The prevalence of obesity, a major modifiable risk factor for cardiovascular disease that is strongly associated with comorbid conditions such as insulin resistance, type 2 diabetes, coronary artery disease, hypertension, stroke, and heart failure (7), has reached epidemic proportions in affluent Western societies and is associated with substantial health care costs (21). The prevalence of adolescent and childhood obesity, which commonly persists into adulthood (24) and predicts a broad range of long-term adverse health effects (22), is also increasing (7).
The pattern of fat distribution in obesity is an important factor in the risk of developing future cardiovascular disease (18,25). The presence of abdominal visceral obesity predisposes individuals to a cluster of metabolic complications, potentially leading to decreased insulin sensitivity, type 2 diabetes, and cardiovascular disease (5). Indeed, several studies have demonstrated significant associations between distribution of body fat in the abdominal viscera and glucose tolerance, plasma insulin levels, and plasma lipoprotein lipids (1,15,28,29). Despite these studies, and the importance of primary prevention in decreasing obesity and limiting its socioeconomic and health impact (7), few studies have investigated the optimal measure of "obesity" for use in longitudinal intervention studies in obese children and adolescents, or the strength of association between different indices of obesity.
A variety of methods exist to quantify body composition and regional body fat distribution in humans. Indices of body weight such as body mass index (BMI) and total body weight are useful at a population level, but do not discriminate between lean mass and fat mass in an individual. Anthropometric measurement, such as skinfold thickness and segment girths, provide a more accurate assessment of the regional distribution of body fat than gross body weight indices; however, equations used to predict total body fat from subcutaneous skinfold measurement are based on adult cadaver studies of small sample sizes (26) and may not be relevant or valid in an obese pediatric cohort. Furthermore, skinfold measurements only provide an estimate of subcutaneous fat and cannot be used to reliably measure fat stored internally, such as visceral fat, which is more strongly associated with insulin resistance and cardiovascular disease. Despite the conclusion of Martin et al. (19) that, for a limited sample (N = 31) of elderly Belgian cadavers, visceral adipose tissue was strongly related to total body adiposity, trunk subcutaneous adipose tissue weight only explained 49 and 64% of the variance in visceral adipose tissue weight for male and female samples, respectively. Finally, and paradoxically, skinfold assessments may be least accurate in obese cohorts, due to the limitations of the calipers and techniques used.
Recently, the development of dual-energy x-ray absorptiometry (DEXA), a technique that measures whole and segmental body fat and lean body mass, has provided a potentially more sensitive technique of body composition assessment. Therefore, the aim of this study was to directly compare, within subjects, anthropometric and DEXA-based methods of body composition assessment in obese young subjects. Furthermore, the relationship between changes in these estimates of body composition with exercise training was investigated.
Subjects and Screening Measures
Thirty-eight obese subjects (21 male, 17 female), aged 12.7 ± 2.1 (SD) yr (Tanner stages 3-5), were recruited. Obesity was defined according to recently published criteria (4), which specify age- and gender-specific BMI cutoffs. Exclusion criteria included SBP > 140 or DBP > 85, previous cigarette smoking, total cholesterol > 5.5 mmol·L−1, or LDL-cholesterol > 3.0 mmol·L−1. These exclusion criteria were selected because our aim was to assess the effect of exercise training in obese subjects without concomitant CV risk factors. Consent was obtained from all children, adolescents, and their parents; parents provided written consent. Participation in the study was voluntary and subjects were free to withdraw at any time. The research protocol was approved by the Royal Perth Hospital ethics committee.
Subjects participated in an 8-wk exercise training program, and the anthropometric measures detailed below were assessed at entry and at 8 wk. Subjects were requested not to modify their dietary behaviors across the 8-wk program, and parental-report diet diaries confirmed this.
Body weight (Seca, model 770, Germany), height, and BMI were obtained before each exercise test. Skinfold thickness was measured by experienced anthropometrists using spring-loaded calipers (Harpenden, UK) at six standard sites: triceps, subscapular, iliac crest, abdominal, anterior thigh, and medial calf (2). In accordance with standard anthropometric practice (2,23), skinfold measurements were taken on the right side of the body to the nearest 0.1 mm, and all sites were measured in triplicate, with the median score recorded. The CV of these repeated skinfold measures was 0.7%. Waist and hip girth were similarly recorded using standard methodology (2). The same trained and experienced investigator performed all anthropometric measures for each subject.
The equation developed by Slaughter et al. (27) and the generalized skinfold prediction equations presented by the American College of Sports Medicine (16) were selected as being most applicable for use with subjects in this study. The latter equations utilize either three or four skinfolds, as presented below:
Slaughter equation (27) (triceps, calf)
Three-Site Formula (abdomen, iliac crest, triceps)
Four-Site Formula (abdomen, iliac crest, triceps, thigh)
Body composition assessment.
On a separate visit, a DEXA scan (Norland XR36 pencil-beam, CT) was obtained for the assessment of whole-body and regional fat and lean tissue mass. DEXA allows for more precise measurements of fat mass and fat-free mass, the latter being comprised of fat-free soft tissue and bone mineral content. DEXA involves the use of very low-dose x-rays, the effective dose from a total-body scan being approximately 0.4-2.0 μSv. The scanner was calibrated daily in a two-part process. The first used a calibration standard to test system diagnostics. The second part involved a scan of the quality control (QC) spine phantom to ascertain the precision and accuracy of the scanner. Reliability of the repeat DEXA measures was enhanced by ensuring that bone mineral content values for each body segment (head, upper limbs, lower limbs, thorax, abdomen), which would not be expected to change across the 8-wk training period, were statistically comparable before and after training. The CV associated with DEXA scanning is approximately 2.5%.
Exercise Training Regimen
The 8-wk regimen consisted of three 1-h sessions of whole-body exercise each week, concentrating on the large muscle groups. Exercise involved a combination of cycle ergometer and resistance training, the latter performed on weight-stack machines (Pulsestar, UK). Intensity and duration of the exercise program were progressively increased as individually tolerated. Cycle ergometry was maintained at 65-85% of maximum HR, and resistance training intensity was maintained at 55-70% of pretraining maximum voluntary contraction. We used circuit training in an attempt to combine the well-established cardiopulmonary fitness benefits of aerobic training with the body composition and lean body mass effects of resistance training. All group exercise sessions were supervised by experienced exercise physiologists, with a maximum subject to staff ratio of 6:1. Exercise intensity (HR and weights lifted) was maintained within the target zones by intermittent checking undertaken in all subjects throughout each session.
Analysis of Data
Results are expressed as means ± SE. Pearson product-moment correlation coefficients were initially performed to determine the degree of association between all anthropometric and body composition variables versus DEXA-derived total body fat and DEXA abdominal fat in the 38 complete data sets collected on subjects at entry to the study. Significant coefficients were reported at the 5% probability level. Subsequent analysis involved stepwise hierarchical linear regression to identify a group of independent variables that best predicted DEXA-derived total fat and, separately, DEXA abdominal fat. The measures that predicted the greatest amount of variance were then removed from the model to reveal subsequent strong predictive variables. In this way the regression model was used to identify those variables that were strongly predictive of DEXA body and abdominal fat scores, as against those that were in the model merely as "support" variables.
Student's paired t-tests were performed to assess the significance of difference between all data collected before and after exercise training, and correlation and regression analyses were repeated for the sample following the exercise intervention to determine the sensitivity of anthropometric and skinfold variables to changes in adiposity in this population.
All subjects completed at least 90% of the 24 exercise sessions, and no significant adverse events occurred.
Correlations between anthropometric methods and DEXA.
There were strong relationships (all P < 0.01) between DEXA total body fat and body weight (r = 0.83), BMI (r = 0.86), waist girth (r = 0.81), hip girth (r = 0.88), sum of six skinfolds (sum6, r = 0.79), and sum of four skinfolds (sum4, r = 0.69). Significant, though less strong, relationships were also evident between DEXA total body fat and the sum of three skinfolds, (sum3; r = 0.60; P < 0.01), percent body fat calculated using the three- and four-skinfold equations (EQ4, r = 0.61, P < 0.01; EQ3, r = 0.53; P < 0.01), as well as the Slaughter equation (r = 0.51, P < 0.01).
DEXA abdominal fat was correlated strongly with body weight (r = 0.79), waist girth (r = 0.71), hip girth (r = 0.69), and height (r = 0.71; all P < 0.01). Less strong relationships were also evident between DEXA abdominal fat and BMI (r = 0.61, P < 0.01), and skinfold sum (sum6, r = 0.44, P < 0.05).
Stepwise hierarchical linear regression analysis performed to determine the best model to predict DEXA total fat mass initially selected hip girth with support from EQ4 and waist:hip (Y = −108688.9 + 751.5 × hip girth + 566.4EQ4 + 49909.4 waist:hip; F = 49.36 P < 0.001, R2 = 0.851). When hip girth was removed from this model, BMI was then selected as the primary independent variable, with support from EQ4 and waist girth. This was associated with a consequent change in R2 to 0.844 (F = 53.2, P < 0.001). However, with hip girth and BMI both removed from the analysis, the final model selected body weight as the primary predictor, with EQ4 (includes sum4 skinfolds and age) and the Slaughter equation (includes triceps and calf skinfolds) as support variables, which together accounted for 85% of the variance.
Similar regression analysis to predict DEXA abdominal fat mass selected body weight alone in the first model (Y = 426.1 + 87.1 × weight; F = 44.8 P < 0.001, R2 = 0.615). With body weight excluded, the final model selected waist girth as the primary variable, with support from height. Some 59% of the variance in DEXA abdominal fat at baseline was predicted by this final equation.
Neither skinfold sums nor skinfold prediction equations were selected as primary predictors in any of the stepwise linear regression models. EQ4 was selected only as a "support variable" and was not a strongly predictive variable of DEXA total or abdominal fat. These data therefore indicate that, although skinfolds were correlated with DEXA measures, the relationships were limited and not strongly predictive.
Effects of Exercise Training
As a result of the 8-wk exercise training intervention, DEXA total body fat decreased by an average of 248 g (41.46 ± 1.86 vs 41.21 ± 1.85 kg, P = NS) and abdominal fat also decreased by an average of 175 g (8.14 ± 0.39 vs 7.96 ± 0.34 kg, see Table 1).
Waist girth significantly decreased with training (101.2 ± 1.8 vs 98.8 ± 2.0 cm, P < 0.05) as did the sum6 (256.8 ± 5.0 vs 239.7 ± 6.2 mm, P < 0.01), sum4 (178.6 ± 3.9 vs 166.9 ± 4.4 mm, P < 0.01), sum3 (129.7 ± 3.3 vs 123.8 ± 3.1 mm, P< 0.05), EQ4 (37.9 ± 1.0 vs 36.3 ± 1.0%, P < 0.001), and Slaughter (52.3 ± 1.8 vs 47.3 ± 1.9%, P < 0.01). However, no significant training effect was evident for total body mass or BMI; DEXA total lean mass increased significantly (43.0 ± 2.1 vs 43.7 ± 2.1%, P < 0.05). Further analysis revealed that changes in sum6, sum4, sum3, and consequently EQ4 and Slaughter, were primarily driven by a significant decrease in the thigh (48.8 ± 1.3 vs 43.1 ± 1.7 mm, P < 0.001) and tricep (37.6 ± 1.6 vs 33.5 ± 1.3 mm, P< 0.01) skinfolds; no significant training effects were evident in other individual skinfold measures. Furthermore, the sum of skinfold measures taken specifically from the "central" regions (abdominal, iliac crest, and subscapular skinfolds) did not change with training (133.7 ± 3.0 vs 130.0 ± 3.3 mm). These findings extend our previously published data, which emphasized the effects of training on vascular endothelial function (31). A small number (N = 6) of the previously studied subjects are included in the present analysis of 38 subjects, which, in contrast to the previous publication (31), provides sufficient power to allow analysis of the relationships between measures of body composition, not previously reported.
Correlations between Anthropometric Methods and DEXA
There were no significant correlations between the change in DEXA total fat mass and change in the sum of skinfolds (sum6, r = 0.04, sum4, r = 0.04, sum3, r = 0.05), or any of the skinfold prediction equations (EQ4, r = −0.02, EQ3, r = −0.05, Slaughter, r = 0.21). The training-induced reduction in DEXA abdominal fat mass was correlated with the sum of skinfolds (sum6, r = 0.34; sum4, r = 0.35, both P < 0.05), the skinfold prediction equations (EQ4, r = 0.36; EQ3, r = 0.37, P < 0.05), and skinfold measures taken specifically from the "central" regions (r = 0.39, P < 0.05), although all of these correlations were low.
Stepwise hierarchical linear regression analysis performed on the post- and pretraining delta scores indicated that no model was sufficiently powerful to predict the changes in either DEXA total fat mass or DEXA abdominal fat mass. That is, no measures of skinfolds or skinfold-based equations predicted changes in DEXA measures of total or abdominal fat mass.
This is the first study, to our knowledge, to directly compare anthropometric and skinfold methods to DEXA assessments of body composition in obese children and adolescents. The principal finding of the present study is that traditional indices of body fat derived from anthropometric measures were poorly predictive of abdominal and total body fat derived from DEXA in this cohort. Cross-sectional comparison of body fat measures collected at baseline indicated that measures of body weight, BMI, and waist and hip girth were more highly correlated with DEXA-derived total and abdominal fat than skinfold measures or skinfold-based equations. Furthermore, neither skinfold sums nor percent body fat calculated from skinfold equations were selected as primary independent predictors of DEXA total or abdominal fat by stepwise hierarchical linear regression. Similarly, changes in DEXA total and abdominal fat as a result of training were not predicted by changes in skinfolds or percent body fat calculated from skinfolds. Taken together, these data highlight the limitations, and question the validity, of the use of skinfolds to assess changes in body composition with interventions such as exercise training in obese subjects.
There is no "gold standard" for body composition analysis and, as previously stated, studies are essentially a comparison between different methodologies (10). However, evidence suggests that DEXA measures are highly correlated with other imaging-based methods of body composition assessment. A recent study compared DEXA abdominal fat mass measures to adipose tissue mass measured by computed tomography (CT) at L1-L4 vertebral levels, the most accurate method of measuring abdominal adipose tissue (11). This study concluded that DEXA was both a valid and reliable method to determine abdominal obesity, which was largely independent of operator error and sensitive to changes in abdominal fat (11). Furthermore, DEXA may be a preferable alternative to CT scanning, as the latter is expensive and multislice abdominal imaging exposes the patient to large effective doses of ionizing radiation (>5000 μSv) (30), compared with the dose associated with a DEXA total-body scan (<2μSv (J. Thwaites, University of Western Australia Radiation Protection Officer, personal communication, August 3, 2004)). These considerations effectively render repeated use of CT scanners in longitudinal studies of children inappropriate.
Other investigators have suggested that the four-compartment (4C) model, which incorporates independent measurement of the individual components of fat-free mass, is the gold standard for body composition assessment (10). However, this approach involves complex and time-consuming assessments of body density, the derivation of which is limited by several key assumptions (26), and assessment of whole-body water using radio-isotopes over a 6-h period. One study that compared DEXA measures to this 4C model concluded that the cost, time, and reliance on several different techniques rendered the 4C method unsuitable for wide-scale implementation, and that DEXA may be the most promising alternative method of body composition assessment because it is quick, easy to administer, and requires minimal subject compliance (10).
There are over 100 equations in the literature that predict body fat from skinfold measurements (20). In the present study, we selected a population-specific equation developed for use in children and youth (27), and two commonly used generalized equations (16). These generalized equations have been developed using data from several sources, and it has been suggested they can be substituted for several population-specific equations without a loss in predictive accuracy (17). Although this is the first study to directly compare skinfold measures and skinfold-based equations with DEXA measures in obese children and adolescents, previous studies have undertaken comparisons in heterogeneous groups of healthy children and adolescents. Our finding that both skinfold sums and percent body fat calculated from skinfold equations were poorly predictive of abdominal and total body fat derived from DEXA, are consistent with previous studies that suggest that skinfold equations may not accurately predict fat mass in nonobese children and adolescents. Indeed, several (6,8,12,14), but not all (9), studies suggest that estimates of percent body fat by skinfold measurement may be inaccurate in children and adolescents.
There may be various methodological and technical reasons for the poor predictive accuracy of the skinfold equations we used. Although one experienced anthropometrist measured skinfolds in triplicate throughout the full series of tests, in accordance with anthropometry accreditation guidelines (13), skinfold compressibility is inconsistent, even within individuals (3). In addition, the structural limitations of typically used calipers (e.g., Harpenden), which allow maximal diameters up to 60 mm, often do not accommodate true skinfold thicknesses in some obese subjects. Measurement error can therefore contribute significantly to the variability associated with skinfold assessments, and this may have contributed to the lack of association between changes in skinfold and DEXA data. Finally, many of the regression equations used to predict body fat from anthropometric measurements are ultimately based on assumptions related to dissection analysis of cadaver studies involving small sample sizes (26), which are not specific to obese pediatric subjects. These limitations to the use of skinfold equations may mask the real impact of interventions such as exercise training, especially as accumulating evidence indicates that gross measures of body composition may not change with exercise training, despite significant countervailing regional changes in fat and lean tissue mass (31). Exercise training is a stimulus that not only decreases adiposity, but also potentially increases skeletal muscle mass, and we would therefore strongly suggest that future interventions, in which body composition is a principal outcome measure, adopt more sophisticated measures of body composition than skinfolds, particularly in obese cohorts.
A final important finding from this study was that changes in skinfolds collected from the abdominal region did not predict changes in DEXA-derived abdominal fat mass, and the correlation between these measures was modest (r = 0.39). Visceral adiposity is strongly associated with insulin resistance, type 2 diabetes, and cardiovascular risk, and our findings have important implications for the use of subcutaneous skinfolds to predict changes in visceral fat.
In conclusion, this study indicates for the first time in obese adolescent subjects, that traditional indices of body fat derived from anthropometric measures are poorly predictive of abdominal and total body fat derived from DEXA. These data highlight the importance of comprehensive assessment of body composition in exercise training studies and question the validity of either total body mass or skinfolds to assess changes in body composition with exercise training interventions in an obese cohort.
This study was supported by a National Heart Foundation of Australia Project Grant.
1. Anderson, A. J., K. A. Sobocinski, D. S. Freedman, J. J. Barboriak, A. A. Rimm, and H. W. Gruchow. Body fat distribution, plasma lipids and lipoproteins. Arteriosclerosis
2. Bloomfield, J., T. R. Ackland, and B. C. Elliott. Applied Anatomy and Biomechanics in Sport
, Melbourne: Blackwell Scientific Publications, 1994, p. 310.
3. Clarys, J. P., A. D. Martin, D. T. Drinkwater, and M. J. Marfell-Jones. The skinfold: myth and reality. J. Sports Sci
. 5:3-33, 1987.
4. Cole, T. J., M. C. Bellizzi, K. M. Flegal, and W. M. Deitz. Establishing a standard definition for child overweight and obesity
worldwide; international survey. BMJ
5. Despres, J. P. Abdominal obesity
as important component of insulin-resistance syndrome. Nutrition
6. Dezenberg, C. V., T. R. Nagy, B. A. Gower, R. Johnson, and M. I. Goran. Predicting body composition
from anthropometry in pre-adolescent children. Int. J. Obes. Relat. Metab. Disord
. 23:253-259, 1999.
7. Eckel, R. H., and R. M. Krauss. American Heart Association call to action: obesity
as a major risk factor for coronary heart disease. Circulation
8. Elberg, J., J. R. McDuffie, N. G. Sebring, et al. Comparison of methods to assess change in children's body composition
. Am. J. Clin. Nutr
. 80:64-69, 2004.
9. Eliakim, A., S. Ish-Shalom, A. Giladi, B. Falk, and N. Constantini. Assessment of body composition
in ballet dancers: correlation among anthropometric measurements, bio-electrical impedance analysis, and dual-energy X-ray absorptiometry
. Int. J. Sports Med
. 21:598-601, 2000.
10. Gately, P. J., D. Radley, C. B. Cooke, et al. Comparison of body composition
methods in overweight and obese children. J. Appl. Physiol
. 95:2039-2046, 2003.
11. Glickman, S. G., C. S. Marn, M. A. Supiano, and D. R. Dengel. Validity and reliability of dual-energy X-ray absorptiometry
for the assessment of abdominal obesity
. J. Appl. Physiol
. 97:509-514, 2004.
12. Goran, M. I., P. Driscoll, R. Johnson, T. R. Nagy, and G. Hunter. Cross-calibration of body-composition techniques against dual-energy X-ray absorptiometry
in young children. Am. J. Clin. Nutr
. 63:299-305, 1996.
13. Gore, C., T. Olds, L. Carter, and K. Norton. Accreditation in Anthropometry
, Canberra: Australian Sports Commission, 1998, pp. 5-40.
14. Gutin, B., M. Litaker, S. Islam, T. Manos, C. Smith, and F. Treiber. Body-composition measurement in 9-11-y-old children by dual-energy X-ray absorptiometry
, skinfold-thickness measurements, and bioimpedance analysis. Am. J. Clin. Nutr
. 63:287-292, 1996.
15. Haffner, S. M., D. Fong, H. P. Hazuda, J. A. Pugh, and J. K. Patterson. Hyperinsulinemia, upper body adiposity and cardiovascular risk factors in non-diabetics. Metabolism: Clin. Exp
. 37:338-345, 1988.
16. Jackson, A. S., and M. L. Pollock. Practical assessment of body composition
. Physician Sports Med
. 13:76-90, 1985.
17. W. L. Kenney, R. H. Humphrey, and C. X. Bryant >(Eds.).> ACSM's Guidelines for Exercise Testing and Prescription
, 6 ed. Baltimore, MD: Williams and Wilkins, 1995, pp. 55-59.
18. Kissebah, A. H., and J. Krakower. Regional adiposity and morbidity. Physiol. Rev
. 74:761-811, 1994.
19. Martin, A. D., V. Janssens, D. Caboor, J. P. Clarys, and M. J. Marfell-Jones. Relationship between visceral, trunk and whole-body adipose tissue weights by cadaver dissection. Annals Human Biol
. 30:668-677, 2003.
20. Martin, A. D., W. D. Ross, D. T. Drinwater, and J. P. Clarys. Prediction of body fat by skinfold caliper: assumptions and cadaver evidence. Int. J. Obes
. 9:31-39, 1986.
21. Mokdad, A. H., B. A. Bowman, and E. S. Ford, et al. The continuing epidemics of obesity
and diabetes in the United States. JAMA
22. Must, A., P. F. Jacques, G. E. Dallal, C. Bajema, and W. Dietz. Long-term morbidity and mortality of overweight adolescents: a follow up of the Harvard Growth Study of 1922-1935. N. Engl. J. Med
. 327:1350-1355, 1992.
23. Norton, K., and T. Olds. Anthropometrica
, Sydney: University of New South Wales Press, 1996, p. 187.
24. Schonfeld-Warden, N., and C. H. Warden. Pediatric obesity
. An overview of etiology and treatment. Ped. Endocrin
. 44:339-360, 1997.
25. Serrano Rios, M. Relationship between obesity
and the increased risk of major complications in non-insulin-dependent diabetes mellitus. Eur. J. Clin. Invest
. 28:14-18, 1998.
26. Siri, W. E. Body composition
from fluid spaces and density: analysis of methods. Techniques for Measuring Body Composition
, Washington, DC: National Academy of Sciences, 1961, p. 244.
27. Slaughter, M. H., T. G. Lohman, R. A. Boileau, et al. Skinfold equations for the estimation of body fatness in children and youth. Human Biol
. 60:709-723, 1988.
28. Sparrow, D., G. A. Borkan, S. G. Gerzof, C. Wisniewski, and C.K. Silbert. Relationship of fat distribution to glucose tolerance. Results of computed tomography in male participants of the Normative Aging Study. Diabetes
29. Stern, M. P., and S. M. Haffner. Body fat distribution and hyperinsulinemiaas risk factors for diabetes and cardiovascular disease. Arteriosclerosis
30. United Nations Scientific Committee on the Effects of Atomic Radiation UNSCEAR 2000 Report to the General Assembly, with scientific annexes. Sources and effects of ionizing radiation. Volume 1 (Annex D), 2000, 374-377.
31. Watts, K., P. Beye, A. Siafarikas, et al. Exercise
training normalises vascular dysfunction and improves central adiposity in obese adolescents. J. Am. Coll. Cardiol
. 43:1823-1827, 2004.