draft - university of toronto t-space · participants were required to meet at least one of the...

Draft

Resting metabolic rate in muscular physique athletes: validity of existing methods and development of new

prediction equations

Journal: Applied Physiology, Nutrition, and Metabolism

Manuscript ID apnm-2018-0412.R1

Manuscript Type: Article

Date Submitted by the Author: 22-Aug-2018

Complete List of Authors: Tinsley, Grant; Texas Tech University, Department of Kinesiology & Sport ManagementGraybeal, Austin; Texas Tech University, Department of Kinesiology & Sport ManagementMoore, M.; Texas Tech University, Department of Kinesiology & Sport Management

Keyword: resting energy expenditure, bodybuilders, metabolism, Harris-Benedict, Mifflin, body composition, indirect calorimetry, Cunningham, ten Haaf

Is the invited manuscript for consideration in a Special

Issue? :Not applicable (regular submission)

https://mc06.manuscriptcentral.com/apnm-pubs

Applied Physiology, Nutrition, and Metabolism

Draft

1

Resting metabolic rate in muscular physique athletes: validity of existing methods and development of new prediction equations

Grant M. Tinsley*, Austin J. Graybeal, M. Lane Moore

Affiliation for all authors: Department of Kinesiology & Sport Management, Texas Tech University, Lubbock, TX, USA

*Corresponding author: Grant M. Tinsley, Department of Kinesiology & Sport Management, Texas Tech University, Lubbock, TX, 79424, USA. [email protected] (806) 834-5895.

Page 1 of 37



mailto:[email protected]

Draft

2

Abstract

Estimation of resting metabolic rate (RMR) is an important step for prescribing an individual’s

energy intake. The purpose of this study was to evaluate the validity of portable indirect

calorimeters and RMR prediction equations in muscular physique athletes. Twenty-seven males

(n=17; BMI: 28.8±2.0 kg/m2; body fat: 12.5±2.7%) and females (n=10; BMI: 22.8±1.6 kg/m2;

body fat: 19.2±3.4%) were evaluated. The reference RMR value was obtained from the

ParvoMedics TrueOne® 2400 indirect calorimeter, and the Cosmed FitmateTM and Breezing®

Metabolism Tracker provided additional RMR estimates. Existing RMR prediction equations

based on body weight (BW) or dual-energy x-ray absorptiometry (DXA) fat-free mass (FFM)

were also evaluated. Errors in RMR estimates were assessed using validity statistics, including t-

tests with Bonferroni correction, linear regression, and calculation of the SEE, total error, and

95% limits of agreement. Additionally, new prediction equations based on BW (RMR [kcal/d] =

24.8*BW [kg] + 10) and FFM (RMR [kcal/d] = 25.9*FFM [kg] + 284) were developed using

stepwise linear regression and evaluated using leave-one-out cross-validation. Nearly all existing

BW- and FFM-based prediction equations, as well as the Breezing® Tracker, did not exhibit

acceptable validity and typically underestimated RMR. The ten Haaf (2014) and Cunningham

(1980) FFM-based equations may produce acceptable RMR estimates, although the Cosmed

FitmateTM and newly developed BW- and FFM-based equations may be most suitable for RMR

estimation in male and female physique athletes. Future research should provide additional

external cross-validation of the newly-developed equations in order to refine the ability to predict

RMR in physique athletes.

Keywords: resting energy expenditure, metabolism, bodybuilders, Harris-Benedict, Mifflin, body composition, Cunningham, ten Haaf, indirect calorimetry

Page 2 of 37



Draft

3

Introduction

Measurement or estimation of resting metabolic rate (RMR) is frequently the first step in

prescribing energy intake, both in the general population and athletes (Thomas et al. 2016). Even

in highly active individuals, RMR represents a substantial contribution to total daily energy

expenditure (TDEE). While laboratory methods, namely indirect calorimetry, are commonly

utilized for RMR measurement, most individuals rely on prediction equations to estimate RMR.

Prediction equations based on body weight (BW) have been utilized for over 100 years (Harris

and Benedict 1918), and numerous distinct equations are presently employed (Flack et al. 2016).

However, due to the known differences in the metabolic activity of fat mass and fat-free mass

(FFM), several other equations predict RMR based on FFM rather than BW (Hayes et al. 2002).

A detailed analysis of energy expenditure at the organ/tissue level of the body demonstrated a

linear relationship between FFM and RMR within the range of FFM typically observed in

humans, and equations typically possess a slope that ranges from 19.7 to 24.5 and a positive

intercept of approximately 200 to 700 kcal/day (Wang et al. 2000).

In athletes, some advocate the use of FFM-based prediction equations due to the relatively

greater proportion of FFM in these individuals (ten Haaf and Weijs 2014). Several investigations

have examined the validity of BW- or FFM-based equations in athletic populations, with some

leading to the development of new athlete-specific equations. These investigations have

examined a variety of athletic groups, including endurance athletes (Thompson and Manore

1996), a mixed group of athletes including waterpolo, judo and karate (De Lorenzo et al. 1999),

rowers and canoeists (Carlsohn et al. 2011), a variety of team sport athletes (i.e. football, track

and field, baseball, swimming and soccer) (Jagim et al. 2017) and a mixed group of individual

and team sport athletes (ten Haaf and Weijs 2014). While some of the aforementioned prediction

Page 3 of 37



Draft

4

equations were developed in narrowly defined groups of athletes, others included substantial

heterogeneity in an attempt to produce a generalizable equation. It is recognized that the

accuracy of RMR prediction equations may be population-specific, indicating that these

equations cannot be indiscriminately applied groups that are dissimilar to those in which they

were developed (da Rocha et al. 2005). Additionally, although generalizable equations are

convenient, they may mask actual differences between specific sub-populations of athletes, who

may vary in body composition and training practices that could impact RMR. For example, the

groups of athletes included in the aforementioned equations do not typically exhibit the degree of

muscularity observed in competitive physique athletes, whereas physique athletes may have

lower energy intake and TDEE than athletes in some traditional sports (Slater and Phillips 2011).

Proper prescription of energy intake to facilitate fat loss, while promoting the retention of

FFM and physical performance, is a major goal of physique athletes preparing for competition

(Helms et al. 2014). However, most of these athletes do not have access to traditional indirect

calorimeters for measurement of RMR. Currently, there are several portable indirect calorimeters

available, which may be a more accessible option to this athletic population. However, limited

information is available concerning the validity of these devices for RMR estimation in athletes,

and the price and complexity of the devices varies widely. Despite the existence of these portable

devices, the most common method of RMR estimation remains the use of prediction equations. It

is unclear if FFM-based equations, which necessitate the estimation of body composition, are

superior to BW-based equations in this population. Theoretically, FFM-based equations could be

advantageous, although this has not previously been examined.

To our knowledge, no previous investigations have examined the validity of portable indirect

calorimetry and BW- or FFM-based RMR prediction equations in physique athletes. Therefore,

Page 4 of 37



Draft

5

the purpose of this study was to evaluate the utility of several practical methods of RMR

estimation in male and female physique athletes and to produce preliminary RMR prediction

equations for this population.

Materials and Methods

At a single research visit, the RMR of physique athletes was measured via three indirect

calorimeters. Body weight and composition were also assessed to allow for estimation of RMR

using prediction equations. Twenty-seven physique athletes volunteered to participate in this

study as previously described (Graybeal et al. 2018). To be eligible for inclusion in this analysis,

participants were required to meet at least one of the following criteria: 1) have competed in a

bodybuilding or physique competition within the last year; 2) have plans to compete within the

next year; or 3) self-identify as a bodybuilder and exhibit a physique commensurate with

competitive physique athletes, as evaluated by study investigators. Additionally, prospective

participants were required to be between the ages of 18 and 50, generally healthy, and report the

completion of ≥ 3 sessions per week of resistance training, continuously for ≥ 3 years, prior to

screening. This study was approved by the Texas Tech University institutional review board, and

all participants signed the informed consent document prior to participation.

Participants reported to the laboratory in the morning after an overnight (≥ 8 hours)

abstention from food, fluid, supplement or medication ingestion, and exercise. Body weight and

height were assessed using a digital scale and stadiometer (Seca 769, Hamburg, Germany). FFM

was estimated via dual-energy x-ray absorptiometry (DXA). DXA scans were performed on a

calibrated GE Lunar Prodigy scanner with enCORE software (v. 16.2), and participant

positioning was conducted according to manufacturer recommendations. Due to the large body

Page 5 of 37



Draft

6

size of many participants, it was necessary to perform the reflection scanning technique in which

the unobserved portion of the body (i.e. the left arm) is estimated from the observed portion of

the body. This technique was performed in accordance with manufacturer recommendations and

has been reported to induce minimal error (Moco et al. 2018; Tinsley et al. 2018b). In order to

further reduce potential errors caused by this procedure, scans were conducted in duplicate and

averaged for analysis. DXA lean soft tissue and bone mineral content were summed to provide

an estimate of FFM.

Following body composition assessment, RMR was assessed by three indirect

calorimetry devices. The TrueOne® 2400 (ParvoMedics, Sandy, UT, USA) was designated as

the reference method, and additional methods were a portable research-grade device (FitmateTM,

Cosmed, Rome, Italy) and a portable consumer-grade device (Breezing® Metabolism Tracker,

Breezing, Tempe, AZ, USA). The TrueOne® 2400 was selected as the reference method due to

its demonstrated accuracy (Cooper et al. 2009; Kaviani et al. 2018). A recent study evaluated the

accuracy and reliability of 12 indirect calorimeters using methanol combustion (Kaviani et al.

2018). Of the 12 devices, two separate TrueOne® 2400 systems were ranked 1st and 2nd for CO2

recovery, 2nd and 5th for O2 recovery and 2nd and 4th for RER accuracy. Furthermore, both of the

evaluated TrueOne® 2400 units measured CO2 recovery, O2 recovery and RER within 2% of

theoretical values, unlike most other devices. The TrueOne® 2400 unit used in this study was

less than 2 years old at study commencement, and regular maintenance and calibrations were

performed according to manufacturer instructions throughout this time period. A new Cosmed

FitmateTM device was purchased for this study, and the oxygen sensor remained in the “optimal”

state throughout data collection. A new Breezing® Tracker device was purchased from the

manufacturer approximately 3 months prior to study commencement, and all testing was

Page 6 of 37



Draft

7

completed in less than 7 months after receipt of the device and its associated supplies (e.g. sensor

cartridges).

Pre-assessment standardization and testing were conducted according recommended

procedures (Compher et al. 2006). Briefly, the participant was rested and fasted overnight prior

to each assessment and was instructed to remain motionless, but awake, throughout testing. Each

participant was offered a blanket at each assessment in order to promote a comfortable body

temperature, and all testing took place in the same climate-controlled room with the lights

dimmed. Due to previous laboratory assessments conducted at the research visit, each participant

rested in the supine position for approximately 30 minutes prior to the commencement of the

first RMR assessment. The order of RMR assessments was randomly determined using the

random integer set generator available at random.org. For each device, manufacturer procedures

were followed. RMR via TrueOne® 2400 (RMRPARVO) and FitmateTM (RMRCOSMED) was

assessed in the supine position, while RMR via Breezing® (RMRBREEZING) was assessed in the

seated position per manufacturer instructions. Regardless of assessment order, each participant

moved from the supine to seated position for a period of approximately two minutes between

RMR assessments.

Prior to TrueOne® 2400 assessments, daily gas and flow calibrations were performed.

Prior to FitmateTM assessments, daily flow and oxygen sensor calibrations were performed. For

both the TrueOne® 2400 and FitmateTM assessments, the first five minutes of each test were

discarded, and the assessment continued until there was a period of 5 consecutive minutes with a

coefficient of variation (CV) for RMR of ≤ 10%. Using 1-minute averaging, the average CVs in

this study were 4.2 ± 1.5% and 4.6 ± 1.9% for RMRPARVO and RMRCOSMED, respectively. The

Breezing® device utilizes a sensor cartridge and flow meter to evaluate expired air as previously

Page 7 of 37



Draft

8

described (Xian et al. 2015). The device is synced with a phone or tablet, and each single-use

sensor has a QR code that is scanned by the associated phone or tablet to provide calibration

information for the sensor. During each assessment, the participant breathes through a disposable

mouthpiece for 1 to 2 minutes, until 6 L of air has been expired during the assessment. RMR is

then estimated from VO2 and VCO2 using the Weir equation (Xian et al. 2015) and reported in

kcal/d.

In addition to the indirect calorimetry assessments, RMR was predicted via five BW-

based equations, five FFM-based equations and one organ/tissue-based equation (Table 1). BW-

based equations (Harris and Benedict 1918; FAO 1985; Mifflin et al. 1990; De Lorenzo et al.

1999; ten Haaf and Weijs 2014) utilized BW obtained on a digital scale (Seca 769, Hamburg,

Germany), while FFM-based equations (Cunningham 1980; Owen et al. 1987; Mifflin et al.

1990; Cunningham 1991; ten Haaf and Weijs 2014) utilized DXA FFM, and the organ/tissue

model used various components of DXA output as previously described (Hayes et al. 2002).

Statistical Analysis

Potential differences in RMR between the reference method and alternative methods

were analyzed using dependent t-tests with a Bonferroni-adjusted alpha level due to multiple

comparisons (p ≤ 0.0033). The constant error (CE) was determined as the mean difference

between an alternate RMR assessment and the reference method (e.g., RMRALTERNATE –

RMRPARVO). Additionally, the Pearson product moment correlation coefficient (r), coefficient of

determination (R2), standard error of the estimate (SEE), and total error (TE) were calculated.

The TE, also known as the root mean square error (RMSE), was calculated as:

𝑇𝐸 = Σ(𝑅𝑀𝑅𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 ― 𝑅𝑀𝑅𝑃𝐴𝑅𝑉𝑂)2/𝑛

Page 8 of 37



Draft

9

The TE represents the average deviation of individual scores from the line of identity

between the reference method and each alternative method, whereas the SEE indicates the

deviation of individual data points around the line of best fit for the reference method and each

alternative method (Heyward and Wagner 2004). The following thresholds were used to describe

the r values: trivial (<0.1), small (0.1 to 0.29), moderate (0.30 to 0.49), large (0.50 to 0.69), very

large (0.70 to 0.89), and extremely large or “near perfect” (0.90 to 1.00) (Hopkins et al. 2009).

The effect size (ES) of the differences between methods was determined using Cohen’s d. The

magnitude of the ES was interpreted as: very small (<0.2), small (0.2 – 0.59), moderate (0.6 –

1.19), large (1.2 – 2.0), and very large (>2.0) (Hopkins, Marshall et al. 2009). The Bland-Altman

method (Bland and Altman 1986) was used to identify the 95% limits of agreement (LOA)

between the reference and alternative methods. The 95% LOA indicate the individual predictive

accuracy of a method based on a 95% confidence interval. Linear regression was utilized to

evaluate proportional bias between the reference method and alternative methods (i.e. varying

discrepancies between reference and alternative methods based on RMR values) as previously

described (Tinsley 2017). Additionally, stepwise linear regression was utilized to develop RMR

prediction equations from relevant variables (i.e. BW, FFM, age, sex and height). Due to the

relatively small sample size, leave-one-out cross-validation was utilized to evaluate the newly

developed equations (Ivanescu et al. 2016). This procedure involves sequentially removing each

participant’s data, developing linear regression equations using the remaining data, and

calculating the error produced when the regression equations are applied to the excluded data.

The TE (i.e. RMSE) of the leave-one-out analysis was calculated using the prediction errors (i.e.

CE) observed when regression equations were applied to excluded data. These leave-one-out TE

values were compared to the TE values of the regression equations developed in the entire

Page 9 of 37



Draft

10

sample (Lohman et al. 2000). Data were analyzed using IBM SPSS (v. 25) and Microsoft Excel

(v. 16.11).

Results

All participants self-identified as bodybuilders, with 48% reporting participation in a

physique contest in the past year. All participants reported practicing high-volume resistance

training for ≥ 3 years, with current training of 5.7 ± 0.9 days per week. The rates of self-reported

anabolic androgenic steroid (AAS) use were 26% (M: 35%, F: 10%) for current use and 41%

(M: 59%, F: 10%) for use in the previous 3 years. However, it is believed that under-reporting of

AAS usage may have occurred. The DXA fat-free mass index (FFMI) of male participants (24.2

± 1.3 kg/m2) was approximately 2 SD greater than reference values from the National Health and

Nutrition Examination Survey (NHANES), and the DXA FFMI of female participants (17.7 ±

0.9 kg/m2) was approximately 1 SD greater than NHANES reference values (Kelly et al. 2009).

Conversely, the DXA fat-mass index (FMI) of male and female participants (M: 3.6 ± 0.9 kg/m2;

F: 4.4 ± 1.0 kg/m2) was approximately 1 SD below NHANES reference values (Kelly et al.

2009). Participant characteristics are displayed in Table 2.

Validity of the evaluated RMR methods for males and females combined are presented in

Table 3, while individual results for males and females are presented in Tables 4 and 5,

respectively. RMRCOSMED was not significantly different from RMRPARVO in males, females or

males and females combined (+3.0 to 4.3%; trivial to small ES). Additionally, proportional bias

was not present, and the LOAs were narrow relative to other methods (Figure 1A). In contrast,

RMRBREEZING was 14.5% lower (moderate ES) in males and females combined when compared

to RMRPARVO. The level of disagreement was much larger in males (-22.0%; large ES) than in

Page 10 of 37



Draft

11

females (+4.3%; small ES). However, in both males and females, LOAs for RMRBREEZING were

wide relative to other methods (Figure 1B).

In general, BW-based equations underestimated RMR in males and females combined by

4.5 to 15.1% (small to moderate ES) (Figure 2; Table 3). In males, all BW-based equations

underestimated RMR by ≥ 10.1% (medium to large ES), with the exception of the ten Haaf

equation, which did not differ significantly from the reference method (-6.2%; small ES).

However, all BW-based equations demonstrated statistically significant negative proportional

bias with regression coefficients of ≥ -0.65. In females, RMR estimates from BW equations were

not statistically different from the reference method. However, three BW-based equations

underestimated RMR by ≥ 7.1% (moderate to large ES), while one overestimated RMR by 7.1%

(DeLorenzo; moderate ES) and one (ten Haaf) displayed no CE (trivial ES). Although not

statistically significant, BW-based equations demonstrated possible negative proportional bias,

with regression coefficients varying from -0.28 to -1.0.

In males and females combined, three FFM-based equations underestimated RMR by

10.2 to 17.4% (small to moderate ES), while two equations (Cunningham [1980] and ten Haaf)

did not differ significantly from RMRPARVO (-0.8 to -2.6%; trivial ES) and exhibited relatively

low TE (Figure 3). In males, three FFM-based equations underestimated RMR by 10.9 to 15.5%

(moderate ES), while two other equations (ten Haaf, Cunningham [1980]) exhibited

underestimations of RMR relative to RMRPARVO (-2.0 to 3.9%; trivial to small ES) that were not

statistically significant. Statistically significant negative proportional bias was seen for all FFM-

based equations in males (Table 4). In females, two FFM-based equations significantly

underestimated RMR by 11.7 to 16.8% (moderate to very large ES), while three equations

exhibited deviations (-8.6 to +2.2%; trivial to moderate ES) that were not statistically significant.

Page 11 of 37



Draft

12

Although not statistically significant, all FFM-based equations demonstrated possible negative

proportional bias, with regression coefficients ≥ -0.24 (Table 5). For males, females, and the

entire sample, RMR estimated by the organ/tissue equation (RMRHAYES; Figure 1C) did not

outperform several predictive equations based solely on FFM or BW (Tables 3 – 5).

New BW- and FFM-based RMR prediction equations were developed within the entire

sample (Figure 4). In the BW-based model (BWTINSLEY), BW was the only statistically

significant predictor of RMR (B: 24.78; p < 0.0001). Excluded variables included sex (B: 0.094;

p = 0.57), age (B: -0.035; p = 0.66) and height (B: -0.070; p = 0.68). Overall, BWTINSLEY had an r

of 0.921 and R2 of 0.849. The final BWTINSLEY equation is: , with RMR 𝑅𝑀𝑅 = 24.8 ∗ 𝐵𝑊 + 10

calculated in kcal/d and BW in kg. In the FFM-based model (FFMTINSLEY), DXA FFM was the

only statistically significant predictor of RMR (B: 25.94; p < 0.0001). Excluded variables

included sex (B: -0.066; p = 0.73), age (B: 0.005; p = 0.95), height (B: -0.115; p = 0.52) and

weight (B: 0.417; p = 0.37). Overall, the FFM-based model had an r of 0.923 and R2 of 0.851.

The final FFMTINSLEY equation is: , with RMR calculated in kcal/d 𝑅𝑀𝑅 = 25.9 ∗ 𝐹𝐹𝑀 + 284

and FFM in kg. In males, females and the entire sample, the average leave-one-out TE for both

the BW- and FFM-based equations were ≤ 15 kcal/d higher than the TEs when linear regression

was performed using the entire dataset (Table 6). The newly developed BWTINSLEY equation

generally minimized CE, ES, TE, and LOA relative to other BW-based equations. Additionally,

it was the only model without statistically significant proportional bias (Figure 2F). The newly

developed FFMTINSLEY equation also generally minimized CE, ES, TE, and LOA relative to other

FFM-based equations and exhibited less proportional bias (Figure 3F).

Additional sub-analysis with the newly developed equations was performed on

participants reporting current AAS usage (6 M, 1 F) versus those reporting no current usage of

Page 12 of 37



Draft

13

AAS. Although these results should be interpreted with caution due to the self-reported nature of

AAS usage and the substantial differences in sample size between those reporting current AAS

usage (n=7) and those reporting no current AAS usage (n=20), the newly developed equations

produced CE and r values of similar magnitudes in users and non-users when compared to the

reference method. When using the BWTINSLEY equation, the CE was 29 kcal/d (r = 0.93) in non-

users vs. – 32 kcal/d (r = 0.88) in users. Additionally, there was no difference in CE when users

and non-users were compared via independent samples t-test (p = 0.44). When using the

FFMTINSLEY equation, the CE was 9 kcal/d (r = 0.92) in non-users vs. -36 kcal/d (r = 0.91) in

users. Additionally, there was no difference in CE when users and non-users were compared via

independent samples t-test (p = 0.57).

Discussion

The purpose of this study was to evaluate the utility of several practical methods of RMR

estimation in male and female physique athletes, including two portable indirect calorimeters and

several commonly used BW- and FFM-based prediction equations. Additionally, preliminary

RMR prediction equations based on BW or FFM were developed for this population. The major

finding of this study was that the Cosmed FitmateTM portable indirect calorimeter and the newly

developed BW- and FFM-based prediction equations produced less error than all other methods

for RMR estimation in male and female physique athletes. No other BW-based RMR prediction

equations performed acceptably in males, although the ten Haaf (2014) BW equation produced

relatively low errors in females. The existing Cunningham (1980) and ten Haaf (2014) FFM-

based equations produced fairly accurate RMR estimates in males and females, although

substantial proportional bias was present, particularly when examining males alone. All other

Page 13 of 37



Draft

14

RMR prediction equations evaluated in this study are apparently unsuitable for use in muscular

physique athletes.

For this investigation, the ParvoMedics TrueOne® 2400 indirect calorimeter was

designated as the reference method of assessment due to its documented accuracy and reliability

(Cooper et al. 2009; Kaviani et al. 2018). Of the methods evaluated in this study, the Cosmed

FitmateTM portable indirect calorimeter produced RMR estimates with the narrowest 95% LOA

and highest correlation when compared to the reference RMR. The Cosmed FitmateTM and the

newly developed BW- and FFM-based equations exhibited relatively low TE and generally

exhibited smaller regression coefficients than most existing prediction equations, indicative of

less proportional bias. The CE for the FitmateTM was similar in males in females (~69 kcal/d;

trivial to small ES), and RMR estimates were not significantly different than the reference

method. Previous research also supports the validity and reliability of the FitmateTM system

(Nieman et al. 2006; Vandarakis et al. 2013). The other portable indirect calorimetry device

evaluated in this study (Breezing® Tracker) utilizes a portable metabolic analyzer that was

previously deemed to produce valid RMR estimates relative to the Douglas bag method (CE: 59

± 31 kcal/d; 95% LOA: ±215 kcal/d) in a sample of 17 healthy adults, ranging from underweight

to obese (Zhao et al. 2014). Another report derived from the same study found a CE of 6 kcal/d

when >300 tests were performed in 12 adults across a range of energy expenditures from 1,000

to 4,000 kcal/d (Xian, Quach et al. 2015). However, despite these results, the Breezing® device

performed very poorly in the present investigation. The device produced alarming

underestimations of RMR in males (CE: -513 kcal/d; large ES), a trivial correlation with the

reference method (r: 0.02), and wholly unacceptable LOA (±919 kcal/d). In females, Breezing®

produced the same CE as Cosmed (67 kcal/d) but only a moderate correlation (r: 0.38) with the

Page 14 of 37



Draft

15

reference method. Additionally, it produced a large regression coefficient (B: 0.569), indicative

of proportional bias, and the widest LOA of any method in females. Overall, this device

performed worse than all other methods in males, including prediction equations, as well as

performing poorly in females. Based on these results, the use of this device for estimating RMR

in this population is strongly discouraged.

In males, most existing BW- and FFM-based equations substantially underestimated

RMR by up to 393 kcal/d, while also exhibiting unacceptable levels of negative proportional bias

(i.e. greater underestimation of RMR in individuals with higher RMR). Based on the results of

this investigation, none of the existing BW-based equations can be deemed acceptable for use in

male physique athletes. However, when applied to males, the newly developed BW-based

equation demonstrated low error, a high correlation with the reference method, and no

statistically significant proportional bias. Overall, the performance of the new BW-based

equation was superior to other BW-based equations and similar to the Cosmed FitmateTM in

males. When considering FFM-based equations, the Cunningham (1980) and ten Haaf (2014)

FFM-based equations produced less error than other equations in males, with the exception of

the newly-developed FFM-based equation. While FFMTINSLEY demonstrated a smaller magnitude

of proportional bias than the other methods, statistically significant bias was still observed. The

frequent negative proportional bias observed in males may be partially attributed to one

participant with a very high RMR assessed by the reference method (2,998 kcal/d) that was

consistently underestimated by prediction equations. When a FFM-based equation is utilized for

RMR prediction in male physique athletes, the newly developed equation may be suitable.

However, appropriate caution should be employed when interpreting RMR estimates,

particularly in very large individuals with high RMR.

Page 15 of 37



Draft

16

In females, the ten Haaf (2014) equation was the only existing BW-based prediction

equation deemed acceptable for use. Both the ten Haaf and newly-developed BW-based equation

can potentially be used in female physique athletes similar to those examined in this

investigation. As in males, the Cunningham (1980) and ten Haaf (2014) equations produced

lower error than other FFM-based equations. Overall, these two equations, along with the newly-

developed FFM-based equation, may be acceptable for use in female physique athletes similar to

those evaluated in the present study.

It is noteworthy that, although the Cunningham (1980) FFM-based equation produced

relatively low error in males and females, individuals who were trained athletes were specifically

excluded during the development of this equation, and the researchers’ goal was to develop an

equation for a population of normal adults (Cunningham 1980). Additionally, since body

composition was not available for the participants used in equation development, it was predicted

from the participants’ body mass and age. Therefore, the accuracy of this equation is likely

serendipitous rather than due to similarities between the populations. A more recent equation

developed by Cunningham et al. (1991) produced relatively higher error than the older equation

in the present investigation. In contrast, ten Haaf et al. (2014) developed their BW- and FFM-

based equations in a group of male and female athletes from a variety of sports, including long

distance running and cycling, gymnastics, sprinting, rowing, swimming, fitness, hockey, soccer,

volleyball, dancing, martial arts, skating and tennis. Despite the diverse nature of the participants

used in equation development, the ten Haaf FFM-based prediction equation generally performed

well in the present sample of physique athletes. Additionally, the BW-based equation performed

well in females. While these results are promising for the application of the ten Haaf equations in

diverse athletic populations, the new BWTINSLEY and FFMTINSLEY equations were developed for a

Page 16 of 37



Draft

17

specific target population of athletes (i.e. muscular physique competitors) rather than broad

application to athletes in general. Nonetheless, cross validation of the newly developed equations

in sports with the most muscular athletes (e.g. American football) may be warranted.

Within the participants of this study, the new BW- and FFM-based equations performed

very similarly when compared to the reference RMR value. Although this indicates that either

could be used in individuals similar to the participants of the present study, an argument can be

made that the BW-based equation could be preferable for several reasons. First, the new FFM-

based equation was developed using DXA FFM, and DXA is often unavailable for body

composition assessment. The utilization of FFM estimates from other devices could impact this

equations accuracy due to discrepancies between DXA and other methods in athletic populations

(Moon et al. 2009; Graybeal et al. 2018). We performed additional analysis in our sample

indicating that the magnitude of difference in RMR with FFM estimates from alternative

methods (i.e. multi-compartment models, multiple bioelectrical impedance analysis devices and

bioimpedance spectroscopy) ranges from -30 to +74 kcal/d on average (data not shown).

Additionally, much lower assessment error can be expected when evaluating BW within a single

individual as compared to FFM. For both body composition assessment and RMR predictions

based on body composition, caution should be employed when evaluating a single individual due

to the distinct possibility of over- or under-estimation of body compartments (e.g. FFM) in any

given individual. Lastly, performing a simple BW measurement is much more feasible than

accurate body composition assessment in most settings. For these reasons, the most practical

option may be to employ the newly developed BW-based equation, provided that the individual

being evaluated exhibits similar characteristics to those used for equation development in this

population (Table 2).

Page 17 of 37



Draft

18

In the present investigation, all participants except one reported that they were in their

offseason from competitions or not currently in a specific phase of their competitive cycle. This

may be attributable to the fact that the data collection for this study was performed in the fall (i.e.

late August to mid-December). As such, our results are generalizable to physique athletes who

are not currently in the preparatory phase prior to a competition. It has been documented that

decreases in RMR are observed in the competition preparation periods, but that RMR is

recovered relatively quickly after competition as energy intake increases (Trexler et al. 2017;

Tinsley et al. 2018a).

The self-reported nature of AAS usage and limited number of individuals reporting

current AAS usage did not allow for comprehensive evaluation of possible differences between

AAS users and natural competitors. However, our preliminary evaluation did not reveal

appreciable differences in the performance of the newly developed equations in AAS users

versus non-users. Based on the available information, it is believed that the developed equations

can be used in athletes similar to those in our sample regardless of current AAS usage. There are

several other limitations to the present investigation. Our sample size is small for the

development of new predictive equations, although relatively few prospective participants were

available for evaluation due to the special population being assessed. However, we performed

leave-one-out cross-validation in accordance with recommendations for small samples sizes.

Nonetheless, we encourage additional external cross-validation of our newly developed

equations in order to more fully determine their utility. Our analysis did not reveal improved

utility of separate RMR prediction equations for male and female physique athletes (data not

shown). However, future research should continue to investigate whether a sex term improves

prediction of RMR in athletes. Although all participants reported to the laboratory after an

Page 18 of 37



Draft

19

overnight period of fasting and resting, it is possible that the frequent training sessions of the

participants could have altered RMR due to prior exercise. However, this limitation was deemed

necessary due to the unwillingness of prospective participants to abstain from exercise for longer

periods of time. In this respect, we believe our sample is representative of physique athletes in

general.

In conclusion, the Cosmed FitmateTM portable indirect calorimeter and the newly

developed BW- and FFM-based prediction equations may be suitable for RMR estimation in

male and female physique athletes similar to those in the present study. Additionally, the ten

Haaf (2014) and Cunningham (1980) FFM-based equations may be acceptable for use in male

and female physique athletes, with the ten Haaf BW-based equation being suitable in females

only. All other BW- and FFM-based prediction equations that were evaluated, as well as the

Breezing® Tracker, do not exhibit acceptable validity and frequently underestimate RMR in

physique athletes. Future research should externally cross-validate the developed equations in

order to refine the ability to practically produce valid estimates of RMR in this unique athletic

population.

Page 19 of 37



Draft

20

Acknowledgements

The authors would like to thank Megan Cruz, Alfred Kankam, and Michael Villarreal for their

assistance with data collection for this study.

Conflict of Interest Statement

The authors have no conflicts of interest to report.

Page 20 of 37



Draft

21

References

Bland, J.M. and Altman, D.G. 1986. Statistical methods for assessing agreement between two

methods of clinical measurement. Lancet, 1: 307-10.

Carlsohn, A., Scharhag-Rosenberger, F., Cassel, M., and Mayer, F. 2011. Resting metabolic rate

in elite rowers and canoeists: difference between indirect calorimetry and prediction. Ann. Nutr.

Metab. 58: 239-44. DOI: 10.1159/000330119

Compher, C., Frankenfield, D., Keim, N., and Roth-Yousey, L. 2006. Best Practice Methods to

Apply to Measurement of Resting Metabolic Rate in Adults: A Systematic Review. J. Am. Diet.

Assoc. 106: 881-903. DOI: 10.1016/j.jada.2006.02.009

Cooper, J.A., Watras, A.C., O'Brien, M.J., Luke, A., Dobratz, J.R., Earthman, C.P., et al. 2009.

Assessing validity and reliability of resting metabolic rate in six gas analysis systems. J. Am.

Diet. Assoc. 109: 128-32. DOI: 10.1016/j.jada.2008.10.004

Cunningham, J.J. 1980. A reanalysis of the factors influencing basal metabolic rate in normal

adults. Am. J. Clin. Nutr. 33: 2372-4. DOI: 10.1093/ajcn/33.11.2372

Cunningham, J.J. 1991. Body composition as a determinant of energy expenditure: a synthetic

review and a proposed general prediction equation. Am. J. Clin. Nutr. 54: 963-9. DOI:

10.1093/ajcn/54.6.963

da Rocha, E.E., Alves, V.G., Silva, M.H., Chiesa, C.A., and da Fonseca, R.B. 2005. Can

measured resting energy expenditure be estimated by formulae in daily clinical nutrition

practice? Curr. Opin. Clin. Nutr. Metab. Care. 8: 319-28. DOI:

10.1097/01.mco.0000165012.77567.1e

Page 21 of 37



Draft

22

De Lorenzo, A., Bertini, I., Candeloro, N., Piccinelli, R., Innocente, I., and Brancati, A. 1999. A

new predictive equation to calculate resting metabolic rate in athletes. J. Sports Med. Phys.

Fitness, 39: 213-9. PMID: 10573663

FAO 1985. Energy and protein requirements: Report of a Joint FAO/WHO/UNU Expert

Consultation.

Flack, K.D., Siders, W.A., Johnson, L., and Roemmich, J.N. 2016. Cross-Validation of Resting

Metabolic Rate Prediction Equations. Journal of the Academy of Nutrition and Dietetics, 116:

1413-1422. DOI: 10.1016/j.jand.2016.03.018

Graybeal, A.J., Moore, M.L., Cruz, M.R., and Tinsley, G.M. 2018. Body Composition

Assessment in Male and Female Bodybuilders: A 4-Compartment Model Comparison of Dual-

Energy X-Ray Absorptiometry and Impedance-Based Devices. Journal of Strength &

Conditioning Research, Published Ahead of Print. DOI: 10.1519/jsc.0000000000002831

Harris, J.A. and Benedict, F.G. 1918. A Biometric Study of Human Basal Metabolism. Proc.

Natl. Acad. Sci. USA, 4: 370-3. PMID: 16576330

Hayes, M., Chustek, M., Wang, Z., Gallagher, D., Heshka, S., Spungen, A., et al. 2002. DXA:

potential for creating a metabolic map of organ-tissue resting energy expenditure components.

Obes. Res. 10: 969-77. DOI: 10.1038/oby.2002.132

Helms, E., Aragon, A., and Fitschen, P. 2014. Evidence-based recommendations for natural

bodybuilding contest preparation: nutrition and supplementation. J. Int. Soc. Sports Nutr. 11.

DOI: 10.1186/1550-2783-11-20

Heyward, V.H. and Wagner, D. 2004. Applied body composition assessment: Champaign, IL :

Human Kinetics.

Page 22 of 37



Draft

23

Hopkins, W.G., Marshall, S.W., Batterham, A.M., and Hanin, J. 2009. Progressive statistics for

studies in sports medicine and exercise science. Med. Sci. Sports Exerc. 41: 3-13. DOI:

10.1249/MSS.0b013e31818cb278

Ivanescu, A.E., Li, P., George, B., Brown, A.W., Keith, S.W., Raju, D., et al. 2016. The

importance of prediction model validation and assessment in obesity and nutrition research. Int.

J. Obes. (Lond). 40: 887-94. DOI: 10.1038/ijo.2015.214

Jagim, A.R., Camic, C.L., Kisiolek, J., Luedke, J., Erickson, J., Jones, M.T., et al. 2017. The

accuracy of resting metabolic rate prediction equations in athletes. Journal of Strength &

Conditioning Research, DOI: 10.1519/jsc.0000000000002111

Kaviani, S., Schoeller, D. A., Ravussin, E., Melanson Edward, L., Henes Sarah, T., Dugas Lara,

R., et al. 2018. Determining the Accuracy and Reliability of Indirect Calorimeters Utilizing the

Methanol Combustion Technique. Nutrition in Clinical Practice, 33: 206-216. DOI:

10.1002/ncp.10070

Kelly, T.L., Wilson, K.E., and Heymsfield, S.B. 2009. Dual Energy X-Ray Absorptiometry Body

Composition Reference Values from NHANES. PLoS ONE, 4: e7038. DOI:

10.1371/journal.pone.0007038

Lohman, T.G., Caballero, B., Himes, J.H., Davis, C.E., Stewart, D., Houtkooper, L., et al. 2000.

Estimation of body fat from anthropometry and bioelectrical impedance in Native American

children. International Journal Of Obesity, 24: 982. DOI: 10.1038/sj.ijo.0801318

Mifflin, M.D., Jeor, S.T.S., Hill, L.A., Scott, B.J., Daugherty, S.A., and Koh, Y.O. 1990. A new

predictive equation for resting energy expenditure in healthy individuals. The Am. J. Clin. Nutr.

51: 241-247. DOI: 10.1093/ajcn/51.2.241

Page 23 of 37



Draft

24

Moco, A.V., Matias, C.N., Santos, D.A., Rocha, P.M., Minderico, C.S., Cyrino, E.S., et al. 2018.

Usefulness of Reflection Scanning in Determining Whole-Body Composition in Broadly Built

Individuals Using Dual-Energy X-ray Absorptiometry. J. Clin. Densitom. In Press. DOI:

10.1016/j.jocd.2018.03.007

Moon, J.R., Eckerson, J.M., Tobkin, S.E., Smith, A.E., Lockwood, C.M., Walter, A.A., et al.

2009. Estimating body fat in NCAA Division I female athletes: a five-compartment model

validation of laboratory methods. European Journal of Applied Physiology, 105: 119-130. DOI:

10.1007/s00421-008-0881-9

Nieman, D.C., Austin, M.D., Benezra, L., Pearce, S., McInnis, T., Unick, J., et al. 2006.

Validation of Cosmed's FitMate in measuring oxygen consumption and estimating resting

metabolic rate. Res. Sports Med. 14: 89-96. DOI: 10.1080/15438620600651512

Owen, O.E., Holup, J.L., D'Alessio, D.A., Craig, E.S., Polansky, M., Smalley, K.J., et al. 1987.

A reappraisal of the caloric requirements of men. Am. J. Clin. Nutr. 46: 875-85. DOI:

10.1093/ajcn/46.6.875

Slater, G. and Phillips, S.M. 2011. Nutrition guidelines for strength sports: Sprinting,

weightlifting, throwing events, and bodybuilding. Journal of Sports Sciences, 29: S67-S77. DOI:

10.1080/02640414.2011.574722

ten Haaf, T. and Weijs, P.J.M. 2014. Resting Energy Expenditure Prediction in Recreational

Athletes of 18–35 Years: Confirmation of Cunningham Equation and an Improved Weight-

Based Alternative. PLoS ONE, 9. DOI: 10.1371/journal.pone.0108460

Thomas, D.T., Erdman, K.A., and Burke, L.M. 2016. American College of Sports Medicine Joint

Position Statement. Nutrition and Athletic Performance. Med. Sci. Sports Exerc. 48: 543-68.

DOI: 10.1249/MSS.0000000000000852

Page 24 of 37



Draft

25

Thompson, J. and Manore, M.M. 1996. Predicted and measured resting metabolic rate of male

and female endurance athletes. J. Am. Diet. Assoc. 96: 30-4. DOI: 10.1016/s0002-

8223(96)00010-7

Tinsley, G.M. 2017. Proportional bias between dual-energy x-ray absorptiometry and

bioelectrical impedance analysis varies based on sex in active adults consuming high- and low-

carbohydrate diets. Nutr. Res. 42: 85-100. DOI: 10.1016/j.nutres.2017.05.003

Tinsley, G.M., Trexler, E.T., Smith-Ryan, A.E., Paoli, A., Graybeal, A.J., Campbell, B.I., et al.

2018a. Changes in Body Composition and Neuromuscular Performance Through Preparation,

Two Competitions, and a Recovery Period in an Experienced Female Physique Athlete. Journal

of Strength & Conditioning Research, Ahead of Print. DOI: 10.1519/JSC.0000000000002758

Tinsley, G.M., Moore, M.L., and Graybeal, A.J. 2018b. Precision of Dual-Energy X-Ray

Absorptiometry Reflection Scans in Muscular Athletes. Journal of Clinical Densitometry, Ahead

of Print. DOI: 10.1016/j.jocd.2018.09.005

Trexler, E.T., Hirsch, K.R., Campbell, B.I., and Smith-Ryan, A.E. 2017. Physiological Changes

Following Competition in Male and Female Physique Athletes: A Pilot Study. Int. J. Sport Nutr.

Exerc. Metab. 1-25. DOI: 10.1123/ijsnem.2017-0038

Vandarakis, D., Salacinski, A.J., and Broeder, C.E. 2013. A comparison of COSMED metabolic

systems for the determination of resting metabolic rate. Res. Sports. Med. 21: 187-94. DOI:

10.1080/15438627.2012.757226

Wang, Z., Heshka, S., Gallagher, D., Boozer, C.N., Kotler, D.P., and Heymsfield, S.B. 2000.

Resting energy expenditure-fat-free mass relationship: new insights provided by body

composition modeling. Am. J. Physiol. Endocrinol. Metab. 279: E539-45. DOI:

10.1152/ajpendo.2000.279.3.E539

Page 25 of 37



Draft

26

Xian, X., Quach, A., Bridgeman, D., Tsow, F., Forzani, E., and Tao, N. 2015. Personalized

indirect calorimeter for energy expenditure (EE) measurement. Global Journal of Obesity,

Diabetes and Metabolic Syndrome, 2: 004-008. DOI: 10.17352/2455-8583.000007

Zhao, D., Xian, X., Terrera, M., Krishnan, R., Miller, D., Bridgeman, D., et al. 2014. A pocket-

sized metabolic analyzer for assessment of resting energy expenditure. Clin. Nutr. 33: 341-7.

DOI: 10.1016/j.clnu.2013.06.001

Page 26 of 37



Draft

27

Table 1. Resting metabolic rate prediction equations.

1Units for these equations are: RMR (kcal/d), FFM (kg), BW (kg), H (cm), age (y); 2Newly-developed equations presented in this article

Type of equation Reference Equation1

Organ/tissue equation Hayes (2002) 𝑅𝑀𝑅 = 𝑅𝑀𝑅𝐵 + 𝑅𝑀𝑅𝑆𝑀 + 𝑅𝑀𝑅𝑏𝑜𝑛𝑒 + 𝑅𝑀𝑅𝐴𝑇 + 𝑅𝑀𝑅𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙

FFM-based equations Cunningham (1991) 𝑅𝑀𝑅 = 21.6 ∗ 𝐹𝐹𝑀 + 370

Cunningham (1980) 𝑅𝑀𝑅 = 22 ∗ 𝐹𝐹𝑀 + 500

Mifflin (1990) 𝑅𝑀𝑅 = 19.7 ∗ 𝐹𝐹𝑀 + 413

Owen (1987) M: 𝑅𝑀𝑅 = 22.3 ∗ 𝐹𝐹𝑀 + 290F: 𝑅𝑀𝑅 = 19.7 ∗ 𝐹𝐹𝑀 + 334

ten Haaf (2014) 𝑅𝑀𝑅 = 0.239(95.272 ∗ 𝐹𝐹𝑀 + 2026.161)

Tinsley2 𝑅𝑀𝑅 = 25.9 ∗ 𝐹𝐹𝑀 + 284

BW-based equations Mifflin (1990) 𝑅𝑀𝑅 = 9.99 ∗ 𝐵𝑊 + 6.25 ∗ 𝐻 ― 4.92 ∗ 𝑎𝑔𝑒 + 166 ∗ 𝑠𝑒𝑥 ― 161

Harris-Benedict (1918) M: 𝑅𝑀𝑅 = 13.75 ∗ 𝐵𝑊 + 5 ∗ 𝐻 ― 6.76 ∗ 𝑎𝑔𝑒 + 66.47F: 𝑅𝑀𝑅 = 9.56 ∗ 𝐵𝑊 + 1.85 ∗ 𝐻 ― 4.68 ∗ 𝑎𝑔𝑒 + 655.1

FAO (1985) M: 𝑅𝑀𝑅 = 15.3 ∗ 𝐵𝑊 + 679F: 𝑅𝑀𝑅 = 14.7 ∗ 𝐵𝑊 + 496

De Lorenzo (1999) 𝑅𝑀𝑅 = 9 ∗ 𝐵𝑊 + 11.7 ∗ 𝐻 ― 857

ten Haaf (2014) 𝑅𝑀𝑅 = 0.239(49.94 ∗ 𝐵𝑊 + 24.59 ∗ 𝐻 ― 34.014 ∗ 𝑎𝑔𝑒 + 799.257 ∗ 𝑠𝑒𝑥 + 122.502)

Tinsley2 𝑅𝑀𝑅 = 24.8 ∗ 𝐵𝑊 + 10

Page 27 of 37



Draft

28

Table 2. Participant Characteristics

Data presented as mean ± SD

All (n=27) Males (n=17) Females (n=10)

Age (y) 25.9 ± 6.0 26.0 ± 6.5 25.8 ± 5.4Height (cm) 175.6 ± 9.2 180.4 ± 7.2 167.5 ± 5.7Weight (kg) 82.9 ± 17.0 94.0 ± 9.7 63.8 ± 5.7

BMI (kg/m2) 26.6 ± 3.5 28.8 ± 2.0 22.8 ± 1.6

DXA Body fat (%) 15.0 ± 4.4 12.5 ± 2.7 19.2 ± 3.4

DXA FFMI (kg/m2) 21.8 ± 3.4 24.2 ± 1.3 17.7 ± 0.9

Page 28 of 37



Draft

29

Table 3. Validity of resting metabolic rate estimates in males and females combined (n = 27).

Method Mean ± SD p-value (t-test)

CE ± SD Cohen's d

r p-value (correlation)

R2 SEE TE 95% LOA

B p-value (linear

regression)ICPARVO 2051 ± 457 --- --- --- --- --- --- --- --- --- --- ---ICCOSMED 2121 ± 482 0.037 69 ± 164 0.15t 0.94p < 0.0001* 0.88 167 175 ± 321 0.054 0.45ICBREEZING 1753 ± 319 0.003* -298 ± 479 -0.76m 0.28s 0.16 0.08 312 557 ± 940 -0.549 0.067

OTHAYES 1896 ± 398 0.0002* -155 ± 188 -0.36s 0.91p < 0.0001* 0.83 166 241 ± 369 -0.146 0.10FFMCUNN1 1999 ± 358 0.16 -53 ± 188 -0.13t 0.92p < 0.0001* 0.85 141 192 ± 368 -0.253 0.004†

FFMCUNN2 1842 ± 352 <0.0001* -210 ± 190 -0.51s 0.92p < 0.0001* 0.85 138 281 ± 372 -0.272 0.002†

FFMMIFFLIN 1755 ± 321 <0.0001* -296 ± 204 -0.75m 0.92p < 0.0001* 0.85 126 357 ± 399 -0.365 < 0.001†

FFMOWEN 1778 ± 400 <0.0001* -273 ± 179 -0.64m 0.92p < 0.0001* 0.85 158 325 ± 351 -0.14 0.094FFMT-H 2036 ± 371 0.66 -16 ± 184 -0.04t 0.92p < 0.0001* 0.85 146 181 ± 360 -0.218 0.011†

BWMIFF 1741 ± 297 <0.0001* -310 ± 232 -0.80m 0.90p <0.0001* 0.80 134 385 ± 455 -0.446 <0.0001†

BWH-B 1852 ± 343 <0.0001* -199 ± 209 -0.49s 0.90p <0.0001* 0.81 151 286 ± 411 -0.302 0.002†

BWFAO 1848 ± 363 <0.0001* -203 ± 215 -0.49s 0.89p <0.0001* 0.79 170 293 ± 421 -0.244 0.017†

BWDEL 1943 ± 252 0.04 -108 ± 259 -0.29s 0.89v <0.0001* 0.80 116 276 ± 508 -0.611 <0.0001†

BWTH 1960 ± 342 0.04 -92 ± 221 -0.23s 0.89v <0.0001* 0.79 161 235 ± 432 -0.307 0.004†

For RMR prediction, the following equations were used: Hayes et al. (2002), Cunningham (1980) [CUNN1], Cunningham (1991) [CUNN2], Mifflin et al. (1990), Owen et al. (1987), ten Haaf (2014) [TH], Harris and Benedict (1918) [H-B], FAO (1985) and DeLorenzo (1999) [DEL]. *statistically significant at p < 0.0033 (adjustment for multiple comparisons); †statistically significant at p < 0.05Italic superscripts indicate magnitude of effect size or correlation (t: trivial, s: small, m: moderate, l: large, v: very large, p: near perfect).Abbreviations: B (regression coefficient), BW (body weight model), CE (constant error), F (female), FFM (fat-free mass model), IC (indirect calorimetry), LOA (limits of agreement), M (male), OT (organ/tissue model), SEE (standard error of the estimate), TE (total error).

Page 29 of 37



Draft

30

Table 4. Validity of resting metabolic rate estimates in males (n = 17).


CE ± SD Cohen's d


R2 SEE TE 95% LOA

B p-value (linear

regression)ICPARVO 2337 ± 310 --- --- --- --- --- --- --- --- --- --- ---ICCOSMED 2408 ± 350 0.15 71 ± 192 0.21s 0.84v < 0.0001* 0.70 198 199 ± 376 0.134 0.396ICBREEZING 1824 ± 359 0.0003* -513 ± 469 -1.53l 0.02t 0.93 0.00 370 686 ± 919 0.286 0.576

OTHAYES 2166 ± 199 0.005 -172 ± 217 -0.66m 0.72v 0.001* 0.51 143 272 ± 426 -0.502 0.023†

FFMCUNN1 2245 ± 170 0.09 -92 ± 212 -0.37s 0.76v 0.0003* 0.57 115 226 ± 416 -0.655 0.002†

FFMCUNN2 2083 ± 167 0.0002* -254 ± 213 -1.02m 0.76v 0.0003* 0.57 113 328 ± 418 -0.673 0.001†

FFMMIFFLIN 1975 ± 152 < 0.0001* -362 ± 218 -1.48m 0.76v 0.0003* 0.57 103 419 ± 428 -0.763 < 0.001†

FFMOWEN 2058 ± 172 0.0001* -279 ± 212 -1.11m 0.76v 0.0003* 0.57 116 346 ± 415 -0.641 0.002†

FFMT-H 2290 ± 176 0.37 -47 ± 210 -0.19t 0.76v 0.0003* 0.57 119 210 ± 413 -0.620 0.003†

BWMIFF 1944 ± 144 <0.0001* -393 ± 238 -1.63l 0.67l 0.003* 0.45 110 456 ± 467 -0.854 0.0005†

BWH-B 2086 ± 176 0.0004* -251 ± 233 -1.00m 0.67l 0.003* 0.44 136 338 ± 457 -0.650 0.008†

BWFAO 2102 ± 160 0.001* -235 ± 241 -0.95m 0.64l 0.005 0.41 128 332 ± 473 -0.758 0.003†

BWDEL 2032 ± 180 <0.0001* -305 ± 224 -1.20l 0.70v 0.001* 0.49 133 374 ± 438 -0.724 0.001†

BWTH 2192 ± 168 0.02 -145 ± 241 -0.58s 0.63l 0.006 0.40 135 275 ± 473 -0.709 0.006†

See footnotes on Table 3.

Page 30 of 37



Draft

31

Table 5. Validity of resting metabolic rate estimates in females (n = 10).Method Mean ± SD p-value

(t-test)CE ± SD Cohen's

dr p-value

(correlation)R2 SEE TE 95%

LOAB p-value

(linear regression)

ICPARVO 1566 ± 133 --- --- --- --- --- --- --- --- --- --- ---ICCOSMED 1633 ± 182 0.08 67 ± 109 0.42s 0.80l 0.003 0.65 115 124 ± 214 0.343 0.169ICBREEZING 1633 ± 200 0.30 67 ± 194 0.4s 0.38m 0.28 0.14 196 196 ± 380 0.569 0.24

OTHAYES 1438 ± 126 0.01 -128 ± 131 -0.99m 0.49m 0.14 0.24 116 178 ± 257 -0.078 0.855FFMCUNN1 1581 ± 107 0.68 16 ± 116 0.13t 0.55l 0.09 0.30 95 111 ± 228 -0.283 0.471FFMCUNN2 1432 ± 105 0.005 -134 ± 116 -1.12m 0.55l 0.09 0.30 93 173 ± 227 -0.307 0.436FFMMIFFLIN 1381 ± 96 0.0006* -184 ± 113 -1.59l 0.55l 0.09 0.30 85 213 ± 222 -0.421 0.285FFMOWEN 1302 ± 96 < 0.0001* -263 ± 113 -2.27v 0.55l 0.09 0.30 85 284 ± 222 -0.421 0.285FFMT-H 1604 ± 110 0.33 38 ± 117 0.31s 0.55l 0.09 0.30 98 118 ± 230 -0.240 0.542

BWMIFF 1396 ± 95 0.004 -169 ± 139 -1.47l 0.29s 0.41 0.08 96 215 ± 273 -0.519 0.332BWH-B 1454 ± 70 0.02 -111 ± 128 -1.04m 0.33s 0.02 0.11 71 165 ± 252 -0.883 0.076BWFAO 1417 ± 77 0.01 -149 ± 156 -1.37l -0.04t 0.92 0.00 82 210 ± 306 -1.03 0.142BWDEL 1677 ± 107 0.04 111 ± 142 0.92m 0.31s 0.37 0.10 108 175 ± 279 -0.333 0.526BWTH 1566 ± 112 0.99 0 ± 142 0.00t 0.26s 0.46 0.07 114 142 ± 294 -0.277 0.619

See footnotes on Table 3.

Page 31 of 37



Draft

32

Table 6. Preliminary evaluation of new resting metabolic rate prediction equations.


CE ± SD Cohen's d


R2 SEE TE TE (LOO)

95% LOA

B p-value (linear

regression)All (n=27) 2051 ± 457 --- --- --- --- --- --- --- --- --- --- --- ---M (n=17) 2337 ± 310 --- --- --- --- --- --- --- --- --- --- --- ---ICPARVOF (n=10) 1566 ± 133 --- --- --- --- --- --- --- --- --- --- --- ---

All (n=27) 2049 ± 422 0.93 -3 ± 176 -0.01t 0.92p < 0.0001* 0.85 166 173 187 ± 346 -0.084 0.31M (n=17) 2338 ± 200 0.99 1 ± 205 0.00t 0.76v 0.0003* 0.57 135 199 214 ± 402 -0.484 0.018†FFMTINSLEYF (n=10) 1557 ± 126 0.83 -9 ± 123 -0.07t 0.55l 0.09 0.30 111 117 129 ± 241 -0.074 0.852

All (n=27) 2065 ± 422 0.70 13 ± 180 0.03t 0.92p <0.0001* 0.85 169 177 188 ± 352 -0.083 0.32M (n=17) 2342 ± 241 0.92 5 ± 204 0.02t 0.75v 0.0003* 0.56 164 198 211 ± 401 -0.283 0.159BWTINSLEYF (n=10) 1593 ± 140 0.50 28 ± 137 0.20s 0.50s 0.13 0.25 129 132 140 ± 268 0.070 0.868

*statistically significant at p < 0.0033 (adjustment for multiple comparisons); †statistically significant at p < 0.05Italic superscripts indicate magnitude of effect size or correlation (t: trivial, s: small, m: moderate, l: large, v: very large, p: near perfect).Abbreviations: B (regression coefficient), BW (body weight model), CE (constant error), F (female), FFM (fat-free mass model), IC (indirect calorimetry), LOA (limits of agreement), LOO (leave-one-out cross-validation), M (male), OT (organ/tissue model), SEE (standard error of the estimate), TE (total error).

Page 32 of 37



Draft

33

FIGURE LEGENDS

Figure 1. Bland-Altman plots for resting metabolic rate estimation via portable indirect calorimeters and an organ/tissue equation. Results are shown for Cosmed FitmateTM (A), Breezing® (B), and the Hayes et al. (2002) equation (C). The middle horizontal line indicates the constant error (i.e. mean difference), while the upper and lower horizontal lines indicate the 95% limits of agreement. The regression line best fitting the data is shown, with larger slopes indicating greater proportional bias.

Figure 2. Bland-Altman plots for resting metabolic rate estimation via body weight equations. Results are shown for Mifflin (A), Harris-Benedict (B), FAO (C), DeLorenzo (D), ten Haaf (E) and Tinsley (F). The middle horizontal line indicates the constant error (i.e. mean difference), while the upper and lower horizontal lines indicate the 95% limits of agreement. The regression line best fitting the data is shown, with larger slopes indicating greater proportional bias.

Figure 3. Bland-Altman plots for resting metabolic rate estimation via fat-free mass equations. Results are shown for Cunningham (1991) (A), Cunningham (1980) (B), Mifflin (C), Owen (D), ten Haaf (E) and Tinsley (F). The middle horizontal line indicates the constant error (i.e. mean difference), while the upper and lower horizontal lines indicate the 95% limits of agreement. The regression line best fitting the data is shown, with larger slopes indicating greater proportional bias.

Figure 4. Newly developed resting metabolic rate prediction equations. A strong linear relationship was observed between resting metabolic rate and body weight (A) and between resting metabolic rate and fat-free mass (B). The newly developed equations produced strong correlations with reference resting metabolic rate estimates (body weight: r = 0.921, R2 = 0.849 [C]; fat-free mass: r = 0.923, R2 = 0.851 [D]).

Page 33 of 37



Draft

AVERAGED RMR (kcal/d)

30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

5 0 0

0

- 5 0 0

- 1 0 0 0

- 1 5 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0

A B C Page 34 of 37



Draft


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0

D E F

A B C Page 35 of 37



Draft


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0


30002500200015001000

ME

AN

DIF

FE

RE

NC

E (

kc

al/

d)

6 0 0

3 0 0

0

- 3 0 0

- 6 0 0

- 9 0 0

D E F

A B C Page 36 of 37



Draft

BODY WEIGHT (kg)

1 1 01 0 09 08 07 06 05 0

RM

R (

kc

al/

d)

3000

2500

2000

1500

1000

FFM (kg)

1 0 09 08 07 06 05 04 0R

MR

(k

ca

l/d

)

3000

2500

2000

1500

1000

RE

FE

RE

NC

E R

MR

(kc

al/d

)

3000

2500

2000

1500

1000

BW-PREDICTED RMR (kcal/d)

30002500200015001000

FFM-PREDICTED RMR (kcal/d)

30002500200015001000

RE

FE

RE

NC

E R

MR

(kc

al/d

)

3000

2500

2000

1500

1000

A B

C D

Page 37 of 37



draft - university of toronto t-space · participants were required to meet at least one of the...

Documents