Indirect calorimetry in humans: a postcalorimetric evaluation procedure
for correction of metabolic monitor variability1–3
Peter Schadewaldt, Bettina Nowotny, Klaus Straßburger, Jörg Kotzka, and Michael Roden
ABSTRACT
Background: Indirect calorimetry (IC) with metabolic monitors is
widely used for noninvasive assessment of energy expenditure and
macronutrient oxidation in health and disease.
Objective: To overcome deficiencies in validity and reliability of
metabolic monitors, we established a procedure that allowed cor-
rection for monitor-specific deviations.
Design: Randomized comparative IC (canopy mode) with the
Deltatrac MBM-100 (Datex) and Vmax Encore 29n (SensorMedix)
was performed in postabsorptive (overnight fast .8 h) healthy sub-
jects (n = 40). In vitro validation was performed by simulation of
oxygen consumption (VO2) and carbon dioxide output (VCO2) rates
by using mass-flow regulators and pure gases. A simulation-based
postcalorimetric calibration of cart readouts [individual calibration
control evaluation (ICcE)] was established in adults (n = 24).
Results: The comparison of carefully calibrated monitors showed
marked differences in VCO2 and VO2 (P , 0.01) and derived
metabolic variables [resting energy expenditure (REE), respiratory
quotient (RQ), glucose/carbohydrate oxidation (Gox), and fat oxi-
dation (Fox); P , 0.001]. Correlations appeared to be acceptable
for breath gas rates and REE (R2 w 0.9) but were unacceptable for
RQ (R2 = 0.3), Gox, and Fox (R2 = 0.2). In vitro simulation exper-
iments showed monitor-dependent interferences for VCO2 and VO2
as follows: 1) within series, nonlinear and variable deviations of
monitor readouts at different exchange rates; 2) between series,
differences and unsteady variability; and 3) differences in individual
monitor characteristics (eg, rate dependence, stability, imprecision).
The introduction of the postcalorimetric recalibration by ICcE re-
sulted in an adjustment of gas exchange rates and the derived met-
abolic variables with reasonable correlations (R2 . 0.9).
Conclusions: Differential, metabolic, monitor-specific deviations
are the primary determinants for lack of accuracy, comparability,
and transferability of results. This problem can be overcome by
the present postcalorimetric ICcE procedure. Am J Clin Nutr
2013;97:763–73.
INTRODUCTION
Indirect calorimetry (IC)4 is a noninvasive method for the
determination of energy expenditure by respiratory gas exchange
analysis. The principles of IC were established in the 19th century
(1–3). With the use of appropriate devices (metabolic monitor or
cart), the rates of carbon dioxide output (VCO2) and oxygen
consumption (VO2) are typically assessed under resting conditions.
The data are used for calculation of resting energy expenditure
(REE) and the respiratory quotient (RQ). The method also allows
Am J Clin Nutr 2013;97:763–73. Printed in USA. Ó 2013 American Society for Nutrition
estimation of the differential contribution of macronutrient oxi-
dation (ie, carbohydrate, fat, protein) to REE in vivo when protein
oxidation is additionally evaluated. The procedure is well estab-
lished and allows far more accurate estimates of REE than all of
the available formula-based approaches (4–15).
An early clinical application was for diagnosis of thyroid
dysfunction. Today, IC is of interest for evaluation of REE, eg, in
Downloaded from ajcn.nutrition.org by guest on November 30, 2017
intensive care units for assessment of caloric needs in severely ill
patients (14, 16, 17). Experimental applications of IC in humans
are widespread, eg, for studies on macronutrient oxidation, as-
sessment of thermogenesis, measurement of energy expenditure
during rest and muscular exercise, and exploration of patho-
genesis of obesity and diabetes mellitus (5, 7, 18, 19).
Despite its broad use, IC has some limitations. In particular,
commercially manufactured devices may differ in technical
construction and analytic performance. For assessment of REE
and RQ in humans, flow-through respirometry (ventilated hood/
canopy mode) is the method of choice in the clinical and the
experimental setting. Among canopy-based metabolic carts, the
Deltatrac MBM-100 (Datex) is one of the most widespread used
devices. The Deltatrac was supposed to exhibit the greatest
accuracy and the least intercart variability of all such devices
worldwide. Because this cart is no longer produced, there is
a need for replacement by contemporary metabolic monitors
(14, 20–25).
We here report on a comparative evaluation of the validity and
reliability of 2 metabolic carts by applying simulation experi-
ments in vitro. The results showed relevant deficiencies in cart
performances, which led us to develop a specific postcalorimetric
1
From the Institutes of Clinical Biochemistry and Pathobiochemistry (PS
and JK), Clinical Diabetology (BN and MR), and Biometry and Epide-
miology (KS), German Diabetes Center, Düsseldorf, Düsseldorf, Germany,
and the Department of Metabolic Diseases, University Clinic Düsseldorf,
Düsseldorf, Germany (MR).
2
There was no funding for this article.
3
Address correspondence to P Schadewaldt, Institut für Klinische Bioche-
mie & Pathobiochemie, Deutsches Diabetes Zentrum, Auf’m Hennekamp
65, D-40225 Düsseldorf. E-mail: schadewa@uni-duesseldorf.de.
4
Abbreviations used: Fox, fat oxidation; Gox, glucose/carbohydrate oxida-
tion; IC, indirect calorimetry; ICcE, individual calibration control evaluation;
Pox, protein oxidation; REE, resting energy expenditure; RQ, respiratory quo-
tient; VinfusedCO2, known infused mass flow of carbon dioxide; VinfusedN2, known
infused mass flow of nitrogen; VCO2, carbon dioxide output rate; VO2, oxygen
consumption rate.
Received January 12, 2012. Accepted for publication December 28, 2012.
First published online February 27, 2013; doi: 10.3945/ajcn.112.035014.
763
Supplemental Material can be found at:
http://ajcn.nutrition.org/content/suppl/2013/03/10/ajcn.112.0
35014.DCSupplemental.html
764 SCHADEWALDT ET AL

calibration procedure allowing us to control for cart-dependent
inaccuracy and variability in measurements with any canopy-
based metabolic monitor.
SUBJECTS AND METHODS
Subjects
Inclusion criteria for participation in the IC studies were
age .18 y and written informed consent. Exclusion criteria
were the presence of an acute or chronic illness, pregnancy or
breastfeeding in women, and contravention of the requirements
of the study. Initial recruitment started in March 2008.
All subjects received detailed preparatory oral and printed
information on the aims and prerequisites of the study. The
following requirements for REE measurements were specifically
stressed: 1) absence of vigorous physical activity or alcohol
abuse for 20 h, 2) no intake of food or drinks other than water
for at least 8 h, 3) abstention from nicotine for 3 h, 4) restriction
of any physical activity in the morning of the study day to the
metabolic carts were calibrated for each individual in a
direct subsequent in vitro simulation experiment (mass-
flow meter regulated infusion of pure gaseous CO2 and N2).
Cohort 3 (intraindividual variability cohort): Each of 6 healthy
adult subjects underwent multiple REE measurements on
4 different study days. The mean (6SD) total observation
period was 21 6 5 wk (median: 21 wk; range: 14–16 wk),
and the mean (6SD) interval between measurements was
7 6 5 wk (median: 6 wk; range: 4–16 wk).
Metabolic monitors and measurements
Carts
Two flow-through system metabolic carts allowing gas dilution–
based canopy-mode measurements were comparatively evalu-
ated: 1) the Deltatrac MBM-100 (Datex), equipped with an
infrared CO2 sensor and a differential paramagnetic O2 sensor
and fixed mass flow (40 L/min), and 2) the Vmax Encore 29n
(SensorMedix; delivered by Cardinal Health Germany) with

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

absolute necessary minimum, and 5) arrival in the metabolic unit
at 0800 (60.5 h). Adherence to these guidelines was inquired
for just before the measurements. Subsequent to the REE mea-
surements, the participants underwent bioimpedance analysis
(Body Impedance Analyzer Nutrigard-S; Data Inputy) for assess-
ment of fat mass and lean body mass, and a generous breakfast
was served after the tests.
The study protocol was examined and approved by the ethics
committee of the University of Düsseldorf.
The following 3 cohorts participated in this study. The an-
thropometric data are summarized in Table 1.
Cohort 1 (cart comparison cohort): A total of 40 healthy adults
participated in the metabolic monitor comparison study.
Cohort 2 (cart evaluation cohort): This group comprised 22
healthy adults. In this series, the mean rates of CO2 pro-
duction and O2 consumption in vivo as read out by the
an infrared CO2 sensor and an electrochemical O2 sensor and
equipped with a variable mass flow.
Maintenance was performed on the Deltatrac and the Vmax
Encore before the study and the devices were checked extensively
for adequate performance by the authorized technical service
units at GE Health Care and Cardinal Health Germany, respectively.
Before each use in the metabolic unit, the carts were prepared and
calibrated according to the manufacturers’ instructions.
The Deltatrac and Vmax Encore provide VCO2 and VO2 in
milliliters per minute and liters per minute, respectively, with the
readout adjusted to standard temperature pressure dry conditions
at temperature = 273 K and pressure = 1013 hPa.
Energy expenditure measurements
The subjects arrived at w0800; they were allowed to void and
were then asked to lie down in comfortable beds in a quiet special
study room. Overall, study conditions were in accordance with

TABLE 1
Characteristics of subjects undergoing resting energy expenditure measurements1
Cohort 3
Cohort 1 (metabolic Cohort 2 (metabolic (intraindividual
cart comparison) cart evaluation) variability)

Variable Women Men Women Men Women/men

n 28 12 17 5 4/2
Age (y) 38.8 6 13.32 48.1 6 10.1 40.2 6 14.5 47.7 6 14.4 24.7 6 2.1
Median (range) 31.8 (20.8–62.6) 51.1 (27.9–60.0) 32.7 (25.8–63.9) 55.8 (27.9–60.9) 23.6 (23.0–28.0)
Height (cm) 167 6 8 181 6 7 167 6 8 177 6 9 177 6 11
Median (range) 167 (153–181) 184 (165–189) 170 (152–180) 176 (165–186) 175 (162–190)
Weight (kg) 66 6 12 83 6 18 68 6 9 74 6 10 69 6 8
Median (range) 63 (49–93) 78 (68–135) 67 (49–84) 74 (59–85) 66 (62–83)
BMI (kg/m2) 23.6 6 4.1 25.3 6 4.9 24.5 6 3.9 23.7 6 4.4 22.0 6 1.5
Median (range) 22.9 (18.9–36.7) 23.1 (21.6–37.8) 23.3 (18.4–33.4) 22.6 (19.8–31.2) 22.0 (20.3–24.0)
Fat mass (kg) 19 6 7 17 6 11 20 6 6 16 6 7 16 6 3
Median (range) 18 (9–35) 16 (6–46) 20 (8–32) 14 (9–26) 17 (11–18)
Lean mass (kg) 47 6 5 66 6 9 48 6 5 58 6 6 53 6 9
Median (range) 45 (40–58) 64 (56–83) 48 (41–59) 60 (48–65) 50 (46–66)
1
Healthy subjects without stated acute or chronic illness entered the study in the postabsorptive state (overnight fast
.8 h).
2
Mean 6 SD (all such values).
 METABOLIC MONITOR VALIDATION AND STANDARDIZATION
recommended best practice guidelines (16, 17). The room temper-
ature was 20–228C, and blankets were available when requested.
After a resting period (10 min), the canopy of one of the met-
abolic carts was positioned over the participant’s head and IC
measurement was started with an initial 10-min period to ac-
custom the participant to the device and for equilibration.
Subsequently, a 20-min recording period followed while the
subject remained under strict resting conditions. Data readout
(ie, VCO2 and VO2) was set at the highest available frequency
(ie, Deltatrac: 1 readout/min; Vmax Encore, 3 readouts/min).
After a second resting period (10 min), the subject received
a second canopy measurement with the other metabolic cart with
the use of the same time protocol as in the first measurement.
In general, 2 subjects were investigated in parallel with both
carts, and the sequence of cart use was randomly assigned
throughout the trials. Preparatory measures, calibration pro-
cedures, and IC measurements were performed or supervised by
one researcher with special experience and expertise.
Gas-exchange simulation studies
In vitro, CO2 production and O2 consumption were simulated by
infusion of gaseous CO2 (purity: 4.5) and N2 (purity: 5.0), re-
spectively, by using a high-precision mass-flow regulator (series
358; range: 0–2 L/min; Analyt-MTC) calibrated to standard con-
ditions (temperature = 273 K, pressure = 1013 hPa). On the basis of
the known infused mass flow VinfusedCO2 and VinfusedN2, the expected
readouts of the metabolic carts were then calculated as VcalcCO2 and
as VcalcO2, respectively, as given below.
Initial experiments conducted in canopy mode with an artificial
head model showed that steady state conditions were established
within a time period comparable to the in vivo measurements. We
then examined a direct mode. Artificial head and canopy were
omitted and gaseous CO2 and N2 were directly introduced via
the mass-flow regulator into the canopy hose of the metabolic
cart (canopy removed; see Supplemental Scheme 1 under “Sup-
plemental data” in the online issue). In the direct mode, steady
state conditions were reached within a few minutes and the cart
readouts were identical in both modes with either cart. The more
rapid direct mode was therefore applied in the subsequent in vitro
experiments.
In vitro, the experimental conditions required throughout
were as follows: VCO2 and VO2 between 0.1 and 0.5 L/min,
with VCO2:VO2 ratios in the range of 0.6–1.1 and, with the
Vmax Encore, an additional expiratory CO2 (FeCO2) in the
range of 0.6–0.9%. After a 5-min equilibration period, 10-min
data recordings were performed under steady state conditions
with maximal rates of data readout (see above). In these ex-
periments, the mean within-series CVs of the Deltatrac read-
outs were 0.33 6 0.25% (median: 0.28%; range: 0.00–1.82%)
for VCO2 and 0.56 6 0.36% (0.46%; 0.00–1.82%) for VO2.
The respective mean CVs of the Vmax Encore were 0.56 6
0.17% (0.53%; 0.26–0.17%) for VCO2 and 0.88 6 0.31%
(0.82%; 0.27–1.89%) for VO2. Of note, the CVs tended to be
higher at lower gas-flow rates.
Case-related in vitro validation by individual calibration
control evaluation
In a set of IC measurements in vivo, the metabolic cart readouts
of VCO2 and VO2 were obtained as described above and then
765
specifically validated by a subsequent in vitro simulation as follows.
The means of the in vivo rates were calculated. Thereafter,
gaseous CO2 and N2 were infused into the hose of the metabolic
cart and the mass-flow meters adjusted until the mean rates as
observed in an individual subject were exactly met. The resulting
adjustment of the mass-flow regulator (in L/min, standard con-
ditions; temperature = 273 K, pressure = 1013 hPa) was then
taken to calculate the estimate for the true breath gas exchange
rate for the subject (see below).
Data analysis and statistics
In vivo studies
The serial readouts of VCO2 and VO2 (in mL/min) were used
to calculate the RQ
RQ ¼ VCO2 OVO2 ð1Þ
and the REE according to the abbreviated Weir equation (4)
Downloaded from ajcn.nutrition.org by guest on November 30, 2017
REE ¼ ð1:106 3 VCO2 þ 3:941 3 VO2 Þ 3 1:44 in kcal=d ð2Þ;
which is equivalent to
REE ¼ ð1:106 3 VCO2 þ 3:941 3 VO2 Þ 3 6:028 in kJ=d ð3Þ
The data presented are the within-series means of the 20 and 60
data points as obtained by the 20-min recordings with the Del-
tatrac and the Vmax Encore, respectively.
The proportional energy contribution of macronutrient ox-
idation [glucose/carbohydrate oxidation (Gox; %); fat oxida-
tion (Fox; %); protein oxidation (Pox; %)] to REE was
estimated on the basis of the assumption that Pox provided 15%
of total REE (ie, Pox% = 15) in healthy subjects under post-
absorptive conditions (overnight fast .8 h) (26) because uri-
nary nitrogen excretion was not measured in these short-term
experiments.
By using Atwater units of 16.74, 37.24, and 16.74 kJ/g of
substrate (27, 28) as metabolic energy equivalents for carbohydrate,
fat, and protein, respectively, Pox (in g/d) was estimated from REE
(in kJ/d) as
Pox ¼ ð0:15 3 REEÞO16:74 ð4Þ
By using established formulas (5, 9, 24–28), Gox and Fox (in
g/d) were estimated from VCO2 and VO2 (mL/min) as
Gox ¼ ½ð4:55 3 VCO2  3:21 3 VO2 Þ 3 1:44  0:459 3 Pox
ð5Þ
Fox ¼ ½ð1:67 3 VO2 2 1:67 3 VCO2 Þ 3 1:44 2 0:307 3 Pox
ð6Þ
The relative energy contribution of carbohydrate and fat
oxidation (ie, Gox% and Fox% in % REE) to total energy
expenditure is given by
766 SCHADEWALDT ET AL

Goxð%Þ ¼ ðGox 3 16:74=REEÞ 3 100
Foxð%Þ ¼ ðFox 3 37:24=REEÞ 3 100
ð7Þ Ve is known and adjusted by the metabolic cart (= Vadj),
whereas Vi is unknown. The latter is derived by using the
Haldane transformation. It is assumed that FiN2 3 Vi = FeN2 3 Ve.
Then, Vi is calculated by
ð8Þ

Vi ¼ FeN2 3 Ve OFiN2 : ð11Þ

In vitro studies This is equivalent to

Estimates for the expected VCO2 and VO2 in the simulation
experiments in which defined infusion rates of gaseous CO2 and Vi ¼ ½ð1  FeO2  FeCO2 ÞOð1  FiO2  FiCO2 Þ 3 Ve ð12Þ
N2 (VinfusedCO2 and VinfusedN2 in standard liters per minute, nor-
malized to a standard temperature of 08C and pressure of 1013
hPa) were applied to mimic different rates of gas exchange rates Thus, O2 uptake and CO2 output rates are calculated by the
considered the use of the Haldane transformation by the meta- metabolic carts as
bolic carts and were obtained as follows:
O2 uptake (VO2) and CO2 exhalation (VCO2) at a given rate VO2 ¼ FiO2 3 ½ð1  FeO2  FeCO2 ÞOð1  FiO2  FiCO2 Þ
of inspiration (Vi) and expiration (Ve) is calculated by  FeO2 3 Ve ð13Þ

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

and
VO2 ¼ FiO2 3 Vi  FeO2 3 Ve ð9Þ
VCO2 ¼ FeCO2 3 Ve  FiCO2 3 ½ð1  FeO2  FeCO2 ÞOð1  FiO2
 FiCO2 Þ 3 Ve :
VCO2 ¼ FeCO2 3 Ve  FiCO2 3 Vi ð10Þ ð14Þ

Fi and Fe represent the appropriate fractional amount of In the simulation experiments, natural abundance in ambient
O 2 and CO 2 in the inspiratory and expiratory gas mixture, air was assumed and fractional amounts of O2 and CO2 in in-
respectively. spired and expired air are given by the equations

TABLE 2
Comparative Vmax Encore 29n (SensorMedix) and Deltatrac MBM-100 (Datex) indirect calorimetry measurements in
postabsorptive healthy subjects: data readout and derived metabolic variables1
Metabolic cart

Variable Vmax Encore 29n Deltatrac MBM-100 Correlation (R2)2 Difference (P value)3

VCO2 (mL/min) 170 6 384 181 6 32 0.869 3.8 3 1025

Median (range) 162 (122–330) 173 (139–304)
VO2 (mL/min) 208 6 41 211 6 37 0.965 1.4 3 1022
Median (range) 201 (148–370) 203 (151–352)
RQ (ratio) 0.821 6 0.052 0.856 6 0.038 0.344 7.7 3 1028
Median (range) 0.851 (0.710–0.940) (0.860, 0.770–0.980)
REE 0.966 2.6 3 1024
(MJ/d) 6.01 6 1.21 6.16 6 1.09
Median (range) 5.79 (4.29–10.89) (5.79; 4.29–10.89)
(kJ $ h21 $ kg21 LBM) 4.79 6 0.52 4.93 6 0.52
Median (range) 4.74 (4.12–6.31) 4.93 (4.08–6.20)
Substrate oxidation5 (% of REE) 0.141 6.1 3 1025
Gox 37 6 20 50 6 13
Median (range) 40 (2–85) 50 (18–85)
Fox 49 6 20 35 6 13
Median (range) 45 (0–83) 35 (0–67)
1
Forty subjects (cohort 1; see Table 1) underwent immediate subsequent indirect calorimetry measurements with both
metabolic carts in random order. Fox, fat oxidation; Gox, glucose/carbohydrate oxidation; LBM, lean body mass; REE,
resting energy expenditure related to lean body mass; RQ, respiratory quotient; VCO2, carbon dioxide output rate; VO2,
oxygen consumption rate.
2
Coefficient of determination as evaluated by linear regression analysis (least-squares method).
3
As evaluated by Student’s t test.
4
Mean 6 SD (all such values).
5
For estimation, protein oxidation was assumed to be 15% of total REE (see Subjects and Methods).
 METABOLIC MONITOR VALIDATION AND STANDARDIZATION 767

FiO2 ¼ 0:2093 ð15Þ and

VcalcCO2 ¼ VinfusedCO2  0:00024 3 VinfusedN2 : ð20Þ


FeO2 ¼ 0:2093 3 Ve  VinfusedN2  VinfusedCO2 3 ð1OVe Þ
Of note, under the present test conditions, the correction for
ð16Þ influence of infusion of CO2 on VcalcO2 and of N2 on VcalcCO2
is ,,1% and can thus actually be neglected.
and
Results are the within-series means of the 10 and 30 data points
FiCO2 ¼ 0:0004 ð17Þ as obtained by the 10-min recordings with the Deltatrac and
the Vmax Encore, respectively, and RQ data were derived from
the appropriate breath gas exchange variables as in the in vivo

studies.
FeCO2 ¼ VinfusedCO2 þ0:0004 3 Ve  VinfusedN2
Statistical analyses
 VinfusedCO2 3 ð1OVe Þ ð18Þ
Results are presented as means 6 SD (median and range in
parentheses). Linear regression (least-squares method) was ap-
On the basis of on these fractions, estimates of the rates of O2 plied for correlation analysis. Differences between paired or
consumption and CO2 output (Vcalc) due to infusion of gaseous unpaired data sets were assessed by using 2-tailed Student’s
N2 and CO2 were calculated as t test as appropriate. Significant differences required P , 0.05.

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

Inter- and intraindividual CVs (CVi) were calculated from the
VcalcO2 ¼ 0:26463 3 VinfusedN2  0:00016 3 VinfusedCO2 ð19Þ total CV (CVt) and analytic CV (CVa) as

FIGURE 1. Comparison of the Vmax Encore 29n (SensorMedix) and Deltatrac MBM-100 (Datex) breath gas-exchange measurements in healthy subjects.
Postabsorptive adults (n = 40) underwent indirect calorimetry measurements with either cart in random order as detailed in Subjects and Methods (see also
Table 1). Bland-Altman plots of VCO2 (A, C) and VO2 (B, D) data are shown. Mean differences between carts and ranges for 61.96 SD and 95% CIs are
indicated. Abs., absolute; CIlower, lower CI; CIupper, upper CI; DT, Deltratrac MBM-100; Rel., relative; VCO2, carbon dioxide output rate; Vmax, Vmax Encore
29n; VO2, oxygen consumption rate.
768 SCHADEWALDT ET AL
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
for the primary variables VCO2 and VO2 and REE (R2 . 0.86),
CVi ¼ CVt2 2 CVa2 ð21Þ
poor for RQ (R2 , 0.35), and unacceptable for the estimates for
macronutrient oxidation (Gox and Fox, R2 , 0.15).
Variations were essentially due to a distinct variability in
Method comparison was visualized by using Bland-Altman VCO2 and VO2 measurement (see Figure 1). In relation to the
plots. The 95% limits of agreement (CL) were calculated as mean value of both methods, the relative differences of both
CL ¼ mean 6 1:96 3 SD ð22Þ devices (ie, Vmax Encore minus Deltatrac) in VCO2 and VO2
were 27 6 6% (range: 222 to +8%) and 22 6 4% (214 to
+5%), respectively.
The limits of the 95% CI were calculated as These differences propagated differentially into derived meta-

pffiffiffi bolic estimates. With REE, the mean relative difference was small
CI ¼ mean 6 1:96 3 SD þ 1:71 3 tð95%;n1Þ 3 SDO n and amounted to 23 6 4% (213 to +6%). With RQ, the mean
ð23Þ difference was 0.044 6 0.042 (20.15 to +0.04) units (Figure 2).
Macronutrient oxidation estimates were strongly affected. The
and are shown as recommended (29–32). mean relative differences were 247 6 59% (2182 to +122%)
with Gox and 23 6 57% (2200 to +200%) with Fox, and corre-
lation between methods was practically abolished (see Figure 3).
RESULTS

Comparative in vivo assessment of the metabolic carts In vitro evaluation of cart performance characteristics

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

The metabolic carts were compared by performing parallel IC
measurements in 40 fasted subjects under strictly controlled
conditions. A summary of results is presented in Table 2. Sig-
nificant differences between carts were found with all variables
(P , or ,, 0.02). Correlation between carts was only moderate
Toward an explanation of the differences between monitors, in
vitro simulation experiments were performed in parallel with
both carts by using calibrated gas-flow regulators and a pure
mixture of 21% CO2 in N2 to imitate VCO2 and VO2 at a fixed
RQ = 1.00.

FIGURE 2. Comparison of estimates of REE and RQ as calculated from comparative Vmax Encore 29n (SensorMedix) and Deltatrac MBM-100 (Datex)
indirect calorimetry measurements. Results are based on data from postabsorptive adults (n = 40; see Figure 1). Bland-Altman plots of data on REE (A, C) and
RQ (B, D) are shown. Mean differences between the 2 methods and ranges for 61.96 SD and 95% CIs are indicated. Abs., absolute; CIlower, lower CI; CIupper,
upper CI; DT, Deltratrac MBM-100; REE, resting energy expenditure; Rel., relative; RQ, respiratory quotient; Vmax, Vmax Encore 29n.
 METABOLIC MONITOR VALIDATION AND STANDARDIZATION
Accuracy of VCO2 and VO2 readouts
In simulation experiments (n = 9) comprising 20 different flow
rates in series (range: .0.1 to ,0.5 L/min), linear-regression
estimates of slope and intercept between cart readouts and mass-
flow regulator-adjusted flow were 1.0 and within 60.018 L/min,
respectively, and overall R2 was .0.995 with either cart (see
Supplementary Figure 1 under “Supplemental data” in the online
issue). A closer inspection of data showed distinct differences.
Between devices, VCO2 and VO2 were significantly different
(P , or ,, 0.05) in a nonlinear fashion over the whole mea-
suring range and at a flow .0.27 L/min, respectively (see
Supplementary Figure 2 under “Supplemental data” in the on-
line issue).
The deviation of readouts from adjusted values is shown in
Figure 4. With the Deltatrac, differences were significant (P ,
or ,, 0.05) over the whole range and with the Vmax Encore at
flows ,0.22 L/min. Deviations in Deltratrac readouts of VCO2
and VO2 were approximately linear and inverse linear correlated
to the mass-flow regulator-adjusted flow, respectively. A
769
leading to a considerable broadening of confidence limits with
both carts.
Variability and long-term stability
In simulation experiments, the CVs of VCO2 and VO2 read-
outs were ,1% within series and ,3% between series with both
monitors. With the Vmax Encore, within- and between-series
CVs tended to be higher than with the Deltatrac (Table 3).
Long-term stability (10-h period) was excellent with both carts,
although some significant but very low drifts were observed at
higher flow rates (Table 4).
Fractional mean estimates for the impact of analytic variability
on metabolic cart measurements of VCO2 and VO2 in humans
were as follows—within series: ,0.02; between series: ,0.04
and ,0.20 for interindividual and intraindividual assays, re-
spectively (Table 5).
Exploitation of RQ readouts

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

complex nonlinear deviation pattern was found with the Vmax The RQ in simulation experiments was 1.0045 units throughout.
Encore. Of note, interassay variability in readouts was high, The Deltatrac readouts yielded significantly elevated RQ values

FIGURE 3. Comparison of estimated Gox and Fox rates as derived from comparative Vmax Encore 29n (SensorMedix) and Deltatrac MBM-100 (Datex)
indirect calorimetry measurements. Results are based on studies in postabsorptive healthy adults (n = 40; see Figure 1). A, B: Bland-Altman plots of
differences between methods (means and ranges for 61.96 SD and 95% CIs are indicated). C, D: Correlation (linear regression, least-squares method) of
the estimated respective contribution of macronutrient oxidation to total REE. Fasting protein oxidation was assumed to be 0.15 of REE (see Subjects and
Methods for further details). Linear regression equations are as follows—Gox: y = 0.59 (60.23) x + 7.3 (612.1); Fox: y = 0.59 (60.23) x + 27.9 (68.7);
R2 = 0.141, Sy/x = 18.3; dotted lines: 95% prediction intervals. Abs., absolute; CIlower, lower CI; CIupper, upper CI; DT, Deltratrac MBM-100; Fox, fat
oxidation; Gox, glucose/carbohydrate oxidation; REE, resting energy expenditure; Vmax, Vmax Encore 29n.
770 SCHADEWALDT ET AL

over the whole flow range, with a mean increment of +0.090 6
0.026 units (range: 0.025–0.176 units) (n = 180). With the Vmax
Encore, the mean RQ was not significantly different from the
assigned value and amounted to 0.002 6 0.033 units (range:
20.048 to 20.102 units; see Supplementary Figure 3 under
“Supplemental data” in the online issue).
Together, these variabilities point at the physiologically signifi-
cant impact of cart readout errors on derived secondary metabolic
variables.
Impact of cart series
Due to the inherent variability of metabolic monitors, cart
differences might also be suspected in instruments of the same
type. In fact, when series of simulation assays were performed
using 2 Vmax Encore devices of identical construction (ie, Vmax
Encore 29n) in parallel, considerable variability in simultaneously
obtained VCO2 and VO2 readouts was found with but minor ef-
fects on the overall mean values (Supplementary Figure 4 under
“Supplemental data” in the online issue).
VO2 readouts and in derived metabolic variables were compa-
rable to the findings in the first in vivo study, and cart-dependent
differences were confirmed.
By means of ICcE, distinct errors of the individual VCO2 and
VO2 readouts were detected, which were comparable to the
findings in the in vitro simulation studies (see Supplementary
Figure 5 under “Supplemental data” in the online issue). Ac-
cordingly, considerable error was found in the derived estimates
for REE, RQ, Gox, and Fox (see Supplementary Figure 6 under
“Supplemental data” in the online issue).
Adjustment of the individual cart readout for the deviations
detected by the ICcE procedure resulted in a considerably better
agreement of results. Although some minor differences remained,
in particular with the VCO2 values, the correlation between
methods improved significantly. In this data set, the coefficients
of determination R2 for the primary variables, VCO2 and VO2,
were 0.936 and 0.953, respectively, and for the estimates of the
derived variables REE, RQ, Gox, and Fox were 0.922, 0.929,
0.939, and 0.927, respectively.

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

Accomplishment of accurate IC measurements in humans by
individual calibration control evaluation
To reduce interferences of metabolic cart deviations and var-
iability in human studies, the postcalorimetric individual cali-
bration control evaluation (ICcE) procedure was introduced. The
procedure evaluates cart performance for each individual IC
measurement by an appropriate in vitro simulation measurement.
Value of case-related in vitro validation
The relevance of ICcE was evaluated in vivo by performing
comparative IC measurements with Deltratrac and Vmax Encore
devices in postabsorptive subjects (n = 22). A summary of data is
presented in Table 6. Distribution of and variation in VCO2 and
DISCUSSION
Among metabolic monitors, the Deltatrac was supposed to
provide the greatest accuracy and least intercart variability of all
such devices worldwide (ie, a “gold standard” for canopy-based
IC), but it is no longer produced (14, 21, 25). In the present
study in healthy volunteers, we compared the Deltatrac with the
Vmax Encore metabolic monitor. The data were similar to re-
sults reported by Cooper et al (25). The primary data, ie, breath
gas exchange rates, as well as the derived secondary metabolic
results differed significantly between devices, although carefully
designed preparatory measures were imposed on the subjects
under study and strictly controlled experimental conditions were
applied for the measurements. Specifically, mean VCO2 was found

FIGURE 4. Deviation of Deltatrac MBM-100 (Datex) (A, C) and Vmax Encore 29n (SensorMedix) (B, D) metabolic monitor readouts from true values as
evaluated in simulation experiments in vitro. Breath gas exchange was simulated by variable infusion of an appropriate gas mixture (21% CO2 in N2).
Deviation means 6 SEMs as obtained in 9 independent series (20 flow rates each) are shown (see Supplemental Figure 2 under “Supplemental data” in the
online issue). The 95% prediction intervals [PI = mean 6 (SD 3 t(97.5%, n 2 1) 3 O(n + 1/n)] are indicated by dotted lines. Significant deviation from zero
(Student’s t test): Deltatrac at a gas flow ,0.25 L/min throughout, Vmax Encore at VCO2 ,0.35 L/min and VO2 .0.27 L/min (P , or ,, 0.05). VCO2, carbon
dioxide output rate; VO2, oxygen consumption rate.
 METABOLIC MONITOR VALIDATION AND STANDARDIZATION 771
TABLE 3 a generally poor correlation of data and an essentially un-
Variability of metabolic cart measurements in vitro1 acceptable correlation of the estimates of RQ, Gox, and Fox.
Metabolic cart However, because no uniform procedure is available to ensure
accuracy and precision of metabolic carts, virtually no final pre-
Variable Vmax Encore 29n2 Deltatrac MBM-1003 ference can be assigned to either cart. Thus, the causes for the
Within-series CV4 differences remain obscure. We therefore conducted an analysis
VCO2 0.56 6 0.175 0.33 6 0.25 of performance with either cart by in vitro simulation experiments
Median (range) 0.53 (0.26–1.23) 0.28 (0.00–1.82) with the use of pure gases and calibrated gas-flow regulators. The
VO2 0.88 6 0.31 0.56 6 0.37 outcome, however, was somewhat disillusioning. The Deltatrac,
Median (range) 0.82 (0.28–1.89) 0.46 (0.00–1.82) the presumed reference cart, exhibited physiologically significant
Between-series CV4 variation in measurements within series with respect to a rate-
VCO2 3.26 6 0.85 1.03 6 0.22 dependent deficiency of accuracy as well as between series with
Median (range) 3.00 (2.25–4.86) 1.01 (0.67–1.38)
respect to a variable imprecision in repeatability. Similar prob-
VO2 2.91 6 0.89 1.77 6 0.45
Median (range) 2.99 (1.57–4.60) 1.77 (1.06–2.62) lems with accuracy and repeatability were encountered with the
1
Vmax Encore. Interestingly, however, the performance charac-
Data are percentages as evaluated in analyses with a fixed VCO2:VO2 teristics showed distinct differences, eg, with a nearly linear rate
ratio (adjusted by infusion of a calibrated gas mixture of 21% CO2 in N2)
dependence of accuracy with the Deltatrac compared with a
and variable gas-flow rates (between 0.1 and 0.5 L/min adjusted by means of
calibrated mass-flow regulators; see Supplemental Figure 1 under “Supple- nonlinear dependence with the Vmax Encore. Of note, the effect
mental data” in the online issue). Ten-minute recordings were performed under of device-inherent variability on differences in the measurements
was also shown when parallel examinations were performed with

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

steady state conditions by using the maximal rate of data readout (ie, Vmax
Encore, 3 readouts/min; Deltatrac, 1 readout/min). VCO2, carbon dioxide output
rate; VO2, oxygen consumption rate.
2
SensorMedix.
3
Datex.
4
Values are based on successive serial measurements of VCO2 and VO2
at 20 different rates, respectively (see Supplemental Figure 1 under “Sup-
plemental data” in the online issue), with a total of 9 repeats performed on
different working days. Of note, CVs were essentially independent from the
flow rate.
5
Mean 6 SD (all such values).
to be substantially higher with the Deltatrac, with more subtle
differences in VO2 measurements. Accordingly, REE estimates,
which depend primarily on VO2, differed less markedly, but
marked differences were observed with the estimates for RQ,
Gox, and Fox. On the whole, differences in cart readouts led
to a physiologically relevant variability between methods with
2 identical Vmax Encore carts.
Clearly, an inevitable within- and between-series variability in
metabolic carts exists. Thus, deviation of the cart readout from
the true value cannot be predicted reliably. Any given individual
measurement is therefore uncertain to an unknown extent. In
principle, confidence limits for a cart readout can be estimated
from the present experimental data, but these variables are in-
appropriate for evaluation of an individual measurement or repeat
measurements in one subject at different occasions. Variability
within carts of the same series may finally level out in population-
based studies with large numbers of participants but will remain
in studies in which different carts devices are applied. In the latter
case, data may additionally be confounded by device-specific
differences in accuracy.
The foregoing considerations point toward the need for rig-
orous control of intra- and interdevice variability as a necessary

TABLE 4
Long-term stability of metabolic cart measurements in vitro1
Metabolic cart

Variable Vmax Encore 29n2 Deltatrac MBM-1003

VCO2 (mL/h) 20.12 6 0.13IV (20.04 6 0.03) 0.06 6 0.21III,IV (0.009 6 0.063)
VO2 (mL/h) 0.32 6 0.12I,III,IV (0.13 6 0.05) 0.02 6 0.02IV (0.0003 6 0.0837)
RQ (units) 20.0018 6 0.0004I,II,III,IV 0.0008 6 0.0017
REE (kJ/h) 0.28 6 0.10I,III,IV (0.10 6 0.04) 0.14 6 0.40IV (0.0005 6 0.0012)
1
Values are mean (6SD) changes with time of the 4 slope estimates obtained and were evaluated in parallel with both
carts by using a calibrated gas mixture (21% CO2 in N2) at 4 different flow rates [ie, levels I, II, III, and IV at 0.65, 0.95,
1.25, and 1.80 L/min, respectively, which is equivalent to VCO2:VO2 ratios of 141:140, 206:205, 271:270, and 391:389,
respectively]. At each level, time-dependent changes (ie, slope estimates) over a total observation period of 10 h were
evaluated by linear regression analyses on the basis of data from five 20-min recordings. These measurements were equally
distributed over time and were performed under steady state conditions. Time-dependent changes at the different gas flow
rates were in a comparable range. For convenience, means 6 SDs of the relative changes observed at the 4 flow levels
(given in %/h) are shown in parentheses. I–IVSignificant changes with time (linear regression analysis, least squares) at P ,
or ,, 0.01 at gas-flow rates for levels I, II, III, or IV, respectively. REE, resting energy expenditure (as estimated by the
abbreviated Weir equation; see Subjects and Methods); RQ, respiratory quotient; VCO2, carbon dioxide output rate; VO2,
oxygen consumption rate.
2
SensorMedix.
3
Datex.
772 SCHADEWALDT ET AL
TABLE 5
Impact of analytic variability on metabolic cart measurements in humans1
Mean analytic contribution of cart3

Variable Total variation2 Vmax Encore 29n4 Deltatrac MBM-1005

CV in % fraction of total mean CV/upper 95% CL

Within-series CV
VCO2 6.2 6 2.4 ,0.01/0.01 ,0.01/0.01
VO2 4.9 6 1.9 ,0.02/,0.06 ,0.01/,0.05
Between-series CV
Interindividual assays6
VCO2 12.5 ,0.04/0.12 ,0.01/,0.01
VO2 10.8 ,0.04/0.10 ,0.02/0.04
Intraindividual assays7
VCO2 7.2 6 2.9 0.11/0.27 0.01/0.03
VO2 4.9 6 1.9 0.20/0.68 0.07/0.16
1
As evaluated in 20-min standard indirect calorimetry recordings in postabsorptive healthy subjects (cohort 1; see
Table 1 and footnote 1 of Table 2). For comparative purposes, metabolic data were normalized to unit weight of lean body
mass. CL, confidence limit; VCO2, carbon dioxide output rate; VO2, oxygen consumption rate.
2
Means 6 SDs of series with 40 different participants (see footnote 1 of Table 2).
3

Downloaded from ajcn.nutrition.org by guest on November 30, 2017

Estimates were based on the analytic imprecision data compiled in Table 4; the maximum is indicated as the 95% CL.
4
SensorMedix.
5
Datex.
6
Variability of measurements in 40 different subjects (see above).
7
Means 6 SDs from 6 healthy subjects undergoing a series of 4 different intraindividual measurements within a mean
period of 3 mo (see Subjects and Methods).

requirement for a reasonable comparability of data from canopy-
based IC studies in humans. The possible effect of device-specific
variability on the primary readout of metabolic monitors and the
impact on derived metabolic variables have not been appropri-
ately reviewed in the literature (8, 9, 13–17, 20–25).
To overcome these problems, we have now developed a new
postcalorimetric procedure, ie, the ICcE, for quantitative eval-
uation of each individual measurement. To avoid any change in
performance or setup of the instrument, this evaluation procedure
has to be performed in series directly after an IC assessment. The

TABLE 6
Accuracy of indirect calorimetry measurements and derived metabolic variables in healthy subjects1
Vmax Encore 29n2 Deltatrac MBM-1003

Variable Readout Calibrated4 Deviation5 Readout Calibrated4 Deviation5

% %
VCO2 (mL/min) 154 6 21 170 6 22I 29 6 2 171 6 21II 166 6 21I 361
Median (range) 157 (119–197) 170 (138–218) (27/214) 168 (138–214) 163 (134–215) (21/+5)
VO2 (mL/min) 189 6 26 202 6 26I 26 6 2 192 6 24 (NS) 199 6 24I 23 6 2
Median (range) 188 (144–257) 200 (159–270) (24/210) 186 (152–248) 195 (161–259) (21/26)
RQ 0.815 6 0.049 0.841 6 0.046I NA 0.889 6 0.036II 0.834 6 0.039I NA
Median (range) 0.808 (0.726–0.903) 0.832 (0.765–0.937) 0.882 (0.883–0.983) 0.826 (0.788–0.915)
REE (MJ/d) 5.52 6 0.74 5.92 6 0.75I 27 6 2 5.70 6 0.69III 5.83 6 0.69I 22 6 1
Median (range) 5.45 (4.21–7.40) 5.82 (4.70–7.83) (25/211) 5.53 (4.53–7.30) 5.70 (4.72–7.52) (60/24)
Gox (g/h) 4.6 6 2.5 6.4 6 2.6I 232 6 19 8.6 6 2.0II 6.0 6 2.2I 52 6 27
Median (range) 3.8 (0.1–10.3) 5.8 (2.2–13.0) (22/294) 7.8 (6.2–12.7) 4.8 (3.4–12.2) (24/+121)
Fox (g/h) 2.9 6 1.1 2.6 6 1.1I 219 6 24 1.5 6 0.8II 2.7 6 0.9I 246 6 18
Median (range) 3.0 (1.1–5.4) 2.7 (0.6–4.8) (215/+99) 1.5 (0.0–3.0) 2.8 (1.1–4.5) (222/2100)
1
Twenty-two postabsorptive subjects without stated acute or chronic illness received immediate subsequent indirect calorimetry measurements with both
metabolic carts in random order. Results are presented as means 6 SDs; median and range in parentheses). I–IIISignificant differences (Student’s t test)—I: vs
appropriate cart readout, P , or ,, 2 3 1026; II: between carts, P , or ,, 3 3 1029; III: between carts, P , 0.005. Fox, fat oxidation (protein oxidation
was assumed to account for 15% of total REE; see Subjects and Methods for details); Gox, glucose/carbohydrate oxidation; NA, not applicable; NS, no
difference between carts; REE, resting energy expenditure; RQ, respiratory quotient; VCO2, carbon dioxide output rate; VO2, oxygen consumption rate.
2
SensorMedix.
3
Datex.
4
Cart readouts were calibrated for each individual in a subsequent simulation experiment (ie, mass-flow regulator infusion of pure gaseous CO2 and N2 as
detailed in Subjects and Methods).
5
Means 6 SDs of deviation of readouts as related to calibrated estimates (with the minimal/maximal deviation in parentheses).
 METABOLIC MONITOR VALIDATION AND STANDARDIZATION
determination of the appropriate deviations in the cart readout
from the true in vivo values is achieved by mimicking the readout
of the in vivo measurement. This is done in vitro by simulating
exactly the in vivo rate readout by means of mass-flow regulator-
controlled infusion of pure gaseous CO2 and N2 via appropriate
tubes into the hose of the flow-through system of the metabolic
cart. In detail, the flow of gaseous CO2 is upregulated via the
regulator until the cart readout exactly matches the mean VCO2
that was observed in the previous in vivo measurement. The best
estimate for the true value for VCO2 in the patient is now shown
directly on the CO2-flow regulator. Similarly, the N2 flow (VN2)
is upregulated to yield a cart readout that matches the mean VO2
in vivo. The best estimate for O2 consumption rate in vivo is
calculated from VN2 as VO2 = VN2 3 0.2646. The factor accounts
for 1) the fact that N2 infused in the system replaced not only
oxygen but whole room air and 2) the numeric contribution of
the Haldane transformation on the metabolic cart readout.
According to our experience, the mean in vivo values can be
reproduced in vitro on the metabolic cart with an agreement better
than 62 mL/min in w15 min. Of note, for reliable and accurate
performance of the ICcE, it is mandatory that the gas-flow regu-
lators are carefully maintained and (re)calibrated according to the
manufacturers’ recommendations and that the standard temperature
pressure dry conditions applied for regulator calibration are iden-
tical to those that are used in the metabolic cart under evaluation.
The applicability and usefulness of the ICcE procedure were
verified in another Deltatrac compared with Vmax Encore IC
study in vivo. In this human study, the agreement of the metabolic
variable estimates for REE, RQ, and the macronutrient oxidation
rates (Gox and Fox) increased significantly, and only minor
differences between the 2 carts remained.
In conclusion, the ICcE procedure established in this study
appears to be suitable to control for variability in devices based
on flow-through respiratory measurements in canopy mode, in-
dependent of device, time, and place.
We gratefully acknowledge the skillful help of Waldemar Merck with con-
struction and setup of the homemade ICcE unit.
The authors’ responsibilities were as follows—PS: design and supervision
of experiments and primary responsibility for the final content; PS, BN, and
JK: collection and evaluation of data; PS, KS, and JK: data analysis and
interpretation; and PS, BN, KS, JK, and MR: writing of the manuscript. All
authors contributed substantially to the concept, data collection and analysis,
or preparation of manuscript and read and approved the final manuscript.
None of the authors had a conflict of interest to declare.
REFERENCES
1. Von Pettenkofer M. Über einen neuen Respirationsapparat. [On a new
device for respiration analyses.] Munich, Germany: K Akademie in
Commission G Franz, 1861 (in German).
2. Atawer WO, Rosa EB. Description of neo respiration calorimeter and
experiments on the conservation of energy in the human body. Wash-
ington, DC: US Department of Agriculture, Office of Experiment
Stations, Government Printing Office, 1899: Bulletin 63.
3. Frankenfield DC. On heat, respiration, and calorimetry. Nutrition 2010;
26:939–50.
4. Weir JB. New methods for calculating metabolic rate with special
reference to protein metabolism. J Physiol 1949;109:1–9.
5. Jéquier E, Acheson K, Schutz Y. Assessment of energy expenditure and
fuel utilization in man. Annu Rev Nutr 1987;7:187–208.
6. Elia M, Livesey G. Theory and validity of indirect calorimetry during
net lipid synthesis. Am J Clin Nutr 1988;47:591–607.
7. Ferrannini E. The theoretical bases of indirect calorimetry: a review.
Metabolism 1988;37:287–301.
773
8. Livesey G, Elia M. Estimation of energy expenditure, net carbohydrate
utilization, and net fat oxidation and synthesis by indirect calorimetry:
evaluation of errors with special reference to the detailed composition
of fuels. Am J Clin Nutr 1988;47:608–28.
9. Simonson DC, DeFronzo RA. Indirect calorimetry: methodological
and interpretative problems. Am J Physiol 1990;258:E399–412.
10. Schutz Y. On problems of calculating energy expenditure and substrate uti-
lization from respiratory exchange data. Z Ernahrungswiss 1997;36:255–62.
11. Frankenfield D, Roth-Yousey L, Compher C. Comparison of predictive
equations for resting metabolic rate in healthy nonobese and obese
adults: a systematic review. J Am Diet Assoc 2005;105:775–89.
12. Henry CJ. Basal metabolic rate studies in humans: measurement and
development of new equations. Public Health Nutr 2005;8:1133–52.
13. da Rocha EEM, Alves VGF, Fonsenca RBV. Indirect calorimetry:
methodology, instruments and clinical application. Curr Opin Clin Nutr
Metab Care 2006;9:247–56.
14. Schoeller DA. Making indirect calorimetry a gold standard for pre-
dicting energy requirements for institutionalized patients. J Am Diet
Assoc 2007;107:390–2.
15. Melanson EL, Ingebrigtsen JP, Bergouignan A, Ohkawara K, Kohrt
WM, Lighton JR. A new approach for flow-through respirometry
measurements in humans. Am J Physiol Regul Integr Comp Physiol
2010;298:R1571–9.
16. Compher C, Frankenfield D, Keim N, Roth-Yousey L; Evidence
Downloaded from ajcn.nutrition.org by guest on November 30, 2017
Analysis Working Group. Best practice methods to apply to mea-
surement of resting metabolic rate in adults: a systematic review. J Am
Diet Assoc 2006;106:881–903.
17. Haugen HA, Chan LN, Li F. Indirect calorimetry: a practical guide for
clinicians. Nutr Clin Pract 2007;22:377–88.
18. Perseghin G. Pathogenesis of obesity and diabetes mellitus: insights
provided by indirect calorimetry in humans. Acta Diabetol 2001;38:7–21.
19. Kaiyala KJ, Schwartz MW. Toward a more complete (and less con-
troversial) understanding of energy expenditure and its role in obesity
pathogenesis. Diabetes 2011;60:17–23.
20. Pennock BE, Donahoe M. Indirect calorimetry with a hood: flow
requirements, accuracy, and minute ventilation measurement. J Appl
Physiol 1993;74:485–91.
21. Littlewood RA, White MS, Bell KL, Davies PS, Cleghorn GJ, Grote R.
Comparison of the Cosmed K4 b(2) and the Deltatrac II metabolic cart
in measuring resting energy expenditure in adults. Clin Nutr 2002;21:
491–7.
22. Foss Ø, Hallén J. Validity and stability of a computerized metabolic
system with mixing chamber. Int J Sports Med 2005;26:569–75.
23. Roffey DM, Byrne NM, Hills AP. Day-to-day variance in measurement
of resting metabolic rate using ventilated-hood and mouthpiece and
nose-clip indirect calorimetry systems. JPEN J Parenter Enteral Nutr
2006;30:426–32.
24. Haman F, Péronnet F, Kenny GP, Massicotte D, Lavoie C, Weber JM.
Partitioning oxidative fuels during cold exposure in humans: muscle
glycogen becomes dominant as shivering intensifies. J Physiol 2005;
566:247–56.
25. Cooper JA, Watras AC, O’Brian JM, Luke A, Dobratz JR, Earthman CP,
Schoeller DA. Assessing validity and reliability of resting metabolic rate
in six gas analysis systems. J Am Diet Assoc 2009;109:128–32.
26. Tentolouris N, Pavlatos S, Kokkinos A, Perrea D, Pagoni S, Katsi-
lambros N. Diet-induced thermogenesis and substrate oxidation are not
different between lean and obese women after two different isocaloric
meals, one rich in protein and one rich in fat. Metabolism 2008;57:
313–20.
27. Buchholz AC, Schoeller DA. Is a calorie a calorie? Am J Clin Nutr
2004;79(suppl):899S–906S.
28. FAO Corporate Document Repository. Chapter 3: calculation of the
energy content of food energy conversion factors. Available from:
http://www.fao.org/DOCREP/006/Y5022E/y5022e04.htm (cited Octo-
ber 2011).
29. Bland JM, Altman DG. Statistical methods for assessing agreement
between two methods of clinical measurement. Lancet 1986;1:307–10.
30. Bland JM, Altman DG. Measuring agreement in method comparison
studies. Stat Methods Med Res 1999;8:135–60.
31. Hamilton C, Stamey J. Using Bland-Altman to assess agreement
between two medical devices—don’t forget the confidence intervals!
J Clin Monit Comput 2007;21:331–3.
32. Hanneman SK. Design, analysis, and interpretation of method-comparison
studies. AACN Adv Crit Care 2008;19:223–34.