Body composition in children and adults by air displacement plethysmography

Transcription

1 European Journal of Clinical Nutrition (1999) 53, 382±387 ß 1999 Stockton Press. All rights reserved 0954±3007/99 $ Body composition in children and adults by air displacement plethysmography CNunÄez 1 *, AJ Kovera 1, A Pietrobelli 1, S Heshka 1, M Horlick 1, JJ Kehayias 2, Z Wang 1 and SB Heyms eld 1 1 Department of Medicine, Obesity Research Center, St. Luke's=Roosevelt Hospital, Columbia University, College of Physicians and Surgeons, New York, NY, and 2 USDA Human Nutrition Research Center on Aging at Tufts University, Boston, MA, USA Objectives: Air displacement plethysmography (ADP) may provide a partial alternative to body density (B d ) and therefore body composition measurement compared to conventional hydrodensitometry (H d ) in children. As there are no evaluation studies of ADP in children, this study had a two-fold objective: to compare B d estimates by ADP and H d ; and to compare fat estimates by both ADP and H d to fat estimates by another reference method, dual energy X-ray absorptiometry (DXA). Setting: Obesity Research Center, St. Luke's=Roosevelt Hospital, New York, USA. Subjects: One hundred and twenty subjects (66 females=54 males) who ranged in age from 6 ± 86 y and in body mass index (BMI, kg=m 2 ) from 14.1 ± 40.0 kg=m 2 met study entry criteria. Study Design: Cross-sectional study of healthy children (age 19 y) and adult group for comparison to earlier studies. Each subject completed ADP, H d, and DXA studies on the same day. Only subjects with subjectivelyjudged successful H d studies were entered into the study cohort. Results: There was a high correlation between B d by ADP and H d (B d Hd ˆ B d ADP; r ˆ 0.93, SEE ˆ g=cm 3, P < ), although the regression line slope and intercept differed signi cantly from 1 and 0, respectively. Additional analyses localized a small-magnitude B d bias in the child (n ˆ 48) subgroup. Both ADP and H d %fat estimates were highly correlated (r > 0.9, P < ) with %fat by DXA in child and adult subgroups. Bland ± Altman analyses revealed no signi cant %fat bias by either ADP or H d vs DXA in either children or adults, although a bias trend (P ˆ 0.11) was detected in the child subgroup. Conclusion: With additional re nements, the air displacement plethysmography system has the potential of providing an accurate and practical method of quantifying body fat in children as it now does in adults. Sponsorship: This study was in-part supported by NIH Grants RR00645, NIDDK and NIDDK Descriptors: air displacement plethysmography; body composition; hydrodensitometry; children Introduction At present there are several different body fat measurement methods that can be applied in the clinical setting (Wang et al, 1998). One traditional research approach is to evaluate a subject's body fat by H d based on a two-compartment model (Siri, 1961). The two compartment H d model is formulated on assumed constant densities of fat and fat-free body mass (Siri, 1961). Measuring the subject's body weight on land and in water permits B d and hence body fat estimation. However, the H d method is impractical for clinical use, particularly as many subjects have dif culty with the breath-hold procedure. An alternative to the H d method, and one that has potential clinical value, is B d evaluation by air displacement plethysmography (Garrow et al, 1979; Dempster & Aitkens, 1995; McCrory et al, 1995). Based on classic gas laws, this method is simple to perform and does not require water submersion. However, there have been relatively few published validation studies of this potentially important *Correspondence: Dr C NunÄez, St. Luke's=Roosevelt Hospital, Weight Control Unit, 1090 Amsterdam Avenue, 14th oor, New York, NY Contributorship: All authors were co-investigators on the project and contributed equally to the development of the manuscript. Both C NunÄez and AJ Kovera performed all of the technical measurements. Received 23 September 1998; revised 30 November 1998; accepted 7 December 1998 method (Garrow et al, 1979; Dempster & Aitkens, 1995; McCroy et al, 1995) particularly, in children. The aim of this present study was to carry out a prospective study of air displacement plethysmography as a method of quantifying body composition in children and adults. Methods Design The study was a cross-sectional evaluation of healthy children (age 19 y) and adults. Inclusion of adults allowed a comparison of present study results to the earlier ADP validation study (McCrory et al, 1995). The air displacement plethysmograph used in the present study was the BODPOD 1 Body Composition System (Life Measurement Instruments, Concord, CA) which costs approximately $ in the United States. The system was rst evaluated by comparing measured B d with B d quanti ed by H d. This allowed evaluation of ADP's accuracy in assessing B d. Comparing body fat estimates by ADP and H d provides no additional new information as both methods use the two compartment Siri equation (Siri, 1961) to convert B d to fat fraction. Body fat estimates by the two densitometric methods were therefore compared to body fat estimates by another independent method, DXA. We reasoned that in some subjects, B d estimates by ADP may be more reliable than by H d as the hydrodensitometry proce-

2 dure requires substantial subject participation. Our working hypothesis was that, even in a subject group capable of full water submersion, B d estimates by ADP may be more reliable than those by H d. Subjects and protocol All subjects were healthy children over the age of 5 y and adults age 20 y or older. Children were recruited from local schools in the New York City area and adult subjects were recruited through yers posted in the local community. The study protocol was approved by the Institutional Review Board of St. Luke's=Roosevelt Hospital and all subjects, including the children's parents, gave written consent prior to participation. Adult subjects reported to the Body Composition Laboratory of the Obesity Research Center after a 12 h overnight fast. All subjects completed the air displacement plethysmograph and DXA measurements and, since subject wetness may affect air displacement results H d was carried out last. Body composition measurements Air displacement plethysmography. A complete description of the principles of the ADP system is provided elsewhere (Dempster & Aitkens, 1995; McCrory et al, 1995). Brie y, the measurement of body composition is accomplished usually within ve minutes by applicaiton of gas laws within a two-chambered plethysmograph (McCory et al, 1995). Small volume changes are produced within the chamber and the corresponding pressure change is measured. The subject's body volume is determined by subtraction of empty chamber volume. Measured body volume is used in estimating B d with the equation of Goldman and Buskirk (Goldman & Buskirk, 1961) and fat mass is then calculated from body weight (Siri, 1961). We also examined Lohman's equaqtions (Lohman, 1989) for converting B d to percent fat in children and estimates were not measurably different from those by the Siri equation. Results are therefore presented for percent fat estimates by the Siri equation in order to maintain consistency between adults and children. Subjects, evaluated with minimal clothing, were weighed to the nearest 0.01 kg using the ADP system's electronic scale (Tanita Corp., Tokyo, Japan). Prior to body volume assessment, a two-point calibration was performed on the ADP with a standard l calibration cylinder. Since hair contributes to measurement error (Dempster & Aitkens, 1995; McCrory et al, 1995), subjects were required to wear an acrylic bathing cap. Body volume was determined as the subject sat quietly inside the ADP chamber for 60 s. An average of two trials was used in estimating body volume. The subject's thoracic gas volume or V TG was estimated during normal tidal breathing using a tube connected to the ADP breathing circuit system. Thoracic gas volume was determined as the subject puffed gently into the tube while the tube was mechanically obstructed. A detailed description of the procedure is given in earlier reports (Dempster & Aitkens, 1995; McCrory et al, 1995). Since McCrory et al found no signi cant difference between measured and predicted V TG in adults (McCrory et al, 1998), a predicted estimate of V TG based on age, sex, and height was used in cases where an accurate measurement of V TG was not possible. The between-day CV for B d estimates by ADP in our laboratory was determined in four adult subjects measured on four consecutive days. It was not possible to derive similar between-day CVs in children because of school and scheduling dif culties. Therefore, the betweenday CV for known water phantoms ranging from L ± 50.0 L was determined on four consecutive days. Hydrodensitometry. Body density was determined using a four-point platform scale system (Akers & Buskirk, 1969) (Precision Biomedical System Inc., University Park, PA). Subjects were measured in bathing suits and asked to maximally expel as much air as possible from their lungs during complete submersion. After 5 ± 10 trials were performed, an underwater weight was recorded as the average of the highest three values (Heyms eld et al, 1990). Residual lung volume was determined prior to underwater weighing using the nitrogen washout technique (Wilmore et al, 1969). Fat mass was then calculated from body weight and B d (Siri, 1961). The between-day CV for measurement of D d in our laboratory is g/cm 2 (Heyms eld et al, 1990). Subjects were subjectively scaled from 1 ± 10 by a single highly trained observer as a measure of under water performance, with 10 corresponding to the best performance. Subjects with a seven or lower score were excluded from further evaluation to insure the reliability of B d and estimates of total body fat. Dual energy X-ray absorptiometry. Total body mass was measured using DXA (Lunar DPX Corp., Madison, WI, software version 3.6 y and pediatric software version 1.5z in children ages 6 ± 18 y). The DXA system used in the present study generates photons at two principal energy levels (40 and 70 KeV) which allows measurements of bone mineral and soft tissue. Soft tissue can be further evaluated for fat and lean soft tissue mass. The betweenday CV for measurement of %fat by DXA in our laboratory is 3.1 % (Russel-Aulet et al, 1991). Statistical methods All statistical analyses were carried out using Statview (Abacus Concepts, Inc., Berkeley, CA, 1996) and group results are presented as mean standard deviation. The between-day CV for B d, %fat estimates and phantom experiments by ADP in our laboratory were calculated as SD=mean6100 %. Statistical analyses were performed for the total study population using paired t-tests to assess any differences between B d estimates by H d and ADP. Simple linear regression analysis was used to compare B d estimates by H d and ADP. Multiple linear regression analysis and residual plots were used to examine whether the relationship between H d and ADP is affected by gender, ethnicity, body weight, and age. Additional analyses were carried out for adult and child (age 19 y) subgroups. Paired t-tests were performed for each group to assess any difference between mean B d estimated by H d and ADP. Simple linear regression analysis was used to compare B d estimates by H d and ADP and multiple linear regression analysis was used to test whether the relationship of H d and ADP is affected by gender. Correlational analysis and Bland ± Altman plots (Altman & Bland, 1983) were used to compare %fat estimates by DXA with both H d and ADP estimates of %fat. Statistical signi cance was set at P <

3 384 Figure 1 Body density by ADP vs body density by H d for the total study population [B d Hdˆ (B d ADP ); r ˆ 0.93, SEE ˆ g=cm 3, P < ]. Results Subjects The baseline subject characteristics are presented for children in Table 1 and for adults in Table 2. For the total study population, there were 66 females (24 Caucasian, 18 African ± American, 10 Hispanic, 10 Asian and 4 multiracial) and 54 males (23 Caucasians, 4 African ± American, 12 Hispanic, and 11 Asian and 4 multiracial). There were no signi cant differences in age (P ˆ 0.06) or body weight (P ˆ 0.19) between males and females for children. There also was no signi cant difference in age (P ˆ 0.35) between males and females in the adults, although there was a signi cant (P < ) between-gender weight difference. Body-density comparisons The between-day CV for B d estimates by ADP in subjects ages 22 ± 33 y and weighing between 52.0 kg and 95.0 kg was g/cm 3. The results of the phantom measurements ranged from 0.07 ± 2.1 % with a group mean of %. An examination of the total study population revealed that mean B d estimates by H d and ADP were not signi cantly different (P ˆ 0.70) and the two B d estimates were Table 1 Baseline characteristics and results in children Table 2 Baseline characteristics and results in adults Females Males Females Males Number Age (y) Weight (kg) Height (cm) Body mass index (kg=m 2 ) B dhd (g=cm 3 ) * B d ADP (g=cm 3 ) * %fat Hd * %fat ADP * %fat DXA * *Signi cant gender differences at P < ; B d ˆ Body density; H d ˆ Hydrodensitometry; ADP ˆ Air Displacement Plethysmograph; DXA ˆ dual energy X-ray absorptiometry. Number Age (y) Weight (kg) * Height (cm) * Body mass index (kg=m 2 ) B dhd (g=cm 3) * B d ADP (g=cm 3) * %fat Hd * %fat ADP * %fat DXA * * Signi cant gender differences at P < ; B d ˆ Body density; H d ˆ Hydrodensitometry; ADP ˆ Air Displacement Plethysmograph; DXA ˆ dual energy X-ray absorptiometry.

4 highly correlated [Figure 1, B dhd ˆ (B d ADP ); r ˆ 0.93, SEE ˆ g=cm 3, P < ]. The slope, however, was signi cantly different from (t ˆ 28.3, P < ) and the intercept of the regression line was signi cantly different from 0 (t ˆ 3.35, P < 0.01). Multiple linear regression analysis showed that the relationship between B d estimates by H d and ADP was not affected by gender (P ˆ 0.60) or ethnicity P ˆ 0.94). The addition of body weight (P < 0.01) or age (P < 0.001), however, made a signi cant contribution to the regression model of H d vs ADP. When both body weight and age were entered into the model, body weight was removed as a variable from the equation and age explained an additional 2 % of the variance. The in uence of age was further evaluated by examining the correlation of age with residuals from the equation estimating H d from ADP. A signi cant negative correlation was observed between the calculated residuals and age Figure 2, r ˆ , P < 0.001). At low age ranges, ADP underestimated density in comparison to H d. There was also greater variability in B d estimates by ADP at the lower age ranges. When we examined the adult group separately, mean B d estimates by both H d and ADP were not signi cantly different (P ˆ 0.69) and the two body density estimates were highly correlated [B d Hdˆ (B d ADP ); r ˆ 0.95, SEE ˆ g=cm 3, P < ]. The slope and intercept of the regression line were not signi cantly different from 1 and 0, respectively. Mean B d estimates in children by H d and ADP were also not signi cantly different (P ˆ 0.58) and the two body density estimates were highly correlated [Figure 3, B dhd ˆ (B d ADP ); r ˆ 0.91, SEE ˆ g=cm 3, P < 0.001]. The slope, however, was signi cantly different from 1 (t ˆ 14.9, P < ) and the intercept of the regression line was signi cantly different from 0 (t ˆ 4.39, P < ). This observation suggests an existing bias in the estimation of B d by ADP in children. None of the B d ndings in children or adults were in uenced by removal of subjects with predicted V TG (n ˆ 7 adults and 16 children) from the analyses. Body fat comparisons The between-day CV for %fat estimates by ADP in adults was %. A zero order correlation matrix for % fat measurements is presented in Table 3 with the upper triangle representing correlation coef cients (that is, r values) for adults and the lower triangle presents correlation coef cients for children. There was a high correlation in adult subjects between %fat estimates by DXA and ADP [ %fat DXA ˆ ( %fat ADP) ; r ˆ 0.94, SEE ˆ 3.53, P < ] and between DXA and H d [ %fat DXA ˆ (%fat Hd ; r ˆ 0.91, SEE ˆ 4.20, P < ]. A high correlation was also observed in children for %fat estimates by DXA and ADP [ %fat DXA ˆ ( %fat ADP ); r ˆ 0.90, SEE ˆ 4.05, P < ] and DXA and H d [%fat DXA ˆ (%fat Hd ); r ˆ 0.91, SEE ˆ 3.84, P < ]. Bland ± Altman analysis in the adult subgroup did not reveal any systematic differences between %fat by DXA vs ADP P ˆ 0:53 and %fat by DXA vs H d P ˆ 0:45. Bland ± Altman analyses in the children subgroup also did not reveal a systematic difference between %fat by DXA vs H d (P ˆ 0.24). The analysis for %fat by DXA vs ADP as also non-signi cant in children but showed a trend (P ˆ 0.11) for an under prediction of %fat by ADP at low fat ranges and an overprediction of %fat at high fat ranges. 385 Figure 2 Body density residuals for H d vs ADP as a function of age [r ˆ , P < 0.001].

5 386 Figure 3 Body density by ADP vs body density by H d for child subgroup [B dhdˆ (B d ADP ; r ˆ 0.91, SEE ˆ g=cm 3, P < 0.001]. Table 3 Zero-order correlation matrix for percent fat estimates %fat DXA %fat H d %fat ADP %fat DXA * 0.935* %fat H d 0.910* * %fat ADP 0.899* Upper-right panel adults and lower left panel children. *P < DXA ˆ dual energy X-ray absorptiometry H d ˆ Hydrodensitometry; ADP ˆ Air Displacement Plethysmograph. Discussion This present study was designed to test the air displacement plethysmography (ADP) body composition system for the rst time in children and to also compare ndings in adults to the original publication (McCrory et al, 1995) which was also carried out in adults. Firstly, body density measured by ADP was compared to the accepted reference method, H d. Body density estimates by ADP and H d are used to estimate body fat using one of several two-compartment models. The second phase of the experiment was to compare similarly calculated body fat, using Siri's two compartment model equation (Siri, 1961), from ADP and H d to another independent method of quantifying total body fat, DXA. Observation in adults Our ndings strongly support ADP body composition estimates in adults and con rm earlier observations (McCrory et al, 1995). Speci cally, we observed high correlations and similar mean values for ADP body density (r ˆ 0.95, SEE ˆ g=cm 3 ) and body fat (r ˆ 0.93, SEE ˆ 3.53) estimates compared to corresponding estimates by H d and DXA, respectively. McCrory et al, 1995, also reported in adults a strong correlation n ˆ 68, r ˆ 0.96, SEE ˆ 1.81) with no observed bias between ADP and H d %fat estimates. We also observed a between-day CV of g=cm 3 for repeated measures of B d by ADP, which is under 1 % and comparable to reported CVs for H d (McCrory et al, 1995). Additionally, a between-day CV of 2 % for repeated measures of total body fat by ADP is slightly higher than reported by McCrory in their earlier study (McCrory et al, 1995), but acceptable given the range in body weights of our four adults subjects (52.0 kg ± 95.0 kg). Taken collectively, these results strongly support ADP as an alternative to H d in adult body composition assessment. Observation in children Results in children differed from those observed in adults. A high correlation between ADP body density and H d was also observed (r ˆ 0.91, SEE ˆ g=cm 3, although a small but statistically signi cant bias was detected. Similarly, a trend (P ˆ 0.11) toward bias was observed between ADP %fat estimates compared to corresponding %fat estimates by DXA. Additionally, a plot of residuals showed greater variability of B d estimates by ADP at low age ranges. The observation of a small, but consistent bias in both B d and %fat estimates by ADP suggests a relatively minor system calibration problem. Our initial analyses indicate that V TG prediction, rather than measurement, is not the source of observed B d and %fat bias. We would anticipate that as subject volume decreases relative to chamber volume, random measurement error would increase. The ADP manufacturer also recommends

6 system use only at body weights above 40 kg and our study consisted of 12 children with body weights below 40 kg with the lowest weight child at 24 kg. However, our experience is that many investigators are applying the ADP method in young children who weight less than the recommended 40 kg. While increasing random ADP measurement error can be expected in smaller subjects, the cause of observed B d bias remains uncertain. The possibility exists, however, in future studies to identify and correct the source of ADP measurement bias in children, since the results of our phantom experiments indicate high reproducibility by the ADP system with a CV of under 1 %. Moreover, smaller air displacement plethysmograph systems can be designed speci cally for pediatric populations. The importance of the ADP system for pediatric patients is that body composition estimates can be obtained without the need for water submersion as with H d or exposure to ionizing radiation, however minimal, as with DXA. Additionally, other than the small observed bias, an examination of the correlation between all three estimates of %fat revealed a slightly higher correlation between ADP and DXA (r ˆ 0.94) and a corresponding lower standard error of estimate (SEE ˆ 3.53) compared with H d vs DXA (r ˆ 0.91, SEE ˆ 4.20) in adults subjects. The data for children showed a slightly higher correlation for H d vs DXA (r ˆ 0.91, SEE ˆ 3.84) than for ADP vs DXA (r ˆ 0.90, SEE ˆ 4.05) which may be a re ection of the error involved in ADP estimates of body composition in this group. An important consideration in this present study is that subjects with sub-optimal under water weighing tests were not entered into the protocol. An expectation is that subjects who are unable to complete a successful submersion procedure would have an accurate measurement of B d. Our hydrodensitometry results are therefroe optimized and may not re ect results obtained in the general population. Our clinical impression is that about ve percent of adults have H d scores below eight and the proportion of unsuccessful tests is even higher in children. These observations underscore the relevance of correcting any minor calibration errors in ADP as this method does not require familiarity or comfort with water submersion. Conclusion This present study results con rm earlier reports of ADP validity in adult subjects and highlight the conceptual validity of air displacement plethysmography as a means of quantifying B d. Our results, extended for the rst time to a pediatric sample, af rm ADPs overall validity but also reveal a small and consistent measurement bias. The nonsigni cant %fat error by ADP compared to DXA would have little consequence for evaluation of the individual child, particularly in the clinical setting. The potential likely exists for correction of the ADP B d measurement bias with identi cation of the speci c error source. Acknowledgements ÐThe authors thank Karen Rosenbaum, Maike Rahn, Ben Schneider, Jack Wang, David Yu, and Dr. Dympna Gallagher for their technical support on this project. References Akers R & Buskirk ER (1969): An underwater weighing system utilizing `force cube' transducers. J. Appl. Physiol. 26, 649 ± 652. Altman DG & Bland JM (1983): Measurement in Medicine; the analysis of method comparison studies. Statistician 32, 307 ± 317. Dempster P & Aitkens S (1995): A new air displacement method for the determination of human body composition. Med. Sci. Sports Exerc. 27, 1692 ± Garrow JS, Stally S, Diethelm R, Pittet P, Hesp R & Halliday D (1979): A new method for measuring the body density of obese adults. B. J. Nutr. 42, 173 ± 183. Goldman D & Buskirk ER (1961): A method for underwater weighting and the determination of body density. In: Techniques for Measuring Body Composition, eds. J Brozek, A Henschel, pp 78 ± 106. Washington DC: National Academy of Sciences. Heyms eld SB, Lichtman S, Baumgartner RN, Wang J, Kamen Y, Aliprantis A & Pierson RN Jr (1990): Body composition of humans: comparison of two improved four compartment models that differ in expense, technical complexity, and radiation exposure. Am. J. Clin. Nutr. 52, 52 ± 58. Lohman TG (1989): Assessment of body composition in children. Ped. Exer. Sci. 1, 19 ± 30. McCrory, MA, Gomez TD, Bernauer EM & Mole PA (1995): Evaluation of a new air displacement plethysmograph for measuring human body composition. 27, 1686 ± McCrory MA, Mole PA, Gomez TD, Dewey KG & Bernauer EM (1998): Body composition by air-displacement plethysmography by using predicted and measured thoracic gas volumes. J. Appl. Physiol. 84, 1475 ± Russel-Aulet M, Wang J, Thornton J & Pierson RN Jr (1991): Comparison of dual-photon absorptiometry systems for total body bone and soft tissue measurements: dual energy X-rays vs Gadolinium 153. J. Bone Min. Res. 6, 411 ± 415. Siri WE (1961): Body Composition from uid spaces and density: Analysis of methods. In: Techniques for Measuring Body Composition, eds. J Brozek, A Henschel, pp 223 ± 244. Washington DC: National Academy of Science. Wang ZM, Deurenberg P, Guo SS, Pietrobelli A, Wang J, Pierson RN Jr & Hemys eld SB (1988): Six compartment body composition model: Inter-method comparisons of total body fat measurement. Int. J. Obes. 22, 329 ± 337. Wilmore JA (1969): A simpli ed method for determation of residual lung volume. J. Appl. Physiol. 27, 96 ±