- Email: [email protected]
Predictive Validity of Four Bioelectrical Impedance Equations in Determining Percent Fat Mass in Overweight and Obese Children
RESEARCH Research and Professional Briefs
Predictive Validity of Four Bioelectrical Impedance Equations in Determining Percent Fat Mass in Overweight and Obese Children JANE CLEARY, MSc; SUZIE DANIELLS, MSc; ANTHONY D. OKELY, EdD; MARIJKA BATTERHAM, PhD; JESSIE NICHOLLS
ABSTRACT Bioelectrical impedance equations are frequently used by food and nutrition professionals to estimate percent fat mass in overweight and obese children. However, it is not known whether they are accurate for such children, as they have been primarily developed for children of varying body weights. The aim of this cross-sectional study was to evaluate the predictive validity of four previously published prediction equations developed for the pediatric population, among a sample of overweight and obese children. Thirty overweight or obese children (mean age⫽7.57⫾1.28 years) underwent measurement of fat mass, percent fat mass, and fat-free mass using dual-energy x-ray absorptiometry (DEXA) and bioelectrical impedance analysis (BIA). Impedance values from the BIA were entered into the four prediction equations and Pearson correlations used to determine the significance of associations between each of the BIA prediction equations and DEXA for percent fat mass, fat mass, and fat-free mass. For percent fat mass, paired t tests were used to assess differences between the methods and the technique of Bland and Altman was used to determine bias and error. Results showed that the mean percent fat mass as determined by DEXA for this age group was 40.79%. In comparison with other BIA prediction equations, the Schaefer equation had the closest mean value of 41.98%, and was the only equation not to significantly differ from the DEXA (P⫽0.121). This study suggests that the Schaefer J. Cleary is a senior pediatric dietitian and J. Nicholls is a dietitian, Department of Clinical Nutrition, Wollongong Hospital, Wollongong, NSW, Australia. S. Daniells is a dietitian and deputy manager, Department of Nutrition & Dietetics, Prince of Wales Hospital, Randwick, NSW, Australia; at the time of the study, she was a pediatric dietitian, Department of Clinical Nutrition, Wollongong Hospital, Wollongong, NSW, Australia. A.D. Okely is senior lecturer, Faculty of Education and director, Child Obesity Research Centre, University of Wollongong, NSW, Australia. M. Batterham is a senior lecturer, School of Health Sciences, University of Wollongong, Wollongong, NSW, Australia. Address correspondence to: Jane Cleary, MSc, Department of Clinical Nutrition, Wollongong Hospital, Crown Street, Wollongong NSW 2500, Australia. Copyright © 2008 by the American Dietetic Association. 0002-8223/08/10801-0013$34.00/0 doi: 10.1016/j.jada.2007.10.004
136
Journal of the AMERICAN DIETETIC ASSOCIATION
equation is the only accurate BIA prediction equation for assessing percent fat mass in this sample of overweight and obese children from primarily white backgrounds. J Am Diet Assoc. 2008;108:136-139.
C
urrent estimates are that one in five children in the developed world is overweight or obese and that this has increased between 0.5% and 1% per year over the past 2 decades (1). With such a high and increasing prevalence, it is important that ways of effectively managing overweight and obesity are found. To do this, researchers and practitioners need to have access to precise and yet practical instruments to measure body composition in overweight and obese children. A variety of methods have been used to assess the percent fat mass in children. These include body mass index (BMI; calculated as kg/m2), skinfold thickness, magnetic resonance imaging, computerized tomography, air displacement plethysmography, dual-energy x-ray absorptiometry (DEXA), isotope dilution, underwater weighing, and bioelectrical impedance (2). Due to cost, portability, ease of use, and other advantages, such as no exposure to radioactivity or underwater submersion, bioelectrical impedance analysis (BIA) has become a method of choice in field and clinical settings involving children (3). Accordingly, a number of equations have been developed and published to convert output from BIA (resistance, reactance, impedance) into percent fat mass, fat mass, and fat-free mass among children: Deurenberg and colleagues (4,5): Fat-free mass (kg)⫽0.64⫻(height [cm]2/impedance)⫹4.83 (Deurenberg 1) Fat-free mass (kg)⫽0.406⫻(height2/impedance)⫹(0.36⫻ weight [kg])⫹(0.56⫻sex)⫹(0.0558⫻height [cm])⫺6.5 (Deurenberg 2) Schaefer and colleagues (6): Fat-free mass (kg)⫽0.65⫻(height [cm]2/impedance)⫹ (0.68⫻age [y])⫹0.15 Houtkooper and colleagues (7): Fat-free mass (kg)⫽0.61 (height [cm]2/impedance)⫹ (0.25⫻weight [kg])⫹1.31 While these equations have been developed and validated among children of varying body weights, they have not been simultaneously cross-validated in an independent sample of overweight and obese children. Given that
© 2008 by the American Dietetic Association
prediction equations derived from BIA tend to underestimate percent fat mass in overweight and obese children (8) and that BIA is commonly used by food and nutrition professionals to assess body composition, it was the purpose of this study to determine which of these BIA equations is most accurate in predicting percent fat mass in overweight and obese children when compared to DEXA. METHODS Participants Using a cross-sectional design, 33 overweight or obese children aged 5 to 9 years were recruited. To be included in the study, children needed to be prepubescent; not taking long-term steroids, antipsychotics, or other medications that can influence weight gain; and not have any of the following conditions: type 1 diabetes mellitus, Prader-Willi syndrome, celiac disease, cystic fibrosis, multiple food allergies, or a substantial physical or developmental disability. Overweight and obesity was defined by BMI (according to the International Obesity Task Force cutpoints for age and sex (9)). Children this age were selected because they had not yet entered puberty. Participants and their parents were approached and recruited from two sources: (a) those already recruited and enrolled in a randomized controlled trial of a communitybased weight management program for overweight and obese children, and (b) from the general pediatric nutrition outpatient clinic at Wollongong Hospital. The study was approved by the University of Wollongong Human Research Ethics Committee and all parents or caregivers of participants provided written informed consent. Child consent was acquired verbally but was not required by ethics as they were younger than 10 years of age. Body Composition Assessment Weight, height, and body composition assessments occurred on-site at Wollongong Hospital, Department of Nuclear Medicine. Height was measured to 0.1 cm using a freestanding (mounted to scales) stadiometer (Seca 220 model, Seca, Hamburg, Germany) and weight was measured to 0.1 kg using electronic digital scales (Tanita BWB 600 model, Tanita Corporation, Tokyo, Japan). All measures took approximately 15 minutes and were obtained by trained operators in the early morning after an overnight fast. BIA Impedance (50 kHz) was measured by BIA (Bodystat 1500 body composition monitoring unit, Bodystat Ltd, Douglas, Isle of Man). Each child was instructed to remove his or her left shoe and sock, and lie supine with arms and legs away from the body. Alcohol was used to clean the skin on the left hand and foot where the electrode would be placed. The BIA was measured according to recommended protocol (10). Output from the BIA was transferred to Bodystat 1500 computer software (Bodystat1500 Body Manager, version 3.16, 2002, Bodystat Ltd) for the application of the default child-specific regression equation which then estimated percent fat mass, fat mass (kg), and fat-free mass (kg) from total body water. The technical manual for the Bodystat 1500 indicates that Houtkooper and colleagues’
equation (7) is the default equation. Impedance and other relevant variables (such as height, weight, age, and sex) were then entered into the other three child-specific estimation prediction equations and percent fat mass, fat mass, and fat-free mass were calculated. These equations (Deurenberg 1, Deurenberg 2, Schaffer equations) were selected for the study for two reasons: First, they were developed in a child population; and second, they calculated fat-free mass based on impedance, rather than resistance and reactance, because the BIA measuring unit used produced only a single impedance value. DEXA The reference method in this study was performed by a trained technician using DEXA (Hologic QDR 4500 whole body scanner, Hologic Inc, Bedford, MA). Participants were asked to remove any metal objects, such as zippers, snaps, and belts. Participants lay supine with their arms at their side, a block was placed between their feet to hold them in an inverted position, and they were required to lay motionless during the scan. Output from the DEXA scan was analyzed through pediatric DEXA computer software (Hologic Inc, version 12.3, Pediatric Whole body 2004) to produce percent fat mass, fat mass, and fat-free mass values. Statistical Analyses In terms of sample size, a difference ⬎2% with a standard deviation of 4% was determined as the upper limit of clinical acceptability. Sample size calculations indicated 34 participants would be required to prevent a type II error. All statistical analyses were undertaken using SPSS statistical package (version 12.0.1, 1998-99, SPSS Inc, Chicago IL). Data were checked for normality using a Shapiro-Wilk test. Paired t tests and Pearson correlation coefficients were calculated for data that were normally distributed. Wilcoxon signed rank tests and Spearman’s were used when data were not normally distributed. The technique of Bland and Altman (11) was utilized to determine the similarity between assessments of DEXA and BIA and to calculate the mean bias and the limits of agreement. RESULTS AND DISCUSSION Three of the 33 children were excluded due to technical faults with the BIA machine (n⫽2) or breaching fasting criteria (n⫽1), resulting in 30 children (91%; 12 boys, 18 girls) included in the analyses. Mean age was 7.57⫾1.28 years, mean weight 43.78⫾10.88 kg, and mean BMI 22.80⫾3.47. Girls had greater mean weights and BMI values (44.65⫾9.59 kg and 23.51⫾2.79, respectively) compared with boys (42.48⫾12.92 kg and 21.75⫾4.20, respectively). Conversely, boys were slightly taller than girls (138.31⫾12.74 cm, and 137.11⫾9.21 cm, respectively). None of the differences between boys and girls were statistically significant. Table 1 shows the comparison of the BIA prediction equations with DEXA. Each equation showed a highly significant correlation with the DEXA for percent fat mass, fat mass, and fat-free mass. BIA-based mean values of percent fat mass all varied in differing degrees from those of DEXA (40.79%⫾7.94%). The Houtkooper and Deurenberg 2 equations both showed a mean under-
January 2008 ● Journal of the AMERICAN DIETETIC ASSOCIATION
137
— 0.966 0.965 0.939 0.938 0.001c ⬍0.000c ⬍0.000c 0.010 25.77⫾5.97 26.67⫾5.86 24.02⫾4.06 30.95⫾7.12 24.78⫾4.77
Equationa BIA BIA BIA BIA
Error Error (ⴙ2 SD)c (ⴚ2 SD)c P valued
Houtkooper (7) ⫺2.62 3.51 Deurenberg 1 (5) 2.70 11.15 Deurenberg 2 (6) ⫺12.10 ⫺4.93 Schaefer (7) 1.18 9.30
⫺8.75 ⫺5.74 ⫺19.27 ⫺6.93
⬍0.000 0.001 ⬍0.000 0.121
Predictive equation used to assess percent fat mass using bioelectrical impedance analysis (BIA). b Bias refers to the mean of the differences between the field method (BIA) and the reference method, dual energy x-ray absorptiometry (11). c Error refers to the 2 standard deviations (SD) of the difference between the two methods and reflects the precision of the technique (11). d Two-tailed statistical significance from the paired t test for the difference within-group for the individual method compared with dual energy x-ray absorptiometry.
— ⫺3.569 ⫺2.561 ⫺4.237 ⫺1.635 18.31⫾6.39 17.14⫾6.16 19.77⫾7.74 12.83⫾4.47 19.00⫾7.38 b
a
DEXA⫽dual energy x-ray absorptiometry. SD⫽standard deviation. c Indicates a significant difference between method and DEXA (P⬎0.05).
— 0.856 0.834 0.836 0.835 ⬍0.000c 0.001c ⬍0.000c 0.121 — ⫺4.675 ⫺3.508 18.483 ⫺1.597 40.79⫾7.94 38.18⫾7.45 43.50⫾9.23 28.69⫾5.56 41.98⫾9.23 DEXAa Houtkooper (7) Deurenberg 1 (5) Deurenberg 2 (6) Schaefer (7)
January 2008 Volume 108 Number 1
Biasb
a
⬍0.000c 0.010c ⬍0.000c 0.102
— 0.972 0.961 0.971 0.966
— 3.664 4.237 13.131 2.765c
Correlation coefficient P value t Value MeanⴞSD Correlation coefficient P value MeanⴞSDb Method
t Value
P value
Correlation coefficient
MeanⴞSD
z Score
Fat Mass Fat-Free Mass %Fat Mass
Table 1. A comparison of four different bioelectrical impedance analysis equations vs dual-energy x-ray absorptiometry for percent fat mass, fat mass, and fat-free mass in white children aged 5 to 9 years 138
Table 2. Mean differences between four predictive equations for bioelectrical impedance analysis and the reference method, dual energy x-ray absorptiometry, in white children aged 5 to 9 years
estimation of percent fat mass, whereas the Deurenberg 1 and Schaefer equations showed a mean overestimation of percent fat mass. The Schaefer equation provided the most accurate estimate of percent fat mass. Other studies have shown this equation to be accurate when used with children (12,13), and the same finding in this study can be attributed to the close similarity in age and weight of the participants. Schaefer also included young participants (3.9 to 19.3 years) with a similar mean weight of 42.8 kg. The Schaefer equation was also the only equation tested that included age as one of its predictor variables. Inclusion of age was found to reduce errors by 18%. Although all BIA equations studied had strong correlations with DEXA for determining percent fat mass, it is the bias and limits of agreement that allow comparison of different techniques. Bias data (Table 2) highlights the limits of applicability for BIA in determining fat mass in overweight and obese children. This can then be used to judge accuracy for use of BIA in clinical settings. A small systematic overestimation of fat mass can be seen in the Schaefer prediction equation bias value, as it produced a small positive result (Table 2) and two main outliers can account for this. The equation itself appears to have the greatest strength for producing valid results in children with fat mass of 35% to 50%, even though one of the outliers falls within this average fat mass region. These data do not allow extrapolation of a sound conclusion above and below these fat masses, as the power of the study is held in the clustering of most of the participants in this region. Thirty-three subjects were recruited and the data for three were unable to be used; therefore, the possibility of a type II error cannot be excluded. The bias estimate with the 30 subjects was smaller than the limit for the study’s sample size calculation (1.18%⫾4.06%). Bland and Altman (11) emphasize that decisions of acceptable bias should focus on clinical rather than statistical assessment. Therefore, a sample of 30 giving a small bias with the Schaeffer and larger and statistically significant bias with the other equations provides evidence that the Schaefer equation can be used in clinical practice. The Deurenberg 1 equation tended to substantially
overestimate fat mass. While the equation was developed for a similarly aged study population of 7- to 9-year-olds, the fact that it was comprised of predominantly nonoverweight participants (mean weight⫽28.3 kg, compared with 43.78 kg in this study) is probably the reason for the overestimation. This equation cannot be appropriate for overweight and obese prepubertal children. The Deurenberg 2 equation, developed in a child population ranging up to 15 years of age, produced highly inaccurate results with an error of 18.48% (Table 1), possibly due to the exclusion of age as a variable. If this equation was used in a large study the true level of body fat would be substantially underestimated. This finding is consistent with those of Deurenberg, who has previously reported that child equations developed in normal pediatric populations underestimate fat mass in overweight and obese children (14). Deurenberg attributed this to differences in body geometry, body water distribution, and hydration of the lean mass compared to the fat mass in children. DEXA was used as the reference method in the present study because it was accessible and the equipment is found in most hospitals in Australia. Use of DEXA as a gold standard technique is controversial because of software differences and assumptions made with hydration factors (15). However, extensive validation studies with other criterion body composition assessment methods including hydrodensitometry have found the magnitude of errors to be small. As a result, DEXA has been found to be acceptable for use as a reference method in overweight and obese children (1,2). Limitations with the study are the use of a single frequency BIA machine, which does not have the capacity to measure impedance at varying intensities. Use of a multifrequency BIA will provide as many values of impedance as there are frequencies and this can provide a more accurate estimation of fat-free mass. The production of only one impedance measure limits the choice of BIA prediction equations used. This excludes use of other prediction equations that include resistance and reactance and might be just as accurate in the prediction of fat-free mass. CONCLUSIONS In this study it was found that the Schaefer equation was the most valid predictor of percent fat mass, fat mass, and fat-free mass in overweight and obese white children aged 5 to 9 years. The poor validity of three of the four BIA prediction equations underscores the necessity of cross-validating equations developed in normal populations, before they are deemed to be applicable for overweight or obese children and other pediatric populations with abnormal body composition (16-18). Children with acquired immunodeficiency syndrome who are taking treatment drugs that cause lipodystrophy, or obese children with spina bifida, are examples of children with abnormal body composition. This study would not allow a conclusion that the Schaefer equation would be valid for such children. The use of validated BIA results in the clinical setting still needs to be interpreted with the knowledge that they will not be exact measurements, and that some variability between BIA and reference methods will exist. It remains to be tested whether the Schaefer equation is sensitive to changes in adiposity over time in this population.
No funding was received by the research team. All costs of the research were absorbed by the South East Sydney and Illawarra Area Health Service. The authors would like to acknowledge Ambrer Nowland and the Department of Nuclear Medicine, Wollongong Hospital, NSW Australia for the provision of equipment and expertise for this study. Participants were recruited through a National Health and Medical Research Council of Australia funded project (354101)—the Hunter Illawarra Kids Challenge Using Parent Support (HIKCUPS). The authors also thank the research team for their support during participant recruitment. References 1. Lobstein T, Baur L, Uauy R. Obesity in children and young people: A crisis in public health. Obes Rev. 2004;5:4-104. 2. Goran MI. Measurement issues related to studies of childhood obesity: Assessment of body composition, body fat distribution, physical activity, and food intake. Pediatrics. 1998;101:505-518. 3. Newton RL, Alfonso A, White MA, York-Crowe E, Walden H, Ryan D, Bray GA, Williamson D. Percent body fat measured by BIA and DXA in obese, African-American adolescent girls. Int J Obes Relat Metab Disord. 2005;29: 560-594. 4. Deurenberg P, Van Der Kooy K, Leenan R, Weststrate JA, Seidell JC. Sex and age specific prediction formulas for estimating body composition from bioelectrical impedance: A cross validation study. Int J Obes Relat Metab Disord. 1991;15:17-25. 5. Deurenberg P, Kusters CSL, Smit H. Assessment of body composition by bioelectrical impedance in children and young adults is strongly age-dependent. Eur J Clin Nutr. 1989;44:261-268. 6. Schaefer F, Georgi A, Zieger A, Scharer K. Usefulness of bioelectric impedance and skinfold measurements in predicting fat-free mass derived from total body potassium in children. Pediatr Res. 1994;35:617-624. 7. Houtkooper LB, Going SB, Lohman TG, Roche AF, Van Loan M. Bioelectrical impedance estimation of fat free body mass in children and youth: A cross validation study. J Appl Physiol. 1992;72:366-373. 8. Okasura K, Takaya R, Tokuda M, Fukunaga Y, Oguni T, Tanaka H, Konishi K, Tamai H. Comparison of bioelectrical impedance analysis and dual energy X-ray absorptiometry for assessment of body composition in children. Pediatr Int. 1999;69:904-912. 9. Cole T, Bellizzi M, Flegal K, Dietz W. Establishing a standard definition for child overweight and obesity worldwide:International survey. BMJ. 2000;320:1240-1243. 10. Kyle UG, Bosaeus I, DeLorenzo A, Deurenberg P, Elia M, Gomez JM, Heitmann BL, Kent-Smith, L, Melchior JC, Pirlich M, Scharfetter H, Schols AMWJ, Pichard C. Bioelectrical impedance analysis—Part II: Utilization in clinical practice. Clin Nutr. 2004;23:1430-1453. 11. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307-310. 12. Schaefer F, Wuhl E, Feneberg R, Mehls O, Scharer K. Assessment of body composition in children with chronic renal failure. Pediatr Nephrol. 2000;49:33-36. 13. Schaefer F, Georgi M, Wuhl E, Scharer K. Body mass index and percentage fat mass in healthy German School children and adolescents. Int J Obes Relat Metab Disord. 1998;22:461-469. 14. Deurenberg P. Limitations of the bioelectrical impedance method for the assessment of body fat in severe obesity. Am J Clin Nutr. 1996; 64(suppl):449S-452S. 15. Mazess RB, Barden HS, Bisek JP, Hanson J. Dual energy x-ray absorptiometry for total-body and regional bone-mineral and softtissue composition. Am J Clin Nutr. 1990;51:1106-111. 16. Houtkooper LB, Lohman TG, Going SB, Hall MC. Validity of bioelectrical impedance for body composition assessment in children. J Appl Physiol. 1989;66:814-821. 17. Horlick M, Arpadi SM, Bethel J, Wang J, Moye J Jr, Cuff P, Pierson Jr RN, Kotler D. Bioelectrical impedance analysis models for prediction of total body water and fat-free mass in healthy and HIV-infected children and adolescents. Am J Clin Nutr. 2002;76:991-999. 18. Wabitsch M, Braun U, Heinze E, Muche R, Mayer H, Teller W, Fusch C. Body composition in 5-18-y-old obese children and adolescents before and after weight reduction as assessed by deuterium dilution and bioelectrical impedance analysis. Am J Clin Nutr. 1996;64:1-6.
January 2008 ● Journal of the AMERICAN DIETETIC ASSOCIATION
139
Predictive Validity of Four Bioelectrical Impedance Equations in Determining Percent Fat Mass in Overweight and Obese Children JANE CLEARY, MSc; SUZIE DANIELLS, MSc; ANTHONY D. OKELY, EdD; MARIJKA BATTERHAM, PhD; JESSIE NICHOLLS
ABSTRACT Bioelectrical impedance equations are frequently used by food and nutrition professionals to estimate percent fat mass in overweight and obese children. However, it is not known whether they are accurate for such children, as they have been primarily developed for children of varying body weights. The aim of this cross-sectional study was to evaluate the predictive validity of four previously published prediction equations developed for the pediatric population, among a sample of overweight and obese children. Thirty overweight or obese children (mean age⫽7.57⫾1.28 years) underwent measurement of fat mass, percent fat mass, and fat-free mass using dual-energy x-ray absorptiometry (DEXA) and bioelectrical impedance analysis (BIA). Impedance values from the BIA were entered into the four prediction equations and Pearson correlations used to determine the significance of associations between each of the BIA prediction equations and DEXA for percent fat mass, fat mass, and fat-free mass. For percent fat mass, paired t tests were used to assess differences between the methods and the technique of Bland and Altman was used to determine bias and error. Results showed that the mean percent fat mass as determined by DEXA for this age group was 40.79%. In comparison with other BIA prediction equations, the Schaefer equation had the closest mean value of 41.98%, and was the only equation not to significantly differ from the DEXA (P⫽0.121). This study suggests that the Schaefer J. Cleary is a senior pediatric dietitian and J. Nicholls is a dietitian, Department of Clinical Nutrition, Wollongong Hospital, Wollongong, NSW, Australia. S. Daniells is a dietitian and deputy manager, Department of Nutrition & Dietetics, Prince of Wales Hospital, Randwick, NSW, Australia; at the time of the study, she was a pediatric dietitian, Department of Clinical Nutrition, Wollongong Hospital, Wollongong, NSW, Australia. A.D. Okely is senior lecturer, Faculty of Education and director, Child Obesity Research Centre, University of Wollongong, NSW, Australia. M. Batterham is a senior lecturer, School of Health Sciences, University of Wollongong, Wollongong, NSW, Australia. Address correspondence to: Jane Cleary, MSc, Department of Clinical Nutrition, Wollongong Hospital, Crown Street, Wollongong NSW 2500, Australia. Copyright © 2008 by the American Dietetic Association. 0002-8223/08/10801-0013$34.00/0 doi: 10.1016/j.jada.2007.10.004
136
Journal of the AMERICAN DIETETIC ASSOCIATION
equation is the only accurate BIA prediction equation for assessing percent fat mass in this sample of overweight and obese children from primarily white backgrounds. J Am Diet Assoc. 2008;108:136-139.
C
urrent estimates are that one in five children in the developed world is overweight or obese and that this has increased between 0.5% and 1% per year over the past 2 decades (1). With such a high and increasing prevalence, it is important that ways of effectively managing overweight and obesity are found. To do this, researchers and practitioners need to have access to precise and yet practical instruments to measure body composition in overweight and obese children. A variety of methods have been used to assess the percent fat mass in children. These include body mass index (BMI; calculated as kg/m2), skinfold thickness, magnetic resonance imaging, computerized tomography, air displacement plethysmography, dual-energy x-ray absorptiometry (DEXA), isotope dilution, underwater weighing, and bioelectrical impedance (2). Due to cost, portability, ease of use, and other advantages, such as no exposure to radioactivity or underwater submersion, bioelectrical impedance analysis (BIA) has become a method of choice in field and clinical settings involving children (3). Accordingly, a number of equations have been developed and published to convert output from BIA (resistance, reactance, impedance) into percent fat mass, fat mass, and fat-free mass among children: Deurenberg and colleagues (4,5): Fat-free mass (kg)⫽0.64⫻(height [cm]2/impedance)⫹4.83 (Deurenberg 1) Fat-free mass (kg)⫽0.406⫻(height2/impedance)⫹(0.36⫻ weight [kg])⫹(0.56⫻sex)⫹(0.0558⫻height [cm])⫺6.5 (Deurenberg 2) Schaefer and colleagues (6): Fat-free mass (kg)⫽0.65⫻(height [cm]2/impedance)⫹ (0.68⫻age [y])⫹0.15 Houtkooper and colleagues (7): Fat-free mass (kg)⫽0.61 (height [cm]2/impedance)⫹ (0.25⫻weight [kg])⫹1.31 While these equations have been developed and validated among children of varying body weights, they have not been simultaneously cross-validated in an independent sample of overweight and obese children. Given that
© 2008 by the American Dietetic Association
prediction equations derived from BIA tend to underestimate percent fat mass in overweight and obese children (8) and that BIA is commonly used by food and nutrition professionals to assess body composition, it was the purpose of this study to determine which of these BIA equations is most accurate in predicting percent fat mass in overweight and obese children when compared to DEXA. METHODS Participants Using a cross-sectional design, 33 overweight or obese children aged 5 to 9 years were recruited. To be included in the study, children needed to be prepubescent; not taking long-term steroids, antipsychotics, or other medications that can influence weight gain; and not have any of the following conditions: type 1 diabetes mellitus, Prader-Willi syndrome, celiac disease, cystic fibrosis, multiple food allergies, or a substantial physical or developmental disability. Overweight and obesity was defined by BMI (according to the International Obesity Task Force cutpoints for age and sex (9)). Children this age were selected because they had not yet entered puberty. Participants and their parents were approached and recruited from two sources: (a) those already recruited and enrolled in a randomized controlled trial of a communitybased weight management program for overweight and obese children, and (b) from the general pediatric nutrition outpatient clinic at Wollongong Hospital. The study was approved by the University of Wollongong Human Research Ethics Committee and all parents or caregivers of participants provided written informed consent. Child consent was acquired verbally but was not required by ethics as they were younger than 10 years of age. Body Composition Assessment Weight, height, and body composition assessments occurred on-site at Wollongong Hospital, Department of Nuclear Medicine. Height was measured to 0.1 cm using a freestanding (mounted to scales) stadiometer (Seca 220 model, Seca, Hamburg, Germany) and weight was measured to 0.1 kg using electronic digital scales (Tanita BWB 600 model, Tanita Corporation, Tokyo, Japan). All measures took approximately 15 minutes and were obtained by trained operators in the early morning after an overnight fast. BIA Impedance (50 kHz) was measured by BIA (Bodystat 1500 body composition monitoring unit, Bodystat Ltd, Douglas, Isle of Man). Each child was instructed to remove his or her left shoe and sock, and lie supine with arms and legs away from the body. Alcohol was used to clean the skin on the left hand and foot where the electrode would be placed. The BIA was measured according to recommended protocol (10). Output from the BIA was transferred to Bodystat 1500 computer software (Bodystat1500 Body Manager, version 3.16, 2002, Bodystat Ltd) for the application of the default child-specific regression equation which then estimated percent fat mass, fat mass (kg), and fat-free mass (kg) from total body water. The technical manual for the Bodystat 1500 indicates that Houtkooper and colleagues’
equation (7) is the default equation. Impedance and other relevant variables (such as height, weight, age, and sex) were then entered into the other three child-specific estimation prediction equations and percent fat mass, fat mass, and fat-free mass were calculated. These equations (Deurenberg 1, Deurenberg 2, Schaffer equations) were selected for the study for two reasons: First, they were developed in a child population; and second, they calculated fat-free mass based on impedance, rather than resistance and reactance, because the BIA measuring unit used produced only a single impedance value. DEXA The reference method in this study was performed by a trained technician using DEXA (Hologic QDR 4500 whole body scanner, Hologic Inc, Bedford, MA). Participants were asked to remove any metal objects, such as zippers, snaps, and belts. Participants lay supine with their arms at their side, a block was placed between their feet to hold them in an inverted position, and they were required to lay motionless during the scan. Output from the DEXA scan was analyzed through pediatric DEXA computer software (Hologic Inc, version 12.3, Pediatric Whole body 2004) to produce percent fat mass, fat mass, and fat-free mass values. Statistical Analyses In terms of sample size, a difference ⬎2% with a standard deviation of 4% was determined as the upper limit of clinical acceptability. Sample size calculations indicated 34 participants would be required to prevent a type II error. All statistical analyses were undertaken using SPSS statistical package (version 12.0.1, 1998-99, SPSS Inc, Chicago IL). Data were checked for normality using a Shapiro-Wilk test. Paired t tests and Pearson correlation coefficients were calculated for data that were normally distributed. Wilcoxon signed rank tests and Spearman’s were used when data were not normally distributed. The technique of Bland and Altman (11) was utilized to determine the similarity between assessments of DEXA and BIA and to calculate the mean bias and the limits of agreement. RESULTS AND DISCUSSION Three of the 33 children were excluded due to technical faults with the BIA machine (n⫽2) or breaching fasting criteria (n⫽1), resulting in 30 children (91%; 12 boys, 18 girls) included in the analyses. Mean age was 7.57⫾1.28 years, mean weight 43.78⫾10.88 kg, and mean BMI 22.80⫾3.47. Girls had greater mean weights and BMI values (44.65⫾9.59 kg and 23.51⫾2.79, respectively) compared with boys (42.48⫾12.92 kg and 21.75⫾4.20, respectively). Conversely, boys were slightly taller than girls (138.31⫾12.74 cm, and 137.11⫾9.21 cm, respectively). None of the differences between boys and girls were statistically significant. Table 1 shows the comparison of the BIA prediction equations with DEXA. Each equation showed a highly significant correlation with the DEXA for percent fat mass, fat mass, and fat-free mass. BIA-based mean values of percent fat mass all varied in differing degrees from those of DEXA (40.79%⫾7.94%). The Houtkooper and Deurenberg 2 equations both showed a mean under-
January 2008 ● Journal of the AMERICAN DIETETIC ASSOCIATION
137
— 0.966 0.965 0.939 0.938 0.001c ⬍0.000c ⬍0.000c 0.010 25.77⫾5.97 26.67⫾5.86 24.02⫾4.06 30.95⫾7.12 24.78⫾4.77
Equationa BIA BIA BIA BIA
Error Error (ⴙ2 SD)c (ⴚ2 SD)c P valued
Houtkooper (7) ⫺2.62 3.51 Deurenberg 1 (5) 2.70 11.15 Deurenberg 2 (6) ⫺12.10 ⫺4.93 Schaefer (7) 1.18 9.30
⫺8.75 ⫺5.74 ⫺19.27 ⫺6.93
⬍0.000 0.001 ⬍0.000 0.121
Predictive equation used to assess percent fat mass using bioelectrical impedance analysis (BIA). b Bias refers to the mean of the differences between the field method (BIA) and the reference method, dual energy x-ray absorptiometry (11). c Error refers to the 2 standard deviations (SD) of the difference between the two methods and reflects the precision of the technique (11). d Two-tailed statistical significance from the paired t test for the difference within-group for the individual method compared with dual energy x-ray absorptiometry.
— ⫺3.569 ⫺2.561 ⫺4.237 ⫺1.635 18.31⫾6.39 17.14⫾6.16 19.77⫾7.74 12.83⫾4.47 19.00⫾7.38 b
a
DEXA⫽dual energy x-ray absorptiometry. SD⫽standard deviation. c Indicates a significant difference between method and DEXA (P⬎0.05).
— 0.856 0.834 0.836 0.835 ⬍0.000c 0.001c ⬍0.000c 0.121 — ⫺4.675 ⫺3.508 18.483 ⫺1.597 40.79⫾7.94 38.18⫾7.45 43.50⫾9.23 28.69⫾5.56 41.98⫾9.23 DEXAa Houtkooper (7) Deurenberg 1 (5) Deurenberg 2 (6) Schaefer (7)
January 2008 Volume 108 Number 1
Biasb
a
⬍0.000c 0.010c ⬍0.000c 0.102
— 0.972 0.961 0.971 0.966
— 3.664 4.237 13.131 2.765c
Correlation coefficient P value t Value MeanⴞSD Correlation coefficient P value MeanⴞSDb Method
t Value
P value
Correlation coefficient
MeanⴞSD
z Score
Fat Mass Fat-Free Mass %Fat Mass
Table 1. A comparison of four different bioelectrical impedance analysis equations vs dual-energy x-ray absorptiometry for percent fat mass, fat mass, and fat-free mass in white children aged 5 to 9 years 138
Table 2. Mean differences between four predictive equations for bioelectrical impedance analysis and the reference method, dual energy x-ray absorptiometry, in white children aged 5 to 9 years
estimation of percent fat mass, whereas the Deurenberg 1 and Schaefer equations showed a mean overestimation of percent fat mass. The Schaefer equation provided the most accurate estimate of percent fat mass. Other studies have shown this equation to be accurate when used with children (12,13), and the same finding in this study can be attributed to the close similarity in age and weight of the participants. Schaefer also included young participants (3.9 to 19.3 years) with a similar mean weight of 42.8 kg. The Schaefer equation was also the only equation tested that included age as one of its predictor variables. Inclusion of age was found to reduce errors by 18%. Although all BIA equations studied had strong correlations with DEXA for determining percent fat mass, it is the bias and limits of agreement that allow comparison of different techniques. Bias data (Table 2) highlights the limits of applicability for BIA in determining fat mass in overweight and obese children. This can then be used to judge accuracy for use of BIA in clinical settings. A small systematic overestimation of fat mass can be seen in the Schaefer prediction equation bias value, as it produced a small positive result (Table 2) and two main outliers can account for this. The equation itself appears to have the greatest strength for producing valid results in children with fat mass of 35% to 50%, even though one of the outliers falls within this average fat mass region. These data do not allow extrapolation of a sound conclusion above and below these fat masses, as the power of the study is held in the clustering of most of the participants in this region. Thirty-three subjects were recruited and the data for three were unable to be used; therefore, the possibility of a type II error cannot be excluded. The bias estimate with the 30 subjects was smaller than the limit for the study’s sample size calculation (1.18%⫾4.06%). Bland and Altman (11) emphasize that decisions of acceptable bias should focus on clinical rather than statistical assessment. Therefore, a sample of 30 giving a small bias with the Schaeffer and larger and statistically significant bias with the other equations provides evidence that the Schaefer equation can be used in clinical practice. The Deurenberg 1 equation tended to substantially
overestimate fat mass. While the equation was developed for a similarly aged study population of 7- to 9-year-olds, the fact that it was comprised of predominantly nonoverweight participants (mean weight⫽28.3 kg, compared with 43.78 kg in this study) is probably the reason for the overestimation. This equation cannot be appropriate for overweight and obese prepubertal children. The Deurenberg 2 equation, developed in a child population ranging up to 15 years of age, produced highly inaccurate results with an error of 18.48% (Table 1), possibly due to the exclusion of age as a variable. If this equation was used in a large study the true level of body fat would be substantially underestimated. This finding is consistent with those of Deurenberg, who has previously reported that child equations developed in normal pediatric populations underestimate fat mass in overweight and obese children (14). Deurenberg attributed this to differences in body geometry, body water distribution, and hydration of the lean mass compared to the fat mass in children. DEXA was used as the reference method in the present study because it was accessible and the equipment is found in most hospitals in Australia. Use of DEXA as a gold standard technique is controversial because of software differences and assumptions made with hydration factors (15). However, extensive validation studies with other criterion body composition assessment methods including hydrodensitometry have found the magnitude of errors to be small. As a result, DEXA has been found to be acceptable for use as a reference method in overweight and obese children (1,2). Limitations with the study are the use of a single frequency BIA machine, which does not have the capacity to measure impedance at varying intensities. Use of a multifrequency BIA will provide as many values of impedance as there are frequencies and this can provide a more accurate estimation of fat-free mass. The production of only one impedance measure limits the choice of BIA prediction equations used. This excludes use of other prediction equations that include resistance and reactance and might be just as accurate in the prediction of fat-free mass. CONCLUSIONS In this study it was found that the Schaefer equation was the most valid predictor of percent fat mass, fat mass, and fat-free mass in overweight and obese white children aged 5 to 9 years. The poor validity of three of the four BIA prediction equations underscores the necessity of cross-validating equations developed in normal populations, before they are deemed to be applicable for overweight or obese children and other pediatric populations with abnormal body composition (16-18). Children with acquired immunodeficiency syndrome who are taking treatment drugs that cause lipodystrophy, or obese children with spina bifida, are examples of children with abnormal body composition. This study would not allow a conclusion that the Schaefer equation would be valid for such children. The use of validated BIA results in the clinical setting still needs to be interpreted with the knowledge that they will not be exact measurements, and that some variability between BIA and reference methods will exist. It remains to be tested whether the Schaefer equation is sensitive to changes in adiposity over time in this population.
No funding was received by the research team. All costs of the research were absorbed by the South East Sydney and Illawarra Area Health Service. The authors would like to acknowledge Ambrer Nowland and the Department of Nuclear Medicine, Wollongong Hospital, NSW Australia for the provision of equipment and expertise for this study. Participants were recruited through a National Health and Medical Research Council of Australia funded project (354101)—the Hunter Illawarra Kids Challenge Using Parent Support (HIKCUPS). The authors also thank the research team for their support during participant recruitment. References 1. Lobstein T, Baur L, Uauy R. Obesity in children and young people: A crisis in public health. Obes Rev. 2004;5:4-104. 2. Goran MI. Measurement issues related to studies of childhood obesity: Assessment of body composition, body fat distribution, physical activity, and food intake. Pediatrics. 1998;101:505-518. 3. Newton RL, Alfonso A, White MA, York-Crowe E, Walden H, Ryan D, Bray GA, Williamson D. Percent body fat measured by BIA and DXA in obese, African-American adolescent girls. Int J Obes Relat Metab Disord. 2005;29: 560-594. 4. Deurenberg P, Van Der Kooy K, Leenan R, Weststrate JA, Seidell JC. Sex and age specific prediction formulas for estimating body composition from bioelectrical impedance: A cross validation study. Int J Obes Relat Metab Disord. 1991;15:17-25. 5. Deurenberg P, Kusters CSL, Smit H. Assessment of body composition by bioelectrical impedance in children and young adults is strongly age-dependent. Eur J Clin Nutr. 1989;44:261-268. 6. Schaefer F, Georgi A, Zieger A, Scharer K. Usefulness of bioelectric impedance and skinfold measurements in predicting fat-free mass derived from total body potassium in children. Pediatr Res. 1994;35:617-624. 7. Houtkooper LB, Going SB, Lohman TG, Roche AF, Van Loan M. Bioelectrical impedance estimation of fat free body mass in children and youth: A cross validation study. J Appl Physiol. 1992;72:366-373. 8. Okasura K, Takaya R, Tokuda M, Fukunaga Y, Oguni T, Tanaka H, Konishi K, Tamai H. Comparison of bioelectrical impedance analysis and dual energy X-ray absorptiometry for assessment of body composition in children. Pediatr Int. 1999;69:904-912. 9. Cole T, Bellizzi M, Flegal K, Dietz W. Establishing a standard definition for child overweight and obesity worldwide:International survey. BMJ. 2000;320:1240-1243. 10. Kyle UG, Bosaeus I, DeLorenzo A, Deurenberg P, Elia M, Gomez JM, Heitmann BL, Kent-Smith, L, Melchior JC, Pirlich M, Scharfetter H, Schols AMWJ, Pichard C. Bioelectrical impedance analysis—Part II: Utilization in clinical practice. Clin Nutr. 2004;23:1430-1453. 11. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307-310. 12. Schaefer F, Wuhl E, Feneberg R, Mehls O, Scharer K. Assessment of body composition in children with chronic renal failure. Pediatr Nephrol. 2000;49:33-36. 13. Schaefer F, Georgi M, Wuhl E, Scharer K. Body mass index and percentage fat mass in healthy German School children and adolescents. Int J Obes Relat Metab Disord. 1998;22:461-469. 14. Deurenberg P. Limitations of the bioelectrical impedance method for the assessment of body fat in severe obesity. Am J Clin Nutr. 1996; 64(suppl):449S-452S. 15. Mazess RB, Barden HS, Bisek JP, Hanson J. Dual energy x-ray absorptiometry for total-body and regional bone-mineral and softtissue composition. Am J Clin Nutr. 1990;51:1106-111. 16. Houtkooper LB, Lohman TG, Going SB, Hall MC. Validity of bioelectrical impedance for body composition assessment in children. J Appl Physiol. 1989;66:814-821. 17. Horlick M, Arpadi SM, Bethel J, Wang J, Moye J Jr, Cuff P, Pierson Jr RN, Kotler D. Bioelectrical impedance analysis models for prediction of total body water and fat-free mass in healthy and HIV-infected children and adolescents. Am J Clin Nutr. 2002;76:991-999. 18. Wabitsch M, Braun U, Heinze E, Muche R, Mayer H, Teller W, Fusch C. Body composition in 5-18-y-old obese children and adolescents before and after weight reduction as assessed by deuterium dilution and bioelectrical impedance analysis. Am J Clin Nutr. 1996;64:1-6.
January 2008 ● Journal of the AMERICAN DIETETIC ASSOCIATION
139