Assessment of Change in Body Fat Percentage with DXA and... : Medicine & Science in Sports & Exercise (original) (raw)
BMI is a simple and cheap method for indirect assessment of fatness, but it cannot distinguish fat mass from fat-free mass, nor can it inform about fat distribution. In clinical obesity treatment programs, it is important to have detailed measures of body composition, especially when evaluating change. Examples of reference methods include dual x-ray absorptiometry (DXA), computed tomography, and magnetic resonance imaging (9). These technologies are accurate, but they are also expensive, and they inflict exposure to radiation. Cheaper methods, without radiation, also exist. One example is bioelectrical impedance (BIA), which is widely used in clinical settings. BIA has been found to agree well with reference methods in several studies (1,2,13,17,18), but it has also been found to provide overestimates (3,16,19) as well as underestimates (5,8,19,20) in some populations. The validity of BIA in the obese population is lower than in nonobese populations. A foot-to-foot, four-electrode BIA equipment has previously been shown to underestimate body fat percentage (%BF) in overweight or obese populations, or display larger discrepancies with increasing adiposity (4,6,11,19,20).
We have previously reported significant differences between DXA and an eight-electrode BIA equipment (BIA8) (14) for assessment of total and trunkal body fat percentage (%BF) in abdominally obese women (12). The discrepancy was, on average, 5.0%BF units. However, the main interest when assessing change over time may not be that the absolute %BF estimates are correct, but that the different methods agree in the relative change in %BF over time. Hereby, it is possible to establish the balance between losses in fat mass and fat-free mass for different intervention strategies. If the under- or overestimation is consistent between technologies, then there will be agreement in the change in %BF (Δ%BF) over time.
Therefore, the aim of this study was to assess the agreement between DXA and BIA8 regarding Δ%BF over 6 months in abdominally obese women participating in an exercise program (walking and bicycling). Furthermore, potential deviations between methods were explored by investigating whether the degree of discrepancy varies with the degree of obesity.
SUBJECTS AND METHODS
Data were collected from participants in an ongoing, randomized exercise intervention study evaluating methods for increasing lifestyle-oriented physical activity in unfit, abdominally obese women (defined as a waist circumference ≥ 88 cm, according to WHO criteria) (21). The exercise intervention consisted of two main activities: walking and bicycling. All participants were given a general recommendation to increase walking by 5000 steps per day above baseline levels. One group was randomized to minimum counseling (one 2-h session with a health educator just after baseline). The other group was randomized to receive a bicycle, and a physician prescription to bicycle to and from work, whenever appropriate. This group also received two additional 2-h group sessions with the health educator. The emphasis for all participants was on building a stable, daily physical activity routine that could be maintained long term.
The subjects were recruited from an advertisement in a free-of-charge newspaper distributed in the Stockholm area. The number of centrally obese (WC ≥ 88 cm) women who were screened before randomization was 139. During screening, three women did not complete DXA measurement and were excluded from analyses, resulting in complete data from 136 women. Of these, 124 were randomized, and two women dropped out before the start of the intervention. A baseline comparison of DXA and BIA8 has been published previously (12). At the 6-month follow-up visit, complete data on body composition were collected on 106 women. All measurements were carried out in a fasting condition on the same day with less than 2 h between the different body composition-measurement procedures.
Fatness measurements.
Body composition was estimated both by use of DXA and by BIA8. DXA measurements were performed by using a total-body scanner (Lunar Prodigy; GE Lunar Corporation; software version 8) employing fan beam and scanning at slow speed. The same two DXA operators measured the subjects lying in a supine position, dressed in underwear, without any metal items in their clothing or elsewhere. Fat-free soft tissue excluding bone (i.e., including muscle, body water, internal organs), fat soft tissue, and bone mineral densities were measured. Percent body fat (%BF) was calculated by dividing fat soft-tissue mass with entire body mass, by use of the software supplied by the manufacturer.
The BC-418 eight-contact electrode system (Tanita Corp., Tokyo, Japan) was used for BIA measurements. The system consists of four stainless-steel rectangular foot-pad electrodes fastened to a metal platform set on force transducers for weight measurement, and two handgrips with an anterior and posterior electrode. Hereby, the system has eight electrodes: two for each foot and two for each hand. Measurements are carried out at 50 kHz with a 0.8-mA sinus wave constant current, and the impedance across the tissues of the subjects is measured by receiver electrodes after the injector electrodes pass through an electrical signal. A research nurse entered information into the equipment of the subjects' age and height, and body type (all were classified as "standard body type," and none as "athletic"). BIA is a double indirect method used to measure body fat by first estimating total body water and thereafter converting this estimate to %BF by use of a prediction equation. The prediction equation supplied in the software was used in this study. The same operator performed all BIA measurements. The coefficient of variation (CV = SD/mean) was determined by scanning one person seven times on the same day, with repositioning between each scan. The CV was estimated to 2.1% and 2.2% for DXA and BIA, respectively.
The medical research ethics committee in Stockholm, Sweden, approved the study. All participants provided written informed consent.
Statistics.
Statistical analyses were conducted using SPSS (version 14.0; SPSS Inc., Chicago, IL) and Microsoft Excel, including the application Analyse-It for graphing the Bland-Altman plots. Summary statistics used for central tendency and dispersion are means and SD. Paired _t_-tests were used to compare estimates from the two methods. Methods used to assess agreement were Bland-Altman pairwise comparisons, Pearson's correlation coefficient, and regression analysis. The 95% limits of agreement for the mean differences in the Bland-Altman plots were also calculated. The association between degree of adiposity, as determined by DXA, and the discrepancy between DXA and BIA8 estimates, was investigated both unadjusted and adjusted for age and BMI by use of multiple regression analysis. Statistical significance was defined as P values < 0.05.
RESULTS
Subject characteristics.
Subject characteristics are presented in Table 1. The subjects ranged in age from 27 to 60 yr. During the 6 months of follow-up, they experienced significant reductions in body weight, BMI, and %BF, although of small magnitude (Table 1). There was also a significant change in obesity status for the group from baseline to follow-up (χ2 = 99.9; P < 0.0001); at baseline, 2/106 were normal weight (7 at follow-up), 46/106 were overweight (44 at follow-up), and 58/106 were obese (55 at follow-up).
Subject characteristics (N = 106).
Agreement between methods.
The agreement between methods is shown in Figure 1 by use of Bland-Altman plots. %BF differed significantly between the BIA8 and DXA estimates, both at baseline and at follow-up, and the Δ%BF also differed (Table 2). Compared with DXA, the BIA8 equipment underestimated fatness and fatness change. BIA significantly underestimated %BF by 5.0 (baseline) and 4.4 % (follow-up). The difference in Δ%BF was 0.6%BF. Indications of magnitude bias, with greater discrepancies between methods with greater adiposity levels, were found when investigating the association between degree of adiposity, as assessed by DXA, and the mean difference between the two methods (Fig. 2).
Bland-Altman plots with mean bias (central line) and 95% limits of agreement (outer lines) for baseline, follow-up, and change in body fat percentage (%BF; N = 106).
Scatterplot displaying the association between increasing degree of total adiposity and mean differences between DXA and BIA (Δ%BF = %BFDXA − %BFBIA; N = 106).
Mean differences between methods and 95% limits of agreement (N = 106).
In regression analysis, variation in %BFDXA explained 23.0% of the variation in differences between methods at baseline, 11.7% at follow-up, and 25.6% for Δ%BF (all P < 0.0001). Because BMI and age are input variables for the BIA8 equipment, these variables were added to the regression models, resulting in further strengthening of the association between degree of adiposity and discrepancy between methods (Table 3). The adjusted _R_2 improved, indicating better model fit, and they varied between 38.3% for the follow-up measurement to 48.7% for the difference in Δ%BF between DXA and BIA8. At both baseline and follow-up, each additional unit %BFDXA resulted in 0.5%BF unit greater underestimation by BIA8, after adjustment for BMI and age (P < 0.001) (Table 3). Correspondingly, for each additional one-unit reduction in %BFDXA over time, the underestimation of the change increased by 0.7%BF units.
Regression analyses of the difference between dual x-ray absorptiometry (DXA) and bioelectrical impedance (BIA) in measurements of total %BF using %BFDXA, age, and BMI as explanatory variables (N = 106).
DISCUSSION
We investigated the agreement between two body composition-measurement techniques in assessing change in %BF during 6 months in abdominally obese women. Compared with DXA, the BIA8 equipment significantly underestimated both the absolute %BF at baseline and follow-up, as well as changes in %BF. However, although the difference regarding change in %BF was statistically significant, it was of small magnitude. On the other hand, the discrepancies between methods were larger with greater adiposity or adiposity changes. This indicates that the BIA underestimation of both absolute fatness and fatness change will be even greater in the severely obese, or when weight losses are greater. Our data therefore indicate that BIA data on fatness change in obese people participating in weight loss programs have limitations.
Previously, we have reported on significant differences between estimates of %BF from DXA and BIA8 in this same sample at baseline, but with no data on their agreement regarding change over time (12). The results in this study are based on a smaller number of subjects (106 vs 136) because of dropout during follow-up. However, the baseline and follow-up differences were fairly consistent, indicating an absolute underestimation by BIA8 of approximately 5%BF units compared with DXA. The discrepancy between the two methods increased with greater adiposity, as has previously been shown for an older, four-electrode BIA equipment reported by Sun et al. (19). For each additional unit increase of %BFDXA, a further underestimation of approximately 0.5% by BIA8 was seen in the range of fatness under study, after adjustment for BMI and age.
It seems from the evidence presented here and elsewhere that the discrepancy between the methods is acceptable at a low degree of body fat (12,19). For those with a higher percentage body fat, however, the estimates were less accurate. Suggestions for improving fatness estimates include adjusting the body fat prediction equations or using more obese people in the validation studies. This could be implemented by adding a third option for the input of body type, where the current choices are either "standard" or "athletic," whereas no "obese" option exists. Until appropriate adjustments are made, it is probably wise to treat BIA estimates of fatness with caution when used on obese people.
The strengths of this study were the longitudinal design and relatively large sample size for the design including DXA, extending previous findings of method disagreement in obese people to assessment of changes in %BF (9,10,12,19). In addition, we examined a patient group that is likely to be targeted for more detailed body composition measures than BMI for screening and intervention purposes. Therefore, it is of interest to clarify how well a fast, simple, and relatively cheap BIA equipment system performs in assessing change in overall fatness in comparison with a measurement technique considered a reference method.
The study also has a number of limitations. Firstly, we do not know how accurate the reference method used is for this category of patients, although DXA has been suggested to display little bias according to degree of fatness, age, sex, and physical activity levels (15). However, accuracy problems with increasing tissue depth have been reported in animal as well as human studies (7). Secondly, this study only investigated abdominally obese women, and no information about men was obtained. Thirdly, the small adiposity changes from baseline to follow-up may falsely indicate that the differences between the methods in assessment of adiposity change are of little clinical significance. If the changes were greater, greater discrepancy magnitude may have been observed. The finding of increasing discrepancies with greater adiposity change supports this argument.
In summary, we found that there were small but significant differences between BIA8 and DXA when assessing fatness change. The discrepancy was independently associated with fatness change, BMI, and age. Given the potential and widespread use of BIA in clinical practice, our findings should be an impetus for adjusting the prediction equations so that more accurate estimates of fatness can be obtained from the obese.
The data-collection phase of this study was funded by Cycleurope Inc. Martin Neovius was funded by Arbetsmarknadens Forsakrings-och Aktiebolag (AFA).
M.N. conceived the hypothesis for the analysis, conducted the statistical analyses, and drafted the manuscript. E.H. is the principal investigator of the study, provided critical input in all phases of the manuscript production, and helped revise the manuscript. J.U. was responsible for medical examinations and provided critical input. All authors participated in the interpretation of the results and approved the final version of the manuscript. Special thanks to research nurse Birgitta Spetz, and DXA operators Mai Andersson, Ninni Qvist and Mia Svedin, as well as Dr. Bo Freyschuss, who was responsible for the DXA equipment.
REFERENCES
1. Aleman-Mateo, H., J. Esparza Romero, N. Macias Morales, G. Salazar, J. Wyatt, and M. E. Valencia. Determination of body composition using air displacement plethysmography, anthropometry and bio-electrical impedance in rural elderly Mexican men and women. J. Nutr. Health Aging 8:344-349, 2004.
2. Bolanowski, M., and B. E. Nilsson. Assessment of human body composition using dual-energy x-ray absorptiometry and bioelectrical impedance analysis. Med. Sci. Monit. 7:1029-1033, 2001.
3. Brummer, R., L. Lonn, B. A. Bengtsson, H. Kvist, I. Bosaeus, and L. Sjostrom. Comparison of different body composition models in acromegaly. Growth Regul. 6:191-200, 1996.
4. Eisenkolbl, J., M. Kartasurya, and K. Widhalm. Underestimation of percentage fat mass measured by bioelectrical impedance analysis compared to dual energy X-ray absorptiometry method in obese children. Eur. J. Clin. Nutr. 55:423-429, 2001.
5. Erselcan, T., F. Candan, S. Saruhan, and T. Ayca. Comparison of body composition analysis methods in clinical routine. Ann. Nutr. Metab. 44:243-248, 2000.
6. Fors, H., L. Gelander, R. Bjarnason, K. Albertsson-Wikland, and I. Bosaeus. Body composition, as assessed by bioelectrical impedance spectroscopy and dual-energy x-ray absorptiometry, in a healthy paediatric population. Acta Paediatr. 91:755-760, 2002.
7. 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 adiposity. J. Appl. Physiol. 97:509-514, 2004.
8. Kitano, T., N. Kitano, T. Inomoto, and M. Futatsuka. Evaluation of body composition using dual-energy x-ray absorptiometry, skinfold thickness and bioelectrical impedance analysis in Japanese female college students. J. Nutr. Sci. Vitaminol. (Tokyo) 47:122-125, 2001.
9. Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, et al. Bioelectrical impedance analysis-part I: review of principles and methods. Clin. Nutr. 23:1226-1243, 2004.
10. Kyle, U. G., I. Bosaeus, A. D. De Lorenzo, et al. Bioelectrical impedance analysis-part II: utilization in clinical practice. Clin. Nutr. 23:1430-1453, 2004.
11. Lazzer, S., Y. Boirie, M. Meyer, and M. Vermorel. Evaluation of two foot-to-foot bioelectrical impedance analysers to assess body composition in overweight and obese adolescents. Br. J. Nutr. 90:987-992, 2003.
12. Neovius, M., E. Hemmingsson, B. Freyschuss, and J. Udden. Bioelectrical impedance underestimates total and truncal fatness in abdominally obese women. Obesity (Silver Spring) 14:1731-1738, 2006.
13. Pietrobelli, A., M. S. Faith, D. B. Allison, D. Gallagher, G. Chiumello, and S. B. Heymsfield. Body mass index as a measure of adiposity among children and adolescents: a validation study. J.Pediatr. 132:204-210, 1998.
14. Pietrobelli, A., F. Rubiano, M. P. St-Onge, and S. B. Heymsfield. New bioimpedance analysis system: improved phenotyping with whole-body analysis. Eur. J. Clin. Nutr. 58:1479-1484, 2004.
15. Prior, B. M., K. J. Cureton, C. M. Modlesky, et al. In vivo validation of whole body composition estimates from dual-energy x-ray absorptiometry. J. Appl. Physiol. 83:623-630, 1997.
16. Ravaglia, G., P. Forti, F. Maioli, F. Boschi, A. Cicognani, and G. Gasbarrini. Measurement of body fat in healthy elderly men: a comparison of methods. J. Gerontol. A Biol. Sci. Med. Sci. 54:M70-M76, 1999.
17. Spungen, A. M., W. A. Bauman, J. Wang, and R. N. Pierson, Jr. Measurement of body fat in individuals with tetraplegia: a comparison of eight clinical methods. Paraplegia 33:402-408, 1995.
18. Stewart, S. P., P. N. Bramley, R. Heighton, et al. Estimation of body composition from bioelectrical impedance of body segments: comparison with dual-energy x-ray absorptiometry. Br. J. Nutr. 69:645-655, 1993.
19. Sun, G., C. R. French, G. R. Martin, et al. Comparison of multifrequency bioelectrical impedance analysis with dual-energy x-ray absorptiometry for assessment of percentage body fat in a large, healthy population. Am. J. Clin. Nutr. 81:74-78, 2005.
20. Wattanapenpaiboon, N., W. Lukito, B. J. Strauss, B. H. Hsu-Hage, M. L. Wahlqvist, and D. B. Stroud. Agreement of skinfold measurement and bioelectrical impedance analysis (BIA) methods with dual energy X-ray absorptiometry (DEXA) in estimating total body fat in Anglo-Celtic Australians. Int. J. Obes. Relat. Metab. Disord. 22:854-860, 1998.
21. World Health Organization. Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ. Tech. Rep. Ser. 894:i-xii, 1-253, 2000.
Keywords:
AGREEMENT; BIOELECTRICAL IMPEDANCE; BODY COMPOSITION; OBESITY
©2007The American College of Sports Medicine