Endocrinology and Metabolism

Body-size dependence of resting energy expenditure can be attributed to nonenergetic homogeneity of fat-free mass

Steven B. Heymsfield, Dympna Gallagher, Donald P. Kotler, Zimian Wang, David B. Allison, Stanley Heshka


An enduring enigma is why the ratio of resting energy expenditure (REE) to metabolically active tissue mass, expressed as the REE/fat-free mass (FFM) ratio, is greater in magnitude in subjects with a small FFM than it is in subjects with a large FFM. This study tested the hypothesis that a higher REE/FFM ratio in subjects with a small body mass and FFM can be explained by a larger proportion of FFM as high-metabolic-rate tissues compared with that observed in heavier subjects. REE was measured by indirect calorimetry, FFM by dual-energy X-ray absorptiometry (DEXA), and tissue/organ contributions to FFM by whole body magnetic resonance imaging (MRI) in healthy adults. Four tissue heat-producing contributions to FFM were evaluated, low-metabolic-rate fat-free adipose tissue (18.8 kJ/kg), skeletal muscle (54.4 kJ/kg), and bone (9.6 kJ/kg); and high-metabolic-rate residual mass (225.9 kJ/kg). Initial evaluations in 130 men and 159 women provided strong support for two key, developed models, one linking DEXA FFM with MRI FFM estimates and the other linking REE predicted from the four MRI-derived components with measured REE. There was an inverse association observed between measured REE/FFM and FFM (r 2 = 0.17, P < 0.001). Allometric models revealed a similar pattern of tissue change relative to body mass across males and females with greater proportional increases in fat-free adipose tissue and skeletal muscle than in FFM and a smaller proportional increase in residual mass than in FFM. When examined as a function of FFM, positive slopes were observed for skeletal muscle/FFM and pooled low-metabolic-rate components, and a negative slope for residual mass. Our linked REE-body composition models and associations strongly support the hypothesis that FFM varies systematically in the proportion of thermogenic components as a function of body mass and FFM. These observations have important implications for the interpretation of between-individual differences in REE expressed relative to metabolically active tissue mass.

  • body composition
  • metabolic rate
  • phenotyping

an intense interestof early workers in the field of energy metabolism (7,28), one maintained to the present time (1, 6, 9,24), is establishing the factors that account for between-individual differences in resting heat production. Exploring these multiple factors required adjusting for observed differences in body size, and typically an individual's resting energy expenditure (REE) was normalized for body mass and surface area (7, 20, 21,28). An important observation, however, was that the ratio of REE to body mass or related body surface area was not constant but rather decreased with greater human body weight (7).

Because the proportion of body mass as “energetically inert” fat increases with greater body mass in adults, later investigators advanced body composition compartments such as fat-free mass and body cell mass as improved measures of “metabolically active” tissue (1, 6, 9, 19). However, an enduring enigma is why REE is still not observed as constant across adult humans when expressed as a ratio to fat-free mass (or body cell mass) but rather decreases with greater fat-free mass (27, 31, 32). Subjects with a small fat-free mass have a greater REE-to-fat-free mass ratio than do subjects with a large fat-free mass, suggesting a body size difference in relative energy expenditure and requirements. The recognition of the “fat-free mass dependence” of the REE/fat-free mass ratio extends back more than a decade, and investigators now typically use alternative statistical methods (e.g., regression analysis) to adjust REE for fat-free mass when exploring between-individual thermogenic differences (9, 17, 27). However, the question remains: why do small subjects have a greater REE/fat-free mass ratio than do large subjects?

Animal studies suggest that with greater mammal size, the proportion of both body mass and fat-free mass as high-metabolic-rate organs and tissues (e.g., brain) decreases (3, 5). In contrast, with greater mammal size, the proportion of both body mass and fat-free mass as low-metabolic-rate tissues (e.g., bone, adipose tissue, etc.) increases (3, 5). However, the means of exploring on a large scale in humans the interrelations between body weight, heat-producing organs and tissues, and fat-free mass in vivo was lacking until the introduction of magnetic resonance imaging (MRI) in the mid-1980s (11). With MRI, investigators can quantify the volumes of all major heat-producing tissues and organs in healthy subjects without exposure to ionizing radiation (13).

A higher REE/fat-free mass ratio and thus a relatively higher metabolic rate in low-body-weight human subjects could be explained by a larger proportion of fat-free mass as high-metabolic-rate tissues compared with heavy subjects with a greater fat-free mass (31, 32). The aim of the present study was to test this hypothesis in a large cohort of healthy adults with the use of advanced imaging methods.


Experimental design.

Our strategy was first to quantify with whole body MRI the volumes of major high- and low-metabolic-rate compartments in healthy adult men and women. In a subsequent analysis phase, we established whether these collective measured components could account for measured REE with the use of previously reported tissue-specific heat production rates (7, 8). We also sought to confirm the close association between the molecular level component fat-free mass and the corresponding tissue/organ level component (measured by MRI) adipose tissue-free mass, as defined by the formula: fat-free mass = adipose tissue-free mass + 0.15 × adipose tissue mass. This model assumes that 85% of adipose tissue is fat (16) and 15% of adipose tissue is the remaining calculated fat-free component. The ratio of fat to adipose tissue is variable (16), but the impact of observed differences in this model have only a small influence on the analyses that follow. Hence, fat-free mass can be calculated as: fat-free mass = adipose tissue-free mass + fat-free adipose tissue. This confirmatory procedure was required to subsequently consider fat-free mass in relation to REE and body composition compartments by using MRI-derived adipose tissue-free mass and adipose tissue mass.

After this initial evaluation phase, we tested the hypothesis that the proportion of fat-free mass as high- and low-metabolic-rate tissues and organs varies as a function of fat-free mass. Our experiment thus critically evaluated the long-standing but often challenged assumption that fat-free mass is a metabolically homogeneous compartment.

Four major tissue/organ-level compartments were evaluated with MRI and dual-energy X-ray absorptiometry (DEXA): adipose tissue, skeletal muscle, bone, and residual mass. Residual mass, the difference between body mass and other measured components (i.e., adipose tissue, skeletal muscle, and bone), includes all of the high-metabolic-rate tissues and organs such as heart, brain, liver, kidneys, spleen, and gastrointestinal tract. In this study, we consider residual mass a high-metabolic-rate compartment (225.9 kJ/kg), whereas low-metabolic-rate compartments include adipose tissue (18.8 kJ/kg), skeletal muscle (54.4 kJ/kg), and bone (9.6 kJ/kg). The four-compartment REE values are based on the tissue/organ-specific metabolic rates reported earlier by Elia (7, 8), Holliday et al. (18), and Grande (14, 15) along with REE and body composition data from Reference Man (30). These specific metabolic rate values reflect literature-derived coefficients based upon limited human and animal studies. The specific metabolic rate values do not consider the metabolic effects of over- and underfeeding, various hormonal/genetic factors that might alter tissue energy flux, or purported interactive metabolic effects of tissues and organs. Earlier studies from our laboratory support the use of these tissue/organ-specific metabolic rates in healthy, weight-stable, young adults (13).

Fat-free mass was calculated as the difference between body mass and fat mass as measured by DEXA and related adipose tissue-free mass as the difference between body mass and adipose tissue mass as measured by MRI. As noted, fat-free mass was also calculated from adipose tissue and adipose tissue-free mass. The associations between the molecular level fat and fat-free mass components and the tissue/organ calculated fat-free mass component are summarized in Fig.1.

Fig. 1.

Associations between molecular and tissue/organ-level components evaluated in the present study. AT, adipose tissue; FFAT, fat-free AT component; ATFM, MRI-measured adipose tissue-free mass; FFM, fat-free mass; RM, residual mass; SM, skeletal muscle.


Subjects were healthy adults with body mass index (BMI) <40 kg/m2. Each subject, recruited through newspaper advertisements and from a local university, completed a medical history and physical examination to ensure his/her good health. Subjects were excluded from the study if they were age <18 yr, were involved in a structured physical activity program (2), had medical conditions or medication use known to affect body composition, or reported recent weight loss or weight gain (>10% of body weight within past year). The study was approved by the Institutional Review Board of St. Luke's-Roosevelt Hospital, and each subject signed an informed consent before participation.

REE and body composition studies were carried out either on the same day or within 1 day of each other, and subjects were asked to fast overnight before each evaluation day.

Body composition.

Body weight and height were measured using a digital scale and a wall-mounted stadiometer, respectively.

MRI of the whole body was carried out as previously reported by Gallagher et al. (13). Whole body MRI scans were prepared using a 1.5 Tesla scanner (General Electric, 6X Horizon, Milwaukee, WI). The MRI data were obtained using a T1-weighted, spin-echo sequence with 210-ms repetition time and a 17-ms echo time. Subjects rested quietly in the magnet bore in a prone position with their arms extended overhead. With the use of the intervertebral space between the fourth and fifth lumbar vertebrae as the origin point, transverse images with 10-mm slice thickness were obtained every 40 mm from hand to foot, resulting in a total of ∼40 images for each subject. A 26-s breath hold was required during abdominal slice imaging.

All MRI scans were segmented into the four components by highly trained analysts using image analysis software (Tomovision, Montreal, QC, Canada). A multiple-step procedure was used to identify specific tissue areas (cm2) for a given MRI image. A threshold was selected for adipose tissue and lean tissue on the basis of the image gray-level histogram, or a filter was used to distinguish between different gray-level regions on the images, and lines were drawn around the selected regions by use of a Watershed algorithm. Thereafter, the analyst labeled the tissues of interest by assigning them different color codes. Images were then reviewed by an interactive slice editor program that allowed for verification and, where necessary, correction of the segmented component results. The original gray level was superimposed on the binary-segmented image by means of a transparency mode to facilitate the corrections. The respective tissue areas in each image were automatically calculated by summing the specific tissue pixels and then multiplying by the individual pixel surface area. The volume per slice (in cm3) of each selected tissue was calculated by multiplying tissue area (cm2) by slice thickness. The volume of each tissue for the space between two consecutive slices was calculated via a mathematical algorithm reported by Ross (26). Volume estimates were converted to mass units (kg) by taking the product of volume (liters) and reported tissue density (i.e., adipose tissue = 0.92 kg/l; skeletal muscle = 1.04 kg/l) (30).

DEXA was used to measure fat, fat-free mass, and bone mineral mass. Subjects were scanned using a whole body DPX (Lunar Radiation, Madison, WI; version 3.6 software) with a cerium filtered X-ray source. The DEXA system software first divides pixels into bone mineral mass and soft tissue compartments. Soft tissue is then further separated by system software into lean soft tissue and fat (22, 23, 25). Bone mass was calculated from bone mineral mass with theassumption of a stable proportion of bone mass as mineral (bone mineral mass/bone = 0.54) (30).


REE was measured in postabsorptive subjects by means of the Columbia Respiratory Chamber-Indirect Calorimeter (13). After entering the metabolic chamber, the supine subject rested quietly in the thermoneutral environment, and a transparent plastic ventilated hood was placed over the head for 40–60 min. Rates of oxygen consumption and carbon dioxide production were analyzed using magnetopneumatic oxygen (Magnos 4G) and carbon dioxide (Magnos 3G) analyzers (Hartmann & Braun, Frankfurt, Germany), respectively. Gas exchange results were evaluated during the stable measurement phase and converted to REE in megaJoules per day using the Weir equation (33).

REE (kJ/day) was also calculated from measured adipose tissue, skeletal muscle, bone, and residual mass (kg) on the basis of published tissue/organ-specific metabolic rates as follows (7, 8, 18,30): REE = 18.8 × adipose tissue + 54.4 × skeletal muscle + 9.6 × bone + 225.9 × residual mass, where adipose tissue and skeletal muscle mass are from MRI analysis, bone is DEXA bone mineral mass/0.54, and residual mass is calculated as body weight − (skeletal muscle + adipose tissue + bone).

Statistical methods.

Descriptive statistics are reported as means ± SD. Between-gender differences were evaluated via t-tests at the two-tailed 0.05 α-level.

Regression analysis was used to examine the relationships between measured and calculated REE and between DEXA-measured fat-free mass and fat-free mass from MRI-derived adipose tissue-free mass and adipose tissue mass. Potential measurement bias was explored as suggested by Bland and Altman (4). Mean differences between measured and calculated REE and between measured and calculated fat-free mass were evaluated using paired t-tests.

Associations between relative change in the measured components (e.g., skeletal muscle) and change in body mass, sometimes referred to as differential growth, were examined in our cross-sectional sample by use of the standard allometric model: component = a × body weightb, where b is the scaling exponent (5, 29). Men and women were analyzed separately, because large between-gender differences in adiposity are recognized and the relationship between body fat (i.e., adipose tissue) and body weight differs significantly between males and females (10,12).

The hypothesis was tested in the last analysis phase by examining the associations between high- and low-metabolic-rate compartments and fat-free mass with the use of linear regression methods. Specifically, we sought to establish whether the fraction of fat-free mass as high- and low-metabolic-rate components varies significantly as a function of fat-free mass. Men and women were pooled in these analyses as is the practice for exploring REE/fat-free mass associations (27).



As shown in Table 1, there were 130 men and 159 women in four ethnic groups: African-American (n = 82), Asian (n = 42), Caucasian (n = 122), and Hispanic (n = 43). Men had greater body weight, skeletal muscle, bone, and residual mass than women (all P < 0.01). Women were shorter and had a larger adipose tissue mass (both P < 0.01) than men.

View this table:
Table 1.

Subject characteristics

Model validations.

The group mean predicted REE was 6.57 ± 1.35 MJ/day and was not significantly different from the measured REE of 6.49 ± 1.36 MJ/day. The group mean predicted REEs for females and males were 5.64 ± 0.77 and 7.70 ± 1.28 MJ/day, respectively, whereas their measured counterparts were 5.72 ± 0.82 and 7.42 ± 1.49 MJ/day, respectively [both P = not significant (NS)]. Predicted and measured REE were highly correlated (r 2 = 0.56, P < 0.001; Fig. 2), and there was no significant bias detected by Bland-Altman analysis (4). However, multiple regression analysis revealed a small but statistically significant (P = 0.004) age contribution to the measured vs. predicted REE relationship (r 2increase from 0.56 to 0.58). The age effect can be seen when the residual REE (i.e., the difference between measured and predicted REE) is plotted against age, as shown in Fig.3. The data, fit with a polynomial model, suggests a close association between measured and predicted REE up to ∼50 yr of age and an increasing small discrepancy thereafter reaching ∼0.42 MJ/day by age 80 yr. These observations suggest that, after the four tissue compartments are accounted for, older subjects have a lower REE than their younger counterparts. However, the small age effect does not measurably influence the results that follow, and we therefore present findings for the composite group of 289 subjects.

Fig. 2.

Predicted vs. measured resting energy expenditure (REE) in 289 study subjects (P < 0.001). Units are MJ/day.

Fig. 3.

Residual REE vs. age in 289 study subjects (y = −0.0001x 2 + 0.0006x + 0.110; r 2 = 0.027, P = 0.005). REE is expressed as MJ/day.

Measured and calculated fat-free mass (52.2 ± 11.9 vs. 53.5 ± 11.9 kg) were not significantly different, and the two estimates were highly correlated with each other (r 2= 0.97, P < 0.001; Fig.4). No bias was detected by Bland-Altman analysis, and no age or sex effects were observed with multiple regression analysis. All data analyses that follow involving fat-free mass are based on MRI-derived fat-free mass, as this component is synchronous with the MRI-based calculated REE estimates.

Fig. 4.

Calculated vs. measured FFM (kg) in 289 study subjects (P < 0.001). Dashed line, the line of identity; white line, the regression line.

REE vs. fat-free mass.

Measured REE is plotted against fat-free mass in Fig.5 A. There was a strong association, as expected, between REE and fat-free mass (r 2 = 0.64, P < 0.001). The REE/fat-free mass ratio is plotted against fat-free mass in Fig.5 B, and the association was negative in slope and statistically significant (r 2 = 0.17,P < 0.001). The predicted REE/fat-free mass for a subject with a 50-kg fat-free mass is 0.123 MJ · kg−1 · day−1 and for a subject with an 80-kg fat-free mass is 105 MJ · kg−1 · day−1. The results of the present study thus demonstrate the previously reported lowering of REE relative to metabolically active tissue, as defined by fat-free mass, in subjects with greater metabolically active tissue mass (31, 32).

Fig. 5.

A: measured REE (MJ/day) vs. FFM (kg) (P < 0.001). B: REE/FFM (MJ · kg−1 · day−1) vs. FFM (P < 0.001) in 289 study subjects.

The fractional contributions of adipose tissue, skeletal muscle, bone, and residual mass to body mass and model-predicted REE are shown for men and women in Fig. 6, A and B, respectively. Collectively, skeletal muscle, bone, and adipose tissue contributed to 69.8 and 73.4% of body mass in men and women, whereas respective contributions to REE were 30.9 and 31.7%, respectively. The small residual mass, 30.1% of body mass in men and 26.6% in women, contributed to 69.1 and 68.3% of predicted REE, respectively.

Fig. 6.

Four tissue/organ components in females and males expressed as a fraction of body mass (A) and as their respective fractional contributions to REE (B).

Body composition: relationship to body mass.

All of the measured components and REE increased as a function of body mass but did so with different scaling exponents (Table2). The pattern of component increase relative to body weight (i.e., the scaling exponent b of body weightb) was identical in males and females with the following sequence: adipose tissue > skeletal muscle > REE = fat-free mass > residual mass.

View this table:
Table 2.

Allometric relationships between measured components and body mass (in kg) *

In other words, with greater body mass, the relative increase in adipose tissue and skeletal muscle exceeded that of REE; relative increases in REE and fat-free mass were similar, and the relative increase in residual mass was smaller than that of all other components and REE. The one exception was the small bone component, which scaled differently compared with the other components in males and females. Hence, with increasing body mass, both males and females responded similarly with relatively greater increases in low-metabolic-rate tissues (i.e., adipose tissue and skeletal muscle) compared with the high-metabolic-rate residual mass.

Fractional contributions to fat-free mass.

With increasing fat-free mass there were corresponding increases in skeletal muscle, bone, residual mass, and fat-free adipose tissue for pooled subjects, as shown in Fig.7 A. When each component was expressed as a fraction of fat-free mass, the four components responded differently, as shown in Fig. 7 B. The proportion of fat-free mass as skeletal muscle increased with greater fat-free mass (slope = 0.0021) whereas there were corresponding reductions in the proportions of fat-free mass as residual mass (slope = −0.001), bone (slope = −0.0002), and fat-free adipose tissue (slope = −0.0009). Thus, relative to fat-free mass, greater fat-free mass was associated with a larger proportion of low-metabolic-rate tissues, skeletal muscle, and bone and a smaller proportion of high metabolic residual mass. This effect can be seen in Fig. 8, in which the fraction of fat-free mass as low-metabolic-rate tissues (i.e., skeletal muscle + bone + fat-free adipose tissue) is plotted against fat-free mass. With increasing fat-free mass, there was a corresponding increase in low-metabolic-rate components.

Fig. 7.

Four tissue/organ components in 289 study subjects plotted as absolute mass against fat-free mass (A) (SM = 0.59 × FFM − 6.33, r 2 = 0.93; RM = 0.32 × FFM + 2.67, r 2 = 0.78; Bone = 0.08 × FFM + 0.70,r 2 = 0.78; FFAT = 0.006 × FFM + 3.00, r 2 = 0.0017) and as a fraction of FFM (B) (SM/FFM = 0.0021 × FFM + 0.35, r 2 = 0.35; RM/FFM = −0.001 × FFM + 0.43, r 2 = 0.081; Bone/FFM = −0.0002 × FFM + 0.11,r 2 = 0.078; FFAT/FFM = −0.0009x + 0.11, r 2 = 0.11). For all correlations, P < 0.001, except FFAT vs. FFM, P = not significant.

Fig. 8.

Ratio of low-metabolic-rate (LMR) tissues and organs to FFM (LMR/FFM) vs. FFM in 289 study subjects. (P < 0.01).


The present study strongly supports the hypothesis that fat-free mass is not an energetically homogeneous compartment but varies systematically in heat-producing components as a function of body mass and fat-free mass. Our allometric analyses in a cross-sectional sample indicate that, with increasing body mass, there is a corresponding relatively large increase in the adipose tissue compartment and, to a lesser extent, the skeletal muscle compartment. In contrast, our empirical cross-sectional models suggest that increasing body mass is accompanied by a minimal expansion of the high-metabolic-rate residual mass component. The result is that composite fat-free mass experiences a change in composition with expansion of body mass as a whole: skeletal muscle and fat-free adipose tissue increase to a relatively greater extent than residual mass. The net result is that, with greater body mass, there is an increase in the proportion of fat-free mass as low-metabolic-rate tissues and a decrease in the proportion as high-metabolic-rate tissues. The anticipated effect, one that is actually observed in vivo, is a lowering of REE relative to fat-free mass with increasing body size and fat-free mass.

Our analysis is supported by close agreement between measured and calculated values for both REE and fat-free mass. A reasonable qualitative group estimate of REE was obtained from only four tissue/organ components: adipose tissue, skeletal muscle, bone, and residual mass. This four-component REE model may be useful for roughly estimating how relative changes in one component or another influences overall resting heat production. In an earlier report, Gallagher et al. (13) confirmed the validity of a more complex tissue/organ model in young adults. The reported model included brain, liver, kidneys, and heart in addition to adipose tissue, skeletal muscle, and bone. The residual mass was thus smaller in magnitude than the residual mass as calculated in the present study. In accordance with the present study, Gallagher et al., in a follow-up investigation, applied their REE model to a group of older subjects and also found a lower-than-expected resting heat production (12). These observations extend earlier reports suggesting that, after controlling for fat-free mass, the elderly have a reduced REE of unknown mechanism. Our estimate of fat-free mass from adipose tissue mass and adipose tissue-free mass was nearly identical to the fat-free mass measured by DEXA. This observation demonstrates the strong and quantifiable links that exist between body composition levels, in this case the molecular and tissue/organ levels.

An immediate implication of the present study is that REE “adjusted for fat-free mass” should be interpreted with caution. A high or low REE after adjusting for fat-free mass is usually interpreted as meaning a high- or low-energy flux rate through metabolically active tissues. Our observations suggest an equally likely possibility: the existence of a relatively large or small proportion of fat-free mass as high-metabolic-rate tissues and organs. This finding leads to the question of what the ideal method is of adjusting REE for between-individual differences in body size. Unfortunately, all potential combined body composition compartments (e.g., adipose tissue-free mass) share with fat-free mass the same problem as representing a metabolically heterogeneous compartment. Ideally, future technological advances will allow direct measurement of a tissue or organ's resting heat production while corresponding tissue/organ mass is estimated using methods such as multi-slice MRI.

Study limitations.

An important assumption in the present study is that tissue/organ-specific metabolic rates are known and constant in healthy young adults. These coefficients reflect compiled summaries (7,8, 14, 15, 18, 30) of small-scale animal and human experiments with limited validation, as in this and earlier studies (12,13), and should be interpreted and applied with caution. Many factors may be responsible for causing individual variation in tissue/organ-specific metabolic rates, and some of these factors can now be studied in vivo with the use of newly developed metabolic study methods.

In conclusion, the present study results, taken collectively, suggest that tissue/organ components associated with fat-free mass vary relative to fat-free mass as a function of body size and fat-free mass. The observed pattern of changes suggests that the greater magnitude REE/fat-free mass ratio observed in low-body mass subjects can be attributed to the high proportion of fat-free mass as residual mass and low proportion as fat-free adipose tissue, skeletal muscle, and bone. These observations have important implications for the evaluation and interpretation of between-individual differences in resting heat production and energy requirements.


This study was supported by National Institutes of Health Grant RR-00645 and National Institute of Diabetes and Digestive and Kidney Diseases Grants DK-42618, DK-51716, and DK-26687.


  • Address for reprint requests and other correspondence: S. B. Heymsfield, Obesity Research Center, 1090 Amsterdam Ave., 14th Floor, New York, NY 10025 (E-mail: SBH2{at}Columbia.edu).

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract