from 86.2 to 100.0, and specificity, from 77.1 to 95.2.The comparison of our study with their findings shows that both found that the two criteria had a good performance in identifying overweight children.However, our study found a greater sensitivity for the Brazilian criterion and a higher specificity for the international criterion.
The differences in the results of the two studies may be explained by the choice of standard criterion to identify overweight and the geographic region where data were collected.Similarly to our findings, the study conducted by Silva et al (16) was the only one in the literature that evaluated sensitivity and specificity of the Brazilian and international BMI criteria for Brazilian children in separate age groups.
In our study, the criterion standard to diagnose overweight was the SSF measurement, an efficient index of truncal fat (28) with a robust associated with cardiovascular risk factors in young individuals (4) .Moreover, SSF has a strong correlation with measurements of body adiposity using ultrasound, CT, dual-energy x-ray absorptiometry and underwater weighing (29,30) .However, skinfold values tend to underestimate adiposity in thinner children and overestimate it in those with greater overweight, which suggests that our study results should be interpreted carefully.
The cut-off point used to classify SSF was the 85th percentile of the reference distribution for sex and age according to the reference curve of the NCHS (21) .This percentile has been used to describe overweight in studies conducted with children (22) and adolescents (11) .The NCHS reference (21) was developed based on a sample in the 1970s, which may generate concerns when used as a reference for current data.However, we believe that the choice of this reference was adequate because: (1) reference data were not affected by obesity, seen in recent data; and (2) mean SSF for ages six to nine years for both sexes in our study was very similar to that of North American children at that time.Therefore, we believe that the results were neither overestimated, nor underestimated.
One of the limitations of our study was the fact that it did not determine the technical measurement error, responsible for the greatest incidence of variability in the measurement of anthropometric parameters.However, the examiners were trained before data collection, and the chance of measurement errors was minimized.Another limitation was the fact that the study did not evaluate the biological stage of maturation of the children, considered to be a variable that affects the amount and distribution of body fat in young individuals.However, as no children 10 years or older participated in the study, we believe that this might not have influenced results, except in cases of early maturation that tend to affect a small percentage of the population, particularly in the age group under analysis in this study.
The major contributions of this study are: the advancement of knowledge about BMI classification criteria to evaluate overweight in Brazilian children, as few studies have compared the international and Brazilian criteria for Brazilian children; and the separation into groups according to age, a factor that has been neglected in many studies, because changes occur as a result of physical growth and development within age groups.However, this study sample was representative of a midsized city in southern Brazil, which makes it difficult to generalize results.Therefore, studies with samples from other Brazilian regions should be conducted to confirm our findings.
In conclusion, both the international and the Brazilian criteria yielded satisfactory results for the diagnosis of overweight in Brazilian children.The Conde & Monteiro criterion (9) , however, was more sensitive and yielded a smaller number of false negative results.

Table 2 -
Subscapular skinfold (SSF) of the population under study and of the NCHS reference

Table 1 -
for sex and age Body mass, height and body mass index (BMI) in boys and girls according to age group Mean±SD.p sex : two-way ANOVA significance level for differences between sexes for body mass, height and BMI; p age : two-way ANOVA significance level for differences between ages for body mass, height and BMI; p sex x age : two-way ANOVA significance level for differences between the interaction of sex and age for body mass, height and BMI (21)an±SD.NCHS(21): National Center for Health Statistics.SSF: subscapular skinfold #