Skip to main content Accessibility help
Hostname: page-component-78bd46657c-lnsrr Total loading time: 0.279 Render date: 2021-05-09T14:55:30.712Z Has data issue: true Feature Flags: { "shouldUseShareProductTool": true, "shouldUseHypothesis": true, "isUnsiloEnabled": true, "metricsAbstractViews": false, "figures": false, "newCiteModal": false, "newCitedByModal": true }

A comparative study of methods for diagnosis of obesity in an urban mixed-race population in Minas Gerais, Brazil

Published online by Cambridge University Press:  01 September 2007

Silvia Nascimento de Freitas
Department of Social and Clinical Nutrition, School of Nutrition, Federal University of Ouro Preto (UFOP), University Campus, Morro do Cruzeiro, CEP 35400/000 Ouro Preto, MG, Brazil
Waleska Teixeira Caiaffa
Department of Social and Preventive Medicine, School of Medicine, Federal University of Minas Gerais (UFMG), Belo Horizonte, Brazil
Cibele Comini César
Department of Statistics, Exact Science Institutes, UFMG, Belo Horizonte, Brazil
Ana Paula Carlos Cândido
Department of Social and Clinical Nutrition, School of Nutrition, Federal University of Ouro Preto (UFOP), University Campus, Morro do Cruzeiro, CEP 35400/000 Ouro Preto, MG, Brazil
Valéria Andrade Faria
Scientific Initiation Grant, UFOP, Ouro Preto, Brazil
Raimundo Marques do Nascimento Neto
Institute of Hypertension, Belo Horizonte, Brazil
George Luiz Lins Machado-Coelho
Department of Pharmacy, School of Pharmacy, UFOP, Ouro Preto, Brazil
E-mail address:
Rights & Permissions[Opens in a new window]



Obesity is defined as an excess of total body fat and may be assessed by different methods. The objective of the present study was to establish the discriminatory power of anthropometric data in determining obesity.


The subjects comprised 685 individuals, aged 20–79 years, sampled from a population-based survey. The following indicators were used: body mass index (BMI), waist circumference (WC) and total body fat percentage estimated with both Siri's equation (%BF Siri) and foot-to-foot bioelectrical impedance analysis (%BF BIA). Sensitivity and specificity of different cut-off points for each method were determined using %BF BIA as reference.


Of 685 participants, 57.6% were aged ≥ 40 years, 69.9% were women and 72.6% self-referred themselves as non-white. To classify obesity based on sex and age among women aged < 40 years, the cut-off points were BMI of 26.0 kg m− 2, WC of 84.0 cm and %BF Siri of 34.0%; in those aged ≥ 40 years, the cut-off points were 28.0 kg m− 2, 90.0 cm and 37.4%, respectively. The cut-off points among men aged < 40 years were BMI of 26.3 kg m− 2, WC of 86.0 cm and %BF Siri of 22.5%, and in those aged ≥ 40 years, 26.3 kg m− 2, 89.0 cm and 24.5%, respectively. BMI was the method with the largest area under the curve (AUC) independent of sex and sex/age, yet no differences were observed in AUC between BMI and WC (P>0.05). Classifying according to skin colour did not change cut-off points in any indicator.


BMI and WC better discriminate obesity among women and men aged ≥ 40 years from a mixed-race population.

Research Paper
Copyright © The Authors 2007

Total body fat, which is one part of body composition, plays an important role in epidemiological studies because it is a known risk factor for non-communicable diseasesReference Zhu, Wang, Shen, Heymsfield and Heshka1, Reference Misra and Vikram2. Several different methods have been used for its assessment, among which hydrostatic weighing, dual-energy X-ray absorptiometry (DEXA), body mass index (BMI), waist circumference (WC), skinfold thickness and bioelectrical impedance stand out.

Hydrostatic weighing and DEXA are considered the gold standards. However, they have the disadvantages of low applicability and high cost for population studies. On the other hand, BMI, WC, skinfold thickness and bioelectrical impedance are cheaper and quite reliableReference Lohman3Reference Bray5 but provide different information about body composition. BMI assesses individuals based on their total body mass; WC classifies central adiposity; and skinfold thickness estimates body fat based on the assumption that subcutaneous fat thickness comprises a constant proportion of total body fatReference Lohman3, Reference Bray5, Reference Prentice and Jebb6.

Bioelectrical impedance has also been used recently to estimate body composition. It is based on the knowledge that lean body mass conducts electricity better than adipose tissue, showing an inverse relationship with body resistance under an electric currentReference Lohman3. Two techniques are used: bipolar (conventional) and portable bioelectrical impedance analysis (BIA).

In the first method, the evaluator should be trained in order to ensure the correct placement of electrodes on the subject's feet and hands, a factor of paramount importance for accuracy and reproducibility of body component estimatesReference Lohman3. However, the latter technique costs less, does not require much time for measurements, and recent studies have shown it to be valid and applicable to epidemiological investigationsReference Nunez, Gallagher, Visser, Pi-Sunyer, Wang and Heymsfield7Reference Jebb, Cole and Doman9.

Although DEXA is considered the gold standard, among portable instruments the Tanita® excels in providing foot-to-foot BIA and has shown a high correlation with both conventional bipolar BIA and DEXA. Furthermore, the classification of body fat percentage takes into account sex and age differencesReference Gallagher, Heymfield, Heo, Jebb, Murgatroyd and Sakamoto10.

The literature reports that body composition can be determined by several factors characterising a population, such as age, sex and raceReference Prentice and Jebb6Reference Weinsier, Hunter, Gower, Schutz, Darnell and Zuckerman11Reference Santos, Albala, Lera, Garcia, Arroyo and Perez-Bravo13. Studies on body composition according to sex and age in mixed-race populations are scarceReference Chór, Faerstein, Kaplan, Lynch and Lopes14Reference Baltrus, Lynch, Everson-Rose, Raghunathan and Kaplan16. Ouro Preto City has a mixed-race population, probably resulting from a high concentration of African Diaspora descendents brought to the city from the 16th to the 18th century, and it has a high percentage of non-white people (63.14%) living in the urban area17. Therefore, its inhabitants present a unique opportunity to investigate indicators for the diagnosis of obesity in population groups with such characteristics.

In the light of diverse methodologies available to assess body composition and total body fatReference Lohman3, Reference Nunez, Gallagher, Visser, Pi-Sunyer, Wang and Heymsfield7, Reference Jebb, Cole and Doman9, the objective of the present study was to verify the sensitivity, specificity and discriminatory power of anthropometric measurements obtained by different methods in determining obesity in a mixed-race population group.

Materials and methods

Study population and sample

This study included 685 participants aged 20–79 years from a population sample of 930 residents aged 15 years or more, randomly sampled in a medium-sized town – Ouro Preto – in the Southeastern region in Brazil during 2001. It is part of a cross-sectional study aimed to investigate the prevalence of risk factors for cardiovascular disease in this population.

The sampling frame was based on the following assumptions: infinite population, systemic arterial hypertension prevalence estimation at 25%Reference Freitas, Carvalho and Neves18, 3% precision, 95% confidence level and estimated loss of 20%.

Anthropometric and body composition measurements

The measurements were performed according to standard recommendations on subjects with an empty bladder who had fasted for 12 h. Measurements were taken between 07.00 and 10.00 hours by the same previously trained team.

Total body fat percentage (%BF) was measured by means of foot-to-foot BIA using Tanita® BF542 scales with an electric current of 500 μA and a fixed frequency of 50 kHz, with 0.5% precision. BMI (kg m− 2) was calculated as weight (in kg), obtained on the scales mentioned above (which had a capacity of 136 kg and weight increments of 0.2 kg), divided by the square of height (in m)Reference Lohman3, Reference Costa19. Height was measured without shoes to the nearest 0.1 cm using a portable stadiometer with fixed tape, with the head in the Frankfurt plane. WC (cm) was measured at the natural waist, which is the midpoint between the lower costal arch and the iliac crest, using a non-elastic measuring tape with a precision of 0.1 cmReference Costa19. Triceps, biceps, subscapular and suprailiac skinfold thickness (mm) was measured on the non-dominant side of the body by means of a CESCORF® apparatus, with a precision of 0.1 mm. Three measurements were carried out on each skinfold, and the mean was considered for analysisReference Costa19. Body density was estimated according to sex and age using the Durnin and Womersley equation. %BF was also calculated based on body density by means of Siri's equation (referred to as %BF Siri below). All participants (n = 685) who complied with all anthropometric measures were included in the analysis.

Definition of obesity

Excess body fat was defined by means of the criterion established by Gallagher et al. for African American populationsReference Gallagher, Heymfield, Heo, Jebb, Murgatroyd and Sakamoto10, following the recommendations of the manufacturer of the Tanita® BF542. Men and women with body fat excess were classified by age, and their %BF was, respectively: ≥ 38% and ≥ 26% for age 20–30 years; ≥ 39% and ≥ 27% for age 40–59 years; and ≥ 41% and ≥ 29% for age 60–79 years.

Statistical analysis

Analysis of correlation, using the Pearson chi-square test, was carried out between Tanita® %BF and the anthropometric and body composition variables. Using bioelectrical impedance measured by the Tanita® scales as the reference (referred to as %BF BIA below), we calculated the sensitivity (probability of detecting truly obese individuals) and specificity (probability of detecting truly not obese individuals) for various obesity discriminating points by plotting on the receiver operating characteristic (ROC) curves by sex, sex and age, and sex and skin colour. Age was categorised according to the median distribution of the dataset ( < 40 years: younger adults, ≥ 40 years: older adults) and self-referred skin colour as white and non-white.

The areas under the curves (AUC), 95% confidence intervals and standard errors were tested using the Wilcoxon testReference Alonso and Gonzáles-Suárez20. The method of Hanley and McNeilReference Hanley and McNeil21, within the 95% confidence interval, was used to compare the areas under the ROC curves for each anthropometric variable.

The project was approved by the Institutional Review Board of the Federal University of Ouro Preto under the protocol 26/2001.


Of 685 participants, 479 (69.9%) were females and 206 (30.1%) males, 57.6% were ≥ 40 years old, and 72.6% self-referred themselves as non-white. Table 1 describes their demographic, anthropometric and body composition characteristics by sex and age.

Table 1 Demographic, anthropometric and body composition features according to sex, Ouro Preto City, Brazil

SD – standard deviation; BMI – body mass index; %BF BIA – percentage total body fat estimated by foot-to-foot bioelectrical impedance analysis; %BF Siri – percentage total body fat estimated by Siri's equation; WC – waist circumference.

Females had the following anthropometry (mean ± standard deviation, SD): age, 46.0 ± 15.2 years; height, 157.1 ± 7.1 cm; weight, 64.8 ± 13.5 kg; BMI, 26.3 ± 5.3 kg m− 2. Stratified by age, younger women presented significantly lower mean BMI (24.6 kg m− 2) than older women (27.4 kg m− 2). %BF Siri (35.4 ± 5.9) was higher than %BF BIA (33.5 ± 8.2), especially among older females. Regarding WC, among all females the mean value (85.2 ± 13.7 cm) was above normal levels, although younger subjects presented mean average WC (79.8 cm) near to normal limits.

In males, anthropometry (mean ± SD) was as follows: age, 42.7 ± 15.4 years; height, 170.8 ± 7.1 cm; weight, 70.7 ± 12.7 kg; BMI, 24.2 ± 3.9 kg m− 2. Stratified by age, men aged < 40 years had a mean BMI of 23.5 kg m− 2, significantly lower than that (24.9 kg m− 2) of men aged ≥ 40 years. %BF BIA (20.0 ± 7.9) was similar to %BF Siri (20.4 ± 6.6); slight but non-significant differences (P = 0.11) were observed when mean %BF Siri was compared with mean %BF BIA among younger male adults. Mean WC for all men (85.9 ± 11.8 cm) was below normal levels but a significant increase with age could be observed: from 81.7 cm in younger males to 89.6 cm among older males.

Table 2 shows the correlation between anthropometric measurements and %BF BIA, by sex and sex and age. The correlations ranged from 0.74 to 0.87 for females, and from 0.67 to 0.93 for males. BMI was the method that presented the best correlations with the reference method, except for females aged < 40 years, whose WC correlation (r = 0.87) was greater than with the reference method. Nevertheless, stratifying by sex independent of age verified that the correlation found between WC and %BF BIA showed no significant differences between sexes (r = 0.80).

Table 2 Correlation between anthropometric measurements and %BF BIA according to sex, Ouro Preto City, Brazil

CI – confidence interval; BMI – body mass index; %BF Siri – percentage total body fat estimated by Siri's equation; WC – waist circumference.

Sensitivity and specificity of several cut-off points for obesity, according to the reference method used and by sex and age, are presented in Table 3. For diagnosis of obesity, it was observed that the cut-off points showing greatest sensitivity and specificity for BMI were 27.5 kg m− 2 for females (sensitivity = 90.3%, specificity = 82.6%) and 26.3 kg m− 2 for males (sensitivity = 90.5%, specificity = 86.6%). The cut-off point of ≥ 30 kg m− 2 suggested by the World Health Organization (WHO)22 and the US National Institutes of Health (NIH)23 showed low sensitivity (60.4% and 33.3% for women and men, respectively) but high specificity (94.2% for women and 98.8% for men).

Table 3 Values of sensitivity (Sens) and specificity (Spec) of the cut-off points for obesity, according to reference method and reference standards, Ouro Preto City, Brazil

*World Health Organization22.

†National Institutes of Health23.

Stratified by age, for young females the BMI cut-off point of 26.0 kg m− 2 corresponded to sensitivity = 97.9% and specificity = 88.5%; and for older females the cut-off point of 28.0 kg m− 2 had sensitivity = 90.8% and specificity = 77.7%. For young and older males, the BMI cut-off point of 26.3 kg m− 2 corresponded to sensitivity and specificity of 95.0% and 93.5%, and 86.4% and 81.6%, respectively.

For WC, the discriminating point for diagnosis of total body fat excess in women was 86 cm (sensitivity = 91.0%, specificity = 75.7%), whereas in men it was 89.5 cm (sensitivity = 92.9%, specificity = 79.1%). Using the values recommended by the NIH (88 cm and 102 cm), sensitivity of 82.7% and 33.3%, and specificity of 79.5% and 98.8%, was calculated for women and men, respectively.

Stratifying by age, the WC cut-off with best discrimination for women aged < 40 years was 84.0 cm (sensitivity = 89.1%, specificity = 93.7%) and for those aged ≥ 40 years was 90.0 cm (sensitivity = 87.4%, specificity = 75.7%). For men, cut-offs of 86.0 cm (sensitivity = 100%, specificity = 86.8%) and 89.0 cm (sensitivity = 100%, specificity = 65.5%) were found, respectively.

Considering %BF Siri, the best discriminating points were 37.0% (sensitivity = 78.9%, specificity = 73.6%) for all females and 21.9% (sensitivity = 92.9%, specificity = 66.5%) for all males, which varied with age in both sexes. The cut-offs were: for women < 40 years, 34.0% (sensitivity = 97.8%, specificity = 67.2%) and ≥ 40 years, 37.4% (sensitivity = 82.8%, specificity = 67.1%); and for men < 40 years, 22.5% (sensitivity = 100%, specificity = 75.3%) and ≥ 40 years, 24.5% (sensitivity = 68.2%, specificity = 79.3%). These levels were very similar to those recommended by Lohman, i.e. 35% for women and 25% for men.

Figure 1 shows the ROC curves and their respective AUC by sex using %BF BIA as the reference method. It can be observed that AUC was larger for BMI and WC and smaller for %BF Siri, for both sexes. Comparing the curves, in females BMI had better discrimination than WC or %BF Siri (P < 0.05), while WC and %BF Siri were quite similar (P = 0.10). In males BMI also showed larger AUC, and this was similar to that of WC (P = 0.31) and statistically different from that of %BF Siri (P = 0.01). On the other hand, AUC for WC and %BF Siri were similar (P = 0.11).

Fig. 1 Receiver operating characteristic curves of anthropometric predictors in the urban population of Ouro Preto City, Brazil aged 20–79 years, by sex (AUC – area under the curve; CI – confidence interval; BMI – body mass index; WC – waist circumference; %BF Siri – percentage total body fat estimated with Siri's equation)

Stratifying according to age, the methods that presented greater AUC were BMI and WC independently of sex. Among females aged < 40 years and ≥ 40 years, it was observed that AUC for BMI was similar to that for WC (P = 0.16 and 0.15, respectively) but differed from AUC for %BF Siri (P = 0.01 for both); whereas WC was similar to %BF Siri in AUC (P = 0.13 and 0.16, respectively). On the other hand, among males aged < 40 years, no differences were observed between AUC for BMI and WC (P = 0.88), BMI and %BF Siri (P = 0.14) and WC and %BF Siri (P = 0.14). Among males aged ≥ 40 years, AUC for BMI was similar to that for WC (P = 0.95) and greater than AUC for %BF Siri (P = 0.055), while AUC for WC and AUC for %BF Siri were different (P = 0.04) (Fig. 2).

Fig. 2 Receiver operating characteristic curves of anthropometric predictors in the urban population of Ouro Preto City, Brazil aged 20–79 years, by sex and age (AUC – area under the curve; CI – confidence interval; BMI – body mass index; WC – waist circumference; %BF Siri – percentage total body fat estimated with Siri's equation)

Defining the balance points based on sensitivity and specificity, we observed similarities between cut-off points by either sex or sex and skin colour. This may be explained by the high percentage of non-whites found in this study population (72.6%). Regarding morbid obesity, prevalent in 0.99% of our population, the exclusion of this group did not change any cut-off point indicators.


In the present study, BMI, WC and %BF using Siri's equation showed good discriminatory power for diagnosis of obesity with specific cut-off points for age and sex/age in an urban mixed-race population. Compared with the reference method (%BF determined by foot-to-foot BIA), BMI and WC were the methods revealing greatest accuracy for the entire population, except for males aged < 40 years, whose estimates did not differ regarding the method adopted. BMI and WC had similar discriminatory power, according to the ROC areas under the curve, independently of age and sex.

Age and sex are well-known and relevant factors to determine anthropometric indicator cut-off points, especially if we take into account body fat composition and distribution. Marked changes in body composition are observed in men and women as they get older, and there is a trend to increase fat mass to the detriment of lean massReference Hortobagyi, Israel and O'Brein24Reference Kuk, Lee, Heymsfield and Ross30 in populations of different racesReference Fernández, Heo, Heymsfield, Pierson, Pi-Sunyer and Wang31Reference Bell, Adair and Popkin33. Accumulation of fat particularly in the abdominal region has been observed in postmenopausal womenReference Prentice and Jebb6, Reference Pichard, Kyle, Bracco, Slosman, Morabia and Schutz34, Reference Jackson, Stanforth and Gagnon35; hence, in our study, BMI and WC had similar discriminatory power in this population. In other words, higher BMI could be attributed to increased adiposity in the abdominal region.

Nevertheless, lower estimates of BMI cut-off points compared with international standards were observed in this study, according to sex and sex/age. Similar results were reported by other authors in studies of specific population groups, in whom the cut-off points varied according to sex and ageReference Stanforth, Jackson, Green, Gagnon, Rankinen and Despres15Reference Movsesyan, Tankó, Larsen, Christiansen and Svendsen28Reference Aronne and Segal36Reference Pitanga and Lessa38. When the values recommended by the NIH23 and WHO22 were used to define obesity, we observed an increase in specificity for men and women (98.8% and 94.2%, respectively) and a significant drop in sensitivity (33.3% and 60.4%, respectively). The rise in the number of false-negatives, i.e. obese individuals who were not considered as such, led to an underestimated rate of obesity of 57.2% and 29.9% for men and women, respectively, regardless of age. Similar underreporting of estimates was found by Frankenfield et al. Reference Frankenfield, Rowe, Cooney, Smith and Becker26 using the same cut-off points recommended by the WHO to estimate obesity in white individuals.

WC is a fat centralisation predictor, deemed a robust body adiposity indicatorReference Misra and Vikram2, and in our study it showed good correlation with total body fat (r ≥ 0.80) for both sexes, except for individuals aged ≥ 40 years (r = 0.77 for females and r = 0.73 for males). Similar results were obtained by Lemos-Santos et al. Reference Lemos-Santos, Valente, Gonçalves-Silva and Schieri39 in a Brazilian male adult population (r = 0.83–0.89). Also, WC was the anthropometric variable that ranked second in discriminatory power for both sex and age; and no differences were observed in the AUC of BMI and WC. These findings may suggest that the excess of body fat could be a result of increased adiposity in the waist in our population.

Similar to BMI, WC showed a tendency to lower cut-off points compared with those recommended by the NIH23, as also observed by Taylor et al. Reference Taylor, Jones, Williams and Goulding40. We found in men that the recommended cut-off point of 102 cm was highly specific and not very sensitive, whereas in women the cut-off point was slightly lower than that recommended by the literature (88 cm), thus again leading to underreported obesity estimates.

With regard to %BF estimated by Siri's equation, we verified a smaller correlation (r = 0.67–0.79) with the foot-to-foot BIA method in both sexes and according to sex/age, when compared with BMI and WC. This estimator also presented a smaller AUC. %BF estimated by means of skinfold thickness using the equation based on the two-compartment body model has been criticised by some authors regarding its agreement with other methodsReference Moreno, Gandoy and González41, Reference Peterson, Czerwinski and Siervogel42, as well as for methodological suitability in its generalised use of equations for body density prediction from Caucasian and Asian populations applied to other populationsReference Deurenberg and Deurenberg-Yap12, Reference Peterson, Czerwinski and Siervogel42, Reference Wagner and Heyward43. Some authors discuss about a possible systematic underreporting of obesity. These claims could corroborate our finding of underreporting of obesity, even using a cut-off point recommended by Gallagher et al. Reference Gallagher, Heymfield, Heo, Jebb, Murgatroyd and Sakamoto10 for the African American population per sex and age. It is important to emphasise that our population group had ethnic characteristics and body distribution features that were distinct from those of so-called African American individuals. However, it is worth mentioning that there are no specific equations for body density prediction in Brazil, which justifies this study.

Moreover, the %BF Siri cut-off points that maximised the combination of sensitivity and specificity for females (37.0%) and males (21.9%) were higher for women and lower for men, with high sensitivity and low specificity, compared with those recommended in the literatureReference Lohman3.

Several issues related to methods to assess body composition, the standards used and the ethnic diversity of population groups studied could be discussed. Body composition assessment methods are indirect procedures that are not free from errors, but technical and operational feasibility should be a decisive factor when choosing methods in epidemiological studies. In the present study, the reference standard adopted was foot-to-foot BIA because it presents a good correlation with DEXA, the gold standard. Furthermore, it is considered the best reference method in nutritional epidemiology for use in population-based studies because of its high accuracy and low costReference Ellis4, Reference Bray44. However, limitations to its use have to be acknowledged and concern some clinical conditions of subjects such as hydration status, exercise level, physiological or pathological status. All of them can lead to an over- or underestimation of fat massReference Kyle, Bosaeus, De Lorenzo, Deurenberg, Elia and Manuel Gomez45, Reference Coppini, Waitzberg and Campos46. This may be a particular problem in subjects with morbid obesityReference Coppini, Waitzberg and Campos46, Reference Krupa47. In our study we tried to minimise all of these factors by the strict study protocol during data collection and in the analysis. For the latter we excluded all patients with morbid obesity (0.99% of our sampled population) and the cut-points did not change by age and sex.

Ethnic diversity may also be a determining factor for body composition and, naturally, for the standard used for comparisonsReference Deurenberg and Deurenberg-Yap12, Reference Fernández, Heo, Heymsfield, Pierson, Pi-Sunyer and Wang31. Asian, black and Hispanic populations apparently have a higher fat deposit in the trunk than in the limbs, and more subcutaneous fat in the upper part of the body, than Caucasian individualsReference Stanforth, Jackson, Green, Gagnon, Rankinen and Despres15, Reference Fernández, Heo, Heymsfield, Pierson, Pi-Sunyer and Wang31, Reference Gallagher, Kuznia, Heshka, Albu, Heymsfield and Goodpaster48. That we observed no significant differences in the cut-off points of the diverse methods used when classifying the sample according to skin colour might be explained by the high percentage of non-whites in the population of Ouro Preto (72.6%). This fact is due to colonisation of the city during the gold extraction period, therefore hindering comparisons, but enabling an anthropometric study in a population known for its multi-race characteristic17, Reference Campolina, Melo and Andrade49. It should be mentioned that skin colour definition and race in Brazil should not be dissociated from social conditions and schooling, as appropriately reported in the literatureReference Chór, Faerstein, Kaplan, Lynch and Lopes14, Reference Maio, Monteiro, Chor, Faerstein and Lopes50.

It must be pointed out that developing reference standards to define obesity in the field of epidemiology is still a problem. It is recommended that the cut-off points should be specific for a given population, due to variations in body composition related to age, sex, level of physical activity and ethnic groupReference Prentice and Jebb6, Reference Deurenberg and Deurenberg-Yap12, Reference Movsesyan, Tankó, Larsen, Christiansen and Svendsen28, Reference Fernández, Heo, Heymsfield, Pierson, Pi-Sunyer and Wang31, Reference Hugles, Frontera, Roubenoff, Evans and Singh51. Indeed, these factors may too have influenced our results, since the reference standard used to classify %BF BIA in the population of Ouro Preto was that proposed by Gallagher et al. Reference Gallagher, Heymfield, Heo, Jebb, Murgatroyd and Sakamoto10.

In conclusion, the present results suggest that the discriminatory capacity of tests for obesity varies according to sex and age. BMI and WC had better discriminatory power for obesity among women, regardless of age, as well as among older adult males. The methods did not differ among young adult men. These findings corroborate the claim that it is necessary to use different reference standards for body fat for each sex and ageReference Prentice and Jebb6, Reference Gallagher, Heymfield, Heo, Jebb, Murgatroyd and Sakamoto10, Reference Aronne and Segal36. Thus, more accurate studies aiming to define body fat prediction formulas for multi-race populations, like that of Ouro Preto, Brazil, are required.


Sources of funding: The study was supported by research grants from the Federal University of Ouro Preto, UNIMED Inconfidentes, Alcan Alumínio do Brasil and Secretaria Municipal de Saúde de Ouro Preto. W.T.C. is the recipient of a scholarship from the Brazilian National Council for Scientific and Technological Development (CNPq).

Conflict of interest declaration: None.

Authorship responsibilities: All authors have participated sufficiently in the conception and design of this work, analysis of the data and writing of the manuscript.

Acknowledgements: We sincerely thank the sponsors of this research, the residents of Ouro Preto City, Dr Aline Cristine de Sousa Lopes and Dr Claudia Marlière for critical review of the manuscript.


1Zhu, S, Wang, ZM, Shen, W, Heymsfield, S, Heshka, S. Percentage body fat ranges associated with metabolic syndrome risk: results based on the third National Health and Nutrition Examination Survey (1988-1994). American Journal of Clinical Nutrition 2003; 78: 228–35.CrossRefGoogle Scholar
2Misra, A, Vikram, NK. Clinical and pathophysiologial consequences of abdominal adiposity and abdominal adipose tissue depots. Nutrition 2003; 19: 457–66.CrossRefGoogle Scholar
3Lohman, TG. Advances in Body Composition Assessment. Champaign, IL: Human Kinetics Publishers, 1992.Google Scholar
4Ellis, KJ. Selected body composition methods can be used in field studies. Journal of Nutrition 2001; 131, (Suppl.): 1589S–95S.CrossRefGoogle ScholarPubMed
5Bray, GA. Evaluation of obesity: who are the obese? Postgraduate Medicine 2003; 114: 1927.CrossRefGoogle ScholarPubMed
6Prentice, AM, Jebb, SA. Beyond body mass index. Obesity Reviews 2001; 2: 141–7.CrossRefGoogle ScholarPubMed
7Nunez, C, Gallagher, D, Visser, M, Pi-Sunyer, FX, Wang, Z, Heymsfield, SB. Bioimpedance analysis: evaluation of leg-to-leg system based on pressure contact food-pad electrodes. Medicine and Science in Sports and Exercise 1997; 29: 524–31.Google Scholar
8Austin, MA, Heymsfield, SB, Nieman, DC. Body Composition Measurement in Females with Leg-to-leg Bioelectrical Impedance Analysis compared to DEXA. New York: Department of Health, Leisure & Exercise Science, Obesity Research Center, St Luke's Columbia University, 1998.Google Scholar
9Jebb, SA, Cole, TJ, Doman, D. Evaluation of the novel Tanita body-fat analyser to measure body composition by comparison with a four-compartment model. British Journal of Nutrition 2000; 83: 115–22.CrossRefGoogle ScholarPubMed
10Gallagher, D, Heymfield, SB, Heo, M, Jebb, SA, Murgatroyd, PR, Sakamoto, Y. Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. American Journal of Clinical Nutrition 2000; 72: 694701.CrossRefGoogle ScholarPubMed
11Weinsier, RL, Hunter, GR, Gower, BA, Schutz, Y, Darnell, BE, Zuckerman, PA. Body fat distribution in white and black women: different patterns of intraabdominal and subcutaneous abdominal adipose tissue utilization with weight loss. American Journal of Clinical Nutrition 2001; 74: 631–6.CrossRefGoogle ScholarPubMed
12Deurenberg, P, Deurenberg-Yap, M. Validity of body composition methods across ethnic population groups. Acta Diabetologica 2003; 40: 5246–9.CrossRefGoogle ScholarPubMed
13Santos, JL, Albala, C, Lera, L, Garcia, C, Arroyo, P, Perez-Bravo, F, et al. . Anthropometric measurements in the elderly population of Santiago, Chile. Nutrition 2004; 20: 452–7.CrossRefGoogle ScholarPubMed
14Chór, D, Faerstein, E, Kaplan, GA, Lynch, JVV, Lopes, CS. Association of weight change with ethnicity and life course socioeconomic position among Brazilian civil servants. International Journal of Epidemiology 2004; 33: 100–6.CrossRefGoogle ScholarPubMed
15Stanforth, PR, Jackson, AS, Green, JS, Gagnon, J, Rankinen, T, Despres, JP, et al. . Generalized abdominal visceral fat prediction models for black and white adults aged 17–65 y: the HERITAGE family study. International Journal of Obesity and Related Metabolic Disorders 2004; 28: 925–32.CrossRefGoogle Scholar
16Baltrus, P, Lynch, JVV, Everson-Rose, S, Raghunathan, TE, Kaplan, GA. Race/ethnicity, life-course socioeconomic position, and body weight trajectories over 34 years: the Alameda Country Study. American Journal of Public Health 2005; 95: 1595–601.CrossRefGoogle Scholar
17Instituto Brasileiro de Geografia e Estatística. Censo Demográfico [online], 2000. Available at Accessed 14 December 2005.Google Scholar
18Freitas, OC, Carvalho, RF, Neves, JM. Prevalência da hipertensão arterial sistêmica na população urbana de Catanduva, SP. Arquivos Brasileiros de Cardiologia 2001; 77: 915.Google Scholar
19Costa, RF. Composição Corporal, 1st ed; São Paulo: Manole, 2001.Google Scholar
20Alonso, ED, Gonzáles-Suárez, R. Análisis de las curvas receiver-operating characteristic: un método útil para evaluar procederes diagnósticos. Revista Cubana de Endocrinologiá 2002; 13: 169–76.Google Scholar
21Hanley, JA, McNeil, BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983; 148: 839–43.CrossRefGoogle ScholarPubMed
22World Health Organization (WHO). Obesity. Preventing and Managing the Global Epidemic. Report of a WHO Consultation. WHO Technical Report Series No. 894. Geneva: WHO, 2000.Google Scholar
23National Institutes of Health (NIH). The Practical Guide: Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. Bethesda, MD: NIH, 2000.Google Scholar
24Hortobagyi, T, Israel, RG, O'Brein, KF. Sensitivity and specificity of the Quetelet index to assess obesity in men and women. European Journal of Clinical Nutrition 1994; 48: 369–81.Google ScholarPubMed
25Forbes, G. Longitudinal changes in adult fat-free mass: influence of body weight. American Journal of Clinical Nutrition 1999; 70: 1025–31.CrossRefGoogle ScholarPubMed
26Frankenfield, DC, Rowe, WA, Cooney, RN, Smith, JS, Becker, D. Limits of body mass index to detect obesity and predict body composition. Nutrition 2001; 17: 2630.CrossRefGoogle ScholarPubMed
27Mei, Z, Grummer-Strawn, LMG, Pietrobelli, A, Goulding, A, Goran, MI, Dietz, WH. Validity of body mass index compared with other body-composition screening indexes for the assessment of body fatness in children and adolescents. American Journal of Clinical Nutrition 2002; 75: 978–85.CrossRefGoogle ScholarPubMed
28Movsesyan, L, Tankó, LB, Larsen, PJ, Christiansen, C, Svendsen, OL. Variations in percentage of body fat within different BMI groups in young, middle-aged and old women. Clinical Physiology and Functional Imaging 2003; 23: 130–3.CrossRefGoogle ScholarPubMed
29Stewart, A. Fat patterning: indicators and implications. Nutrition 2003; 19: 559–60.CrossRefGoogle ScholarPubMed
30Kuk, J, Lee, S, Heymsfield, SB, Ross, R. Waist circumference and abdominal adipose tissue distribution: influence of age and sex. American Journal of Clinical Nutrition 2005; 81: 1330–4.CrossRefGoogle ScholarPubMed
31Fernández, JR, Heo, M, Heymsfield, SB, Pierson, RN Jr, Pi-Sunyer, FX, Wang, ZM, et al. . Is percentage body fat differentially related to body mass index in Hispanic Americans, African Americans, and European Americans? American Journal of Clinical Nutrition 2003; 77: 71–5.CrossRefGoogle ScholarPubMed
32Mott, JW, Wang, J, Thornton, JC, Allison, DB, Heymsfield, SB, Pierson, RN Jr. Relation between body fat and age in 4 ethnic groups. American Journal of Clinical Nutrition 1999; 69: 1007–13.CrossRefGoogle ScholarPubMed
33Bell, AC, Adair, LS, Popkin, BM. Ethnic differences in the association between body mass index and hypertension. American Journal of Epidemiology 2002; 155: 346–53.CrossRefGoogle Scholar
34Pichard, C, Kyle, UG, Bracco, D, Slosman, DO, Morabia, A, Schutz, Y. Reference values of fat-free and fat masses by bioelectrical impedance analysis in 3393 healthy subjects. Applied Nutritional Investigation 2000; 16: 245–54.Google ScholarPubMed
35Jackson, AS, Stanforth, P, Gagnon, J. The effect of sex, age, and race on estimating percentage body fat from body mass index: the Heritage Family Study. International Journal of Obesity and Related Metabolic Disorders 2002; 26: 789–96.CrossRefGoogle ScholarPubMed
36Aronne, LJ, Segal, KR. Adiposity and fat distribution outcome measures: assessment and clinical implications. Obesity Research 2002; 10, (Suppl. 1): 14S21S.CrossRefGoogle ScholarPubMed
37Sánchez-Castillo, CP, Velázquez-Monroy, O, Berber, A, Lara-Esqueda, A, Tapia-Conyer, R, James, WP; Encuesta Nacional de Salud (ENSA) 2000 Working Group. Anthropometric cutoff points for predicting chronic diseases in the Mexican National Health Survey 2000. Obesity Research 2003; 11: 442–51.CrossRefGoogle Scholar
38Pitanga, FJG, Lessa, I. Indicadores antropométricos de obesidade como instrumento de triagem para risco coronariano elevado em adultos na cidade de Salvador–Bahia. Arquivos Brasileiros de Cardiologia 2005; 85: 2631.CrossRefGoogle Scholar
39Lemos-Santos, MGF, Valente, JG, Gonçalves-Silva, RMV, Schieri, R. Waist circumference and waist-to-hip ratio as predictors of serum concentration of lipids in Brazilian men. Nutrition 2004; 20: 857–62.CrossRefGoogle ScholarPubMed
40Taylor, RW, Jones, IE, Williams, SM, Goulding, A. Evaluation of waist circumference, waist-to-hip ratio, and the conicity index as screening tools for high trunk fat mass, as measured by dual-energy X-ray absorptiometry, in children aged 3–19 y. American Journal of Clinical Nutrition 2000; 72: 490–5.CrossRefGoogle ScholarPubMed
41Moreno, VM, Gandoy, JBG, González, MJA. Medición de la grasa corporal mediante impedancia bioeléctrica, pliegues cutáneos y ecuaciones a partir de medidas antropométricas, análisis comparativo. Revista Española de Salud Pública 2001; 75: 221–36.Google Scholar
42Peterson, MJ, Czerwinski, SA, Siervogel, RM. Development and validation of skinfold-thickness prediction equations with a 4-compartment model. American Journal of Clinical Nutrition 2003; 77: 1186–91.CrossRefGoogle ScholarPubMed
43Wagner, DR, Heyward, VH. Measures of body composition in blacks and whites: a comparative review. American Journal of Clinical Nutrition 2000; 71: 1392–402.CrossRefGoogle ScholarPubMed
44Bray, GA. What is the ideal body weight? Journal of Nutritional Biochemistry 1998; 9: 489–92.CrossRefGoogle Scholar
45Kyle, U, Bosaeus, I, De Lorenzo, AD, Deurenberg, P, Elia, M, Manuel Gomez, J, et al. .; ESPEN. Bioelectrical impedance analysis–part II: utilization in clinical practice. Clinical Nutrition 2004; 23: 1430–53.CrossRefGoogle ScholarPubMed
46Coppini, LZ, Waitzberg, DL, Campos, ACL. Limitations and validation of bioelectrical impedance analysis in morbidly obese patients. Current Opinion in Clinical Nutrition and Metabolic Care 2005; 8: 329–32.CrossRefGoogle ScholarPubMed
47Krupa, D. Body composition measurement in severe obesity. Current Opinion in Clinical Nutrition and Metabolic Care 2005; 8: 602–6.Google Scholar
48Gallagher, D, Kuznia, P, Heshka, S, Albu, J, Heymsfield, SB, Goodpaster, B, et al. . Adipose tissue in muscle: a novel depot similar in size to visceral adipose tissue. American Journal of Clinical Nutrition 2005; 81: 903–10.CrossRefGoogle ScholarPubMed
49Campolina, A, Melo, CA, Andrade, MG. Escravidão em Minas Gerais. Belo Horizonte: Secretaria do Estado da Cultura, Arquivo Público Mineiro, COPASA MG, 1988.Google Scholar
50Maio, MC, Monteiro, S, Chor, D, Faerstein, E, Lopes, CS. Cor/raça no Estudo Pró-Saúde: resultados comparativos de dois métodos de autoclassificação no Rio de Janeiro, Brasil. Cadernos de Saúde Pública 2005; 21: 171–80.CrossRefGoogle Scholar
51Hugles, VA, Frontera, WR, Roubenoff, R, Evans, WJ, Singh, MAF. Longitudinal changes in body composition in older men and women: role of body weight change and physical activity. American Journal of Clinical Nutrition 2002; 76: 473–81.Google Scholar
You have Access

Send article to Kindle

To send this article to your Kindle, first ensure is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about sending to your Kindle. Find out more about sending to your Kindle.

Note you can select to send to either the or variations. ‘’ emails are free but can only be sent to your device when it is connected to wi-fi. ‘’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

Find out more about the Kindle Personal Document Service.

A comparative study of methods for diagnosis of obesity in an urban mixed-race population in Minas Gerais, Brazil
Available formats

Send article to Dropbox

To send this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Dropbox.

A comparative study of methods for diagnosis of obesity in an urban mixed-race population in Minas Gerais, Brazil
Available formats

Send article to Google Drive

To send this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Google Drive.

A comparative study of methods for diagnosis of obesity in an urban mixed-race population in Minas Gerais, Brazil
Available formats

Reply to: Submit a response

Your details

Conflicting interests

Do you have any conflicting interests? *