Validation Study of Multi-Frequency Bioelectrical Impedance with Dual-Energy X-ray Absorptiometry Among Obese Patients

OBES SURG (2014) 24:1476 1480 DOI 10.1007/s11695-014-1190-5 OTHER Validation Study of Multi-Frequency Bioelectrical Impedance with Dual-Energy X-ray Absorptiometry Among Obese Patients Silvia L. Faria & Orlando P. Faria & Mariane D. A. Cardeal & Marina Kiyomi Ito Published online: 26 January 2014 # Springer Science+Business Media New York 2014 Abstract Background Body mass index (BMI) is the most common parameter for classifying nutritional status. However, body composition (BC) may vary considerably among individuals with identical BMIs; consequently, we need to assess BC efficiently. Bariatric surgery is the most effective method for treating obesity. To improve quality assessment of postoperative weight loss, it is essential to assess BC. Multi-frequency bioelectrical impedance analysis (BIA) is a practical assessment instrument, though limited when applied among the obese population. Despite dual-energy X-ray absorptiometry (DXA) being the current reference standard, it has physical limitations which restrict its practical application. This study, therefore, sought to correlate the results of BC assessments of same patient population using BIA and DXA. Methods This was a cross-sectional validation study with patients invited to undergo a multi-frequency BIA (Inbody 720 ) and afterwards a DXA examination Statistical analyses were done using the intraclass correlation coefficient (ICC), paired t-test and the Bland Altman plot analysis. Results A total of 108 patients were randomly selected, with 73 meeting the criteria for study inclusion. Most were female (89 %) and had an average BMI of 40.17±4.08 kg/m 2.An almost perfect correlation of fat (kg) and fat-free mass (kg) was found in results from the BIA and DXA examination (ICC=0.832 and ICC=0.899, respectively). A substantial correlation was also found between the percentage of body fat (%BF) and the percentage of fat-free mass (%FFM). The comparison made between the BIA and DXA using the t-test showed significant differences between all parameters. The S. L. Faria: O. P. Faria : M. D. A. Cardeal Gastrocirurgia, Brasilia, Brazil S. L. Faria (*): M. D. A. Cardeal : M. K. Ito University of Brasília, Campus Universidade Darcy Ribeiro, Brasilia, Distrito Federal 70910-900, Brazil e-mail: silvialeitefaria@brturbo.com.br Bland Altman plot showed that the BIA method tends to underestimate the FM and overestimate the LM measurements when compared with DXA. Conclusion BIA proved to be a safe alternative for assessing BC in clinically severely obese patients and thus provides a more accessible evaluation tool for this population. But, consideration should be given to the formula added to the BIA measurement, adjusting the values to differences observed in order to reduce errors when compared with the DXA measurements. Keywords Dual-energy X-ray absorptiometry. Bioelectrical impedance analysis. Obesity. Bariatric surgery Introduction Obesity, defined as a body mass index (BMI) above 30 kg/m 2, increases the risk of death, whatever the actual cause may be [1]. In 2010, obesity among adult men exceeded 20 % in ten countries of the Western Hemisphere [2]. However, the definition of obesity must consider not only weight increase, but even more precisely any increase in body fat. The accumulation of the latter is associated with an increased risk of several diseases, such as type 2 diabetes and cardiovascular diseases [3]. Thus, there is a need for improvement in the clinical evaluation of patients with severe obesity (BMI 35 kg/m 2 with co-morbidities or BMI 40 kg/m 2, independent of comorbidities). Currently, BMI (weight/(height) 2 ) is the most common parameter used to classify the nutritional status of individuals. However, there may be considerable variation in body composition (BC) even between individuals with the same BMI [4]. In this sense, a serious need arises for an adequate assessment of BC of individuals, which can serve as a basis to better evaluate the effectiveness of various treatments of obesity [5]. At present, bariatric surgery is considered the best and most effective treatment for obesity, due to weight loss achieved

OBES SURG (2014) 24:1476 1480 1477 and, principally, the long-term maintenance of this weight loss [6]. To improve the assessment of the quality of weight loss after this surgical procedure, it is essential that the assessment of BC be part of the clinical evaluation of these patients preoperatively and postoperatively, i.e., to assess whether the weight loss post-surgery was fat loss and/or loss of fat-free mass (FFM), with consequent malnutrition. Bioelectrical impedance analysis (BIA) is a rapid, practical, minimally invasive and relatively inexpensive instrument used to analyze BC of individuals based on the electrical conductivity of tissues. The method brings results in the form of body fat percentage (%BF), FFM percentage (%FFM) and distribution of body water. Studies have showed that BIA is a valid instrument for patients with a BMI up to 34 kg/m 2 [7]. The Multi-frequency Bioelectrical Impedance Analysis (Inbody 720, Biospace) uses an exclusive direct, targeted and multi-frequency system that is not based on statistical data of any specific population. It is capable of accurately assessing people with very different physical types, whether obese, elderly or athletic. Equations presented in recognized scientific literature were all developed using mono-frequency equipment and impedance measurement of the whole body. Thus, they are not comparable to results obtained from a segmental multi-frequency system. The trunk of the body provides about 50 % of conductivity, but only 10 % of the body impedance. As a result, studies show that there is a disproportion between body mass and body conductivity. This situation may decrease the accuracy in the results when using this method in obese patients. There is also a change in intra- and extracellular water in these patients, which may also interfere with the results [8]. Thus, validation studies are needed that measure BIA in patients with a BMI greater than 34 kg/m 2. Yet, despite the technological advances and the presence of different electric pulses of low intensity at different locations, which potentially increase the accuracy of the method, the formulas embedded in BIA machines do not offer specific considerations for the obese population. The current reference standard for analysis of BC is the dual-energy X-ray absorptiometry (DXA) examination. However, this examination has its own limitations related to the weight limit of the machine (in most cases, 120 kg) [5]and the cost involved. In view of the need to assess the BC of obese patients starting from the preoperative stage, and considering the cost and difficulty of performing DXA on all patients on a regular and frequent basis, this study sought to correlate the BC results of the same individuals using BIA and DXA. Material and Methods This was a cross-sectional validation study, where patients were first asked to undergo an examination of multi-frequency BIA (Inbody 720 ) to assess BC. Shortly after this examination, patients underwent a DXA examination (the gold standard for BC assessment) to evaluate the same BC data. Inclusion criteria for participation in this study were as follows: person having clinically severe obesity (BMI 35 kg/m 2 with co-morbidities or BMI 40 kg/m 2 ), participant in the preoperative treatment for bariatric surgery in the Gastrocirurgia Clinic of Brasilia (a private practice), weigh less than 120 kg (limitation of the DXA machine) and be between 18 and 65 years of age. Exclusion criteria were: being pregnant and/or with walking difficulties. Patients followed the following protocol for BIA assessment of BC: 8-h fast, as well as avoidance of strenuous physical activity, alcohol, along with food and drinks containing caffeine on the previous day. The intraclass correlation coefficient (ICC) combines random and systematic errors and, therefore, is preferable to the linear correlation coefficient of Pearson. Although confounding results between linear association and agreement does not merit concern, the main problem related to the correlation coefficients resides in the dependence on the amplitude of the measurement. In other words, the more heterogeneous the set of individuals (greater variability intra-individuals) studied is, the greater the value of the ICC. This indicator is a measure of corrected agreement due to chance and ranges from 1 and +1 (plus one). The maximum value indicates perfect reproducibility between measurements. The joint roughness coefficient (JRC) is equivalent to the kappa statistic for continuous variables and their interpretation can be given according to Landis and Koch [9]. The average of the differences between the two values of the pairs under observation is a measure of the degree of systematic differences between the two sets of readings. A paired t-test statistical analysis was performed to calculate the statistical significance of the calculated difference. The Bland Altman plot was used to assess the magnitude of disagreement (including systematic differences), to spot outliers, and to observe any trends. Results Out of a possible total of 104 patients, 73 met criteria for inclusion in this study and agreed to participate. Of these, the majority was female (89 %). The average age of the group was 37.31±10.49 years and their BMI was 40.17±4.08 kg/m 2. Mean values of %BF and %FFM, fat mass (FM) and FFM in kg, using the two methods, compared through paired t-test, are shown in Table 1. From the test results, we have: 1. The FM value (kg) obtained by the BIA method is significantly inferior (2.05 kg), on average, to the FM (kg) value obtained by the DXA method.

1478 OBES SURG (2014) 24:1476 1480 Table 1 Results of weight, height, % body fat, % lean mass, fat mass and lean mass in kg using two methods multi-frequency bioimpedance and DXA Variable DXA BIA Average (±SD) Average (±SD) Difference (BIA-DXA) Paired t value p value Weight (kg) 105.95 (±10.83) 105.19 (±11.45) 0.76 1.88 0.0638 Height (m) 1.63 (±0.08) 1.62 (±0.07) 0.01 0.45 0.6508 BMI (kg/m 2 ) 40.17 (±4.09) 39.89 (±3.97) 0.28 0.58 0.5608 FM (kg) 54.79 (±7.46) 52.74 (±7.37) 2.05 4.46 <0.0001 %BF 51.72 (±4.80) 50.11 (±4.04) 1.61 4.68 <0.0001 Lean mass (kg) 51.17 (±7.55) 52.45 (±7.05) 1.28 3.54 0.0007 Lean mass (%) 48.28 (±4.80) 49.89 (±4.04) 1.61 4.68 <0.0001 %BF percentage of body fat, BMI body mass index, DXA dual-energy X-ray absorptiometry, BIA multi-frequency bioimpedance, FM fat mass 2. The %BF value obtained by the BIA method is significantly inferior (1.61 %), on average, to the %BF obtained by the DXA method. 3. The FFM value (kg), obtained by the BIA method, is significantly superior (1.28 kg), on average, to the FFM value (kg) obtained by the DXA method. 4. The FFM value (%), obtained by the BIA method, is significantly superior (1.61 %), on average, to the FFM value (%) obtained by the DXA method. Applying the Bland Altman plot analysis, a statistically significant bias can be seen in Figs. 1 and 2 (p<0.0001), showing that the BIA method tends to underestimate the FM measurements (kg and %BF) when compared with DXA. A relation is seen between the bias (difference between the measurements) and the mean of the measurements. The bias appears to be greater for mean values larger than 55 kg and 48 %. For mean values smaller or equal to 55 kg and 48 %, the FM (kg and %BF) calculated from the BIA, adding a bias of 1.29 kg and 0.61 %, could be used (excluding the outlier point with bias close to 20 kg and 10 %) as an estimate of FM (kg and %BF) obtained by DXA. For mean values greater than 55 kg and 48 %, the FM (kg and %BF) obtained from BIA, adding a bias of 2.40 kg and 1.71 %, as an estimate of the FM (kg and %BF) obtained by DXA. Three points outside the limit of agreement can be seen, one of which showed a bias difference of nearly 20 kg and 10 % (Figs. 1 and 2). A statistically significant bias is seen in the graph, showing that the BIA method tends to overestimate the FFM measurements (kg and %FFM) when compared with DXA (p<0.0007 and p<0.0001). The spatial distribution of points tends to be homogeneous, in the sense of not indicating relation between the bias (difference between measurements) and the mean of the measurements, leading to the conclusion that the bias is systematic. Therefore, despite the measurements not being consistent, FFM (kg) calculated by BIA, removing a bias of 1.28kgand1.61%,couldbeusedasanestimateofthe FFM (kg and %FFM) obtained by DXA. Three points outside the limit of agreement to FFM in kg can be seen (Fig. 3). But, for the %FFM, there are six points outside of the limit of agreement, one of which showed a bias difference of nearly 10 % (Fig. 4). Fig. 1 Bland-Altman plot analysis showing the Fat Mass (FM) difference using DXA and BIA Fig. 2 Bland-Altman plot analysis showing the %Fat Mass (%FM)

OBES SURG (2014) 24:1476 1480 1479 Table 2 Correlation between parameters of body composition measuredbydxaandbia %BF percentage of body fat, BMIbody mass index a Intra-class coefficient (ICC), with a confidence interval of 95 % (in parentheses) ICC a Weight 0.950 (0.921 0.969) Height 0.860 (0.785 0.910 BMI 0.851 (0.773 0.904) Lean mass (kg) 0.899 (0,819 0.941) Lean mass (%) 0.736 (0.508 0.851) Fat mass (kg) 0.832 (0,670 0.907) %BF 0.736 (0.508 0.851) Fig. 3 Bland-Altman plot analysis showing Lean Mass (LM) in Kg With the application of the two methods, the ICC showed an almost perfect correlation between fat and FFM in kg. In relation to %BF and the amount of fat in kg, a substantial correlation was found (Table 2). Discussion The importance of assessing BC in obese patients can be seen in the evaluation of the actual effectiveness of existing treatments for obesity in favor of quality weight loss, principally of fat. But there are a number of factors that complicate the measurement of BC in these patients who often have significant changes in their body compartments when compared with eutrophic individuals. One major obstacle is the increase in total body water, especially extracellular water in relation to intracellular water [5], which impedes an accurate BC measurement when using various methods. All this makes for scant data on BC and reduced information on the methods Fig. 4 Bland-Altman plot analysis showing % Lean Mass (%LM) that can safely be used for the measurement of this variable among the obese population [2]. The DXA examination is the most valid technique currently used to measure BC in eutrophic patients who are overweight or moderately obese. Despite its detailed and accurate assessment, this method is expensive and cannot be repeated often due to the level of radiation emitted during the examination. Besides this, obese patients often exceed the capacity of the DXA unit [10]. For this reason, an alternative is to use multi-frequency BIA, a non-invasive and relatively inexpensive method that has been used for measuring total body water, FFM and extracellular water. However, studies show that in cases of clinically severe obesity, most prediction equations used with Bioelectral Impedance Analysis are unable to ascertain the static BC and over time are not reproducible for individuals [11]. The disproportion between body mass and conductivity of the body of obese patients seems to reduce the accuracy of BIA [7]. The method of bioelectrical impedance is based on the ability of hydrated tissues to conduct electricity. The measurement of total body impedance allows for the assessment of total body water, assuming a constant state of hydration. From the total body water (in a healthy adult, 60 % of body weight is attributable to water), assuming that water makes up 73 % of muscles and using validated equations and reference values, it is possible to estimate FFM and, by calculating the difference, body fat. With these assumptions, questions are being raised about using the method with the obese population, considered to be hyper-hydrated [4]. However, in this study, BIA presented highly reliable results in the measurement of BC among clinically severe obese patients when compared with the reference method DXA. Strain et al. [12], comparing a single frequency BIA examination (Tanita TBF-310) with the deuterium dilution method, observed a reasonable correlation between the methods (ICC=0.91), and affirmed that the use of bioelectrical impedance is safe for clinically severe obese patients. Donini et al. [4] studied over 100 overweight patients using a model of a single frequency bioelectrical impedance, found a reasonable correlation between body fat measured by DXA (r=0.91) and body fat measured by bioelectrical impedance (r=0.86). The

1480 OBES SURG (2014) 24:1476 1480 study of Donini et al. [4], however, was done with patients who had a BMI 25 and 35 kg/m 2. For this reason, they may have presented better results than those found in the present study. The increase in BMI and in fat distribution tend to decrease the accuracy of the BIA test [5]. The apparatus used in these previous studies was not the multi-frequency modality used in the present study, yet a reasonable correlation was presented. A multi-frequency bioelectrical impedance examination also uses empirical linear regression models. However, unlike the single-frequency bioelectrical impedance, it includes impedances at multiple frequencies (0, 1, 5, 50, 100, 200 to 500 khz) in order to evaluate FFM, the total body water, intracellular water and extracellular water, which can further improve the accuracy of the results [13]. The basic principle of DXA is that the attenuation of X-rays, using high and low energy photons, is measurable and depends on the thickness, density and chemical composition of underlying tissue. Thus, the attenuation of X-ray energies through the fat, bones and muscles varies due to differences in density and chemical composition within these tissues. In this way, the moisture level of the individual can change the result. However, studies show that changes in body hydration of up to 1 kg do not seem to have significant affect on the accuracy of DXA assessment [14]. Lohman et al. [15], after extensive review of the literature, theorized that a 5 % change in water present in FFM could affect the measurement of the % of body fat (%BF) in only 2 % of cases considered [15]. The limitations of this present study relate to external validity of results obtained. This study was undertaken mainly with female patients weighing up to 120 kg. Thus, for individuals with an altered hydration state, as possibly in the case of the obese individuals studied here, DXA appears, in fact, to be a reference for BC measurement, with which the multi-frequency bioelectrical impedance (Inbody 720 ) presented almost perfect correlation, proving itself to be a good method for assessing BC in clinically severe obese patients. Conclusion The multi-frequency BIA proved to be a safe alternative to assess BC in clinically severe obese patients who weigh up to 120 kg, thus providing easier access to this important measurement tool for this population, considering the facilities of use. Consideration should be given to the formula added to the BIA measurement, adjusting the values to differences observed in order to reduce errors when compared with the DXA measurements. Acknowledgments This study received financial support from Gastrocirurgia, Brasília, Brazil. SEPS 710/910, Centro Clínico Via Brasil, Salas 314, 334, 336, 337, 348, 70390-108 Brasília, Brazil. Conflict of interest interest. References All the authors declare that they have no conflict of 1. Wandell PE, Carlsson AC, Theobald H. The association between BMI value and long-term mortality. Int J Obes (Lond). 2009;33(5): 577 82. 2. World Health Organization (WHO). http://www.who.int/en/. Accessed July 2010. 3. Must A, Spadano J, Coakley EH, et al. The disease burden associated with overweight and obesity. Am J Med Assoc. 1999;282: 1523 9. 4. Donini LM, Poggiogalle E, Del Balzo V, et al. How to estimate fat mass in overweight and obese subjects. Int J Endocrinol. 2013;2013: 285680. 5. Das SK. Body composition measurement in severe obesity. Curr Opin Clin Nutr Metab Care. 2005;8(6):602 6. 6. Valezi AC, Junior JM, de Menezes MA. Weight loss outcome after silastic ring Roux-en-Y gastric bypass: 8 years of follow-up. Obes Surg. 2010;20:1491 5. 7. Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis Part II. Utilization in clinical practice. Clin Nutr. 2004;23(6):1430 53. 8. Bracco D, Thiebaud D, Chiolero R, et al. Segmental body composition assessed by bioelectrical impedance analysis and DEXA in humans. J Appl Physiol. 1996;81:2580 7. 9. Szklo M, Nieto, JF. Epidemiology beyond the basics. 2000; 377. 10. Kehayias JJ, Heymsfield SB, Dilmanian FA, et al. Measurement of body fat by neutron inelastic scattering: comments on installation, operation and error analysis. Basic Life Sci. 1990;55: 339 46. 11. Hendel HW, Gotfredsen A, Hojgaard L, et al. Change in fat-free mass assessed by bioelectrical impedance, total body potassium and dual energy X-ray absorptiometry during prolonged weight loss. Scand J Clin Lab Invest. 1996;56(8):671 9. 12. Strain GW, Wang J, Gagner M, et al. bioelectral impedance for severe obesity: comparing research methods for total body water and resting energy expenditure. Obesity (Silver Spring). 2008;16(8):1953 6. 13. Heyward VH, Wagner DR. Applied body composition assessment. Champaign: Human Kinetics Pub Incorporated; 2004. 268 p. 14. Pietrobelli A, Wang Z, Formica C, et al. Dual-energy X-ray absorptiometry: fat estimation errors due to variation in soft tissue hydratation. Am J Physiol. 1998;274:808 16. 15. Tg L. Harris M, Teixeira P J, et al. Assessing body composition and changes in body composition: another look at dual-energy X-ray absorptiometry. Ann N Y Acad Sci. 2000;904:45 54.