Journal of Parenteral and Enteral

Nutrition http://pen.sagepub.com/

Validation of 2 approaches to predicting resting metabolic rate in critically ill patients

D Frankenfield, JS Smith and RN Cooney
JPEN J Parenter Enteral Nutr 2004 28: 259
DOI: 10.1177/0148607104028004259

The online version of this article can be found at:

http://pen.sagepub.com/content/28/4/259

Published by:

http://www.sagepublications.com

On behalf of:

The American Society for Parenteral & Enteral Nutrition

Additional services and information for Journal of Parenteral and Enteral Nutrition can be found at:

Email Alerts: http://pen.sagepub.com/cgi/alerts

Subscriptions: http://pen.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Jul 1, 2004

What is This?

Downloaded from pen.sagepub.com at UQ Library on October 12, 2014

0148-6071/04/2804-0259$03.00/0
JOURNAL OF PARENTERAL AND ENTERAL NUTRITION
Copyright © 2004 by the American Society for Parenteral and Enteral Nutrition
Original Communications
The following article is one of two articles offered for continuing education credit in this
issue. Please see page 285 for details.
Validation of 2 Approaches to Predicting Resting Metabolic Rate in
Critically Ill Patients
David Frankenfield, MS, RD*; J. Stanley Smith, MD†; and Robert N. Cooney, MD†
From the *Department of Clinical Nutrition and †Division of Trauma, Department of Surgery, College of Medicine, Pennsylvania State University,
Milton S. Hershey Medical Center, Hershey, Pennsylvania
ABSTRACT. Background: Indirect calorimetry is the crite-
rion method for determining resting metabolic rate for nutri-
tion support in critically ill patients. However, calculation
equations are more commonly used. In the current study we
tested the validity of 2 such calculation systems. Methods:
Indirect calorimetry was performed with an open-circuit
device in mechanically ventilated surgical, trauma, and med-
ical patients at rest. Feedings were not stopped for the mea-
surements. Two predictive equations by Ireton-Jones and 3
versions of a multivariate equation developed at our institu-
tion (referred to as Penn State equations) were then used to
estimate resting metabolic rate. These estimates were com-
pared on a percentage basis with the measured value of
resting metabolic rate. Estimated resting metabolic rate
within 10% of measured was considered accurate, whereas
estimations ⬎15% different from measured were considered
Although metabolic rate can be measured using indi-
rect calorimetry in most critically ill patients, the
equipment is expensive, and therefore most nutrition
support patients receive feedings according to esti-
mated metabolic rate. Methods of prediction can be
divided into 2 types. In the first, healthy resting metab-
olism is calculated and then adjusted by a stress factor
according to injury type or severity. The second
approach is to calculate metabolic rate from a regres-
sion equation that incorporates healthy resting metab-
olism (or its determinants) and clinical markers of
illness (such as body temperature and minute ventila-
tion) as dependent variables. Numerous such equa-
tions have been published,1– 6 but validation studies of
these equations are few.3,7,8
In the current study, we undertook a validation of
one of the more well-known calculation equations3,4
and a regression equation developed at our own insti-
tution from previous indirect calorimeter studies6 to
determine how reliably resting metabolic rate can be
Received for publication June 27, 2003.
Accepted for publication February 26, 2004.
Correspondence: David Frankenfield, MS, RD, Department of Clin-
ical Nutrition, H-1307, Penn State’s Milton S. Hershey Medical
Center, PO Box 850, Mail Code H124, Hershey, PA 17033. Electronic
mail may be sent to dfrankenfield@psu.edu.
259
Downloaded from pen.sagepub.com at UQ Library on October 12, 2014
Vol. 28, No. 4
Printed in U.S.A.
large errors. Results: Forty-seven subjects were measured. A
larger percentage of subjects were estimated accurately by
the Penn State equations (72% in the best equation) than by
the Ireton-Jones equations (60% in the best equation; not
significant). The incidence of errors ⬎15% of measured was
significantly lower in the Penn State equation (11% of sub-
jects) compared with the Ireton-Jones equation (32% of sub-
jects) (p ⬍ .05). Conclusions: The Penn State equation for
resting metabolic rate in mechanically ventilated intensive
care patients receiving nutrition support appears to be a
valid clinical tool for determining energy goals in the absence
of or as a supplement to indirect calorimetry. The Ireton-
Jones equation performed less well, especially in that a
higher number of large errors occurred. ( Journal of Parenteral
and Enteral Nutrition 28:259 –264, 2004)
calculated. Two questions were posed: (1) Which equa-
tion resulted in the highest percentage of accurate
estimations (wherein calculated metabolic rate was
ⱕ10% different from measured); and (2) Which equa-
tion had the smallest number of large errors (wherein
calculated metabolic rate was ⬎15% different from
measured)? Special attention was paid to performance
of these equations in young vs elderly and nonobese vs
obese subjects.
MATERIALS AND METHODS
The data used in the current study were compiled
during the course of routine clinical care in the surgical
and medical intensive care units of a tertiary medical
center that includes a Level I trauma center (data
collected between 2000 and 2003). Our Human Volun-
teer Protection Office approved release of these data in
an anonymous form. Measurements of resting meta-
bolic rate, ordered by critical care teams, were
obtained, along with supporting, clinically available
data (eg, body temperature, minute ventilation, admis-
sion height and weight, history of hospitalization). All
subjects were mechanically ventilated.
The indirect calorimeter used in this study was the
Deltatrac 101 Metabolic Monitor (open-circuit calorim-
eter; Sensormedics, Yorba Linda, CA).9 All tests were
260 FRANKENFIELD ET AL
performed by one of the investigators with experience
performing indirect calorimetry (DCF). The device was
calibrated before each test. Studies were performed if
the patient was at rest, there were no air leaks, and the
fraction of inspired oxygen was 60% or less. Rest was
defined as a period of time where the subject was not
moving and had not been significantly disturbed (eg,
physical therapy, bathed, gotten out of bed) for the
previous 20 minutes. Per clinical protocol at our insti-
tution, continuous feeding methods were used, and
these feedings were not stopped before undertaking
calorimeter measurements. Individual tests were
accepted for inclusion in the current study if the coef-
ficient of variation for oxygen consumption and carbon
dioxide production on a 30-minute test was ⬍10%, or if
this standard was not met, a 5% coefficient of variation
on a 5-minute test was accepted.10 The first 5 minutes
of data were automatically discarded in all tests.
Upon completion of each study, data were entered
into a computer database (Minitab Release 12;
Minitab, Inc, State College, PA). Standard resting met-
abolic rates for health were calculated using the Harris
Benedict and Mifflin St. Jeor equations. Predicted rest-
ing metabolic rates for illness were then calculated
using several multivariate predictive equations. The
first 2 equations were those of Ireton-Jones, both as
originally published in 19923 and as amended in 1997.4
The original equation3 was as follows:
RMR 共kcal/d兲 ⫽ wt 共5兲 ⫺ age 共10兲 ⫹ male 共281兲
⫹ trauma 共292兲 ⫹ burn 共851兲 ⫹ 1925 (1)
The amended 1997 equation4 was:
RMR 共kcal/d兲 ⫽ wt 共5兲 ⫺ age 共11兲 ⫹ male 共244兲
⫹ trauma 共239兲 ⫹ burn 共804兲 ⫹ 1784 (2)
For both equations, RMR was resting metabolic rate,
wt was weight in kilograms, age was in years, and
male, trauma, and burn were 1 for yes and 0 for no.
The next set of equations consisted of an equation
developed at our institution from earlier data6 and 2
permutations thereof. As originally described in 1998,6
the equation (Penn State 1998) was:
RMR 共kcal/d兲 ⫽ Harris-Benedict 共1.1兲 ⫹ VE 共32兲
⫹ TMax 共140兲 ⫺ 5340 (3)
In this equation as originally described, Harris-Bene-
dict was healthy resting metabolic rate calculated from
the Harris-Benedict equations11 using actual weight in
nonobese patients and adjusted body weight in obese
patients (obesity was defined as a body weight ⬎125%
of ideal; adjusted body weight was 25% of the actual
less the ideal weight added to the ideal weight). VE was
minute ventilation in liters/minute (read from the ven-
tilator, not the calorimeter), and TMax was maximum
body temperature (degrees centigrade) in the previous
24 hours read from clinical flow sheets. Recent
research from our institution13 suggested that the
Mifflin St. Jeor equation for resting metabolic rate in
healthy people14 was more accurate than that of Harris
and Benedict11 and that use of obesity-adjusted body
Downloaded from pen.sagepub.com at UQ Library on October 12, 2014
Vol. 28, No. 4
weight in the Harris-Benedict equation was not appro-
priate (leading to routine underestimation of metabolic
rate in healthy obese people). We therefore recomputed
our regression equation from the original database6
using actual body weight in all Harris-Benedict calcu-
lations (Penn State 2003a), and again substituting the
Mifflin St. Jeor equation for the Harris-Benedict equa-
tion (Penn State 2003b):
RMR 共kcal/d兲 ⫽ Harris-Benedict 共0.85兲 ⫹ VE 共33兲
⫹ TMax 共175兲 ⫺ 6433 (4)
R 2 ⫽ .67
RMR 共kcal/d兲 ⫽ Mifflin 共0.96兲 ⫹ VE 共31兲
⫹ TMax 共167兲 ⫺ 6212 (5)
R 2 ⫽ .69
For all equations, the body weight used in the calcu-
lation was current body weight if current weight was
less than or equal to admission weight, and admission
body weight if current body weight was greater than
admission (implying excess body water).
Bias and precision of each equation were evaluated
after the methods of Sheiner and Beal.15 An equation
was considered unbiased (not prone to underestima-
tion or overestimation) if the 95% confidence interval of
the mean difference between predicted and measured
resting metabolic rate included zero. Precision was
defined as a 95% confidence interval for absolute dif-
ference between measured and predicted resting met-
abolic rate ⱕ10% of the measured resting metabolic
rate.
On a more clinical level, each calculation was eval-
uated against measured resting metabolic rate by tak-
ing the difference (calculated-measured) as a percent-
age of measured resting metabolic rate. A calculated
value ⱕ10% of measured metabolic rate (rounded to
the nearest whole number) was considered accurate,
whereas a value ⬎15% of measured was considered a
large error. Differences in accuracy rates among equa-
tions were analyzed with analysis of proportions
(2-tailed test; Minitab Release 12.21; Minitab Corp).
RESULTS
Forty-seven subjects were measured (28 men, 19
women). Thirty percent of studies were performed on
trauma patients, 57% on surgical patients, and 13% on
medical patients. There were no burn patients in the
study sample. Thirty-eight percent of subjects were
obese and 6% were underweight. Forty-three percent of
the subjects were at least 65 years old, with a mean age
of 77 ⫾ 7 years and a range of 65 to 87 years. The
remaining subjects had a mean age of 44 ⫾ 15 years
(range 16 to 64 years). Among the elderly subjects, 30%
were obese, whereas in the younger subjects, incidence
of obesity was 44% (not significant). The mean day of
hospitalization at the time of study was 19.8 ⫾ 18.7,
with a range of 2 to 105 days and a median of 15 days.
Fifty-eight percent of subjects were febrile within the
July–August 2004 TWO APPROACHES TO PREDICTING RESTING METABOLIC RATE 261

TABLE I often than did the revised 1997 Ireton-Jones equation.4

Demographic data for 28 men and 19 women entered into study
Variable Mean ⫾ SD
Height, cm 170 ⫾ 9
Body weight, kg 82 ⫾ 26
Ideal body weight, kg 65 ⫾ 11
Percent ideal body weight 127 ⫾ 39
Body mass index, kg/m2 28 ⫾ 9
Age, y 58 ⫾ 20
Oxygen consumption, mL/min 287 ⫾ 70
Carbon dioxide production, mL/min 252 ⫾ 62
Respiratory quotient 0.88 ⫾ 0.11
Resting metabolic rate, kcal/day 2013 ⫾ 479
Maximum body temperature, °C 38.1 ⫾ 0.7
Minute ventilation, L/min 12.4 ⫾ 3.8
24 hours before study. Other descriptive data for the
subjects are located in Table I.
All equations were linearly associated with mea-
sured resting metabolic rate. The unmodified Harris-
Benedict and Mifflin St. Jeor equations correlated with
resting metabolic rate with an R2 of 0.69. The Penn
State equations, which use the Harris-Benedict equa-
tion, increased the R2 to 0.81. The Ireton-Jones equa-
tion correlations were R2 0.57 (1992 equation)3 and
0.54 (1997 equation)4 (p ⬍ .05 for all).
Mean differences and mean absolute differences
between each calculation and measured resting meta-
bolic rate are shown in Table II. Mean differences were
low for all equations except Ireton-Jones 19974 equa-
tions. Absolute differences were lowest for the Penn
State equations (p ⬍ .05). As with mean difference, the
mean absolute difference for the original 1992 form of
the Ireton-Jones equation3 was lower than that for the
revised 1997 equation.4 Bias and precision were calcu-
lated from differences and absolute differences
between measured and calculated resting metabolic
rate (Table II). The 1998 and 2003a Penn State equa-
tions were unbiased and precise (Table II). The Ireton-
Jones (1992)3 equation was unbiased but not precise
(95% confidence interval for absolute difference was
10% to 16% of measured resting metabolic rate),
whereas the 1997 Ireton-Jones equation4 was neither
unbiased nor precise (Table II).
Table III shows the distribution of errors for each
equation for all subjects. The 1992 Ireton-Jones equa-
tion3 and all of the Penn State equations accurately
predicted resting metabolic rate significantly more
The 1997 Ireton-Jones equation was particularly prone
Range to underestimating measured resting metabolic rate.
147–188 The percentage of large errors (ie, errors in excess of
45–177 15% of measured) was significantly lower in the Penn
43–86 State equations than in the Ireton-Jones equations.
84–288
19–61 Division of the study sample by obesity status (Table
16–87 III) showed that the Ireton-Jones 1992 equation3 was
189–472 accurate significantly more often in the obese than in
161–396 the nonobese subjects. The Penn State 1998 equation
0.67–1.17
1360–3260 accuracy rate and large error rate was similar in non-
35.7–39.3 obese and obese subjects. The other two Penn State
6.6–21.1 equations had higher accuracy rates and lower large
error rates in nonobese compared with obese subjects.
Table IV shows the distribution of errors among
young (⬍65 years old) and elderly (at least 65 years
old) subjects. The Ireton-Jones equations tended to be
accurate more often in young than in elderly subjects
and also tended toward more large errors in the elderly
than in the young. The Penn State equations on the
other hand tended to accurately predict resting meta-
bolic rate more often in elderly than young subjects,
with somewhat lower incidence of large errors in the
elderly. The accuracy rate in elderly subjects using the
Ireton-Jones 1997 equation was significantly lower
than for all other equations. Large errors were statis-
tically more common using the Ireton-Jones 1997 equa-
tion than either Penn State 2003a or 2003b equations
in young subjects. The 2003a Penn State equation sig-
nificantly reduced large errors in the elderly compared
with both Ireton-Jones equations, whereas the 1998
and 2003b Penn State equations significantly reduced
large errors compared with Ireton-Jones 1997 and
tended to do the same compared with the 1992 Ireton-
Jones equation.3,4
A further division of the sample by age and obesity is
shown in Table V. Only Ireton-Jones 1992 and Penn
State 2003a equations are shown, as they were the
most accurate forms of each equation set. The n-values
within each subdivision were small, so these data can
only be considered trends. The Ireton-Jones 1992 equa-
tion was accurate more often than the Penn State
2003a equation in young obese subjects, whereas the
Penn State 2003a equation was accurate more often
than Ireton-Jones in young nonobese, elderly non-
obese, and elderly obese subjects. Similarly, in terms of
large errors, the Ireton-Jones equation outperformed

TABLE II
Bias and precision parameters for each predictive equation*
Precision, 95% CI for
Bias, 95% CI for Absolute difference, Absolute difference,
Equation Difference, kcal absolute difference,
difference, kcal kcal % of RMR
% of RMR

Ireton-Jones (1992)3 3 ⫾ 316 (⫺814 to 760) ⫺90 to 96 257 ⫾ 197 13 ⫾ 10 10–16

Ireton-Jones (1997)4 ⫺237 ⫾ 326a (⫺1051 to 487) ⫺333 to ⫺142 316 ⫾ 248 15 ⫾ 10 12–18
Penn State (1998) 4 ⫾ 208 (⫺606 to 439) ⫺57 to 65 156 ⫾ 136b 8 ⫾ 6b 6–9
Penn State (2003a) ⫺15 ⫾ 211 (⫺433 to 468) ⫺78 to 46 166 ⫾ 128b 8 ⫾ 6b 6–10
Penn State (2003b) ⫺13 ⫾ 226 (⫺408 to 503) ⫺79 to 54 183 ⫾ 131b 9 ⫾ 6b 7–11

CI, confidence interval; RMR, resting metabolic rate.

*An equation is considered unbiased if the 95% CI for the difference between predicted and measured RMR includes zero. The equation is
considered precise if the 95% CI for absolute difference as a percentage of measured RMR is ⱕ10%. ap ⬍ .05 vs all other equations; bp ⬍ .05
vs Ireton-Jones 1992 and 1997.

Downloaded from pen.sagepub.com at UQ Library on October 12, 2014

262 FRANKENFIELD ET AL Vol. 28, No. 4

TABLE III
Percentage of all subjects and subjects subdivided by obesity status whose predicted metabolic rate was within 10% or 15% of measured resting metabolic
rate
Percent of predictions within 10% of measured Percent of predictions ⬎ 15% different from measured
All (n ⫽ 47) Nonobese (n ⫽ 29) Obese (n ⫽ 18) All (n ⫽ 47) Nonobese (n ⫽ 29) Obese (n ⫽ 18)

Ireton-Jones (1992)3 60a 52 72d 32 34 28

Ireton-Jones (1997)4 36 41 28 40 34 50
Penn State (1998) 68a 69b 67d 15a 14c 17d
Penn State (2003a) 72a 79c 61d 11e 0c 28d
Penn State (2003b) 57a 66 44 13e 7c 22d
a
p ⬍ .05 vs Ireton-Jones 1997, all subjects.
b
p ⬍ .05 vs Ireton-Jones 1997, nonobese group.
c
p ⬍ .05 vs Ireton-Jones 1992 and 1997, nonobese group.
d
p ⬍ .05 vs Ireton-Jones 1997, obese group.
e
p ⬍ .05 vs Ireton-Jones 1992 and 1997, all subjects.

the Penn State equation in young obese subjects, but in
all other categories the Penn State equations had a
lower rate of large errors.3
DISCUSSION
Prediction of metabolic rate is a clinical necessity,
even though measurements are more accurate. This is
so because the calorimetry equipment for measure-
ment is expensive, and the procedure can be time-
consuming. Therefore, many institutions do not use
measurement of resting metabolic rate during nutri-
tion assessment and nutrition support. Given the lim-
itation on calorimetry use, there is a need to develop as
accurate and safe predictive equations as possible. No
predictive equation will accurately predict resting met-
abolic rate in every patient. The question is which
equation will predict resting metabolic rate accurately
in the most patients with the smallest number of large
errors.
There are 2 approaches to calculating resting meta-
bolic rate in sick patients. The first involves calculating
healthy resting metabolic rate and multiplying by a
stress factor for the patient’s diagnosis (eg, sepsis,
trauma, burns, surgery). The earliest calculation sys-
tems are of this type,2 and this approach is probably
the one that clinically is most often used.1 The second
method is to use a multivariate regression equation
that includes healthy resting metabolic rate or param-
eters associated with resting metabolic rate plus clin-
ical variables that relate to the degree of inflammatory
response independent of the patient’s diagnosis. The
Ireton-Jones equations are in regression form, but the
stress modifiers are categorical (ie, trauma yes/no,
burn yes/no) and therefore act more like stress factors
than ␤-coefficients. The Penn State equations are true
examples of the multivariate regression method in that
they use noncategorical dependent variables (body
temperature, minute ventilation) that correlate with
the degree of hypermetabolism and that can change
from day to day as hypermetabolism fluctuates.
The 1992 Ireton-Jones equation3 was developed from
a sample of 65 ventilator-dependent intensive care unit
patients, 52% of whom were burn patients. The coeffi-
cient of determination (R2) for this equation against
the measured data were 0.43 (ie, 43% of the variation
in the measured metabolic rate was explained by the
equation). Ireton-Jones validated this equation in 36
other ventilator-dependent intensive care unit
patients, finding no statistically significant difference
between measured and predicted resting metabolic
rate. The mean difference was 160 ⫾ 138 kcal/day (note
that this is a mean difference and not a mean absolute
difference, which almost certainly would have been
higher). In 1997, Ireton-Jones published corrections in
the equation, changing the ␤-coefficients but not the
dependent variables.4 In a validation study in 1999,
Flancbaum et al7 found a low coefficient of determina-
tion for the original 1992 Ireton-Jones equation3
against measured resting metabolic rate (R2 ⫽ .07) and
noted that the equation underestimated resting meta-
bolic rate in 89% of the 36 subjects measured. Mean
absolute difference between calculated and measured

TABLE IV
Percentage of subjects subdivided by age group whose predicted metabolic rate was within 10% or ⬎15% of measured resting metabolic rate
Percent of predictions within 10% of measured Percent of predictions ⬎ 15% different from measured
Young (n ⫽ 27) Elderly (n ⫽ 20) Young (n ⫽ 27) Elderly (n ⫽ 20)

Ireton-Jones (1992)3 70 45 26 40
Ireton-Jones (1997)4 48 15a,b 33c 50
Penn State (1998) 63 75 19 10d
Penn State (2003a) 63 85 15 5d,e
Penn State (2003b) 52 65 11 15d
a
p ⬍ .05 vs Young group, Ireton-Jones 1997.4
b
p ⬍ .05 vs all other equations in the Elderly group.
c
p ⬍ .05 vs Penn State (2003b), Young group.
d
p ⬍ .05 vs Ireton-Jones (1997), Elderly group.
e
p ⬍ .05 vs Ireton-Jones (1992), Elderly group.

Downloaded from pen.sagepub.com at UQ Library on October 12, 2014

July–August 2004 TWO APPROACHES TO PREDICTING RESTING METABOLIC RATE
Percentage of subjects subdivided by age and body composition group whose predicted metabolic rate was within 10% or ⬎15% of measured resting
Percent of predictions within 10% of measured
Young Young
nonobese obese
(n ⫽ 15) (n ⫽ 12)
Ireton-Jones (1992)3 60 83
Penn State (2003a) 67 58
a
p ⬍ .05 vs Penn State 2003a equation.
resting metabolic rate was 386 kcal/day (19% of mea-
sured resting metabolic rate). Dickerson et al16 per-
formed a validation on the burn component of the Ire-
ton-Jones equation and found an absolute difference
between predicted and measured resting metabolic
rates of 458 ⫾ 356 calories/day, which was 20% ⫾ 20%
of measured. MacDonald and Hildebrandt8 found in
the Ireton-Jones equation a tendency to underestimate
resting metabolic rate (mean prediction value was 85%
⫾ 16% of measured) and a low association with mea-
sured resting metabolic rate (R2 ⫽ .23). Only 63% of
estimates were within ⫾20% of measured. It should be
noted that MacDonald compared predicted resting
metabolic rate to measured total metabolic rate (via
continuous gas exchange monitoring), and so system-
atic underestimation by an equation to estimate rest-
ing metabolic rate is to be expected. The current vali-
dation study was more favorable to the Ireton-Jones
equation than was the study by Flancbaum et al,7
Dickerson et al,16 or MacDonald and Hildebrandt.8 The
R2 between calculated and measured resting metabolic
rate was 0.57 (actually better than in Ireton-Jones’
original data), but not as high as the unmodified Har-
ris-Benedict equation in the current study. We found a
lower mean absolute difference between calculated and
measured resting metabolic rate than did Flancbaum
et al,7 (257 ⫾ 197 kcal/day, or 13% of measured resting
metabolic rate). The percentage of subjects accurately
predicted by the Ireton-Jones equation (60% of sub-
jects) was higher than in the study by Flancbaum et
al,7 or MacDonald and Hildebrandt.8 Of note, the 1997
revisions of the Ireton-Jones equation made the pre-
diction of resting metabolic rate in critically ill patients
less accurate that the originally published version of
the equation.
The original Penn State equation (1998) for resting
metabolic rate was derived from retrospective analysis
of indirect calorimetry measurements from 169
mechanically ventilated, surgical, trauma, and medical
intensive care unit patients. This equation was pub-
lished in book form in 19986 and explained 70% of the
variation in resting metabolic rate in the original data-
base. MacDonald and Hildebrandt’s validation study8
showed that this equation had an association with
measured total metabolic rate (R2 ⫽ .66) but a system-
atic tendency to underestimate total metabolic rate
(87% ⫾ 11% of measured total metabolic rate). Only
72% of subjects were estimated to within 20% of total
metabolic rate. As noted above, MacDonald and Hilde-
brandt8 compared estimates of resting metabolic rate
with measurements of total metabolic rate; so system-
Downloaded from pen.sagepub.com at UQ Library on October 12, 2014
263
TABLE V
metabolic rate
Percent of predictions ⬎ 15% different from measured
Elderly Elderly Young Young Elderly Elderly
nonobese obese nonobese obese nonobese obese
(n ⫽ 14) (n ⫽ 6) (n ⫽ 15) (n ⫽ 12) (n ⫽ 14) (n ⫽ 6)
43a 50 33a 17 36a 50
93 67 0 33 0 17
atic underestimation is not surprising. In the current
validation study, the original Penn State equation
explained 81% of the variation in measured resting
metabolic rate and was shown to be accurate to within
10% of measured resting metabolic rate in 64% of sub-
jects. This was essentially the same as the success rate
of the 1992 Ireton-Jones equation3 (accurate 60% of the
time), but the occurrence of large errors (estimate met-
abolic rate ⬎15% of measured) was significantly lower
in the 1998 Penn State equation than the Ireton-Jones
equations (15 vs 32% of subjects). Mean absolute dif-
ference between calculated and measured resting met-
abolic rate (156 ⫾ 136 kcal/day, or 8% of measured)
was significantly lower than that achieved with the
Ireton-Jones equations.3,4
The revised Penn State equations were developed
before undertaking the current study. The decision to
revise the equations was based on separate research in
healthy subjects that showed that the Mifflin St. Jeor
equation for healthy resting metabolic rate was more
accurate than the Harris-Benedict equation in healthy
subjects13 and that the common practice of using obe-
sity-adjusted body weight in the Harris-Benedict equa-
tion underestimates resting metabolic rate in other-
wise healthy obese people.13 The Harris-Benedict
equation was used in the development of the 1998
Penn State equation, as was obesity-adjusted body
weight in the obese members of the database. We
therefore felt compelled to recompute the Penn State
equations from the original data6 using the Mifflin
equation for healthy resting metabolism in one case
and retaining the Harris-Benedict equation but using
actual body weight instead of adjusted weight in the
other. The results were the 2003b and 2003a Penn
State equations of the current study. Each of these
permutations of the original Penn State equation had
as strong a correlation with measured resting meta-
bolic rate as the original equation. The Penn State
2003a equation was accurate in more subjects than
was the original 1998 Penn State equation (72% vs 64%
of subjects predicted to within 10% of measured). The
same was not true for the Penn State equation 2003b,
with only 57% of subjects measured accurately. The
incidence of large errors was reduced to 11% in the
2003a Penn State equation. Mean absolute difference
between measured and calculated resting metabolic
rate was similar to that for the original 1998 equation,
and the equation was statistically precise and unbi-
ased.
More than 50% of Americans are overweight or
obese. Obesity has a negative effect on the accuracy of
264 FRANKENFIELD ET AL
predictive equations for resting metabolic rate in oth-
erwise healthy people.13 Additionally, the proportion of
the population that is elderly is rising. Advancing age
is associated with metabolic and body composition
changes, including decreased metabolic rate largely as
a result of loss of lean body tissue.17 Given the increas-
ing number of elderly and obese people in our society
and the differences these factors can make in predic-
tion of metabolic rate in healthy people, methods pro-
posed for estimating resting metabolic rate should be
assessed for accuracy in these populations when they
become ill. In the current validation study, the Penn
State 2003a equation was successful in predicting rest-
ing metabolic rate in nonobese and obese elderly
patients and in nonobese young patients but was not as
successful in the young obese patient. The Ireton-Jones
equation, on the other hand, had its best success in the
young obese.
Given the evidence that there is no apparent detri-
ment to hypocaloric feeding in the obese18,19 and the
potential detriment of overfeeding, the best approach
in constructing nutrition support regimens in the crit-
ically ill obese patient might be to rely on unmodified
Harris-Benedict or Mifflin equations for resting meta-
bolic rate. For example, in our sample of 18 obese
subjects, 17% were underestimated and 22% were
overestimated by ⬎10% using the Penn State 2003a
equation. The overestimates were by 284, 349, 449, and
468 kcal/day. Substituting the unmodified Harris-
Benedict equation for healthy people instead of the
Penn State 2003a equation in these 4 obese patients
gave caloric differences of ⫺246, ⫺215, 164, and 110
kcal/day, respectively, effectively eliminating large
overestimates.
CONCLUSIONS
The 2003a Penn State equation for estimating rest-
ing metabolic rate in mechanically ventilated, critically
ill patients was the most accurate of the equations
tested and seems to be a valid clinical tool in the
absence of indirect calorimetry measurements. The
equation uses the Harris-Benedict equation using
actual body weight in obese and nonobese subjects,
minute ventilation, and maximum body temperature
in the following form: RMR ⫽ HBE(0.85) ⫹ VE(33) ⫹
Tmax(175) ⫺ 6433. Using this equation, the practitio-
ner can expect to predict resting metabolic rate to
within 10% of actual in about 80% and to within 15% of
actual in about 100% of nonobese patients. However,
because of a tendency toward inaccuracy in the obese,
consideration should be made in obese people for opt-
ing for hypocaloric regimens that avoid use of illness-
modified equations. The 1992 Ireton-Jones equation3
was less accurate than the 2003a Penn State equation
(specifically in nonobese and elderly subjects) but was
Downloaded from pen.sagepub.com at UQ Library on October 12, 2014
Vol. 28, No. 4
more accurate than the 1997 revised Ireton-Jones
equation.4
REFERENCES
1. Barak N, Wall-Alonso E, Sitrin MD. Evaluation of stress factors
and body weight adjustments currently used to estimate energy
expenditure in hospitalized patients. JPEN. 2002;26:231–238.
2. Long CL, Schaffel N, Geiger JW, Schiller WR, Blakemore WS.
Metabolic response to injury and illness: estimation of energy
and protein needs from indirect calorimetry and nitrogen bal-
ance. JPEN. 1979;3:452– 456.
3. Ireton-Jones CS, Turner WW, Liepa GU, Baxter CR. Equations
for estimation of energy expenditures of patients with burns with
special reference to ventilatory status. J Burn Care Rehab. 1992;
13:330 –333.
4. Ireton-Jones CS, Jones JD. Why use predictive equations for
energy expenditure assessment? J Amer Diet Assoc. 1997;97:A-
44.
5. Swinamer DL, Grace MG, Hamilton SM, Jones RL, Roberts P,
King EG. Predictive equation for assessing energy expenditure
in mechanically ventilated critically ill patients. Crit Care Med.
1990;18:657– 661.
6. Frankenfield DC. Energy dynamics. In: Matarese LE,
Gottschlich MM, eds. Contemporary Nutrition Support Practice:
A Clinical Guide. Philadelphia, Pa: WB Saunders; 1998:79 –98.
7. Flancbaum L, Choban P, Sambucco S, Verducci J, Burge JC.
Comparison of indirect calorimetry, the Fick method, and pre-
diction equations in estimating the energy requirements of crit-
ically ill patients. Am J Clin Nutr. 1999;69:461– 466.
8. MacDonald A, Hildebrandt L. Comparison of formulaic equa-
tions to determine energy expenditure in the critically ill. Nutri-
tion. 2003;19:233–239.
9. Phang PT, Rich T, Ronco J. A validation and comparison study of
two metabolic monitors. JPEN. 1990;14:259 –261.
10. Frankenfield DC, Sarson YG, Blosser SA, Cooney RN, Smith JS.
Validation of a 5-minute steady state indirect calorimetry pro-
tocol for resting energy expenditure in critically ill patients.
J Am Coll Nutr. 1996;15:397– 402.
11. Harris JA, Benedict FG. A Biometric Study of Basal Metabolism
in Man. Washington, DC: Carnegie Institute; 1919. Publication
No. 279.
12. Webster JD, Hesp R, Garrow JS. The composition of excess
weight in obese women estimated by body density, total body
water, and total body potassium. Hum Nutr Clin Nutr. 1984;38:
299 –306.
13. Frankenfield DC, Rowe WA, Smith JS, Cooney RN. Validation of
several established equations for resting metabolic rate in obese
and nonobese people. J Am Diet Assoc. 2003;103:1152–1159.
14. Mifflin MD, St. Jeor ST, Hill LA, et al. A new predictive equation
for resting energy expenditure in healthy individuals. Am J Clin
Nutr. 1990;51:241–247.
15. Sheiner LB, Beal SL. Some suggestions for measuring predictive
performance. J Pharmaco Biopharm. 1981;9:503–512.
16. Dickerson RN, Gervasio JM, Riley ML, Scott BJ, Daugherty SA,
Koh YO. Accuracy of predictive methods to estimate resting
energy expenditure of thermally-injured patients. JPEN. 2003;
26;17–29
17. Roberts SB, Fuss P, Heyman MB, Young VR. Influence of age on
energy requirements. Am J Clin Nutr. 1995;62:1053S–1058S.
18. Dickerson RN, Boschert KJ, Kudsk KA, Brown RO. Hypocaloric
enteral tube feeding in critically ill obese patients. Nutrition.
2002;18:241–246.
19. Burge JC, Goon A, Choban PS, Flancbaum L. Efficacy of hypoca-
loric total parenteral nutrition in hospitalized obese patients: a
prospective, double-blind randomized trial. JPEN. 1994;18:203–
207.