Impact of indexing resting metabolic rate against fat-free mass ...

European Journal of Clinical Nutrition (2004) 58, 1132–1141

& 2004 Nature Publishing Group All rights reserved 0954-3007/04 $30.00 www.nature.com/ejcn

ORIGINAL COMMUNICATION Impact of indexing resting metabolic rate against fatfree mass determined by different body composition models J LaForgia1*, GE van der Ploeg2, RT Withers2, SM Gunn2, AG Brooks2 and BE Chatterton3 1 School of Pharmaceutical, Molecular and Biomedical Sciences, University of South Australia, Adelaide, Australia; 2Exercise Physiology Laboratory, School of Education, Flinders University, Adelaide 5001, Australia; and 3Department of Nuclear Medicine, Royal Adelaide Hospital, Adelaide, Australia

Objective: To examine the differences arising from indexing resting metabolic rate (RMR) against fat-free mass (FFM) determined using two-, three- and four-compartment body composition models. Design: All RMR and body composition measurements were conducted on the same day for each subject following compliance with premeasurement protocols. Subjects: Data were generated from measurements on 104 males (age 32.1712.1 y (mean7s.d.); body mass 81.15712.85 kg; height 179.576.5 cm; body fat 20.677.6%). Interventions: Body density (BD), total body water (TBW) and bone mineral mass (BMM) were measured by hydrodensitometry, deuterium dilution and dual energy X-ray absorptiometry (DXA), respectively. These measures were used to determine two (hydrodensitometry: BD; hydrometry: TBW)-, three (BD and TBW)- and four- compartment (BD, TBW and BMM) FFM values. DXA also provided three compartment derived FFM values. RMR was measured using open circuit indirect calorimetry. Results: Three (body fat group: lean, moderate, high) five (body composition determination: hydrodensitometry, hydrometry, three-compartment, DXA, four-compartment) ANOVAs were conducted on FFM and RMR kJ.kg FFM1.d1. Within-group comparisons revealed that hydrodensitometry and DXA were associated with significant (Po0.001) overestimations and underestimations of FFM and RMR kJ.kg FFM1.d1, respectively, compared with four-compartment-derived criterion values. A significant interaction (Po0.001) resulted from DXA’s greater deviations from criterion values in lean subjects. While hydrometric means were not significantly (PZ0.68) different from criterion values intraindividual differences were large (FFM: 1.5 to 2.9 kg; RMR: 6.0 to 3.2 kJ.kg FFM-1.d-1). Conclusion: The relationship between RMR kJ.kg FFM1.d1 and exercise status would best be investigated using three (BD, TBW)- or four (BD, TBW, BMM)-compartment body composition models to determine FFM. Other models either significantly underestimate indexed RMR (hydrodensitometry, DXA) or display large intraindividual differences (hydrometry) compared with four-compartment derived criterion values. Sponsorship: Australian Research Council (small grants scheme).

European Journal of Clinical Nutrition (2004) 58, 1132–1141. doi:10.1038/sj.ejcn.1601941 Published online 31 March 2004 Keywords: hydrodensitometry; hydrometry; DXA; multicompartment body composition models; RMR

Introduction *Correspondence: J LaForgia, School of Pharmaceutical, Molecular and Biomedical Sciences, University of South Australia, Adelaide, 5001, Australia. E-mail: [email protected] Guarantor: J LaForgia. Contributors: JLF, GEV, SMG and AGB recruited the subjects, collected and analysed the data; JLF, RTW and GEV conceived and wrote the paper; BEC supervised the DXA scans. Received 3 July 2003; revised 26 October 2003; accepted 27 November 2003; published online 31 March 2004

Numerous studies have been published over the last two decades regarding the response of the resting metabolic rate (RMR) to exercise training. The impetus for this work stems from the fact that small but chronic changes in the RMR, which comprises B60–70% of the total daily energy expenditure (TDEE) in humans (Poehlman et al, 1988), have a cumulative impact on the maintenance of body mass. While it is clear that physical activity is associated with the

Impact of indexing resting metabolic rate J LaForgia et al

1133 long-term maintenance of fat-free mass (FFM), which is the greatest determinant of the RMR (kJ/day), no clear trend has emerged identifying an increased RMR independent of changes in FFM kJ.kg FFm1.d1. The equivocal findings have been partly attributed to methodological problems. However, among the longitudinal studies that had adequate subject numbers and were relatively well controlled, a clear trend still remained elusive. Some (Poehlman & Danforth, 1991; Berke et al, 1992; Goran & Poehlman, 1992; Poehlman et al, 1992; Pratley et al, 1994) found that the RMR of older subjects (50–80 y) increased in response to moderate aerobic or resistance training independent of changes in FFM, while others (Frey-Hewitt et al, 1990; Broeder et al, 1992; Westerterp et al, 1992) reported no RMR changes in sedentary younger subjects who participated in training interventions. Also, one study (LaForgia et al, 1999) reported that 3 weeks, detraining in young physically active males had no apparent impact on the RMR. While all these studies were generally robust in their experimental designs and controls for the measurement of RMR, the majority have employed twocompartment body composition models that are generally prone to significant errors in the determination of the FFM. Only three studies (Goran & Poehlman, 1992; Westerterp et al, 1992; LaForgia et al, 1999) utilised more valid multicompartment body composition models. The popular two-compartment (fat mass or FM, FFM) models of hydrodensitometry and hydrometry both assume a consistency in the composition of the FFM. Therefore, the validity of determining FFM by these methods is adversely affected by biological variability in the assumed percentages (Brozek et al, 1963) for the four FFM components (total body water or TBW ¼ 73.72%, protein ¼ 19.41%, bone mineral or BM ¼ 5.63%, nonbone mineral or NBM ¼ 1.24%). Withers et al (1999) have estimated that the variability in the TBW component of the FFM within the normal healthy population may lead to percentage body fat (%BF) errors of up to 6.7 and 4.8% BF for hydrodensitometry and hydrometry, respectively. The three (FM, TBW, fat-free dry solid or FFDS)and four (FM, TBW, BM, residual)- compartment models control for biological variability in TBW and both TBW and BM, respectively. These multicompartment models therefore theoretically provide more valid estimates of body composition than the two-compartment models because their densitometric assumptions relate to smaller compartments (FFDS and residual). The four-compartment model currently represents the best in vivo criterion against which other methods may be validated (Elia, 1992; Withers et al, 1999). Comparisons between two-compartment and multicompartment estimates of body composition have been reported for a variety of subjects. Some investigators have found that hydrodensitometry significantly over- and underestimated %BF in older (Baumgartner et al, 1991; Smith et al, 1999) and younger (Withers et al, 1998a) subjects, respectively. Two other studies (Modlesky et al, 1996; Withers et al, 1997) have reported significant overestimations in

hydrodensitometrically determined %BF in weight-trained subjects. While other investigators (Friedl et al, 1992; Cote & Adams, 1993; Penn et al, 1994; Siconolfi et al, 1995; Prior et al, 1997; Visser et al, 1997; Goran et al, 1998; Clasey et al, 1999) have not found significant mean differences, their data have displayed large individual differences between two- and multicompartment techniques. This was exemplified by the hydrodensitometric data of Prior et al (1997) that contained deviations from their four-compartment values ranging from 8.1 to 8.5%BF in young men and women. Some studies have also compared three- and fourcompartment body composition estimates. Three-compartment models involving the measurement of body density and TBW generate body composition values that closely approximate those of four-compartment models (Friedl et al, 1992; Siconolfi et al, 1995; Modlesky et al, 1996; Withers et al, 1998a). However, dual-energy X-ray absorptiometry (DXA), which also divides the body into three compartments (FM, BM and bone mineral free lean), may display significant deviations from four-compartment-derived FFM values (Friedl et al, 1992; Modlesky et al, 1996; Withers et al, 1998a). It has also been established that DXA %BF values significantly underestimate four-compartment estimates in lean individuals (Withers et al, 1998a); Gallagher et al, 2000; van der Ploeg et al, 2001a; van der Ploeg et al, 2003). As relatively small changes (B5–8%) in RMR are regarded as physiologically significant, it is crucial that body composition estimates are as accurate as possible. The impact of discrepancies between two- and multicompartment estimates of body composition on the indexed RMR should therefore be determined. This also has to be established for DXA, which has become a popular method of body composition analysis because it provides rapid measurements that are atraumatic. However, only two studies (LaForgia et al, 1999; Smith et al, 1999) have compared RMR values indexed against FFM determined using two- and multicompartment models. One of the studies (LaForgia et al, 1999) was limited to young physically active males, while the other used older males and both did not include DXA or hydrometry. The aim of this study was therefore to compare RMR values indexed against FFM estimates from two- and multicompartment models in males spanning a wide range for age and %BF.

Methods Subjects The descriptive statistics for the 104 males are presented in Table 1. These data are a combination of our previously published studies on 18–60 y-old male subjects (LaForgia et al, 1997, 1999; van der Ploeg et al, 2001b; van der Ploeg & Withers, 2002) plus one other person. The emphases of these publications were different from that of the present paper. All subjects were nonsmokers and reported good European Journal of Clinical Nutrition


1134 Table 1 Descriptive statistics for the entire sample and groupings segregated according to percentage body fat (%BF)

Age (y) Height (cm) Mass (kg) BMI (kg/m2) %BF (four compartment)

All n ¼ 104

Low %BF (o14% BF) n ¼ 31

Moderate %BF (Z14r24% BF) n ¼ 41

High %BF (424% BF) n ¼ 32

32.1712.1 (18.0–58.7) 179.576.5 (166.0–198.4) 81.15712.85 (57.08–115.21) 25.273.4 (18.1–33.4) 20.677.6 (8.0–36.6)

27.279.8 (18.0–52.4) 177.877.3 (166.0–198.4) 69.7278.25 (57.08–85.87) 22.071.7 (18.1–25.5) 11.571.5 (8.0–13.9)

31.9712.2 (18.9–57.8) 180.776.2 (172.4–192.1) 80.7179.130 (58.85–99.29) 24.772.1 (19.8–29.6) 20.472.6 (15.0–24.0)

37.0712.4 (19.0–58.7) 179.476.0 (170.9–196.6) 92.79710.21 (76.55–115.21) 28.872.5 (23.3–33.4) 29.573.6 (24.2–36.6)

Values are means7s.d., with range in parentheses. BMI, body mass index.

health and mass stability (72 kg) during the preceding 12 months. Ethical approval for these studies was obtained from the Flinders Medical Centre’s Committee on Clinical Investigation prior to the commencement of data collection. Written informed consent was also obtained from the subjects.

Experimental protocol An initial laboratory visit was used to discuss the previously distributed written information regarding the study. This session was also used to perform an RMR habituation trial. The next visit involved a second RMR habituation trial. Experimental RMR measures were conducted on the morning of a third visit and were immediately followed by body composition determinations. Subjects were categorised as low, moderate or high %BF (Table 1).

Body composition Five estimates of body composition were determined using three densitometric models, hydrometry and DXA. The former included two (FM; FFM)-, three (FM; TBW; FFDS)- and four-compartment (FM; TBW; BM; residual) body composition models. Body density (BD), TBW and BM were measured by hydrodensitometry, deuterium dilution and DXA, respectively. The specific measurement techniques have been fully described elsewhere (Withers et al, 1998a). Respiratory system gas volume for BD determination was measured by either oxygen dilution or helium dilution (van der Ploeg et al, 2000). Hydrometric determination of FFM was based on the assumption that water comprises 72% of the FFM (Withers et al, 1998a). DXA FFM values were derived from the scan data by summing lean tissue and bone mineral content (BMC) after the latter, which represented ashed bone mineral, had been adjusted (BM ¼ 1.0436BMC) (Mendez et al, 1960). All measurements were conducted on the same morning when the subjects were 12 h postprandial and euhydrated. European Journal of Clinical Nutrition

RMR Subjects reported to the laboratory at 0720 hours after a 36 h abstention from exercise which is more than sufficient time to allow for dissipation of the excess postexercise oxygen consumption (Mole, 1990). A normal evening meal was ingested by 2000 hours on the preceding day with only water permitted thereafter. Subjects were weighed after voiding and then asked to attach a heart rate transmitter (Polar Electro OY, Kempele, Finland). They then rested quietly on a bed. All RMR measurements were conducted after 50 min of bed rest, while the subject was supine with the head and shoulders slightly elevated, breathing through a Hans Rudolph R2600 respiratory valve and wearing a nose clip. The subjects were covered by a blanket and the ambient temperature in their vicinity was maintained at 24.070.51C. A 150-l Douglas bag (Plysu Industrial, Milton Keynes, Buckinghamshire, UK), which had previously been flushed with the subject’s expirate, was connected to the expiratory port of the respiratory valve via a two-way straight through valve. After sufficient time allowed for the washout of the tubing’s deadspace, the subject was switched into the Douglas bag at the end of an expiration. Following B10 min of expirate collection the subject was disconnected from the Douglas bag at the end of an expiration. A second B10 min collection commenced 5 min after the first. The expirate was analysed using a Sensormedics, LB-2 CO2 analyser (Anaheim, CA, USA) and an AEI Technologies, S-3A O2 analyser (Pittsburgh, PA, USA), which were calibrated before testing and checked afterwards for drift using three gases that had been authenticated by Lloyd-Haldane analyses. The expirate volume was then measured via a 350-l Tissot spirometer (Warren Collins, Braintree, MA, USA) that had been mapped for constant cross-sectional area throughout its elevation. The equation of Elia and Livesey (1992) was implemented to determine RMR (kJ/h) for the two collections that were then averaged. Heart rate was recorded during the RMR measurements. RMR reliability trials using six subjects measured on consecutive days yielded an intraclass


1135 correlation coefficient of 0.909 and a technical error of measurement of 3.9 kJ/h.

Statistical analysis The associations between RMR indexed against the criterion four-compartment estimates of FFM and values derived using the four other FFM estimates were evaluated by linear regression analyses. Significant differences between the body composition models for FFM and indexed RMR values were identified using a three (body fat group: lean, moderate, high) five (body composition method: hydrodensitometry, hydrometry, three -compartment, DXA, four-compartment) ANOVA with repeated measures across methods. Any statistically significant F-ratios arising from the main effects and interactions were adjusted using a Greenhouse–Geisser epsilon value if the sphericity assumption was violated. Tukey’s (HSD) post-hoc tests were applied if significance was reached (Pr0.05). As our primary focus in the aforementioned analysis was related to the within-subject comparisons, there was no expectation that the resultant inferences would suffer a reduction in validity due to mathematical bias (Ravussin & Bogardus, 1989).

Results Table 2 presents the FFM and RMR data for the subjects, who were segregated into three groups, based on fourcompartment %BF (low %BF: o14% BF, n ¼ 31; moderate %BF: X14 p24% BF, n ¼ 41; high %BF: 424% BF, n ¼ 32). Hydrodensitometry and DXA significantly (Pr0.005) overestimated FFM in all three groups compared with the four-compartment criterion model. While the mean differences between hydrodensitometric and four-compartment FFM were similar for all three groups (low %BF ¼ 1.8 kg; moderate %BF ¼ 1.9 kg; high %BF ¼ 1.4 kg), the DXA differences were greatest in the lean group (low %BF ¼ 2.3 kg; moderate % BF ¼ 1.4 kg; high %BF ¼ 0.9 kg). The latter were responsible for a significant group by body composition method interaction (Po0.001). Hydrometry and the three-compartment method produced FFM values that were not significantly (PZ0.68) different from the four-compartment estimates. Figures 1–4 display the relationships between RMR indexed against FFM determined by the criterion four-compartment model and the four other body composition models. The aforementioned FFM differences resulted in hydrodensitometry (Figure 1) and DXA (Figure 4) generating lower correlation coefficients (hydrodensitometry:

Table 2 FFM and RMR data for the three groups of subjects Low %BF (o14% BF) n ¼ 31

Moderate %BF (Z14r24% BF) n ¼ 41

High %BF (424% BF) n ¼ 32

FFM (kg) (hydrodensitometry)

63.5677.39 (52.57–76.52)

66.1177.10 (50.24–77.87)

66.6677.16 (57.24–84.84)

FFM (kg) (hydrometry)

61.8177.51 (49.95–75.17)

64.3777.43 (51.58–79.36)

65.7976.88 (56.28–81.59)

FFM (kg) (three compartment)

61.7477.31 (50.73–74.07)

64.2777.14 (49.71–77.53)

65.3076.85 (57.84–81.90)

FFM (kg) (DXA)

64.0077.31 (52.21–77.90)

65.5876.95 (50.20–78.14)

66.1677.26 (57.47–85.25)

FFM (kg) (four compartment)

61.7477.39 (50.77–74.25)

64.1877.18 (49.57–77.68)

65.2876.95 (57.11–82.10)

RMR, kJ/day

73667935 (5417–9602)

74007808 (5765–9204)

76627759 (6825–9771)

RMR (kJ.kg FFM1.d1) (hydrodensitometry)

116.178.1 (101.8–132.3)

112.278.6 (92.4–137.1)

115.275.8 (100.9–127.3)

RMR (kJ.kg FFM1.d1) (hydrometry)

119.578.8 (104.2–141.2)

115.478.9 (95.2–140.6)

116.776.1 (99.4–127.7)

RMR (kJ.kg FFM1.d1) (three compartment)

119.578.8 (104.1–139.1)

115.578.7 (95.2–140.8)

117.575.8 (101.4–129.2)

RMR (kJ.kg FFM1.d1) (DXA)

115.277.4 (101.4–131.4)

113.178.6 (94.3–135.7)

116.277.4 (96.0–131.7)

RMR (kJ.kg FFM1.d1) (four compartment)

119.578.6 (104.4–139.0)

115.778.7 (95.3–140.9)

117.675.8 (101.6–129.1)

Values are means7s.d., with range in parentheses. %BF, percent body fat; DXA, dual-energy X-ray absorptiometry.

European Journal of Clinical Nutrition


1136 150

y = 0.986x + 4.778 r2 = 0.931 SEE = 2.1 kJ.kg FFM-1.d-1

RMR kJ.kg FFM (4C)-1.d-1

140

Trend line

130 Line of identity 120

110

Low %BF, n = 31 Moderate %BF, n = 41

100

High %BF, n = 32

90 90

100

110

120

130

140

150

RMR kJ.kg FFM (Hydrodensitometry)-1.d-1

Figure 1

Association between RMR indexed against four-compartment (4C) and hydrodensitometric FFM.

150 y = 0.931x + 8.528 r2 = 0.964 SEE = 1.5 kJ.kg FFM-1.d-1


140

130 Trend line 120

110 Low %BF, n = 31 Moderate %BF, n = 41 100 High %BF, n = 32 Line of identity 90 90

100

110

120

130

140

150

RMR kJ.kg FFM (Hydrometry)-1.d-1

Figure 2

Association between RMR indexed against four-compartment (4C) and hydrometric FFM.

r2 ¼ 0.931; hydrometry: r2 ¼ 0.963; three-compartment: r2 ¼ 0.994; DXA: r2 ¼ 0.866) and larger SEEs (hydrodensitometry, 2.1; hydrometry, 1.5; three-compartment, 0.6 and DXA, 2.9 kJ.kg FFM1.d1) compared with the other methods. Comparison of the indexed RMR values reflected the previous results of the FFM analysis with both hydrodensiEuropean Journal of Clinical Nutrition

tometry- and DXA-derived values falling significantly (Po0.001) below those using four-compartment FFM estimates. Hydrometry- and three-compartment-related RMR values were not significantly different (PZ0.58) from the four-compartment derived values. The significant (Po0.001) group by body composition method interaction was due to the DXA-associated RMR differences


1137 150

y = 0.985x + 1.789 r2 = 0.994 SEE = 0.4 kJ.kg FFM-1.d-1


140

Line of identity

130

120

110 Trend line


100

High %BF, n = 32 90 90

100

110

120

130

140

150


Figure 3

Association between RMR indexed against four-compartment (4C) and three-compartment (3C) FFM.

150

y = 0.937x + 9.983 2 r = 0.866 -1 -1 SEE = 2.9 kJ.kg FFM .d

Trend line

130

-1

RMR kJ.kgFFM (4C) .d

-1

140

Line of identity 120

110


100

High %BF, n = 32

90 90

100

110

120

130 -1

140

150

-1

RMR kJ.kgFFM (DXA) .d

Figure 4

Association between RMR indexed against four-compartment (4C) and DXA FFM.

diminishing from the lean to the high fat group (low %BF, 4.3, moderate %BF, 2.5 and high % BF, 1.4 kJ.kg FFM1.d1). Indexed RMR differences between hydrodensitometry and the four compartment model did not display the aforementioned trend (low %BF, 3.5, moderate %BF, 3.4 and high %BF, 2.4 kJ.kg FFM1.d1). The adjustment of indexed RMR values (Ravussin and Bogardus, 1989) to avoid mathematical bias did not alter the trends displayed by the aforementioned results.

Discussion Although Siri (1961) identified the limitations of the twocompartment hydrodensitometric body composition model some time ago, investigators continue to use it extensively. Hydrodensitometry has been the method of choice for the majority of cross-sectional and longitudinal studies that have attempted to determine the link between exercise status and RMR independent of FFM. Only a small number of cross-sectional (Smith et al, 1997, 1999; Withers et al, 1998b) European Journal of Clinical Nutrition


1138 and longitudinal (Goran & Poehlman, 1992; Westerterp et al, 1992; LaForgia et al, 1999) studies have employed the more valid multicompartment models. The robustness of the conclusions reached by those studies, which utilised hydrodensitometry, is therefore diminished given that small but physiologically significant (5–10%) differences in RMR resulting from exercise training may be masked or falsely created due to errors in estimating FFM. Analysis of the data used in this study has revealed significant overestimations of FFM (1.4–1.9 kg, Pr0.005) in each of the three groups by hydrodensitometry compared with the criterion four-compartment estimates. Furthermore, individual variations were large and ranged from 1.8 to 4.0 kg. This finding was consistent with previous work completed in our laboratory involving young men and women (Withers et al, 1998a), who displayed similar mean overestimations in hydrodensitometrically determined FFM values (1.3–1.7 kg) across sedentary and active groups. However, another study conducted in our laboratory (Smith et al, 1999) on older sedentary and active males (54–71 y) found that hydrodensitometry underestimated FFM in both groups (B1.0 kg). This finding concurred with other data on older (65–94 y) men and women (Baumgartner et al, 1991). The reason for the aforementioned over- and underestimations of FFM relate to the FFM hydration values for the older subjects (75.17%) exceeding the hydrodensitometry assumed value of 73.72% and the value for younger subjects (B72.30%) falling below this value. We chose not to segregate our subjects into older and younger groups because none were older than 60 y and only 15 were 50 y or older (Z50r58.7 y). Furthermore, these 15 subjects had a mean FFM hydration (72.9%), which was below the hydrodensitometric constant. The DXA-derived FFM for our subjects was also significantly greater than its four-compartment equivalent. However, unlike the hydrodensitometry overestimations, which were relatively consistent across the three groups, the DXA data indicated that the largest FFM errors occurred in the low %BF group. This trend was consistent with the findings of Gallagher et al (2000). Unfortunately, it is unclear as to why DXA FFM errors are greater in lean subjects; however, beam hardening compensations that account for the influence of tissue thickness are thought to be a contributing factor (Gallagher et al, 2000). While only one RMR/exercise status study has utilised DXA (Ryan et al, 1995), the popularity of this technology for determining body composition is growing rapidly. It is important therefore that critical appraisals continue to be conducted to better establish the utility of DXA in the determination of FFM across a range of subjects. While hydrometric FFM did not differ significantly (PZ0.51) from the four-compartment values, individual deviations were sizeable (1.5 to 2.9 kg) but smaller than those for hydrodensitometry (1.8 to 4.0 kg) and DXA (2.2 to 4.9 kg). This finding was ostensibly due to the FFM hydration values obtained for each group (low %BF ¼ 72.1%; moderate %BF ¼ 72.2%; high %BF ¼ 72.6%) approximating 72.0%, which was the value we assumed to determine FFM European Journal of Clinical Nutrition

by hydrometry (Withers et al, 1999). However, large significant errors would occur if the aforementioned value was applied uncritically to the previously mentioned older subjects, who displayed higher FFM hydrations (B75%). The three-compartment FFM values were essentially identical to the four compartment estimates (Table 2) and individual deviations were small (0.5 to 0.5 kg). This indicated that the total mineral/protein ratio of 0.354 that results in an FFDS density of 1.569 g/cm3, which is assumed by the threecompartment model (Withers et al, 1998a), was not violated. Furthermore, the closeness of the three- and four-compartment FFM estimates also highlights that the hydrodensitometric errors ostensibly resulted from variability in FFM hydration and not BM. The significant overestimations of FFM by hydrodensitometry and DXA compared with the four-compartment estimates translated to significant (Po0.001) underestimations in indexed RMR (kJ.kg FFM1.d1). While the correlations between indexed RMR for these two methods and the four-compartment-derived values (Figures 1 and 4) were high (hydrodensitometry: r2 ¼ 0.931; DXA: r2 ¼ 0.866), considerable intraindividual differences existed across the three groups of subjects (hydrodensitometry: 7.6 to 3.7 kJ.kg FFM1.d1; DXA: 7.3 to 4.2 kJ.kg FFM1.d1) resulting in sizeable SEEs (hydrodensitometry: SEE ¼ 2.1 kJ.kg FFM1.d1; DXA: SEE ¼ 2.9 kJ.kg FFM1.d1). Also, the largest mean difference of 4.3 kJ.kg FFM1.d1, which occurred in the low %BF group with DXA, approached a level (B5.0–6.0 kJ.kg FFM1.d1) that investigators would generally use as an appropriate effect size to establish if the RMR is sensitive to exercise status. The estimates of FFM using hydrometry also resulted in RMR values that were highly correlated with the fourcompartment-derived data (Figure 2). While the intraindividual differences were not markedly less (6.0 to 3.2 kJ.kg FFM1.d1) than those obtained for hydrodensitometry and DXA, the trend line displayed less bias and the SEE was smaller (1.5 kJ.kg FFM1.d1). The mean differences between the hydrometry and four-compartment-indexed RMR values for each of the groups were not significant (PZ0.58); however, the small differences displayed an increasing trend from the low to the high %BF group (low %BF: 0.0; moderate %BF: 0.3; high %BF: 0.9 kJ.kg FFM1.d1). This trend was linked to the progressive drift above the assumed FFM hydration constant of 72.0%, which was identified previously, thereby progressively overestimating the FFM. Figure 3 indicates that the RMR data obtained using the three-compartment estimates of FFM deviated minimally from the four-compartment-derived values. Intraindividual differences only ranged from 1.0 to 0.8 kJ.kg FFM1.d1. The impact of the aforementioned errors on the validity of conclusions drawn from investigations attempting to clarify the association between RMR and exercise status will depend on the groups studied and the influence of treatments on body composition. Hydrodensitometry produced consistent errors over a wide %BF range in the current study and would


1139 therefore not be expected to produce biased results when comparing the indexed RMR of leaner physically active subjects with sedentary, fatter individuals. Smith et al (1999) demonstrated this with active (20.4% BF) and sedentary (29.6% BF) older male subjects. While they did not present their RMR data indexed against hydrodensitometrically determined FFM, they stated that it yielded the same conclusions as for the data generated using their more valid four-compartment FFM values. However, it should be noted that the use of hydrodensitometrically determined FFM values could produce flawed conclusions when making RMR comparisons between elderly and younger individuals. It was noted earlier that hydrodensitometry underestimated and overestimated FFM in elderly and younger subjects, respectively, because of different FFM hydrations. This highlights the need for caution when applying hydrometry across age cohorts. Another concern related to two-compartment models involves their ability to detect body composition changes associated with longitudinal intervention studies. Sedentary subjects will increase their muscular glycogen stores significantly and therefore TBW as a result of exercise training. Evidence for this phenomenon was reported by Goran & Poehlman (1992) who trained elderly subjects for 8 weeks. They found that the significant post-treatment increase in FFM, which was determined using a three-compartment model, was equivalent to the measured increase in TBW. A shift in TBW is unlikely to be detected accurately by hydrodensitometry and hydrometry because they both assume that the composition of the FFM is fixed. We have reported evidence for this in earlier work (LaForgia et al, 1999) related to the detraining of eight physically active young male subjects, who were included in the current larger data set. Our four-compartment estimate of FFM revealed that a significant decrease (B0.70 kg) in FFM following 3 weeks detraining, which was equivalent to the measured loss of TBW, was only partially detected (B0.35 kg) by hydrodensitometry. However, the current cross-sectional data cannot provide further insights into the suitability of the two-compartment body composition methods for longitudinal studies. Apart from providing more expedient body composition measures, DXA could not be regarded currently as a superior replacement for the two-compartment models. The errors discussed previously may be problematical for both crosssectional and longitudinal investigations. Comparisons between active and sedentary individuals, who generally differ in body fatness, would have reduced legitimacy because of the propensity of DXA to overestimate FFM progressively in lean subjects. While our data do not directly evaluate the ability of DXA to detect longitudinal body composition changes accurately it certainly raises concerns. It could be inferred that significant reductions in body fat resulting from exercise interventions may lead to larger DXA FFM estimate errors on post-treatment measures. Nevertheless, some studies have reported favourably on DXA’s

ability to detect body composition changes accurately, following acute alterations in hydration (Prior et al, 1997) and the placement of exogenous fat over various parts of the body (Kohrt, 1995). However, these designs may not represent a suitable model for the body composition changes that result from exercise training. The suitability of DXA for longitudinal exercise training or detraining applications requires further investigation. The preceding discussion indicates that it would be wise to employ three- and four-compartment FFM estimates when indexing the RMR in nonelderly males. Whether this recommendation is valid for female and elderly male subjects should be established by future studies. It would appear that FFM errors associated with hydrodensitometry also apply to female subjects (Withers et al, 1998a). However, female subjects would probably not display the same magnitude of error for DXA FFM estimates because they are generally not as lean as males. Evidence also suggests that hydrodensitometry significantly underestimates FFM in elderly subjects (Baumgartner et al, 1991; Smith et al, 1999), which would lead to overestimates of the indexed RMR. The superior accuracy of the multicompartment models would generally provide greater confidence when comparing the findings of different studies. Furthermore, their accuracy may be pivotal in attempts to explain a larger proportion of the RMR variance in future work focusing on the impact of organ mass/FFM ratios. It should furthermore be noted that these models suffer to some extent from the cumulative errors associated with multiple measurements (BD, TBW, BM). However, the propagated error (Withers et al, 1998a) determined for our threeand four-compartment models (B0.4 kg FFM) is much smaller than the differences identified in the current data between hydrodensitometry, DXA and the four-compartment model. FFM estimates using the three- and fourcompartment models are also relatively insensitive to departures from the assumed densities for the small FFDS and residual compartments, respectively. Violations of the two-compartment assumptions, which pertain to the density and hydration of the entire FFM compartment, result in significant errors. In conclusion, the present findings indicated that hydrodensitometric and DXA FFM estimates generated indexed RMR data that were significantly lower than those for the more accurate FFM calculated using the four-compartment body composition model. While the underestimations associated with hydrodensitometry were consistent across the %BF range, the DXA values displayed a greater bias in the lean subjects. The hydrometry-and three-compartment-associated indexed RMR data were not significantly different from the four-compartment-derived values, but large intraindividual differences existed in the hydrometry data. Given that a small change in RMR independent of FFM would be regarded as physiologically significant, accurate body composition data are crucial for ascertaining if the RMR is truly sensitive to exercise status. It would therefore appear European Journal of Clinical Nutrition


1140 prudent to utilise FFM estimates derived using three- or fourcompartment models. However, further work is required to establish if the biases identified in our cross-sectional data also apply to female and elderly male subjects and whether they would occur longitudinally following a training/ detraining intervention.

References Baumgartner RN, Heymsfield SB, Lichtman SL, Wang J & Pierson J (1991): Body composition in elderly people: effect of criterion estimates on predictive equations. Am. J. Clin. Nutr. 53, 1345– 1353. Berke EM, Gardner AW, Goran MI & Poehlman ET (1992): Resting metabolic rate and the influence of the pretesting environment. Am. J. Clin. Nutr. 55, 626–629. Broeder CE, Burrhus KA, Svanevik LS & Wilmore JH (1992): The effects of either high-intensity resistance or endurance training on resting metabolic rate. Am. J. Clin. Nutr. 55, 802–810. Brozek J, Grande F, Anderson JT & Keys A (1963): Densitometric analysis of body composition: revision of some quantitative assumptions. Ann. NY Acad. Sci. 110, 113–140. Clasey JL, Kanaley JA, Wideman L, Heymsfield SB, Teates CD, Gutgesell ME, Thorner MO, Hartman ML & Weltman A (1999): Validity of methods of body composition assessment in young and older men and women. J. Appl. Physiol. 86, 1728–1738. Cote KD & Adams WC (1993): Effect of bone density on body composition estimates in young adult black and white women. Med. Sci. Sports Exerc. 25, 290–296. Elia M (1992), Organ and tissue contribution to metabolic rate. In Energy metabolism: Tissue Determinants and Cellular Corollaries. eds JM Kinney & HM Tucker pp 63–76. New York: Raven Press Ltd.. Elia M & Livesey G (1992): Energy expenditure and fuel selection in biological systems: The theory and practice of calculations based on indirect calorimetry and tracer methods, In Metabolic Control of Eating, Energy Expenditure and the Bioenergetics of Obesity. World Review of Nutrition and Dietetics. ed. AP Simopoulos, pp 68–131. Basel: Karger. Frey-Hewitt B, Vranizan KM, Dreon DM & Wood PD (1990): The effect of weight loss by dieting or exercise on resting metabolic rate in overweight men. Int. J. Obesity Relat. Metab. Disord. 14, 327–334. Friedl KE, Deluca JP, Marchitelli LJ & Vogel JA (1992): Reliability of body fat estimations from a four-compartment model by using density, body water, and bone mineral measurements. Am. J. Clin. Nutr. 55, 764–770. Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR & Sakamoto Y (2000): Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am. J. Clin. Nutr. 72, 694–701. Goran MI & Poehlman ET (1992): Endurance training does not enhance total energy expenditure in healthy elderly persons. Am. J. Physiol. 263, E950–E957. Goran MI, Toth MJ & Poehlman ET (1998): Assessment of researchbased body composition techniques in healthy elderly men and women using the 4-compartment model as a criterion method. Int. J. Obesity Relat. Metab. Disord. 22, 135–142. Kohrt WM (1995): Body composition by DXA: tried and true? Med. Sci. Sports Exerc. 27, 1349–1353. LaForgia J, Withers RT, Shipp NJ & Gore CJ (1997): Comparison of energy expenditure elevations after submaximal and supramaximal running. J. Appl. Physiol. 82, 661–666. LaForgia J, Withers RT, Williams AD, Murch BJ, Chatterton BE, Schultz CG & Leaney F (1999): Effect of 3 weeks of detraining on the resting metabolic rate and body composition of trained males. Eur. J. Clin. Nutr. 53, 126–133.


Mendez J, Keys A, Anderson JT & Grande F (1960): Density of fat and bone mineral of the mammalian body. Metabolism 9, 472–477. Modlesky CM, Cureton KJ, Lewis RD, Prior BM, Sloniger MA & Rowe DA (1996): Density of the fat-free mass and estimates of body composition in male weight trainers. J. Appl. Physiol. 80, 2085–2096. Mole PA (1990): Impact of energy intake and exercise on resting metabolic rate. Sports Med. 10, 72–87. Penn I-W, Wang Z-M, Buhl KM, Alison DB, Burastero SE & Heymsfield SB (1994): Body composition and two-compartment model assumptions in male long distance runners. Med. Sci. Sports Exerc. 26, 392–397. Poehlman ET & Danforth E (1991): Endurance training increases metabolic rate and norepinephrine appearance rate in older individuals. Am. J. Physiol. 261, E233–E239. Poehlman ET, Gardner AW & Goran MI (1992): Influence of endurance training on energy intake, norepinephrine kinetics, and metabolic rate in older individuals. Metabolism 41, 941–948. Poehlman ET, Melby CL & Badylak SF (1988): Resting metabolic rate and postprandial thermogenesis in highly trained and untrained males. Am. J. Clin. Nutr. 47, 793–798. Pratley R, Nicklas B, Rubin M, Miller J, Smith A, Smith M, Hurley B & Goldberg A (1994): Strength training increases resting metabolic rate and norepinephrine levels in healthy 50- to 65-yr-old men. J. Appl. Physiol. 76, 133–137. Prior BM, Cureton KJ, Modlesky CM, Evans EM, Sloniger MA, Saunders M & Lewis RD (1997): In vivo validation of whole body composition estimates from dual-energy X-ray absorptiometry. J. Appl. Physiol. 83, 623–630. Ravussin E & Bogardus C (1989): Relationship of genetics, age, and physical fitness to daily energy expenditure and fuel utilization. Am. J. Clin. Nutr. 49, 968–975. Ryan AS, Pratley RE, Elahi D & Goldberg AP (1995): Resistive training increases fat-free mass and maintains RMR despite weight loss in postmenopausal women. J. Appl. Physiol. 79, 818–823. Siconolfi SF, Gretebeck RJ & Wong WW (1995): Assessing total body protein, mineral, and bone mineral content from total body water and body density. J. Appl. Physiol 79, 1837–1843. Siri WE ed. (1961): Body composition from fluid spaces and density: analysis of methods, In Techniques for Measuring Body Composition. ed. JBA Henschel, pp 223–224. Washington, DC: National Academy of Sciences National Research Council. Smith DA, Dollman J, Withers RT, Brinkman M, Keeves JP & Clark DG (1997): Relationship between maximum aerobic power and resting metabolic rate in young adult women. J. Appl. Physiol. 82, 156–163. Smith DA, Withers RT, Brinkman M, Tucker RC, Chatterton BE, Schultz CG & Clark DG (1999): Resting metabolic rate, body composition and aerobic fitness comparisons between active and sedentary 54–71 year old males. Eur. J. Clin. Nutr. 53, 434–440. van der Ploeg GE, Brooks AG, Withers RT, Dollman J, Leaney F & Chatterton BE (2001a): Body composition changes in female bodybuilders during preparation for competition. Eur. J. Clin. Nutr. 55, 268–277. van der Ploeg GE, Gunn SM, Withers RT, Modra AC & Crockett AJ (2000): Comparison of two hydrodensitometric methods for estimating percent body fat. J. Appl. Physiol. 88, 1175–1180. van der Ploeg GE, Gunn SM, Withers RT, Modra AC, Keeves JP & Chatterton BE (2001b): Predicting the resting metabolic rate of young Australian males. Eur. J. Clin. Nutr. 55, 145–152. van der Ploeg GE & Withers RT (2002): Predicting the resting metabolic rate of 30–60 year-old Australian males. Eur. J. Clin. Nutr. 56, 701–708. van der Ploeg GE, Withers RT & Laforgia J (2003): Percent body fat via DEXA: comparison with a four-compartment model. J. Appl. Physiol 94, 499–506. Visser M, Gallagher D, Deurenberg P, Wang J, Pierson RN & Heymsfield SB (1997): Density of the fat-free body mass: relationship with race, age, and level of body fatness. Am. J. Physiol. 272, E781–E787.


1141 Westerterp KR, Meijer GAL, Janssen EME, Saris WHM & Hoor FT (1992): Long-term effect of physical activity on energy balance and body composition. Br. J. Nutr. 68, 21–30. Withers RT, LaForgia J & Heymsfield SB (1999): Critical appraisal of the estimation of body composition via two-, three-, and fourcompartment models. Am. J. Hum. Biol. 11, 175–185. Withers RT, LaForgia J, Pillans RK, Shipp NJ, Chatteron BE, Schultz CG & Leaney F (1998a): Comparisons of two-, three-, and four-

compartment models of body composition analysis in men and women. J. Appl. Physiol. 85, 238–245. Withers RT, Noell CJ, Whittingham NO, Chatterton BE, Schultz CG & Keeves JP (1997): Body composition changes in elite male bodybuilders during preparation for competion. Aust. J. Sci. Med. Sport 29, 11–16. Withers RT, Smith DA, Tucker RC, Brinkman M & Clark DG (1998b): Energy metabolism in sedentary and active 49- to 70-yr-old women. J. Appl. Physiol. 84, 1333–1340.
