Portable A-Mode Ultrasound for Body Composition Assessment in Adolescents

Transcription

1 ORIGINAL RESEARCH Portable A-Mode Ultrasound for Body Composition Assessment in Adolescents Wagner Luis Ripka, Leandra Ulbricht, DrEng, Lucas Menghin, MEng, Pedro Miguel Gewehr, PhD Received June , from the Graduate Program in Electrical and Computer Engineering (W.L.R., P.M.G.) and Graduate Program in Biomedical Engineering (L.U., L.M.), Federal University of Technology Paraná, Curitiba, Brazil. Revision requested July 10, Revised manuscript accepted for publication July 18, Address correspondence to Wagner Luis Ripka, Graduate Program in Electrical and Computer Engineering, Federal University of Technology Paraná, Avenida Sete de Setembro 3165, Curitiba, Paraná PR, Brazil. Abbreviations BF%, body fat percentage; DXA, dual-energy x-ray absorptiometry; SEE, standard estimate of error; US, ultrasound doi: /ultra Objectives Ultrasound (US) imaging is a low-cost, highly feasible alternative method for monitoring the nutritional status of a population; however, only a few studies have tested the body composition agreement between US and reference standard methods, especially in adolescents. The purposes of this study were to assess the agreement of portable US with a reference standard method, dual-energy x-ray absorptiometry (DXA), for body fat percentage (BF%) in adolescents and to verify whether the use of a new mathematical model, based on the anatomic thickness obtained by US, is capable of improving BF% prediction. Methods This research was a descriptive study. Measures of total body mass, BF% on DXA, and BF% on US were collected from 105 adolescents. Results The participants included 71 male adolescents (median age ± interquartile range, 14.0 ± 2.0 years) and 34 female adolescents (13.0 ± 2.3 years). Ultrasound yielded significantly lower BF% values than DXA for male (mean ± SD, US, 9.6% ± 6.6%; DXA, 20.0% ± 7.2%; R = 0.848; P <.05) and female (US, 22.5% ± 5.7%; DXA, 30.3% ± 4.9%; R = 0.495; P <.05) participants. In addition, Bland-Altman analysis showed low concordance. When a multivariate regression was tested, the results improved for both sexes (US, 20.3% ± 4.6%; R= 0.848; P=.503) and female participants (US, 29.0% ± 5.7%; R = 0.712; P =.993) with a standard estimate of error of 1.57%. Conclusions This study has shown that US applied in a specific regression for BF% prediction in adolescents has a strong correlation with DXA as well as concordance with Bland-Altman analysis. Key Words adolescents; body composition; dual-energy x-ray absorptiometry; fat mass; ultrasound Monitoring the nutritional status of a population allows the identification of several health risks, such as diabetes, circulatory problems, and hypertension. 1,2 For this monitoring, different techniques are used over time, among them anthropometry, including the skin fold thickness method and body mass index, 3 dual-energy x-ray absorptiometry (DXA), computed tomography, 4 underwater weighing, 5 bioelectrical impedance, 6 and ultrasound (US) imaging. 3,7,8 The US method has attracted attention for body composition assessment, presenting practical advantages such as low cost, noninvasiveness, and low interexaminer and intraexaminer error. 9, by the American Institute of Ultrasound in Medicine J Ultrasound Med 2016; 35:

2 The technical operational principle of A-mode (amplitude mode) US makes use of emitted waves with frequencies higher than 20 khz and the acquisition of corresponding echoes by a piezoelectric transducer. 11 The information about this 1-dimensional mode is provided by reflected waves from different body interfaces: skin-fat, fat-muscle, and muscle-bone. 7 A commercial portable A-mode US system (BodyMetrix; IntelaMetrix, Concord, CA) has been used for body fat percentage (BF%) assessment; however, it has yielded controversial results in young adults, 3,8,9,12 adults, 13 and adolescents. 14,15 Therefore, the purposes of this study were: to assess the agreement of the BodyMetrix portable US system with a reference standard method, DXA, for BF% in adolescents and to verify whether the use of a new mathematical model, based on the anatomic fat thickness measured by the equipment is capable of improving BF% prediction. Materials and Methods Participants This research was a cross-sectional descriptive study. The sample consisted of 105 Brazilian adolescents aged between 12 and 17 years. Parents or guardians of all participants signed an informed consent form. This study was approved by the Ethics Committee of the Federal University of Technology Paraná, under protocol All procedures were performed in concordance with the Declaration of Helsinki. Protocol Initially, total body mass measurements of the adolescents were performed for DXA and US assessment. For DXA, a Discovery QDR radiologic densitometer (Hologic, Inc, Bedford, MA) was used. The DXA assessment is based on the emission of x-rays with low and high energy (40 and 70kV, respectively). The total scan time was 5 minutes. The BF% was provided by specific software for children in the DXA equipment and referred to total body fat. For US analysis, the BodyMetrix system operating in A-mode with a frequency of 2.5 MHz was used. Local skin fold measurements were performed at the triceps, subscapular, abdominal, suprailiac, midaxillary, chest, and thigh sites. Ultrasound gel was applied on each skin point and on the transducer s head. The US transducer was placed perpendicular to the measuring site with minimal circular movement (±5 mm) and pressure. The BodyView software made it possible to obtain BF% values among 12 formula options, as follows: 1 at only 1 anatomic site; 3 at 2 sites; 3 at 3 sites; 3 at 4 sites; 1 at 7 sites; 1 at 9 sites, respectively. For this research, the 7-site option was the chosen, since previous research indicated that equations using more anatomic points tended to be more precise for estimating BF%. The 9-site option was not adopted because of the lack of studies indicating its reliability. 3,9 For each evaluation, at least 2 measurements were made at each point, with the possibility of performing a third one in cases of software requests due to discrepancies in the 2 previous measurements. Statistical Analysis The normality of the variables was verified by the Shapiro- Wilk test. The descriptive data were presented as median ± interquartile range and mean ± standard deviation. The values obtained by DXA and US were compared by the t test for independent samples or the Wilcoxon signed ranks test. The capability of US to estimate BF% was verified by multivariate linear regression. The independent variables were sex, age, and body mass, in addition to subcutaneous fat thickness as determined by US measurements. Data were inserted by a stepwise method at one hierarchical level. The independent variables were inserted in the model in descending order of their correlation with BF%. Once the variable was inserted, the improvement of the estimation by the model was verified by analysis of variance. If there was no improvement, the analysis was stopped, and the last model with improvement was considered. At each level, the standardized coefficients (β) were tested by the t test, and then variables with β coefficients that were not significantly different from 0 were excluded. 16 The US forecast capacity was tested by a square correlation between the predicted values and BF% (R 2 ) and the application capacity in different samples by its adjusted value (R 2 ADJUSTED). The accuracy of predicted values was indicated by the standard estimate of error (SEE). The predicting variables were described by regression coefficients (B), standard error for each parameter (SE B), standardized value for each coefficient (β), and partial correlation coefficient. The results obtained by US (BF% US ) and predicted by the model (BF% US PREDICTED ) were compared to the values obtained by DXA (BF% DXA ) by Bland-Altman graphic analysis. In this analysis, the limits of the method had been defined as the mean difference between methods ± 1.96 SD with the 95% confidence interval. 17 P <.05 was established as significant for all analyses, which were performed with SPSS version 21.0 software (IBM Corporation, Armonk, NY). 756 J Ultrasound Med 2016; 35:

3 Results One hundred five adolescents were evaluated. The values for age, weight, height, body mass, BF% US, and BF% DXA are presented in Table 1. Significant differences were found between the methods. The Bland-Altman analysis found no concordance between US and DXA. Limits around ±7% and ±9% with a significant bias of 10.9 (P <.001) for male participants and ±11% and ±5% with a significant bias of 7.8 (P<.001) to female participants were found (Figures 1 and 2). However, a strong correlation between US and DXA for male participants (R = 0.848; P <.001) and a week correlation for female participants (R = 0.495; P <.001) were also found. Also, a new model was tested to estimate the BF% by multivariate linear regression. The values in Table 2 present an overview of the final model generated. Table 1. Descriptive Data From the Sample Variable Male Female n Age, y a 14.0 ± ± 2.3 Weight, kg b 57.8 ± ± 9.3 Height, m a 1.69 ± ± 0.07 BMI, kg/m² a ± ± 4.31 BF% c US 9.6 ± 6.6 b 22.5 ± 5.7 b BF% c DXA 20.0 ± 7.2 b 30.3 ± 4.9 b a Median ± interquartile range, Wilcoxon signed ranks test. b Mean ± SD, paired t test. c P <.05 between US and DXA. From the multivariate linear regression, again, the Bland-Altman concordance analysis was applied for BF% DXA BF% US PREDICTED by sex (Figures 3 and 4). The new model showed improvement in the correlation for the female participants (R = ; Figure 4A) in addition to improving the agreement for both sexes (Figures 3B and 4B). The mean values for BF% US PREDICTED in male and female participants were 20.3% ± 4.6% (P=.503) and 29.0% ± 5.7% (P =.993), respectively. Discussion With respect to the health status in adolescents, the World Health Organization presented several concerns regarding overweight in this age group because it can result in early medical expenses. 18 Thus, development and improvement of nutritional status assessment technologies are important for monitoring and generating prevention strategies to reduce overweight and obesity. The objective of this study was to analyze the A-mode US method in comparison to DXA in adolescents of both sexes for BF% assessment. Using the 7-sites equation in the BodyView software, the results showed a strong correlation between the methods for male participants (R = 0.848) and a poor correlation for female participants (R = 0.495), since US tended to underestimate the BF% value. Furthermore, we found no concordance between the methods with a significant bias in both genders (male, 10.9 with limits of 1.4 to 17.4; female, 7.8 with limits of 2.8 to 18.3) in the Bland-Altman analysis. The lower correlation values found for girls can be explained by pubertal differences that were Figure 1. A, Plot of BF% on DXA versus US. B, Bland-Altman plot between BF% DXA and BF% US for male participants. LLA indicates lower limit of agreement; and ULA, upper limit of agreement. J Ultrasound Med 2016; 35:

4 not controlled in this study, which can directly influence the body composition in this age group. 19 To our knowledge, only a few studies analyzed the BodyMetrix A-mode US system in adolescents. Goulart et al 14 examined the BF% in 36 adolescents (age, 13.6 ± 1.9 years); they found a high correlation with DXA (r = 0.903) but underestimation of BF% compared to US. Utter and Hager 15 tested the accuracy of portable US for fat-free mass assessment in comparison to hydrostatic weighing in 70 adolescent wrestlers (age, 15.5 ± 1.5 years). The results did not indicate significant differences between the methods, which did not indicate underestimation or overestimation. For young adults, underestimated BF% values on US were also reported by Ulbricht et al, 3 Johnson et al, 8 Neves et al, 9 Loenneke et al, 12 and Lyon et al. 20 Smith-Ryan et al 13 tested the reproducibility and validity of US in 47 overweight and obese adults in comparison to air displacement plethysmography and bioelectrical impedance, again finding underestimated BF% values. These divergences highlight the need for testing portable US technology in different groups of participants to ensure its use in clinical practice. From the multivariate linear regression application, it was possible to test a new equation for predicting BF% in adolescents, which showed significant improvement in the concordance between the methods (R = 0.88; SEE = 1.57%), which was another objective of this study. In a study by Heyward and Wagner, 21 SEE values of 3% or lower were considered very good. These findings suggest that the development of new equations based on measurements obtained with US can minimize measurement errors compared to reference standard methods such as DXA, computed tomography, and underwater weighing. Figure 2. A, Plot of BF% on DXA versus US. B, Bland-Altman plot between BF% DXA and BF% US for female participants. Abbreviations are as in Figure 1. Table 2. Multivariate Linear Regression for BF% US PREDICTED by the Stepwise Method at One Hierarchical Level Variable R 2 R 2 ADJUSTED SEE B SE B β PCC Model Constant Triceps a Sex a Subscapular a Thigh b Age c Chest c PCC indicates partial correlation coefficient. a P <.001. b P <.01. c P < J Ultrasound Med 2016; 35:

5 Indirect assessment methods such as portable US are based on the patient s collected information, which is inserted into mathematical models that predict variable body composition coefficients. In addition, other authors stated that the application of a proper mathematical model for the assessed population is essential to obtain good results. 6,22,23 When using US to compare BF% with a reference standard method such as DXA, it has been shown that US significantly underestimates that variable in adolescents. By using the 7-site equation, a strong correlation for male participants, a poor correlation for female participants, and no agreement for both sexes was found in the Bland-Altman analysis. However, when the data were tested by using a new multivariate regression, the results improved for female participants and remained equal for male participants. These new results also showed agreement between both methods. These data support the use of US in clinical and medical applications in adolescents. However, as for the skin fold method, it is suggested that specific and generalized equations should be developed for different populations. Finally, the possibility of adding new equations in the BodyView software will allow the achievement of more accurate data for estimating BF% in adolescents. Figure 3. A, Plot of BF% on DXA versus US predicted by multivariate regression. B, Bland-Altman plot between BF% DXA and BF% US PREDICTED for male participants. Abbreviations are as in Figure 1. Figure 4. A, Plot of BF% on DXA versus US predicted by multivariate regression. B, Bland-Altman plot between BF% DXA and BF% US PREDICTED for female participants. Abbreviations are as in Figure 1. J Ultrasound Med 2016; 35:

6 References 1. Goldfield GS, Saunders TJ, Kenny GP, et al. Screen viewing and diabetes risk factors in overweight and obese adolescents. Am J Prev Med 2013; 44(suppl 4):S364 S Thibault R, Genton L Pichard C. Body composition: why, when and for who? Clin Nutr 2012; 31: Ulbricht L, Neves EB, Ripka WL, Romaneli EFR. Comparison between body fat measurements obtained by portable ultrasound and caliper in young adults. In Proceedings of the 34th Annual International Conference of the Institute of Electrical and Electronics Engineers and Engineering in Medicine and Biology Society. Piscataway, NJ: Institute of Electrical and Electronics Engineers; 2012: Antoun S, Lanoy E, Albiges-Sauvin L, Escudier B. Clinical implications of body composition assessment by computed tomography in metastatic renal cell carcinoma. Exp Rev Anticancer Ther 2014; 14: Rodrigues MN, Silva SC, Monteiro WD, Farinatti PTV. Estimativa da gordura corporal através de equipamentos de bioimpedância, dobras cutâneas e pesagem hidrostática. Rev Bras Med Esporte 2001; 7: Oeffinger DJ, Gurka MJ, Kuperminc M, Hassani S, Buhr N, Tylkowski C. Accuracy of skinfold and bioelectrical impedance assessments of body fat percentage in ambulatory individuals with cerebral palsy. Dev Med Child Neurol 2014; 56: Wagner DR. Ultrasound as a tool to assess body fat. J Obes 2013; 2013: Johnson KE, Miller B, Juvancic-Heltzel JA, et al. Agreement between ultrasound and dual-energy X-ray absorptiometry in assessing percentage body fat in college-aged adults. Clin Physiol Funct Imaging 2014; 34: Neves EB, Ripka WL, Ulbricht L, Stadnik AMW. Comparação do percentual de gordura obtido por bioimpedância, ultrassom e dobras cutâneas em adultos jovens. Rev Bras Med Esporte 2013; 19: Leahy S, Toomey C, McCreesh K, O Neill C, Jakeman P. Ultrasound measurement of subcutaneous adipose tissue thickness accurately predicts total and segmental body fat of young adults. Ultrasound Med Biol 2012; 38: Webster JG. Medical Instrumentation: Application and Design. New Delhi, India: Wiley-India; Loenneke JP, Barnes JT, Wagganer JD, et al. Validity and reliability of an ultrasound system for estimating adipose tissue. Clin Physiol Funct Imaging 2014; 34: Smith-Ryan AE, Fultz SN, Melvin MN, Wingfield HL, Woessner MN. Reproducibility and validity of A-mode ultrasound for body composition measurement and classification in overweight and obese men and women. PLoS One 2014; 9:e Goulart C, Beraldo L Ulbricht L. Comparativo da estimativa da gordura corporal em adolescentes obtida por DXA e ultrassom portátil. Paper presented at: XXIV Congresso Brasileiro de Engenharia Biomédica; October 13 17, 2014; Uberlândia, Brazil. 15. Utter AC, Hager ME. Evaluation of ultrasound in assessing body composition of high school wrestlers. Med Sci Sports Exerc2008; 40: Maroco J. Statistical Analysis With SPSS Application. Lisboa, Portugal: Edições Silabo; Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 327: World Health Organization (WHO). Population-Based Prevention Strategies for Childhood Obesity: Report of a WHO Forum and Technical Meeting. Geneva, Switzerland: World Health Organization; Hall JE, Guyton AC. Textbook of Medical Physiology. London, England: Saunders; Lyon J, Drew R MacRae H. Comparison of skinfold thickness measures with ultrasound imaging to determine body composition. Paper presented at: 26th Annual Meeting of the Southwest Chapter of the AmericanCollege of Sports Medicine; November 11 12, 2005; Las Vegas, NV. 21. Heyward VH, Wagner DR. Applied Body Composition Assessment. Champaign, IL: Human Kinetics Publishers; Queiroga MR. Testes e Medidas Para Avaliação da Aptidão Física Relacionada à Saúde em Adultos. Rio de Janeiro, Brazil: Guanabara Koogan; Jackson AS, Pollock ML. Generalized equations for predicting body density of men. Br J Nutr 1978; 40: J Ultrasound Med 2016; 35: