![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
,2
* Department of Family Medicine and Community Health, Tufts University School of Medicine, Boston, MA;
Division of Pediatric Gastroenterology and Nutrition, New England Medical Center, Boston, MA;
** Clinical Research Center, Massachusetts Institute of Technology, Cambridge, MA; and
Gerald J. and Dorothy R. Friedman School of Nutrition Science and Policy, Tufts University, Boston, MA
2To whom correspondence should be addressed. E-mail: aviva.must{at}tufts.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT SUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
KEY WORDS: • bioelectrical impedance • adolescence • longitudinal analysis • body composition • isotopic dilution
Bioelectrical impedance analysis (BIA) is a widely used technique for estimating body composition and it is particularly useful in large, population-based studies because it is quick, portable, inexpensive and noninvasive. Numerous cross-sectional studies, using a variety of reference methods, have shown that BIA predicts total body water (TBW) (1 –3 ), fat-free mass (FFM) (4 –7 ) and fat mass or percentage body fat (%BF) (8 ,9 ) well among healthy children of various ages. From these studies, several formulas were developed for the prediction of TBW, FFM and %BF in children. Less is known, however, about their ability to estimate the rate of change in FFM and %BF in a longitudinal setting. In addition to estimation of the rate of change (slope), the ability to estimate levels of FFM and %BF in absolute terms is also of substantial interest.
Previous studies that have used BIA-based estimates to evaluate change in FFM or BF have typically assessed change using a simple pre/post comparison. This approach, by definition, fails to consider whether the pattern of change over time is well approximated by a straight line. Furthermore, the published equations in the literature, although developed cross-sectionally against reference measures, have not been tested longitudinally against them.
To our knowledge, no previous study has examined the ability of BIA to detect long-term changes in body composition among healthy girls using a longitudinal analysis approach. Longitudinally collected data require special techniques for analysis because observational studies often have unbalanced designs and/or missing data and because repeated measurements taken on the same individual are correlated with each other.
Using data from a 10-y longitudinal study of growth and development in adolescent girls, we introduce the use of longitudinal models in the assessment of body composition methodology over the adolescent period and present a new method for evaluating and comparing the results of longitudinal models ("slope-intercept" plots). The objectives of these analyses were as follows: 1) to develop our own equation to predict FFM using a reference measure of TBW collected in our sample; and 2) to use longitudinal data analysis approaches to evaluate our equation as well as other published equations to assess their performance against reference measures of body composition.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
Between September 1990 and June 1993, we enrolled 196 girls in the Massachusetts Institute of Technology (MIT) Growth and Development Study. The criteria for enrollment were premenarcheal status and a triceps skinfold less than the 85th percentile for age and sex (10 ). Premenarcheal girls aged 8–12 y were recruited from the Cambridge and Somerville public schools in Massachusetts, the MIT summer day camp, and friends and siblings of enrolled participants. All participants were healthy, free of disease and were not taking any medication at study entry. Participants were followed annually until 4 y postmenarche. The racial composition of the cohort was 72% white, 15% black and 13% from other races. At baseline, 63% of the girls were prepubertal (Tanner stage 1) and 37% were pubertal but premenarcheal (Tanner stages 2 and 3). The study protocol was approved by both the Committee on the Use of Humans as Experimental Subjects at MIT and the Human Investigation Review Committee of the New England Medical Center.
Study protocol.
TBW measurements at both baseline and exit were available for 69% of participants. In addition, within the cohort, some girls participated in supplementary hypothesis-driven substudies in which TBW was measured at additional time points. In one substudy, girls who were 12 y old at the second follow-up visit were invited to participate in a longitudinal study of energy expenditure. In the other, girls at menarche were invited to participate in a study of changes in visceral fat measured by MRI at menarche and 4 y postmenarche. Girls were admitted to the General Clinical Research Center at MIT for an overnight visit. At the time of their initial visit, a physician obtained a medical history and examined each participant to ensure that she was in good health. On the evening of admission, participants consumed no food or beverages after 1800 h. At 2000 h, the staff obtained baseline urine and each participant was given 0.25g H218O/kg estimated TBW. After administering the isotope, the study staff collected all urine voided until 0600 h the next morning to determine the loss of isotope in the urine. The second void of the morning was collected at 
0800 h to measure 18O enrichment above baseline values. Isotopic analyses for assessment of body composition were conducted at the USDA Human Nutrition Research Center at Tufts University (Boston) on two isotope ratio mass spectrometers (Hydra Gas; PDZ Europa LTD, Northwich, UK, and SIRA 10; Micromass, Altrincham, UK). Urinary enrichment of 18O was determined by mass spectroscopy as described elsewhere (11
). The CV of TBW determined from 18O dilution in urine was 2% (12
). Oxygen dilution space was calculated according to Halliday and Miller (13
). We assumed that the 18O dilution space was 1% higher than TBW (12
).
Height and body weight were measured in the morning. Height was measured to 0.1 cm with a wall-mounted stadiometer. Fasting weight was measured in a hospital gown using a Seca scale (Seca Corporation, Hanover, MD) accurate to 0.1 kg.
Bioelectrical impedance analysis was conducted at the time when TBW was measured. Resistance (R) and reactance were measured after an overnight fast with the participant supine (Bioelectrical impedance analyzer, BIA 101, RJL, Clinton Township, MI). The accuracy of the machine was checked before the measurements with a 500-
resistor supplied by the manufacturer. Electrodes were placed on the dorsal surface of the right foot and ankle, and right wrist and hand. A current was applied at a frequency of 50 kHz. Previous analyses conducted in this cohort indicated that FFM based on a measure of BIA is highly reproducible (3
).
Study variables.
Both isotopic dilution and BIA provide a measure of total body water. Because we were interested in FFM and %BF, it was necessary to convert TBW into FFM, applying the assumption that FFM is 73% water (14
). We will refer to FFM calculated from TBW as FFMref to indicate FFM by the reference method. The percentage body fat estimated from TBW was calculated from the measures of body weight and FFMref according to the following equation: %BFref = [Weight (kg) - FFMref (kg)]/Weight (kg) x 100. BMI was calculated as weight (kg) divided by height (m) squared. A BMI Z-score was calculated using the revised Centers for Disease Control and Prevention growth reference standards (15
). The impedance index (H2/R) was calculated as height (cm squared) divided by resistance ().
|
Statistical analyses |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
Using our data, we first developed a regression equation, which we will refer to as the MIT equation, to predict FFMref from BIA and anthropometric measures. Previous analyses conducted in this cohort suggested that the accretion of both FFM and body fat was related to maturational status (16 ). Therefore, we modeled FFMref separately pre- and postmenarche, resulting in two prediction equations, one for premenarche and one for postmenarche (16 ). Generating separate equations based on maturational status also helped avoid problems of collinearity between age and other model predictors.
Our data were structured using a "long" format, such that a girl’s measurements at a given visit represented one row in the data set. For example, if a girl had three measurements of TBW and BIA conducted at three different visits (and hence three different ages), she would provide three "rows" of data to the analysis. Using this format, the 196 participants provided 422 measurements (or rows) that were available for analysis. The distribution of measurement numbers was as follows: 52 participants had one measure of TBW, 84 participants had two measures of TBW, 42 participants had three measures of TBW, 14 participants had four measures of TBW and 4 participants had five measures of TBW.
In the first step of the model development process, participants were randomly assigned to one of two data sets: a model development data set and a model validation data set. Each participant appeared in only one of the two data sets. Because we had participants with varying numbers of measurements, the selection of participants into one of the two data sets was done randomly and in blocks, so that participants with one, two, three, four or five TBW measurements were equally distributed between the two data sets.
To develop the MIT prediction equation, stepwise multiple linear regression analysis was conducted in the model development and model validation data sets using FFMref as the outcome. Within each data set, separate models were generated pre- and postmenarche. Possible predictors considered included height2/resistance, weight, height and reactance. Race, coded as two dummy variables for black and "other," with white as the reference category, was included in all models. These two equations developed in our data set will be referred to as FFMMITpre and FFMMITpost. It is important to note that although we developed two prediction equations, one for premenarche and one for postmenarche, for a given measurement for a specific girl, a single predicted FFM value was generated. Thus, predicted values for FFM were generated in the model development and model validation data sets. The model parameters, R2 and residual plots were compared from the model development and model validation data sets. An independent-samples t test was used to test for significant differences in predicted FFM between the two data sets.
The data from the two sets were then recombined and the stepwise regression procedure was repeated. The parameters for height2/resistance, height, weight and race (categorized as white, black, "other") generated from this full set (n = 422) constituted our final model. Then, using FFM predicted from our final model, we calculated %BFMIT according to the following equation: [Weight (kg) - FFMMIT (kg)]/Weight (kg) x 100. In addition, predicted FFM and %BF were calculated using equations published by Deurenberg et al. (4 ,6 ,17 ), Houtkooper et al. (7 ), Kushner et al. (18 ) and Schaefer et al. (19 ) (Table 1). FFM calculated by the equation of Houtkooper et al. (7 ) was adjusted by 4% to correct for the overestimation of TBW by deuterium dilution.
|
After our prediction equation was developed, our goal was to apply our equation and other published equations to our data and assess their performance over time using longitudinal data analysis approaches. To accomplish this, we first calculated the difference between the reference estimates of FFMref and %BFref and the estimates of FFM and %BF generated from the BIA prediction equations. Next, this difference was plotted against age using nonparametric robust local smoothing procedures (20 ). Although this technique ignores the correlation structure in repeated measurements, these plots allowed us to visualize the general pattern of the relations between age and error in estimating FFM and %BF. Second, linear mixed-effects modeling (LME) was used to evaluate the change in FFM and %BF over time. In these models, age was treated as a random effect and a fixed effect to express the trend of change over time in the whole population. This approach characterizes individual variation relative to the population mean while taking into account the correlation between repeated measurements on the same participant as well as different numbers of measurements per participant. Using both reference and predicted values of FFM and %BF as the outcome, the LME models generated a set of slopes and intercepts associated with each model, along with their corresponding error measures. The width of the 95% confidence intervals (CI) around the slopes and intercepts reflects the variation around individual slopes and intercepts and was used to assess the precision of the various equations.
The results of the LME models were visualized using slope-intercept plots, which allow one to summarize and compare graphically the model results for both the slope and intercept simultaneously. All analyses were performed using SPSS (Version 10.1, SPSS, Chicago, IL), SAS (Version 8.0, SAS Institute, Cary, NC) and S-PLUS (Version 4.5, MathSoft, Seattle, WA).
|
RESULTS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
At study entry, we had complete data for 196 participants. As expected, FFM and %BF increased between baseline and exit (Table 2; paired t test, P < 0.001). By either H218O or BIA, the mean %BF at study entry was 23% (n = 196) and the mean %BF at 4 y postmenarche was 27% (n = 133) (Table 2). We observed no significant difference between FFM values predicted in the model development and model validation data sets (t test; P > 0.05). We evaluated the predictive ability (R2) and standard error of the estimate (SEE) of four multiple regression models that predicted FFMref from our set of predictors (Table 3). Height2/resistance entered both models first and accounted for 91% of the variation in FFM premenarcheally and 79% of the variation in FFM postmenarcheally. Weight and height entered next in both models, adding a small amount of explanatory capability over height2/resistance alone. Reactance entered the models, but because its inclusion did not substantially improve R2 values, the more parsimonious model was chosen (Table 3). For any given set of explanatory variables, the R2 values were higher and SEE were lower before menarche than after menarche, indicating increased variability in FFM postmenarcheally. The best equations to predict FFMref in our dataset pre- and postmenarche were as follows: FFMMITpre = -5.508 + (0.420 x height2/resistance) + (0.209 x weight) + (0.08593 x height) + (0.515 x black race) - (0.02273 x other race); FFMMITpost = -11.937 + (0.389 x height2/resistance) + (0.285 x weight) + (0.124 x height) + (0.543 x black race) + (0.393 x other race). The pre- and postmenarche equations differ mainly in the intercept, i.e., the premenarche intercept is about –5.5 and the postmenarche intercept is about –12.
|
|
Before assessing the performance of the MIT and other published equations longitudinally, we evaluated the prediction errors cross-sectionally. For each measurement, the difference between FFMref or %BFref and the estimates of FFM and %BF predicted from the various equations was calculated (Table 4). In general, the published predictive equations underestimated FFM and overestimated %BF. As expected, the MIT equation performed the best for both FFM and %BF. In this cross-sectional analysis, the equations of Kushner et al. (18 ) and Houtkooper et al. (7 ) had the smallest mean prediction errors. Overall, the variability in the error estimates was smaller for FFM than for %BF.
|
|
Our final step, as described in the Subjects and Methods section, was to use LME models to generate a set of slopes and intercepts for FFM and %BF (Table 5). These coefficients represent the populational slopes and intercepts, which are the weighted average of the slopes from the 196 participants. As expected, the MIT equation was closest to the reference in terms of both slope and intercept for FFM. Similarly, the MIT equation was closest to the reference in terms of both slope and intercept for %BF. For estimates of FFM, the slope for age was relatively consistent across all of the equations; the values for the intercept, however, varied considerably. For %BF, the values for both intercept and slope varied to differing degrees from the reference estimates. The 95% CI around the slopes and intercepts for both FFM and %BF indicated that the precision around the estimates was relatively good for all of the equations examined (i.e., CI were fairly narrow). The range of CI widths around the age slope was very similar for both FFM (range = 0.13–0.25) and %BF (range = 0.15–0.23). For intercept values, the range of widths was slightly narrower for FFM (range = 1.7–2.7) than for %BF (range = 2.1–3.5). Although the precision of the various equations varied little, the validity of the equations did. For example, for FFM, the Schaefer equation (20 ) has the narrowest CI around the intercept and age slope but it was farthest from the reference estimate of FFM in terms of validity.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
We had an opportunity to examine how well BIA estimates changes in FFM and %BF as measured by H218O in a relatively homogeneous, initially nonobese premenarcheal cohort of 196 girls followed through adolescence. There is very little information concerning whether BIA can detect changes in body composition during the rapid period of adolescent growth. Our findings indicate that BIA is a useful technique for estimating change in body composition over time. Our prediction equation, based on a reference measurement of TBW to estimate FFM, performed significantly better cross-sectionally and over time in this cohort of girls than other published prediction equations. Furthermore, our results are strengthened by the use of longitudinal analysis techniques that allow the changes that occur over time to be characterized most accurately.
The finding that the prediction equation developed in our cohort performed better than other published prediction equations highlights the importance of equation selection when using BIA. Of the published equations examined in our analysis, the equation of Deurenberg et al. (4 ), developed in a large cohort of 827 male and female subjects between 7 and 83 y of age, performed best in terms of both FFM and %BF and did not differ significantly from the reference in estimating FFM. Other equations considered were based on smaller samples, ranging from 94 to 246 male and female subjects. The age ranges in the other studies varied, with some including both children and adults (4 ,18 ), and others including only children and young adults (6 ,7 ,19 ). The discrepancy among published equations may be due to different methodologies used to determine FFM. Schaefer et al. (19 ) based their estimates on 40K counting, Deurenberg et al. (4 ,6 ,17 ) on hydrodensitometry, and Houtkooper et al. (7 ) and Kushner et al. (18 ) on isotopic dilution with deuterium and H218O, respectively. Each of these methods has its own set of assumptions.
Several authors have cautioned that the ability of BIA to accurately assess body composition varies with the prediction equation used. In a longitudinal assessment of body composition change in adults with acquired immune deficiency syndrome, Paton et al. (28 ) found that different conclusions were drawn about the proportion of FFM change depending on the BIA equation used. Our experience lends support to others who have recommended that researchers avoid applying prediction formulas from the literature to their own data without first validating the formulas in their subjects (4 ).
Our study has some noteworthy limitations. Both isotopic dilution and BIA assess TBW, not FFM. We assumed that the hydration of FFM is 73% water. Variability in the hydration constant may introduce errors in the calculation of FFM and %BF. This variability reflects both individual variation and variability due to the age-related decline in hydration, believed to be decreasing during childhood toward the adult value of 0.73. Studies that have used a multicompartmental model to determine the hydration status of fat-free mass have had varying results. Roemmich et al. (29 ) used a four-component model to estimate water content of FFM. They found a hydration factor of 0.755 ± 0.067 in the prepubertal girls (mean age 10.4 y) and a hydration factor of 0.744 ± 0.052 in the pubertal girls (mean age 13.3 y). Another study that used a multicompartmental approach found a FFM water constant of 0.722 ± 0.014 in prepubertal children (mean age 8.5 y) compared with 0.707 ± 0.013 in young women (30 ). Although these hydration factors differ from the 73% determined from cadaver analyses, they also differ among studies. Although it seems clear that hydration factors vary among individuals, it is not clear that the variability within subjects exceeds the variation with age. Thus, the exact hydration factor of children and adolescents has not been established. In a study that included almost 1200 children, the hydration factor appears to be stable over the age range in our study (M. Horlick, Columbia University College of Physicians and Surgeons, personal communication). Therefore, any error in the hydration factor would be constant over adolescence and would not affect the comparison of longitudinal model parameters.
In summary, we developed a specific BIA equation for use in our population of girls throughout the adolescent period, using H218O as the reference standard. We then applied our equation and various other published equations developed among external populations to our data. Using longitudinal analysis techniques, we examined the ability of these equations to estimate changes in FFM and %BF over time compared with measures from the reference method. The slope-intercept plots introduced represent a useful way to summarize and compare the results of longitudinal linear models. Our results show that BIA provides accurate estimates of the pattern of change in FFM and %BF over time, but that the accuracy depends on the equation chosen. Our experience suggests that researchers who use longitudinal designs to study changes in body composition with indirect methods must be certain that their equations accurately reflect change over the period studied. Given accurate equations developed over the time period of interest, estimates of changes in FFM and %BF by BIA will be accurate as well when compared with a gold standard.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
3 Abbreviations used: %BF, percentage body fat; BIA, bioelectrical impedance analysis; FFM, fat-free mass; LME, linear mixed-effects modeling; MIT, Massachusetts Institute of Technology; R, resistance; TBW, total body water.
Manuscript received 23 October 2002. Initial review completed 19 November 2002. Revision accepted 9 January 2003.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSStatistical analysesRESULTSDISCUSSIONLITERATURE CITED |
|---|
1. Davies, P.S.W., Preece, M. A., Hicks, C. J. & Halliday, D. (1988) The prediction of total body water using bioelectrical impedance analysis in children and adolescents. Ann. Hum. Biol. 15:237-240.[Medline]
2. Danford, L. C., Schoeller, D. A. & Kushner, R. F. (1992) Comparison of two bioelectrical impedance analysis models for total body water measurement in children. Ann. Hum. Biol. 19:603-607.[Medline]
3. Bandini, L. G., Vu, D. M., Must, A. & Dietz, W. H. (1997) Body fatness and bioelectrical impedance in non-obese pre-menarcheal girls: comparison to anthropometry and evaluation of predictive equations. Eur. J. Clin. Nutr. 51:673-677.[Medline]
4. Deurenberg, P., van der Kooy, K., Leenen, R., Westrrate, J. A. & Seidell, J. C. (1991) Sex and age specific prediction formulas for estimating body composition from bioelectrical impedance: a cross-validation study. Int. J. Obes. 15:17-25.[Medline]
5. Deurenberg, P., van der Kooy, K., Paling, A. & Withagen, P. (1989) Assessment of body composition in 8–11 year old children by bioelectrical impedance. Eur. J. Clin. Nutr. 43:623-629.[Medline]
6. Deurenberg, P., Kusters, C. & Smit, H. E. (1990) Assessment of body composition by bioelectrical impedance in children and young adults is strongly age-dependent. Eur. J. Clin. Nutr. 44:261-268.[Medline]
7. Houtkooper, L., Going, S. B., Lohman, T. G., Roche, A. F. & Van Loan, M. (1992) Bioelectrical impedance estimation of fat-free body mass in children and youth: a cross-validation study. J. Appl. Physiol. 72:366-373.
8. Houtkooper, L., Lohman, T. G., Going, S. B. & Hall, M. C. (1989) Validity of bioelectric impedance for body composition assessment in children. J. Appl. Physiol. 66:814-821.
9. Young, R. E. & Sinha, D. P. (1992) Bioelectrical-impedance analysis as a measure of body composition in a West Indian population. Am. J. Clin. Nutr. 55:1045-1050.
10. Must, A., Dallal, G. E. & Dietz, W. H. (1991) Reference data for obesity: 85th and 95th percentiles of body mass index (wt/ht2) and triceps skinfold thickness. Am. J. Clin. Nutr. 53:839-846.
11. Wong, W. W., Lee, L. S. & Klein, P. D. (1987) Deuterium and oxygen-18 measurements on microliter samples of urine, plasma, saliva, and human milk. Am. J. Clin. Nutr. 45:905-913.
12. Schoeller, D. A., van Santen, E., Peterson, D. W., Dietz, W., Jaspan, J. & Klein, P. D. (1980) Total body water measurement in humans with 18O and 2H labeled water. Am. J. Clin. Nutr. 33:2686-2693.
13. Halliday, D. & Miller, A. G. (1977) Precise measurement of total body water using trace quantities of deuterium oxide. Biomed. Mass Spectrom. 4:82-87.[Medline]
14. Pace, N. & Rathbun, E. N. (1949) Studies of body composition III. The body water and chemically combined nitrogen content in relation to fat content. J. Biol. Chem. 158:685-691.
15. Ogden, C. L., Kuczmarski, R. J., Flegal, K. M., Mei, Z., Guo, S., Wei, R., Grummer-Strawn, L. M., Curtin, L. R., Roche, A. F. & Johnson, C. L. (2002) Centers for Disease Control and Prevention 2000 growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics 109:45-60.
16. Naumova, E. N., Must, A. & Laird, N. M. (2001) Tutorial in biostatistics: evaluating the impact of ’critical period’ in longitudinal studies of growth using piecewise mixed effects models. Int. J. Epidemiol. 30:1332-1341.
17. Deurenberg, P., Smit, H. E. & Kusters, C. (1989) Is the bioelectrical impedance method suitable for epidemiological field studies?. Eur. J. Clin. Nutr. 43:647-654.[Medline]
18. Kushner, R. F., Schoeller, D. A., Fjeld, C. R. & Danford, L. (1992) Is the impedance index (ht2/R) significant in predicting total body water?. Am. J. Clin. Nutr. 56:835-839.
19. Schaefer, F., Georgi, M., Zieger, A. & Scharer, K. (1994) Usefulness of bioelectric impedance and skinfold measurements in predicting fat-free mass derived from total body potassium in children. Pediatr. Res. 35:617-624.[Medline]
20. Hastie, T. J. & Tibshirani, R. J. (1990) Generalized Additive Modeling 1990 Chapman & Hall St. Edmunds, Suffolk, UK.
21. Kushner, R. F., Kunigk, A., Alspaugh, M., Andronis, P. T., Leitch, C. A. & Schoeller, D. A. (1990) Validation of bioelectrical-impedance analysis as a measurement of change in body composition in obesity. Am. J. Clin. Nutr. 52:219-223.
22. Evans, E. M., Arngrimsson, S. A. & Cureton, K. J. (2001) Body composition estimates from multicomponent models using BIA to determine body water. Med. Sci. Sports Exerc. 33:839-845.[Medline]
23. Powell, L. A., Nieman, D. C., Melby, C., Cureton, K., Schmidt, D., Howley, E. T., Hill, J. O., Mault, J. R., Alexander, H. & Stewart, D. J. (2001) Assessment of body composition change in a community-based weight management program. J. Am. Coll. Nutr. 20:26-31.
24. Vazquez, J. A. & Janosky, J. E. (1991) Validity of bioelectrical-impedance analysis in measuring changes in lean body mass during weight reduction. Am. J. Clin. Nutr. 54:970-975.
25. Deurenberg, P., Weststrate, J. A. & Hautvast, J. (1989) Changes in fat-free mass during weight loss measured by bioelectrical impedance and by densitometry. Am. J. Clin. Nutr. 49:33-36.
26. van der Kooy, K., Leenen, R., Deurenberg, P., Seidell, J. C., Westerterp, K. R. & Hautvast, J. (1992) Changes in fat-free mass in obese subjects after weight loss: a comparison of body composition measures. Int. J. Obes. 16:675-683.
27. Wabitsch, M., Braun, U., Heinze, E., Muche, R., Mayer, H., Teller, W. & Fusch, C. (1996) 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. 64:1-6.
28. Paton, N. I., Macallan, D. C., Jebb, S. A., Noble, C., Baldwin, C., Pazianas, M. & Griffin, G. E. (1997) Longitudinal changes in body composition measured with a variety of methods in patients with AIDS. J. Acquir. Immune Defic. Syndr. Hum. Retrovirol. 14:119-127.[Medline]
29. Roemmich, J. N., Clark, P. A., Weltman, A. & Rogol, A. D. (1997) Alterations in growth and body composition during puberty. I. Comparing multicompartment body composition models. Am. J. Physiol. 83:927-935.
30. Hewitt, M. J., Going, S. B., Williams, D. P. & Lohman, T. G. (1993) Hydration of the fat-free body mass in children and adults: implications for body composition assessment. Am. J. Physiol. 265:E88-E95.
This article has been cited by other articles:
|
|
 |
|
  L. G Bandini, A. Must, S. M Phillips, E. N Naumova, and W. H Dietz Relation of body mass index and body fatness to energy expenditure: longitudinal changes from preadolescence through adolescence Am. J. Clinical Nutrition, November 1, 2004; 80(5): 1262 - 1269. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
