12 Pages • 5,394 Words • PDF • 200.8 KB

Uploaded at 2021-09-24 07:50

This document was submitted by our user and they confirm that they have the consent to share it. Assuming that you are writer or own the copyright of this document, report to us by using this DMCA report button.

267

Prediction of fat and fat-free mass using DXA

Fig. 1. Comparison of bioelectrical impedance analysis and skinfold-predicted fat against DXA percent fat. Predictions indicate linear regression and 95% con® dence intervals. J&P(3) = the three-site calliper prediction of Jackson and Pollock (1978); D&W = the four-site calliper prediction of Durnin and Womersley (1974); Lukaski = the bioelectrical impedance analysis prediction of Lukaski and Bolonchuk (1987); Lohman = the bioelectrical impedance analysis prediction of Lohman (1992).

(coeý cient of determination = 0.81; standard error of the estimate = 1732 g; P < 0.001), where skinfolds are in millimetres and m is total body mass in kilograms. Both methods were used to predict fat-free mass. Bioelectrical impedance analysis predicted fat-free mass best according to the equation: fat-free mass (g) = 294.3h2/R + 662.7m + 71.8Xc + 662.7 (7) (coeý cient of determination = 0.90; standard error of the estimate = 2680 g; P < 0.001), where h is height in metres, R is resistance in ohms, m is total body mass in kilograms and Xc is the reactance in ohms. The equivalent prediction from anthropometric variables was investigated using corrected circumferences, generated as the circumference minus the average of two skinfolds taken along it, similar to the

method of Martin et al. (1990). Stepwise regression analysis was used to select the variables that predicted fat-free mass best. The following equation was produced: fat-free mass (g) = 689m + 285SH + 1025CFG + 534CCG - 473CWG - 20895

(8)

(coeý cient of determination = 0.90; standard error of the estimate = 2653 g; P < 0.001), where m is total body mass in kilograms, SH is sitting height in centimetres, CFG is the corrected forearm girth (forearm girth minus p ´ the mean of forearm (anterior) and radial skinfolds), CCG is the corrected calf girth (calf girth minus p ´ medial calf skinfold), and CWG is the corrected waist girth (waist girth minus p ´ the mean of abdominal and suprailiac skinfolds). All girths and skinfolds are in centimetres.

268

Stewart and Hannan

Corrected girth methodology was designed to estimate dissected skeletal muscle mass and, in addition, DXA lean tissue mass represents considerable nonmuscle mass. The hypothesis that DXA lean tissue mass is predicted better by subtracting fat mass from total mass than with the corrected girths model, was tested by entering a variety of skinfolds into a stepwise regression analysis with DXA fat-free mass as the dependent variable. The following equation was produced: fat-free mass (g) = 888m - 252(abdominal) 382(suprailium) - 335(thigh) + 9120 (9) (coeý cient of determination = 0.96; standard error of the estimate = 1738 g; P < 0.001), where m is mass in kilograms and skinfolds are in millimetres. The standard errors of the estimate and coeý cients of determination of this prediction approach the accuracy of those of the prediction of muscle mass by Martin et al. (1990). The suprailium site is an oblique fold on the anterior axillary line, as used in the equations of Jackson and Pollock (1978), and diþ ers from the suprailiac site, which is a horizontal fold taken in the mid-axillary line, used in the equations of Durnin and Womersley (1974). Once the optimal prediction for determining fat and fat-free mass had been established, these were applied to a further sample of 24 athletes of equivalent competitive standard (including seven international athletes), age and training, but who were heavier (81.0 ± 9.2 kg, P < 0.05) and had a higher body mass index (24.8 ± 2.5 kg ´m- 2; P < 0.01) and higher percent fat (13.2 ± 5.4%; P < 0.01). The total error (TE), calcu-

lated according to the formula TE = [ S(DXA mass predicted mass)2/n], where n is the number of participants, was 2310 g (2.9%) and 2159 g (2.7%) for fat and fat-free mass, respectively. Predicted and measured fat and fat-free mass showed a correlation of 0.86 and 0.96, respectively (P < 0.01). The limits of agreement, calculated according to the method of Bland and Altman (1986), showed a mean bias (± 2 standard deviations) of 1.3 ± 5.4% for percent fat and - 0.8 ± 5.1% for percent fat-free mass, as illustrated in Fig. 2. Subsequent predictions of fat and fat-free mass using anthropometry were made using data from all 106 athletes to maximize the sample size and introduce greater heterogeneity. When fat mass was predicted using the thigh and abdominal sites only, the coeý cient of determination was 0.82 and the standard error of the estimate was 1843 g, equivalent to 2.4% for a typical athlete in the sample. When all the skinfolds were entered into a stepwise regression using all 106 participants, fat and fat-free mass were predicted best from the following formulae: fat mass (g) = 105.2m + 189.5A + 345.2P - 521.1FR + 215.9MC + 258.3T + 293S - 8334.4 (10) (coeý cient of determination = 0.85; standard error of the estimate = 1679 g; total error = 1614 g; P < 0.001) fat-free mass (g) = 890m - 261A 244T - 342S - 237TP + 9966 (11)

Fig. 2. Bland and Altman plots of (a) predicted and measured percent fat and (b) predicted and measured percent fat-free mass in 24 athletes. In (a), percent fat predicted from the formula: fat mass (g) = 331.5A + 356.2T + 111.9m - 9108, where A = abdominal and T = thigh skinfolds in millimetres, m = total mass in kilograms, and converted into a percentage of total mass. In (b), percent fat-free mass predicted from the formula: fat-free mass (g) = 888m - 252A - 382S - 335T + 9120, where m = total mass in kilograms, A = abdominal, S = suprailium and T = thigh skinfolds in millimetres, and converted into a percentage of total mass (equations 10 and 11).

269

Prediction of fat and fat-free mass using DXA Table 3. Stepwise regression analysis predicting fat mass from skinfolds in 106 athletes

Model 1 2 3 4 5 6 7 8 9

R2

SEE (g)

0.667 0.756 0.798 0.823 0.834 0.841 0.837 0.845 0.854

2460 2117 1934 1821 1770 1744 1754 1718 1679

R2 change 0.667 0.089 0.042 0.025 0.011 0.007 - 0.004 0.008 0.009

F change

d.f.

Signi® cant F change

208.4 37.4 21.4 14.0 6.9 4.1 2.2 5.3 5.7

104 103 102 101 100 99 101 99 98

0.000 0.000 0.000 0.000 0.010 0.045 0.138 0.024 0.019

Stepwise criteria: probability of F-to-enter < 0.05; probability of F-to-remove > 0.1. Predictors (constant in all predictions): Model 1: abdominal Model 2: abdominal, medial calf Model 3: abdominal, medial calf, mass Model 4: abdominal, medial calf, mass, thigh Model 5: abdominal, medial calf, mass, thigh, suprailium Model 6: abdominal, medial calf, mass, thigh, suprailium, pectoral Model 7: abdominal, mass, thigh, suprailium, pectoral Model 8: abdominal, mass, thigh, suprailium, pectoral, forearm (radial) Model 9: abdominal, mass, thigh, suprailium, pectoral, forearm (radial), medial calf Note: R2 = coeý cient of determination, SEE = standard error of the estimate, d.f. = degrees of freedom.

Table 4. Stepwise regression analysis predicting fat-free mass from skinfolds in 106 athletes

Model 1 2 3 4 5

R2

SEE (g)

R2 change

F change

d.f.

Signi® cant F change

0.809 0.929 0.951 0.957 0.960

3628 2220 1852 1738 1694

0.809 0.120 0.022 0.006 0.003

441.5 174.6 46.1 14.8 6.4

104 103 102 101 100

0.000 0.000 0.000 0.000 0.013

Stepwise criteria: probability of F-to-enter < 0.05; probability of F-to-remove > 0.01. Predictors (constant in all predictions): Model 1: mass Model 2: mass, abdominal Model 3: mass, abdominal, thigh Model 4: mass, abdominal, thigh, suprailium Model 5: mass, abdominal, thigh, suprailium, thigh-patella Note: R2 = coeý cient of determination, SEE = standard error of the estimate, d.f. = degrees of freedom.

(coeý cient of determination = 0.96; standard error of the estimate = 1694 g; total error = 1645 g; P < 0.001), where m = mass in kilograms, A = abdominal, P = pectoral, FR = forearm radial, MC = medial calf, T = thigh, S = suprailium and TP = thigh-patellar skinfolds, all in millimetres. For both predictions, the standard errors of the estimate and total errors are 2.1± 2.2% for a typical athlete in the sample. Model summary and change statistics are shown in Tables 3 and 4.

Discussion The results for the 82 athletes for whom both anthropometry and bioelectrical impedance analysis were performed indicate that fat and fat-free masses in male athletes are predicted best from skinfolds, especially at the abdominal, thigh and suprailium sites. Application of the prediction equations to fatter athletes resulted in a small mean diþ erence between predicted

270 and measured values. This diþ erence may have resulted from diþ erences in regional fat distribution between athletes of diþ erent adiposity, or the possibility that the smaller validation group of athletes was more heterogeneous. Combining the initial and validation samples introduced greater heterogeneity into the total sample, and the optimal prediction of fat mass used six skinfolds in producing a similar standard error of the estimate (1.7 kg), although this explained a further 4% of the variation in DXA-derived fat. Although the coeý cient of determination (R2) will increase as extra predictor variables are added, simply because of the relationship between the number of predictors and the sample size, this artefact accounted for only 0.5% of the observed 4% increase in R2 as a result of combining samples. For the fat-free mass prediction, combining samples produced no change in the standard error of the estimate or the coeý cient of determination. The present study used a large sample of athletes, whose numbers may exceed the physically active individuals of generalized prediction equations. Because the sample is more homogeneous than that of generalized studies, it is reasonable to assume that its predictions will be more accurate. The use of a large selection of skinfolds in the analysis enabled an optimal prediction of fat mass to accommodate the diþ ering fat `topography’ associated with physical ® tness. The agreement of predicted and measured values in the validation sample was well within 95% con® dence limits, but showed a mean bias of +1.3% for percent fat and of - 0.8% for percent fat-free mass. This suggests that there may be a tendency for the fatter athletes in the validation group to display more heterogeneity in fat distribution, and that increasing error will be found if the prediction equation is applied to increasingly fat athletes. Bioimpedance data were not collected on the validation sample and, therefore, it is unclear if a similar anomaly would occur with fatter athletes using bioelectrical impedance analysis. The agreement between the skinfold prediction equations of Jackson and Pollock (1978) and Durnin and Womersley (1974) and percent fat derived by DXA varied with adiposity. Compared with the DXA value, mean diþ erences were within 3% over ranges of 0± 18% fat for Jackson and Pollock’ s equation and >14% fat for Durnin and Womersley’ s equation. These ranges represent 91% and 23% of the athlete cohort, respectively. Close inspection of the data in both studies reveals that the participants in the study of Jackson and Pollock were taller and leaner, and were closer in body composition to the athletes in the present study than those of Durnin and Womersley. It is also possible that the study of Jackson and Pollock (1978) included a greater proportion of athletic or physically active individuals than the study of Durnin and Womersley (1974).

Stewart and Hannan A diþ erence in fat distribution between athletes and controls would aþ ect the optimal skinfold selection to predict total fatness in each category. In the present study of 82 athletes, the sum of four sites (biceps, triceps, subscapular and suprailiac ± as used by Durnin and Womersley, 1974) explained 63% of the variation in DXA percent fat (standard error of the estimate = 2.8%, P < 0.001), while the sum of three (abdominal + chest + thigh, as used by Jackson and Pollock, 1978) explained 79% (standard error of the estimate = 2.1%, P < 0.001). Including all 106 athletes increased the explained variance in DXA percent fat to 80% and the standard error of the estimate to 2.2%. This suggests that appropriate skinfold selection is a key element in the accuracy of a prediction equation, and that those sites chosen by Jackson and Pollock (1978) include the two which, according to the present study, best describe total fatness in athletes. This supports previous preferred skinfold selection using existing formulae (Stewart and Eston, 1997). Despite its greater precision, bioelectrical impedance analysis oþ ers a less accurate prediction of percent fat than skinfolds. The diþ erence in accuracy may be more closely linked to lean tissue rather than fat tissue distribution. The resistance to electric current in the human body is essentially determined by body water, the electrolytes of which are excellent conductors. Total body water is therefore estimated directly from the impedance measurement, fat-free mass is predicted by assuming a ® xed hydration, and fat mass is predicted by subtraction of fat-free mass from total mass. Errors in fat prediction are likely to be higher than in lean tissue prediction because fat-free mass represents a much greater mass and similar variance. Using densitometry as the reference method, Lohman (1992) postulated that existing equations using bioelectrical impedance analysis, with fat mass as the dependent variable, should have standard errors of the estimate which fall within the range 2.1± 2.9 kg. Those in the present study fall within this range, whereas those based on anthropometry are substantially lower. While anthropometric predictions could be aþ ected by diþ erences in fat patterning between athletes and controls, bioimpedance could also be in¯ uenced by diþ erences in hydration and lean tissue distribution. Hydration of fat-free mass is commonly assumed to be 0.732 (Baumgartner, 1996), although a range of 0.69± 0.77 has been reported (Streat et al., 1985). Close inspection of the data for the predominantly active sample of Prior et al. (1997), measured by deuterium dilution, gives a range 0.68± 0.77. Part of this observed variation in hydration can be attributed to inaccuracies in the estimates of fat-free mass, because there is currently no truly accurate estimate available. Departure from the assumed average will reduce the accuracy

Prediction of fat and fat-free mass using DXA of the body composition prediction by bioelectrical impedance analysis; dehydration is one of several factors that could introduce errors (Heyward and Stolarczyk, 1996). Among the athletes in the present study, whose weekly training averaged 9.3 h, daily ¯ uid balance ± and thus the hydration of fat-free mass ± may have ¯ uctuated widely. It is possible that some individuals may not have become fully rehydrated since exercising the previous day. Dual-energy X-ray absorptiometry is relatively impervious to such shifts in hydration (Pietrobelli et al., 1998). Athletes also have greater lean tissue than nonathletes, and such muscular development ± particularly in the arms ± may dramatically enhance the conductivity. Site-speci® c muscular development could introduce greater variance of conductivity in athletes than in non-exercising controls, rendering bioelectrical impedance analysis less accurate. In this respect, it is perhaps surprising that the equation of Lukaski and Bolonchuk (1987), which was derived on athletes, did not predict fat better. The diþ erence in percent fat prediction between DXA and bioelectrical impedance analysis from either formula was signi® cantly correlated with mesomorphy (r = 0.26; P < 0.05), which suggests that predictions based on bioelectrical impedance analysis may be population-speci® c to the extent that strength athletes may be poorly represented by a predictive formula based on endurance athletes. No equivalent correlation was observed in anthropometric predictions. Although it may be logical to expect subcutaneous fat ± the largest compartment ± to be representative of total body fat, this becomes less tenable as greater understanding emerges of the proportional distribution of fat between compartments (Hawes, 1996). In an investigation of regional fat placement in 165 soldiers, Nindl et al. (1996) reported a hierarchy of regional fat mobilization by DXA of abdomen > arms > legs as a consequence of weight loss, but this was not supported by skinfold evidence in the arms and legs. Such an observation could be explained by a change in the partitioning between subcutaneous and internal fat with weight loss. Evidence on fat partitioning from medical imaging techniques is con¯ icting. Using magnetic resonance imaging, Ross et al. (1994) found no signi® cant relationship between visceral adipose tissue and total adipose tissue in 17 obese men (over 30% fat), while DesprŠs et al. (1991), using computed tomography, found deep abdominal fat correlated with total fat and age (r = 0.76 and 0.63 respectively, P < 0.0001) in 110 men whose fatness ranged from 2.2 to 39.8% (22.9 ± 8.9%). The proportional distribution of fat in athletes is unknown, although a high proportion of total fat situated in the subcutaneous compartment would explain why skinfolds are relatively accurate in predicting total fat.

271 Dual-energy X-ray absorptiometry is calibrated for a normal expected range of tissue thickness, beyond which the accuracy may be reduced (Wahner and Fogelman, 1994). Separate corrections are therefore necessary to accommodate growth studies (Mitchell et al., 1996a). Ryde et al. (1998) found DXA (Lunar DPX model) to under- and overestimate small and large changes, respectively, in fat loss in 10 obese women compared with a four-compartment model. Experimental work with the same machine as that used in the present study indicated that fat measurements were stable up to a total tissue depth of 25 cm (Jebb et al., 1993), which exceeds the antero-posterior thickness of any of the athletes. Inter-manufacturer diþ erences (Tothill et al., 1994) are greater for sub-regions than the whole body, but of the three manufacturers, Hologic (as used in the present study) is the scanner best validated for accuracy in an athletic sample. Prior et al. (1997) found diþ erences between percent fat as derived by the four-compartment model and DXA to be weakly related to body mass index but not mesomorphy, suggesting that increasing fat mass, rather than lean tissue mass, contributes to the discrepancy. Although concern has been expressed about the accuracy of DXA in obese individuals, extremely lean people represent a greater threat to accuracy in the current sample. Nine athletes had fat measured by DXA less than 5% (a suggested minimal weight criterion used by Lohman, 1992), the lowest being only 2.9%. While this might appear plausible by comparison to two participants in the densitometric study of Adams et al. (1982), whose fat was predicted to be - 12%, +2.9% is still a remarkably low value. This ® gure could theoretically represent only essential lipids in the leanest athletes who had reduced their `excess fat’ to zero. Dual-energy X-ray absorptiometry failed to record any fat in the torso region of the leanest three individuals. The regression of torso fat against percent total fat predicted zero torso fat to occur at 3.3% total fat (95% CI = 2.2± 4.4). However, even for these leanest individuals, abdominal skinfold averaged 5.8 ± 0.3 mm. In addition, essential lipids in the liver and other organs suggest that zero fat in the torso is, in practical terms, impossible. A more plausible explanation is that accuracy is diminished at the lean extreme of this sample for several reasons. Poorer reproducibility may be expected in very lean individuals because the variance between replicate measures will represent a greater proportion of the measurement value. In addition, the bone-containing pixels, which DXA must interpolate across to derive a fat content, are most prevalent in the torso. The leanest individuals have proportionately fewer soft tissue pixels to provide the information for this interpolation and accuracy is reduced as a consequence. It is not possible using

272 DXA to measure soft tissue inside the skull, and the software used by the present study assumes a ® xed value of 17% fat. Lastly, the algorithms that predict fat content, which are clearly diþ erent between manufacturers and software versions, are designed to be robust for individuals of a normal range of adiposity, and extremes at both ends of this scale will compromise their stability. The ® ndings of Sinning et al. (1985) indicate that, using the two-compartment method (fat and fat-free mass) as the criterion, most generalized equations were not valid for predicting the fat content of athletes, with the exception of the three equations of Jackson and Pollock (1978). While the three-site equation of Jackson and Pollock is clearly the best of the four predictions compared in the present study, the conclusion of Adams et al. (1982) was that this method is inappropriate for some athletes because of the increased density of fat-free mass. The present results are not altogether surprising because the athletes studied here (n = 82) and those in the study of Sinning et al. (1985) had mean body mass indices within 0.3 kg ´ m- 2 and mean percent fat within 1%. This contrasts with the equivalent ® gures of Prior et al. (1997), whose male sample (67 athletes and 24 non-athletes) had mean body mass indices of 27.0 ± 4.6 kg ´m- 2 and fat of 13.1 ± 5.7% (by DXA). The greater mesomorphy of the athletes in the study of Prior et al. (1997) compared with the present study (6.2 ± 1.7 vs 5.0 ± 1.2) suggests considerable musculoskeletal development and potentially greater variation in the density of fat-free mass. Although good agreement has been reported between DXA and densitometry in 128 men and women (Wellens et al., 1994), in the study of Prior et al. (1997), variability in the density of fat-free mass beyond that expected in a sedentary population could explain the poorer association of percent fat by densitometry and the four-compartment model than by DXA and the four-compartment model. The accuracy of both methods relative to the four-compartment model was compared in 78 young and older men and women by Clasey et al. (1999). Overall, DXA was slightly more accurate: the total error for DXA and the four-compartment model compared with densitometry and the four-compartment model was 5.3% and 5.7% respectively, but the diþ erence was greatest for young men and older women. Although physical activity was not reported, a higher bone density in young men and a lower bone density in older women may have violated the assumption of the constant density for fat-free mass more than in the other groups. The criteria of Lohman (1996) suggest the evaluation of a new method of predicting percent fat accurately should include a standard error of the estimate within 3% of body mass. Using the data of Prior et al. (1997), DXA ful® ls the criteria (standard error of

Stewart and Hannan the estimate = 2.8%), while densitometry does not (standard error of the estimate = 3.4%). Dual-energy X-ray absorptiometry was also more accurate than densitometry compared with a six-compartment model (based on fat, water, protein, glycogen and osseous and non-osseous mineral) in a small heterogeneous sample (Wang et al., 1998). The implications of this are that DXA may be better than densitometry as a reference method. Thus, until a four-compartment criterion or alternative prediction is established for athletes, DXA may be the criterion method of choice. Compared with the best of the existing equations used (Jackson and Pollock, 1978), the standard error of the estimate of the prediction of percent fat in the present study is smaller (2.2% vs 3.3%), and the coeý cients of determination comparable (0.82 for the equation of Jackson and Pollock; 0.81 for the sample of 82 athletes, and 0.85 for all 106 athletes in the present study). This strengthens the case for promoting their use in predicting the fat content of athletes. Fat-free mass predictions using the anthropometric corrected girth model were of equivalent accuracy to bioimpedance, although both were less accurate than using skinfolds. Although fat-free mass was predicted with greater accuracy than fat, it was predicted slightly less accurately than the muscle mass prediction of Martin et al. (1990), which may be of greater utility with athletic groups. The validation of DXA against chemical analysis of pig carcasses (Mitchell et al., 1996b), the development of cross-calibration equations to convert between results from diþ erent manufacturers (Modlesky et al., 1996), and the widespread use of DXA have diminished the scope for earlier criticism of its accurate prediction of fat and fat-free masses (Kohrt, 1993, 1998). Considerable potential remains for inter-laboratory research using DXA, both as a reference method by itself and as an integral part of a four-compartment comparison. As a consequence of the increased availability and reduced scanning times with new fan-beam technology, the use of DXA in body composition studies in groups including athletes is likely to increase.

Acknowledgements The authors wish to thank R. Elton for advice and assistance with the statistical methods.

References Adams, J., Mottola, M., Bagnell, K.M. and McFadden, K.D. (1982). Total body fat content in a group of professional football players. Canadian Journal of Applied Sport Sciences, 7, 36± 40.

Prediction of fat and fat-free mass using DXA Baumgartner, R.N. (1996). Electrical impedance and total body electrical conductivity. In Human Body Composition (edited by A. Roche, S. Heyms® eld and T. Lohman), pp. 79± 108. Champaign, IL: Human Kinetics. Bland, J.M. and Altman, D.G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, i, 307± 310. Carter, J.E.L. and Heath, B.H. (1990). Somatotyping: Development and Applications. Cambridge: Cambridge University Press. Clasey, J.L., Kanaley, J.A., Wideman, L., Heyms® eld, S.B., Teates, C.D., Gutgesell, M.E., Thorner, M.O., Hartman, M.L. and Weltman, A. (1991). Validity of methods of body composition assessment in young and older men and women. Journal of Applied Physiology, 86, 1728± 1738. De Lorenzo, A., Sorge, S.P., Iacopino, L., Andreoli, A., de Luca, P.P. and Sasso, G.F. (1998). Fat-free mass by bioelectrical impedance vs dual-energy X-ray absorptiometry (DXA). Applied Radiation and Isotopes, 49, 739± 741. DesprŠs J.P., Prud’ homme, D., Tremblay, A. and Bouchard, C. (1991). Estimation of deep abdominal adipose tissue anthropometric measurements in men. American Journal of Clinical Nutrition, 54, 471± 477. Durnin, J.V.G.A. and Womersley, J. (1974). Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. British Journal of Nutrition, 32, 77± 97. Fogelman, I. and Blake, G.M. (1998). Clinical use of DXA measurements. In Current Research in Osteoporosis and Bone Mineral Measurement V (edited by E.F.J. Ring, D.M. Elvins and A.K. Bhalla), pp. 74± 75. London: British Institute of Radiology. Fuller, N.J., Laskey, M.A. and Elia, M. (1992). Assessment of the composition of major body regions by dual-energy X-ray absorptiometry (DEXA), with special reference to limb muscle mass. Clinical Physiology, 12, 253± 266. Goran, M.I., Driscoll, P., Johnson, R., Nagy, T.R. and Hunter, G. (1996). Cross-calibration of body composition techniques against dual-energy X-ray absorptiometry in young children. American Journal of Clinical Nutrition, 63, 299± 305. Hawes, M.R. (1996). Human body composition. In Kinanthropometry and Exercise Physiology Manual (edited by R. Eston and T. Reilly), pp. 5± 33. London: E & FN Spon. Heyward, V.H. and Stolarczyk, L.M. (1996). Applied Body Composition Assessment. Champaign, IL: Human Kinetics. Jackson, A.S. and Pollock, M.L. (1978). Generalized equations for predicting body density of men. British Journal of Nutrition, 40, 497± 504. Jebb, S.A., Goldberg, G.R. and Elia, M. (1993). DXA measurements of fat and bone mineral density in relation to depth and adiposity. In Human Body Composition (edited by K.J. Ellis and J.D. Eastman), pp. 115± 119. New York: Plenum Press. Kohrt, W.M. (1993). Body composition by DXA: Tried and true? Medicine and Science in Sports and Exercise, 27, 1349± 1353.

273 Kohrt, W.M. (1998). Preliminary evidence that DEXA provides an accurate assessment of body composition. Journal of Applied Physiology, 84, 372± 377. Lohman, T.G. (1992). Advances in Body Composition Assessment. Current Issues in Exercise Science Series Monograph No. 3. Champaign, IL: Human Kinetics. Lohman, T.G. (1996). Dual energy X-ray absorptiometry. In Human B ody Composition (edited by A. Roche, S. Heyms® eld and T. Lohman), pp. 63± 78. Champaign, IL: Human Kinetics. Lohman, T.G., Roche, A.F. and Martorell, R. (1988). Anthropometric Standardization Reference Manual. Champaign, IL: Human Kinetics. Lukaski, H.C. and Bolonchuk, W.W. (1987). Theory and validation of the tetrapolar bioelectrical impedance method to assess human body composition. In In vivo Body Composition Studies (edited by K.J. Ellis, S. Yasamura and W.D. Morgan), pp. 410± 414. London: The Institute of Physical Sciences in Medicine. Martin, A.D., Drinkwater, D.T., Clarys, J.P. and Ross, W.D. (1986). The inconstancy of the fat-free mass: A reappraisal with implications for densitometry. In Kinanthropometry III: Proceedings of the VII Commonwealth and International Conference on Sport, Physical Education, Dance, Recreation and Health (edited by T. Reilly, J. Watkins and J. Borms), pp. 92± 97. London: E & FN Spon. Martin, A.D., Spenst, L.F., Drinkwater, D.T. and Clarys, J.P. (1990). Anthropometric estimation of muscle mass in men. Medicine and Science in Sports and Exercise, 22, 729± 733. Mazess, R.B., Barden, H.S., Bisek, J.P. and Hanson, J. (1990). Dual-energy X-ray absorptiometry for total and regional bone mineral and soft tissue composition. American Journal of Clinical Nutrition, 51, 1106± 1112. Mitchell, A.D., Conway, J.M. and Scholz, A.M. (1996a). Incremental changes in total and regional body composition of growing pigs measured by dual-energy X-ray absorptiometry. Growth, Development and Aging, 60, 95± 105. Mitchell, A.D., Conway, J.M. and Potts, W.J.E. (1996b). Body composition analysis of pigs by dual-energy X-ray absorptiometry. Journal of Animal Science, 74, 2663± 2671. Modlesky, C.M., Lewis, R.D., Yetman, K.A., Rose, B., Rosskopf, L.B., Snow, T.K. and Sparling, P.B. (1996). Comparison of body composition and bone mineral measurement from two DXA instruments in young men. American Journal of Clinical Nutrition, 64, 669± 676. Nindl, B.C., Friedl, K.E., Marchitelli, L.J., Shippee, R.L., Thomas, C.D. and Patton, J.F. (1996). Regional fat placement in physically ® t males and changes with weight loss. Medicine and Science in Sports and Exercise, 28, 786± 793. Nord, R.H. and Payne, R.K. (1995). DXA vs underwater weighing: Comparison of strengths and weaknesses. Asian Paci® c Journal of Clinical Nutrition, 4, 173± 175. Pietrobelli, A., Faith, M.S., Allison, D.B., Gallacher, D., Chiumello, G. and Heyms® eld, S.B. (1998). Body mass index as a measure of adiposity among children and adolescents: A validation study. Journal of Pediatrics, 132, 204± 210.

274 Prior, B.M., Cureton, K.J., Modlesky, C.M., Evans, E.M., Sloniger, M.A., Saunders, M. and Lewis, R.D. (1997). In vivo validation of whole body composition estimates from dual-energy X-ray absorptiometry. Journal of Applied Physiology, 83, 623± 630. Ross, R., Shaw, K.D., Rissanen, J., Martel, Y., de Guise, J. and Avrush, L. (1994). Sex diþ erences in lean and adipose tissue distribution by magnetic resonance imaging: Anthropometric relationships. American Journal of Clinical Nutrition, 59, 1277± 1285. Ryde, S.J., Eston, R., Laskey, M.A., Evans, C.J. and Hancock, D.A. (1998). Changes in body fat: Measurements by neutron activation, densitometry and dual X-ray absorptiometry. Applied Radiation and Isotopes, 49, 507± 509. Sinning, W.E., Dolny, D.G., Little, K.D., Cunningham, L.N., Racaniello, A., Siconol® , S.F. and Sholes, J.L. (1985). Validity of `generalized’ equations for body composition analysis in male athletes. Medicine and Science in Sports and Exercise, 17, 124± 130. Siri, W.E. (1956). The gross composition of the body. In Advances in Biological and Medical Physics (edited by J. Lawrence and C. Tobias), pp. 239± 280. London: Academic Press.

Stewart and Hannan Stewart, A. and Eston, R. (1997). Skinfold thickness measurement. British Journal of Nutrition, 78, 1040± 1042. Streat, S., Beddoe, A. and Hill, G. (1985). Measurement of body fat and hydration of the fat-free body in health and disease. Metabolism, 34, 509± 518. Tothill, P., Avenell, A., Love, J. and Reid, D.M. (1994). Comparisons between Hologic, Lunar and Norland dualenergy X-ray absorptiometers and other techniques used for whole body soft tissue measurements. European Journal of Clinical Nutrition, 48, 781± 794. Wahner, H.W. and Fogelman, I. (1994). The Evaluation of Osteoporosis: Dual Energy X-Ray Absorptiometry in Clinical Practice. London: Dunitz. Wang, Z.M., Deurenberg, P., Guo, S.S., Pietrobelli, A., Wang, J., Pierson, R.N., Jr. and Heyms® eld, S.B. (1998). Six-compartment body composition model: Inter-method comparisons of total body fat measurement. International Journal of Obesity, 22, 329± 337. Wellens, R., Chumlea, W.C., Guo, S., Roche, A.F., Reo, N.V. and Siervogel, R.M. (1994). Body composition in white adults by dual-energy X-ray absorptiometry, densitometry and total body water. American Journal of Clinical Nutrition, 59, 547± 555.

Prediction of fat and fat-free mass using DXA

Fig. 1. Comparison of bioelectrical impedance analysis and skinfold-predicted fat against DXA percent fat. Predictions indicate linear regression and 95% con® dence intervals. J&P(3) = the three-site calliper prediction of Jackson and Pollock (1978); D&W = the four-site calliper prediction of Durnin and Womersley (1974); Lukaski = the bioelectrical impedance analysis prediction of Lukaski and Bolonchuk (1987); Lohman = the bioelectrical impedance analysis prediction of Lohman (1992).

(coeý cient of determination = 0.81; standard error of the estimate = 1732 g; P < 0.001), where skinfolds are in millimetres and m is total body mass in kilograms. Both methods were used to predict fat-free mass. Bioelectrical impedance analysis predicted fat-free mass best according to the equation: fat-free mass (g) = 294.3h2/R + 662.7m + 71.8Xc + 662.7 (7) (coeý cient of determination = 0.90; standard error of the estimate = 2680 g; P < 0.001), where h is height in metres, R is resistance in ohms, m is total body mass in kilograms and Xc is the reactance in ohms. The equivalent prediction from anthropometric variables was investigated using corrected circumferences, generated as the circumference minus the average of two skinfolds taken along it, similar to the

method of Martin et al. (1990). Stepwise regression analysis was used to select the variables that predicted fat-free mass best. The following equation was produced: fat-free mass (g) = 689m + 285SH + 1025CFG + 534CCG - 473CWG - 20895

(8)

(coeý cient of determination = 0.90; standard error of the estimate = 2653 g; P < 0.001), where m is total body mass in kilograms, SH is sitting height in centimetres, CFG is the corrected forearm girth (forearm girth minus p ´ the mean of forearm (anterior) and radial skinfolds), CCG is the corrected calf girth (calf girth minus p ´ medial calf skinfold), and CWG is the corrected waist girth (waist girth minus p ´ the mean of abdominal and suprailiac skinfolds). All girths and skinfolds are in centimetres.

268

Stewart and Hannan

Corrected girth methodology was designed to estimate dissected skeletal muscle mass and, in addition, DXA lean tissue mass represents considerable nonmuscle mass. The hypothesis that DXA lean tissue mass is predicted better by subtracting fat mass from total mass than with the corrected girths model, was tested by entering a variety of skinfolds into a stepwise regression analysis with DXA fat-free mass as the dependent variable. The following equation was produced: fat-free mass (g) = 888m - 252(abdominal) 382(suprailium) - 335(thigh) + 9120 (9) (coeý cient of determination = 0.96; standard error of the estimate = 1738 g; P < 0.001), where m is mass in kilograms and skinfolds are in millimetres. The standard errors of the estimate and coeý cients of determination of this prediction approach the accuracy of those of the prediction of muscle mass by Martin et al. (1990). The suprailium site is an oblique fold on the anterior axillary line, as used in the equations of Jackson and Pollock (1978), and diþ ers from the suprailiac site, which is a horizontal fold taken in the mid-axillary line, used in the equations of Durnin and Womersley (1974). Once the optimal prediction for determining fat and fat-free mass had been established, these were applied to a further sample of 24 athletes of equivalent competitive standard (including seven international athletes), age and training, but who were heavier (81.0 ± 9.2 kg, P < 0.05) and had a higher body mass index (24.8 ± 2.5 kg ´m- 2; P < 0.01) and higher percent fat (13.2 ± 5.4%; P < 0.01). The total error (TE), calcu-

lated according to the formula TE = [ S(DXA mass predicted mass)2/n], where n is the number of participants, was 2310 g (2.9%) and 2159 g (2.7%) for fat and fat-free mass, respectively. Predicted and measured fat and fat-free mass showed a correlation of 0.86 and 0.96, respectively (P < 0.01). The limits of agreement, calculated according to the method of Bland and Altman (1986), showed a mean bias (± 2 standard deviations) of 1.3 ± 5.4% for percent fat and - 0.8 ± 5.1% for percent fat-free mass, as illustrated in Fig. 2. Subsequent predictions of fat and fat-free mass using anthropometry were made using data from all 106 athletes to maximize the sample size and introduce greater heterogeneity. When fat mass was predicted using the thigh and abdominal sites only, the coeý cient of determination was 0.82 and the standard error of the estimate was 1843 g, equivalent to 2.4% for a typical athlete in the sample. When all the skinfolds were entered into a stepwise regression using all 106 participants, fat and fat-free mass were predicted best from the following formulae: fat mass (g) = 105.2m + 189.5A + 345.2P - 521.1FR + 215.9MC + 258.3T + 293S - 8334.4 (10) (coeý cient of determination = 0.85; standard error of the estimate = 1679 g; total error = 1614 g; P < 0.001) fat-free mass (g) = 890m - 261A 244T - 342S - 237TP + 9966 (11)

Fig. 2. Bland and Altman plots of (a) predicted and measured percent fat and (b) predicted and measured percent fat-free mass in 24 athletes. In (a), percent fat predicted from the formula: fat mass (g) = 331.5A + 356.2T + 111.9m - 9108, where A = abdominal and T = thigh skinfolds in millimetres, m = total mass in kilograms, and converted into a percentage of total mass. In (b), percent fat-free mass predicted from the formula: fat-free mass (g) = 888m - 252A - 382S - 335T + 9120, where m = total mass in kilograms, A = abdominal, S = suprailium and T = thigh skinfolds in millimetres, and converted into a percentage of total mass (equations 10 and 11).

269

Prediction of fat and fat-free mass using DXA Table 3. Stepwise regression analysis predicting fat mass from skinfolds in 106 athletes

Model 1 2 3 4 5 6 7 8 9

R2

SEE (g)

0.667 0.756 0.798 0.823 0.834 0.841 0.837 0.845 0.854

2460 2117 1934 1821 1770 1744 1754 1718 1679

R2 change 0.667 0.089 0.042 0.025 0.011 0.007 - 0.004 0.008 0.009

F change

d.f.

Signi® cant F change

208.4 37.4 21.4 14.0 6.9 4.1 2.2 5.3 5.7

104 103 102 101 100 99 101 99 98

0.000 0.000 0.000 0.000 0.010 0.045 0.138 0.024 0.019

Stepwise criteria: probability of F-to-enter < 0.05; probability of F-to-remove > 0.1. Predictors (constant in all predictions): Model 1: abdominal Model 2: abdominal, medial calf Model 3: abdominal, medial calf, mass Model 4: abdominal, medial calf, mass, thigh Model 5: abdominal, medial calf, mass, thigh, suprailium Model 6: abdominal, medial calf, mass, thigh, suprailium, pectoral Model 7: abdominal, mass, thigh, suprailium, pectoral Model 8: abdominal, mass, thigh, suprailium, pectoral, forearm (radial) Model 9: abdominal, mass, thigh, suprailium, pectoral, forearm (radial), medial calf Note: R2 = coeý cient of determination, SEE = standard error of the estimate, d.f. = degrees of freedom.

Table 4. Stepwise regression analysis predicting fat-free mass from skinfolds in 106 athletes

Model 1 2 3 4 5

R2

SEE (g)

R2 change

F change

d.f.

Signi® cant F change

0.809 0.929 0.951 0.957 0.960

3628 2220 1852 1738 1694

0.809 0.120 0.022 0.006 0.003

441.5 174.6 46.1 14.8 6.4

104 103 102 101 100

0.000 0.000 0.000 0.000 0.013

Stepwise criteria: probability of F-to-enter < 0.05; probability of F-to-remove > 0.01. Predictors (constant in all predictions): Model 1: mass Model 2: mass, abdominal Model 3: mass, abdominal, thigh Model 4: mass, abdominal, thigh, suprailium Model 5: mass, abdominal, thigh, suprailium, thigh-patella Note: R2 = coeý cient of determination, SEE = standard error of the estimate, d.f. = degrees of freedom.

(coeý cient of determination = 0.96; standard error of the estimate = 1694 g; total error = 1645 g; P < 0.001), where m = mass in kilograms, A = abdominal, P = pectoral, FR = forearm radial, MC = medial calf, T = thigh, S = suprailium and TP = thigh-patellar skinfolds, all in millimetres. For both predictions, the standard errors of the estimate and total errors are 2.1± 2.2% for a typical athlete in the sample. Model summary and change statistics are shown in Tables 3 and 4.

Discussion The results for the 82 athletes for whom both anthropometry and bioelectrical impedance analysis were performed indicate that fat and fat-free masses in male athletes are predicted best from skinfolds, especially at the abdominal, thigh and suprailium sites. Application of the prediction equations to fatter athletes resulted in a small mean diþ erence between predicted

270 and measured values. This diþ erence may have resulted from diþ erences in regional fat distribution between athletes of diþ erent adiposity, or the possibility that the smaller validation group of athletes was more heterogeneous. Combining the initial and validation samples introduced greater heterogeneity into the total sample, and the optimal prediction of fat mass used six skinfolds in producing a similar standard error of the estimate (1.7 kg), although this explained a further 4% of the variation in DXA-derived fat. Although the coeý cient of determination (R2) will increase as extra predictor variables are added, simply because of the relationship between the number of predictors and the sample size, this artefact accounted for only 0.5% of the observed 4% increase in R2 as a result of combining samples. For the fat-free mass prediction, combining samples produced no change in the standard error of the estimate or the coeý cient of determination. The present study used a large sample of athletes, whose numbers may exceed the physically active individuals of generalized prediction equations. Because the sample is more homogeneous than that of generalized studies, it is reasonable to assume that its predictions will be more accurate. The use of a large selection of skinfolds in the analysis enabled an optimal prediction of fat mass to accommodate the diþ ering fat `topography’ associated with physical ® tness. The agreement of predicted and measured values in the validation sample was well within 95% con® dence limits, but showed a mean bias of +1.3% for percent fat and of - 0.8% for percent fat-free mass. This suggests that there may be a tendency for the fatter athletes in the validation group to display more heterogeneity in fat distribution, and that increasing error will be found if the prediction equation is applied to increasingly fat athletes. Bioimpedance data were not collected on the validation sample and, therefore, it is unclear if a similar anomaly would occur with fatter athletes using bioelectrical impedance analysis. The agreement between the skinfold prediction equations of Jackson and Pollock (1978) and Durnin and Womersley (1974) and percent fat derived by DXA varied with adiposity. Compared with the DXA value, mean diþ erences were within 3% over ranges of 0± 18% fat for Jackson and Pollock’ s equation and >14% fat for Durnin and Womersley’ s equation. These ranges represent 91% and 23% of the athlete cohort, respectively. Close inspection of the data in both studies reveals that the participants in the study of Jackson and Pollock were taller and leaner, and were closer in body composition to the athletes in the present study than those of Durnin and Womersley. It is also possible that the study of Jackson and Pollock (1978) included a greater proportion of athletic or physically active individuals than the study of Durnin and Womersley (1974).

Stewart and Hannan A diþ erence in fat distribution between athletes and controls would aþ ect the optimal skinfold selection to predict total fatness in each category. In the present study of 82 athletes, the sum of four sites (biceps, triceps, subscapular and suprailiac ± as used by Durnin and Womersley, 1974) explained 63% of the variation in DXA percent fat (standard error of the estimate = 2.8%, P < 0.001), while the sum of three (abdominal + chest + thigh, as used by Jackson and Pollock, 1978) explained 79% (standard error of the estimate = 2.1%, P < 0.001). Including all 106 athletes increased the explained variance in DXA percent fat to 80% and the standard error of the estimate to 2.2%. This suggests that appropriate skinfold selection is a key element in the accuracy of a prediction equation, and that those sites chosen by Jackson and Pollock (1978) include the two which, according to the present study, best describe total fatness in athletes. This supports previous preferred skinfold selection using existing formulae (Stewart and Eston, 1997). Despite its greater precision, bioelectrical impedance analysis oþ ers a less accurate prediction of percent fat than skinfolds. The diþ erence in accuracy may be more closely linked to lean tissue rather than fat tissue distribution. The resistance to electric current in the human body is essentially determined by body water, the electrolytes of which are excellent conductors. Total body water is therefore estimated directly from the impedance measurement, fat-free mass is predicted by assuming a ® xed hydration, and fat mass is predicted by subtraction of fat-free mass from total mass. Errors in fat prediction are likely to be higher than in lean tissue prediction because fat-free mass represents a much greater mass and similar variance. Using densitometry as the reference method, Lohman (1992) postulated that existing equations using bioelectrical impedance analysis, with fat mass as the dependent variable, should have standard errors of the estimate which fall within the range 2.1± 2.9 kg. Those in the present study fall within this range, whereas those based on anthropometry are substantially lower. While anthropometric predictions could be aþ ected by diþ erences in fat patterning between athletes and controls, bioimpedance could also be in¯ uenced by diþ erences in hydration and lean tissue distribution. Hydration of fat-free mass is commonly assumed to be 0.732 (Baumgartner, 1996), although a range of 0.69± 0.77 has been reported (Streat et al., 1985). Close inspection of the data for the predominantly active sample of Prior et al. (1997), measured by deuterium dilution, gives a range 0.68± 0.77. Part of this observed variation in hydration can be attributed to inaccuracies in the estimates of fat-free mass, because there is currently no truly accurate estimate available. Departure from the assumed average will reduce the accuracy

Prediction of fat and fat-free mass using DXA of the body composition prediction by bioelectrical impedance analysis; dehydration is one of several factors that could introduce errors (Heyward and Stolarczyk, 1996). Among the athletes in the present study, whose weekly training averaged 9.3 h, daily ¯ uid balance ± and thus the hydration of fat-free mass ± may have ¯ uctuated widely. It is possible that some individuals may not have become fully rehydrated since exercising the previous day. Dual-energy X-ray absorptiometry is relatively impervious to such shifts in hydration (Pietrobelli et al., 1998). Athletes also have greater lean tissue than nonathletes, and such muscular development ± particularly in the arms ± may dramatically enhance the conductivity. Site-speci® c muscular development could introduce greater variance of conductivity in athletes than in non-exercising controls, rendering bioelectrical impedance analysis less accurate. In this respect, it is perhaps surprising that the equation of Lukaski and Bolonchuk (1987), which was derived on athletes, did not predict fat better. The diþ erence in percent fat prediction between DXA and bioelectrical impedance analysis from either formula was signi® cantly correlated with mesomorphy (r = 0.26; P < 0.05), which suggests that predictions based on bioelectrical impedance analysis may be population-speci® c to the extent that strength athletes may be poorly represented by a predictive formula based on endurance athletes. No equivalent correlation was observed in anthropometric predictions. Although it may be logical to expect subcutaneous fat ± the largest compartment ± to be representative of total body fat, this becomes less tenable as greater understanding emerges of the proportional distribution of fat between compartments (Hawes, 1996). In an investigation of regional fat placement in 165 soldiers, Nindl et al. (1996) reported a hierarchy of regional fat mobilization by DXA of abdomen > arms > legs as a consequence of weight loss, but this was not supported by skinfold evidence in the arms and legs. Such an observation could be explained by a change in the partitioning between subcutaneous and internal fat with weight loss. Evidence on fat partitioning from medical imaging techniques is con¯ icting. Using magnetic resonance imaging, Ross et al. (1994) found no signi® cant relationship between visceral adipose tissue and total adipose tissue in 17 obese men (over 30% fat), while DesprŠs et al. (1991), using computed tomography, found deep abdominal fat correlated with total fat and age (r = 0.76 and 0.63 respectively, P < 0.0001) in 110 men whose fatness ranged from 2.2 to 39.8% (22.9 ± 8.9%). The proportional distribution of fat in athletes is unknown, although a high proportion of total fat situated in the subcutaneous compartment would explain why skinfolds are relatively accurate in predicting total fat.

271 Dual-energy X-ray absorptiometry is calibrated for a normal expected range of tissue thickness, beyond which the accuracy may be reduced (Wahner and Fogelman, 1994). Separate corrections are therefore necessary to accommodate growth studies (Mitchell et al., 1996a). Ryde et al. (1998) found DXA (Lunar DPX model) to under- and overestimate small and large changes, respectively, in fat loss in 10 obese women compared with a four-compartment model. Experimental work with the same machine as that used in the present study indicated that fat measurements were stable up to a total tissue depth of 25 cm (Jebb et al., 1993), which exceeds the antero-posterior thickness of any of the athletes. Inter-manufacturer diþ erences (Tothill et al., 1994) are greater for sub-regions than the whole body, but of the three manufacturers, Hologic (as used in the present study) is the scanner best validated for accuracy in an athletic sample. Prior et al. (1997) found diþ erences between percent fat as derived by the four-compartment model and DXA to be weakly related to body mass index but not mesomorphy, suggesting that increasing fat mass, rather than lean tissue mass, contributes to the discrepancy. Although concern has been expressed about the accuracy of DXA in obese individuals, extremely lean people represent a greater threat to accuracy in the current sample. Nine athletes had fat measured by DXA less than 5% (a suggested minimal weight criterion used by Lohman, 1992), the lowest being only 2.9%. While this might appear plausible by comparison to two participants in the densitometric study of Adams et al. (1982), whose fat was predicted to be - 12%, +2.9% is still a remarkably low value. This ® gure could theoretically represent only essential lipids in the leanest athletes who had reduced their `excess fat’ to zero. Dual-energy X-ray absorptiometry failed to record any fat in the torso region of the leanest three individuals. The regression of torso fat against percent total fat predicted zero torso fat to occur at 3.3% total fat (95% CI = 2.2± 4.4). However, even for these leanest individuals, abdominal skinfold averaged 5.8 ± 0.3 mm. In addition, essential lipids in the liver and other organs suggest that zero fat in the torso is, in practical terms, impossible. A more plausible explanation is that accuracy is diminished at the lean extreme of this sample for several reasons. Poorer reproducibility may be expected in very lean individuals because the variance between replicate measures will represent a greater proportion of the measurement value. In addition, the bone-containing pixels, which DXA must interpolate across to derive a fat content, are most prevalent in the torso. The leanest individuals have proportionately fewer soft tissue pixels to provide the information for this interpolation and accuracy is reduced as a consequence. It is not possible using

272 DXA to measure soft tissue inside the skull, and the software used by the present study assumes a ® xed value of 17% fat. Lastly, the algorithms that predict fat content, which are clearly diþ erent between manufacturers and software versions, are designed to be robust for individuals of a normal range of adiposity, and extremes at both ends of this scale will compromise their stability. The ® ndings of Sinning et al. (1985) indicate that, using the two-compartment method (fat and fat-free mass) as the criterion, most generalized equations were not valid for predicting the fat content of athletes, with the exception of the three equations of Jackson and Pollock (1978). While the three-site equation of Jackson and Pollock is clearly the best of the four predictions compared in the present study, the conclusion of Adams et al. (1982) was that this method is inappropriate for some athletes because of the increased density of fat-free mass. The present results are not altogether surprising because the athletes studied here (n = 82) and those in the study of Sinning et al. (1985) had mean body mass indices within 0.3 kg ´ m- 2 and mean percent fat within 1%. This contrasts with the equivalent ® gures of Prior et al. (1997), whose male sample (67 athletes and 24 non-athletes) had mean body mass indices of 27.0 ± 4.6 kg ´m- 2 and fat of 13.1 ± 5.7% (by DXA). The greater mesomorphy of the athletes in the study of Prior et al. (1997) compared with the present study (6.2 ± 1.7 vs 5.0 ± 1.2) suggests considerable musculoskeletal development and potentially greater variation in the density of fat-free mass. Although good agreement has been reported between DXA and densitometry in 128 men and women (Wellens et al., 1994), in the study of Prior et al. (1997), variability in the density of fat-free mass beyond that expected in a sedentary population could explain the poorer association of percent fat by densitometry and the four-compartment model than by DXA and the four-compartment model. The accuracy of both methods relative to the four-compartment model was compared in 78 young and older men and women by Clasey et al. (1999). Overall, DXA was slightly more accurate: the total error for DXA and the four-compartment model compared with densitometry and the four-compartment model was 5.3% and 5.7% respectively, but the diþ erence was greatest for young men and older women. Although physical activity was not reported, a higher bone density in young men and a lower bone density in older women may have violated the assumption of the constant density for fat-free mass more than in the other groups. The criteria of Lohman (1996) suggest the evaluation of a new method of predicting percent fat accurately should include a standard error of the estimate within 3% of body mass. Using the data of Prior et al. (1997), DXA ful® ls the criteria (standard error of

Stewart and Hannan the estimate = 2.8%), while densitometry does not (standard error of the estimate = 3.4%). Dual-energy X-ray absorptiometry was also more accurate than densitometry compared with a six-compartment model (based on fat, water, protein, glycogen and osseous and non-osseous mineral) in a small heterogeneous sample (Wang et al., 1998). The implications of this are that DXA may be better than densitometry as a reference method. Thus, until a four-compartment criterion or alternative prediction is established for athletes, DXA may be the criterion method of choice. Compared with the best of the existing equations used (Jackson and Pollock, 1978), the standard error of the estimate of the prediction of percent fat in the present study is smaller (2.2% vs 3.3%), and the coeý cients of determination comparable (0.82 for the equation of Jackson and Pollock; 0.81 for the sample of 82 athletes, and 0.85 for all 106 athletes in the present study). This strengthens the case for promoting their use in predicting the fat content of athletes. Fat-free mass predictions using the anthropometric corrected girth model were of equivalent accuracy to bioimpedance, although both were less accurate than using skinfolds. Although fat-free mass was predicted with greater accuracy than fat, it was predicted slightly less accurately than the muscle mass prediction of Martin et al. (1990), which may be of greater utility with athletic groups. The validation of DXA against chemical analysis of pig carcasses (Mitchell et al., 1996b), the development of cross-calibration equations to convert between results from diþ erent manufacturers (Modlesky et al., 1996), and the widespread use of DXA have diminished the scope for earlier criticism of its accurate prediction of fat and fat-free masses (Kohrt, 1993, 1998). Considerable potential remains for inter-laboratory research using DXA, both as a reference method by itself and as an integral part of a four-compartment comparison. As a consequence of the increased availability and reduced scanning times with new fan-beam technology, the use of DXA in body composition studies in groups including athletes is likely to increase.

Acknowledgements The authors wish to thank R. Elton for advice and assistance with the statistical methods.

References Adams, J., Mottola, M., Bagnell, K.M. and McFadden, K.D. (1982). Total body fat content in a group of professional football players. Canadian Journal of Applied Sport Sciences, 7, 36± 40.

Prediction of fat and fat-free mass using DXA Baumgartner, R.N. (1996). Electrical impedance and total body electrical conductivity. In Human Body Composition (edited by A. Roche, S. Heyms® eld and T. Lohman), pp. 79± 108. Champaign, IL: Human Kinetics. Bland, J.M. and Altman, D.G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, i, 307± 310. Carter, J.E.L. and Heath, B.H. (1990). Somatotyping: Development and Applications. Cambridge: Cambridge University Press. Clasey, J.L., Kanaley, J.A., Wideman, L., Heyms® eld, S.B., Teates, C.D., Gutgesell, M.E., Thorner, M.O., Hartman, M.L. and Weltman, A. (1991). Validity of methods of body composition assessment in young and older men and women. Journal of Applied Physiology, 86, 1728± 1738. De Lorenzo, A., Sorge, S.P., Iacopino, L., Andreoli, A., de Luca, P.P. and Sasso, G.F. (1998). Fat-free mass by bioelectrical impedance vs dual-energy X-ray absorptiometry (DXA). Applied Radiation and Isotopes, 49, 739± 741. DesprŠs J.P., Prud’ homme, D., Tremblay, A. and Bouchard, C. (1991). Estimation of deep abdominal adipose tissue anthropometric measurements in men. American Journal of Clinical Nutrition, 54, 471± 477. Durnin, J.V.G.A. and Womersley, J. (1974). Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. British Journal of Nutrition, 32, 77± 97. Fogelman, I. and Blake, G.M. (1998). Clinical use of DXA measurements. In Current Research in Osteoporosis and Bone Mineral Measurement V (edited by E.F.J. Ring, D.M. Elvins and A.K. Bhalla), pp. 74± 75. London: British Institute of Radiology. Fuller, N.J., Laskey, M.A. and Elia, M. (1992). Assessment of the composition of major body regions by dual-energy X-ray absorptiometry (DEXA), with special reference to limb muscle mass. Clinical Physiology, 12, 253± 266. Goran, M.I., Driscoll, P., Johnson, R., Nagy, T.R. and Hunter, G. (1996). Cross-calibration of body composition techniques against dual-energy X-ray absorptiometry in young children. American Journal of Clinical Nutrition, 63, 299± 305. Hawes, M.R. (1996). Human body composition. In Kinanthropometry and Exercise Physiology Manual (edited by R. Eston and T. Reilly), pp. 5± 33. London: E & FN Spon. Heyward, V.H. and Stolarczyk, L.M. (1996). Applied Body Composition Assessment. Champaign, IL: Human Kinetics. Jackson, A.S. and Pollock, M.L. (1978). Generalized equations for predicting body density of men. British Journal of Nutrition, 40, 497± 504. Jebb, S.A., Goldberg, G.R. and Elia, M. (1993). DXA measurements of fat and bone mineral density in relation to depth and adiposity. In Human Body Composition (edited by K.J. Ellis and J.D. Eastman), pp. 115± 119. New York: Plenum Press. Kohrt, W.M. (1993). Body composition by DXA: Tried and true? Medicine and Science in Sports and Exercise, 27, 1349± 1353.

273 Kohrt, W.M. (1998). Preliminary evidence that DEXA provides an accurate assessment of body composition. Journal of Applied Physiology, 84, 372± 377. Lohman, T.G. (1992). Advances in Body Composition Assessment. Current Issues in Exercise Science Series Monograph No. 3. Champaign, IL: Human Kinetics. Lohman, T.G. (1996). Dual energy X-ray absorptiometry. In Human B ody Composition (edited by A. Roche, S. Heyms® eld and T. Lohman), pp. 63± 78. Champaign, IL: Human Kinetics. Lohman, T.G., Roche, A.F. and Martorell, R. (1988). Anthropometric Standardization Reference Manual. Champaign, IL: Human Kinetics. Lukaski, H.C. and Bolonchuk, W.W. (1987). Theory and validation of the tetrapolar bioelectrical impedance method to assess human body composition. In In vivo Body Composition Studies (edited by K.J. Ellis, S. Yasamura and W.D. Morgan), pp. 410± 414. London: The Institute of Physical Sciences in Medicine. Martin, A.D., Drinkwater, D.T., Clarys, J.P. and Ross, W.D. (1986). The inconstancy of the fat-free mass: A reappraisal with implications for densitometry. In Kinanthropometry III: Proceedings of the VII Commonwealth and International Conference on Sport, Physical Education, Dance, Recreation and Health (edited by T. Reilly, J. Watkins and J. Borms), pp. 92± 97. London: E & FN Spon. Martin, A.D., Spenst, L.F., Drinkwater, D.T. and Clarys, J.P. (1990). Anthropometric estimation of muscle mass in men. Medicine and Science in Sports and Exercise, 22, 729± 733. Mazess, R.B., Barden, H.S., Bisek, J.P. and Hanson, J. (1990). Dual-energy X-ray absorptiometry for total and regional bone mineral and soft tissue composition. American Journal of Clinical Nutrition, 51, 1106± 1112. Mitchell, A.D., Conway, J.M. and Scholz, A.M. (1996a). Incremental changes in total and regional body composition of growing pigs measured by dual-energy X-ray absorptiometry. Growth, Development and Aging, 60, 95± 105. Mitchell, A.D., Conway, J.M. and Potts, W.J.E. (1996b). Body composition analysis of pigs by dual-energy X-ray absorptiometry. Journal of Animal Science, 74, 2663± 2671. Modlesky, C.M., Lewis, R.D., Yetman, K.A., Rose, B., Rosskopf, L.B., Snow, T.K. and Sparling, P.B. (1996). Comparison of body composition and bone mineral measurement from two DXA instruments in young men. American Journal of Clinical Nutrition, 64, 669± 676. Nindl, B.C., Friedl, K.E., Marchitelli, L.J., Shippee, R.L., Thomas, C.D. and Patton, J.F. (1996). Regional fat placement in physically ® t males and changes with weight loss. Medicine and Science in Sports and Exercise, 28, 786± 793. Nord, R.H. and Payne, R.K. (1995). DXA vs underwater weighing: Comparison of strengths and weaknesses. Asian Paci® c Journal of Clinical Nutrition, 4, 173± 175. Pietrobelli, A., Faith, M.S., Allison, D.B., Gallacher, D., Chiumello, G. and Heyms® eld, S.B. (1998). Body mass index as a measure of adiposity among children and adolescents: A validation study. Journal of Pediatrics, 132, 204± 210.

274 Prior, B.M., Cureton, K.J., Modlesky, C.M., Evans, E.M., Sloniger, M.A., Saunders, M. and Lewis, R.D. (1997). In vivo validation of whole body composition estimates from dual-energy X-ray absorptiometry. Journal of Applied Physiology, 83, 623± 630. Ross, R., Shaw, K.D., Rissanen, J., Martel, Y., de Guise, J. and Avrush, L. (1994). Sex diþ erences in lean and adipose tissue distribution by magnetic resonance imaging: Anthropometric relationships. American Journal of Clinical Nutrition, 59, 1277± 1285. Ryde, S.J., Eston, R., Laskey, M.A., Evans, C.J. and Hancock, D.A. (1998). Changes in body fat: Measurements by neutron activation, densitometry and dual X-ray absorptiometry. Applied Radiation and Isotopes, 49, 507± 509. Sinning, W.E., Dolny, D.G., Little, K.D., Cunningham, L.N., Racaniello, A., Siconol® , S.F. and Sholes, J.L. (1985). Validity of `generalized’ equations for body composition analysis in male athletes. Medicine and Science in Sports and Exercise, 17, 124± 130. Siri, W.E. (1956). The gross composition of the body. In Advances in Biological and Medical Physics (edited by J. Lawrence and C. Tobias), pp. 239± 280. London: Academic Press.

Stewart and Hannan Stewart, A. and Eston, R. (1997). Skinfold thickness measurement. British Journal of Nutrition, 78, 1040± 1042. Streat, S., Beddoe, A. and Hill, G. (1985). Measurement of body fat and hydration of the fat-free body in health and disease. Metabolism, 34, 509± 518. Tothill, P., Avenell, A., Love, J. and Reid, D.M. (1994). Comparisons between Hologic, Lunar and Norland dualenergy X-ray absorptiometers and other techniques used for whole body soft tissue measurements. European Journal of Clinical Nutrition, 48, 781± 794. Wahner, H.W. and Fogelman, I. (1994). The Evaluation of Osteoporosis: Dual Energy X-Ray Absorptiometry in Clinical Practice. London: Dunitz. Wang, Z.M., Deurenberg, P., Guo, S.S., Pietrobelli, A., Wang, J., Pierson, R.N., Jr. and Heyms® eld, S.B. (1998). Six-compartment body composition model: Inter-method comparisons of total body fat measurement. International Journal of Obesity, 22, 329± 337. Wellens, R., Chumlea, W.C., Guo, S., Roche, A.F., Reo, N.V. and Siervogel, R.M. (1994). Body composition in white adults by dual-energy X-ray absorptiometry, densitometry and total body water. American Journal of Clinical Nutrition, 59, 547± 555.

Prediction of fat and fat-free mass in male athletes using

12 Pages • 5,394 Words • PDF • 200.8 KB

prediction, learning, and games (2006)

407 Pages • 196,585 Words • PDF • 5.5 MB

Functional Movement Screen Scores in a Group of Running Athletes.

5 Pages • 4,332 Words • PDF • 83.8 KB

ADME – PreADMET Prediction of ADMETox

1 Pages • 47 Words • PDF • 24.3 KB

Notes on \'Superforecasting_ The Art and Science of Prediction\'

22 Pages • 5,660 Words • PDF • 139.6 KB

A Comparison of Traditional and Block Periodized Strength Training Programs in Trained Athletes.

8 Pages • 5,132 Words • PDF • 351.8 KB

6th Solucionário - Incropera - Fundamentals of Heat and Mass Transfer 6th

2,313 Pages • 539,535 Words • PDF • 73.4 MB

30x30 Fat Loss Shred 2.0

44 Pages • 5,508 Words • PDF • 11.4 MB

Fundamentals of Heat and Mass Transfer (7th Ed) - Incropera

1,076 Pages • 487,857 Words • PDF • 41.7 MB

Injury incidence and patterns among dutch crossfit athletes

13 Pages • 5,538 Words • PDF • 476.4 KB

Using IP Block List Providers and the Connection Filter agent in Exchange 2007

5 Pages • 1,958 Words • PDF • 649.8 KB

Hydrodynamics, Mass, and Heat Transfer in Chemical Engineering, CRC (2002) - OPET

406 Pages • 139,102 Words • PDF • 19.7 MB