Energy Expenditure of Walking and Running: Comparison with Prediction Equations

Abstract

This study established the published prediction equations for the energy expenditure of walking and running compared with the measured values. To make this comparison we first determined whether differences exist in energy expenditure for 1600 m of walking versus running, and whether energy expenditure differences occur due to being on the track or treadmill.
Energy was measured via indirect calorimetry in 24 subjects while walking (1.41 m.s(-1)) and running (2.82 m.s(-1)) 1600 m on the treadmill. A subgroup also performed the 1600-m run/walk on the track. The measured energy expenditures were compared with published prediction equations.
Running required more energy (P < 0.01) for 1600 m than walking (treadmill: running 481 +/- 20.0 kJ, walking 340 +/- 14 kJ; track: running 480 +/- 23 kJ, walking 334 +/- 14 kJ) on both the track and treadmill. Predictions using the ACSM or Leger equations for running, and the Pandolf equation for walking, were similar to the actual energy expenditures for running and walking (total error: ACSM: -20 and 14.4 kJ, respectively; Legers walking: -10.1 kJ; Pandolf walking: -10.0 kJ). An overestimation (P < 0.01) for 1600 m was found with the McArdle's table for walking and running energy expenditure and with van der Walt's prediction for walking energy expenditure, whereas the Epstein equation underestimated running energy expenditure (P < 0.01).
Running has a greater energy cost than walking on both the track and treadmill. For running, the Leger equation and ACSM prediction model appear to be the most suitable for the prediction of running energy expenditure. The ACSM and Pandolf prediction equation also closely predict walking energy expenditure, whereas the McArdle's table or the equations by Epstein and van der Walt were not as strong predictors of energy expenditure.

Full text

Energy Expenditure of Walking and Running:
Comparison with Prediction Equations
CAMERON HALL, ARTURO FIGUEROA, BO FERNHALL, and JILL A. KANALEY
Department of Exercise Science, Syracuse University, Syracuse, NY
ABSTRACT
HALL, C., A. FIGUEROA, B. FERNHALL, and J. A. KANALEY. Energy Expenditure of Walking and Running: Comparison with
Prediction Equations. Med. Sci. Sports Exerc., Vol. 36, No. 12, pp. 2128 –2134, 2004. Purpose: This study established the published
prediction equations for the energy expenditure of walking and running compared with the measured values. To make this comparison we
first determined whether differences exist in energy expenditure for 1600 m of walking versus running, and whether energy expenditure
differences occur due to being on the track or treadmill. Methods: Energy was measured via indirect calorimetry in 24 subjects while walking
(1.41 m·s
⫺1
) and running (2.82 m·s
⫺1
) 1600 m on the treadmill. A subgroup also performed the 1600-m run/walk on the track. The measured
energy expenditures were compared with published prediction equations. Results: Running required more energy (P⬍0.01) for 1600 m than
walking (treadmill: running 481 ⫾20.0 kJ, walking 340 ⫾14 kJ; track: running 480 ⫾23 kJ, walking 334 ⫾14 kJ) on both the track and
treadmill. Predictions using the ACSM or Le´ger equations for running, and the Pandolf equation for walking, were similar to the actual energy
expenditures for running and walking (total error: ACSM: ⫺20 and 14.4 kJ, respectively; Le´gers walking: ⫺10.1 kJ; Pandolf walking: ⫺10.0
kJ). An overestimation (P⬍0.01) for 1600 m was found with the McArdle’s table for walking and running energy expenditure and with van
der Walt’s prediction for walking energy expenditure, whereas the Epstein equation underestimated running energy expenditure (P⬍0.01).
Conclusion: Running has a greater energy cost than walking on both the track and treadmill. For running, the Le´ger equation and ACSM
prediction model appear to be the most suitable for the prediction of running energy expenditure. The ACSM and Pandolf prediction equation
also closely predict walking energy expenditure, whereas the McArdle’s table or the equations by Epstein and van der Walt were not as strong
predictors of energy expenditure. Key Words: CALORIC COST, PREDICTED ENERGY COST, MODE OF EXERCISE, PREDICTION
TABLES
According to the principles of physics, to move a
specific mass a specific distance requires a given
amount of energy (3,32); thus theoretically it should
require the same amount of energy to walk or run a given
distance. Previous work in quadrupeds has found that the
amount of energy used to run a mile is nearly the same
whether it is run at high speed or at a leisurely pace (19),
whereas research on humans (7,13,15,18) has shown that
humans tend to expend more energy running than walking.
The American College of Sports Medicine (ACSM)
guidelines (1) provide formulas to calculate energy expen-
diture for both running and walking speeds when caloric
expenditure is calculated based on oxygen consumption. For
example, according to this formula, an individual weighing
80 kg and walking on level grade at 1.41 m·s
⫺1
should burn
88 kcal, and, when running at 2.82 m·s
⫺1
, should burn 137
kcal for 1600 m. However, in other tables provided (22), this
same individual would burn 118 kcal walking at 1.41 m·s
⫺1
,
and approximately 145 kcal running at 2.82 m·s
⫺1
, respec-
tively, for 1600 m (22). The inconsistency between these
values and ACSM prediction formulas causes concern about
the accuracy of these equations and how they compare with
actual energy expenditures (1,12,20,22,27,33). Moreover,
these estimations do not consider the potential differences in
energy expenditure during track and treadmill exercise.
These inconsistencies in calculation of energy expenditure
create potential problems when prescribing exercise for
weight loss.
The primary purpose of this study was to establish how well
the published prediction equations for the energy expenditure
of walking and running compare with the measured values. To
accomplish this we reexamined whether differences exist in the
energy expenditure of walking versus running 1600 m, and
whether energy expenditure differences occur due to being on
the track or treadmill, which could confound the estimation of
energy expenditure. It was hypothesized that running 1600 m
would result in a greater energy expenditure than walking, and
a large discrepancy would be found between the measured
value and values obtained from the prediction equations. We
anticipated that the ACSM equation would be the best estimate
of energy expenditure compared with the other tables or equa-
tions we found in the literature.
METHODS
Subjects. Male and female subjects (age range 18 –30
yr) were recruited from the Syracuse University campus and
the surrounding community. Before participation, all sub-
Address for Correspondence: Jill Kanaley, Ph.D., 820 Comstock Ave, Rm
201, Syracuse, NY 13244, E-mail: jakanale@syr.edu.
Submitted for publication November 2003.
Accepted for publication July 2004.
0195-9131/04/3612-2128
MEDICINE & SCIENCE IN SPORTS & EXERCISE
®
Copyright © 2004 by the American College of Sports Medicine
DOI: 10.1249/01.MSS.0000147584.87788.0E
2128

jects signed an informed consent approved by the Syracuse
University Institutional Review Board. All subjects were
recreationally active with no orthopedic limitations, disease
free, taking no medications affecting metabolism, and non-
smokers. Most importantly, all subjects had to be able to run
and walk 1600 m on both the track and treadmill.
Experimental design and protocol. Each subject
had a total of five visits. The first visit entailed an orienta-
tion of the study, informed consent acquisition, collection of
initial demographics (i.e., height and weight), and the mea-
surement of peak aerobic power (V
˙O
2peak
). Visits 2–5 were
randomized and included walking at 1.41 m·s
⫺1
and running
at 2.82 m·s
⫺1
on both a treadmill and indoor track. These
measurements were made on separate days.
V
˙O
2peak
testing. The V
˙
peak
test began with the subject
walking at 1.41 m·s
⫺1
at 0% grade. In 2-min stages, the
speed of the treadmill was increased by 0.47 m·s
⫺1
incre-
ments until 2.82 m·s
⫺1
was achieved. Thereafter, the speed
of 2.82 m·s
⫺1
was held constant while the grade was in-
creased in 2.5% increments every 2 min until volitional
fatigue. Heart rate (HR) and Borg scale rating of perceived
exertion (RPE) values were recorded during the last 20 s of
each stage and at the very end of the test, when the subject
reached exhaustion. V
˙O
2
values were recorded continuously
via breath-by-breath sampling during the test using the
Cosmed Quark b
2
metabolic cart (Rome, Italy). A V
˙O
2peak
test was accepted when at least three of the following
criteria were satisfied: the RPE was greater than 17, the
respiratory quotient (R-value) was greater than 1.10, HR
approached age-predicted max and V
˙O
2
did not increase
more than 150 mL with an increase in work load (1,22).
Study day. On a subsequent day, all subjects were in-
structed to refrain from eating for at least 3 h and refrain
from caffeine at least 6 h before testing. The subjects were
instructed to wear the same footwear each day, as well as
shorts and T-shirts during testing. Although energy would
be dissipated in the shoe cushioning, wearing the same
sneakers should have limited the variability in energy ex-
penditure lost in the dissipating substratum (14). All visits
had a minimum of 24 h between visits to omit carry over
effects on energy expenditure between protocols and the
possible influence of fatigue.
For measurements on the track, an indoor 200-m track
was utilized for walking and running. Before the test, the
track was measured using a meter wheel to ensure the length
of the track was exactly 200 m. An indoor track was selected
to control for wind resistance changes (24,29,30) and to
control the thermal environment (26). On the track, the aid
of a stopwatch and track markings permitted regulation of
running and walking speeds. Four cones divided the track
into 50-m quarters. The time required to run or walk 50 m
was calculated. Walking on the indoor track was at 1.41
m·s
⫺1
(02:30 min per lap for 23 min; a 3-min warm-up
followed by the 1600-m walk), while running on the track
was at 2.82 m·s
⫺1
(01:15 min per lap for 13 min; 3 min
warm up followed by the 1600-m run). Verbal feedback to
the subject to increase or decrease speed was employed to
monitor speed at every track marking. Every effort was
made to minimize pace variation.
For the treadmill exercise, a Quinton Instruments Tread-
mill (Q65 Series 90, Seattle, Washington) was selected for
the 1600-m walk/run. The hard running surface of the tread-
mill aided in limiting energy dissipation (14,21). To induce
convectional cooling as experienced during running or
walking in track environments, a fan was placed to blow a
minimum breeze on the subject during all treadmill testing.
The treadmill was calibrated before each use to ensure the
proper speed was maintained. According to pace calcula-
tion, 1.41 m·s
⫺1
would require the duration of 20 min to
complete 1600 m; and 2.82 m·s
⫺1
would require the dura-
tion of 10 min to complete 1600 m. An additional 3 min
were included to permit the subject to reach steady state
before metabolic measurements.
Energy expenditure was measured by indirect calorimetry
utilizing open circuit spirometry of the Cosmed K4 b
2
metabolic analyzer (Rome, Italy). The K4 b
2
analyzer has
been validated in previous studies (16,23,28). The Cosmed
system was calibrated using known gas concentrations for
the gas analyzers. A calibration syringe (3 L) was used to
calibrate the turbine, which is flow-rate independent. Before
running or walking on the track or treadmill, the subjects
rested quietly for 5 min without any measurements and then
energy expenditure was measured for 3 min while the sub-
ject sat quietly in a chair. The day to day test-retest reli-
ability for V
˙O
2max
and submaximal exercise testing in our
lab is 0.96 and 0.90, respectively.
Body composition. Each subject underwent body
composition analysis by air displacement plethysmography
(BodPod, Concord, California). Subjects were asked to wear
a bathing suit, a swim cap, and a nose clip. Procedures
followed were according to the manufacturer’s guidelines.
Percent body fat and fat-free mass were calculated from
body density using the subject’s body weight divided by the
volume of air displacement minus the measured anatomical
lung volume. The day to day reliability for our system was
0.980 and the BodPod has been validated previously in the
literature (34).
Data analysis. Energy expenditure was determined by
converting the V
˙O
2
to kilojoules by assuming 1 mL of
oxygen consumed produces 20.1 J of energy
(7,13,19,20,31). Whole body oxygen consumption was
measured using breath-by-breath V
˙O
2
and V
˙CO
2
and the
respiratory exchange ratio was calculated from these values.
Using this data, the Cosmed software calculates energy
expenditure using the equations in Elia and Livesey (11).
Total energy expenditure for 1600 m was the sum of the
1-min energy expenditure values for the duration of the
exercise. In addition for comparison with a previous study
(13), it was assumed that V
˙O
2
during exercise above resting
values (sitting) was used for locomotion. The cost of loco-
motion was also calculated by subtracting the preexercise
resting values multiplied by the total minutes of the exercise
from the total energy expenditure. Energy expenditure was
also calculated from the end of the 3-min window that was
MEASURED VERSUS PREDICTED ENERGY EXPENDITURE Medicine & Science in Sports & Exercise姞
2129

permitted for the subject to reach steady state to the end of
the 1600-m walk/run.
Statistical analysis of the data. Descriptive statistics
were used to analyze potential differences between gender
using SPSS 10.0 for Windows. Differences in energy ex-
penditure for 1600 m between intensities (walking or run-
ning) and mediums (treadmill or track) were performed for
both genders using a two (intensity) by two (mode) by two
(gender) analysis of variance (ANOVA) repeated measures
design where significance was accepted at a preset
␣
ⱕ
0.05. In addition, fat-free mass was used as a covariate in the
ANOVA model to adjust for gender differences. Only
steady state data for the 1600-m run/walk were used in the
analysis and are reported as means ⫾standard error. Pear-
son’s bivariate correlations were applied to identify if a
relationship existed between V
˙O
2
and anthropometric mea-
sures. A modified Bland–Altman plot (2) was performed to
compare the measured results with those calculated from the
prediction formula. We used a modified Bland–Altman plot
because we had the actual energy expenditure measurement
and were able to compare it with estimations of energy
expenditure to determine their agreement. Power calcula-
tions were performed to determine the probability that a
statistical difference was missed when no statistical differ-
ence was found. Values are presented as means ⫾SE.
Numerous equations have been used in the literature for
the prediction of energy expenditure. We selected the fol-
lowing tables and equations because they were frequently
cited in the literature. The ACSM equation was used be-
cause most exercise physiologists are familiar with the
ACSM guidelines handbook (1). The McArdle tables for
walking and running were employed because they are found
in a commonly used teaching textbook (22) and they are
frequently used by researchers in the field for estimation of
energy expenditure. The other equations were selected be-
cause they had been cited in the literature and they provided
additional estimates of both walking and running. The pre-
diction formulas that were used are listed below:
ACSM (1):
Running.V
˙O
2
(mL·kg
⫺1
·min
⫺1
)⫽0.2 (m·s
⫺1
)⫹0.9
(m·s
⫺1
) (fractional grade) ⫹3.5
Walking.V
˙O
2
(mL·kg
⫺1
·min
⫺1
)⫽0.1 (m·s
⫺1
)⫹1.8
(m·s
⫺1
) (fractional grade) ⫹3.5
McArdle (22):
McArdle’s tables are available in the referenced text.
Van der Walt (33):
Walking.
V
˙O
2
(L·min
⫺1
)⫽0.00599 M ⫹0.000366 MV
2
Running.
V
˙O
2
(L·min
⫺1
)⫽⫺0.419 ⫹0.03257 M ⫹0.000117
MV
2
Pandolf (27):
W (J·s
⫺1
)⫽1.5 M ⫹2.0 (M ⫹L)(L/M)
2
⫹n(M ⫹
L)[1.5V
2
⫹0.35VG]
M⫽body mass (kg), L ⫽load carried, V ⫽velocity
(m·s
⫺1
),G⫽grade, and n is the terrain factor. For unloaded,
level walking on a track or treadmill, the following formula
is used: W (1 J·s
⫺1
)⫽1.5 W ⫹1.5V
2
W
Le´ ger (21):
V
˙O
2
(mL·kg
⫺1
·min
⫺1
)⫽2.209 ⫹3.1633 (running speed
in km·h
⫺1
)
Epstein (12):
Mr ⫽Mw ⫺0.5 (1-0.01L) (Mw ⫺15L-850)
Mr ⫽metabolic cost of running, Mw ⫽metabolic cost of
walking, L ⫽clothing weight
In all equations, V
˙O
2
predictions were expressed in
kilojoules (kJ).
RESULTS
Twenty-eight subjects (15 males and 13 females) were
recruited for this investigation, with only 24 subjects (12
males and 12 females) meeting all the inclusion criteria.
Subjects were excluded if their RER was above 1.0 while
running 1600 m on the track or treadmill. The mean age of
the women and men was 21.4 ⫾1.1 yr and 23.2 ⫾1.0 yr,
respectively. The women were a mean height and weight of
1.69 ⫾0.02 m and 63.9 ⫾3.1 kg, and the men were 1.80
⫾0.02 m and 76.6 ⫾3.0 kg, respectively. The women had
a significantly higher percent body fat (23.7 ⫾2.2%) and
had a smaller fat-free mass (48.0 ⫾1.5 kg) than was found
in the men (11.6 ⫾1.9% and 67.4 ⫾2.3 kg, respectively,
P⬍0.05). The V
˙O
2peak
was significantly lower in the
women (48.0 ⫾1.5 mL·kg
⫺1
·min
⫺1
) than was observed in
the men (53.0 ⫾1.6 mL·kg
⫺1
·min
⫺1
,P⬍0.05). The
subjects ranged in BMI from 20.0 to 27.4 kg·m
⫺2
.
Energy expenditure. Running elicited a significantly
greater total energy expenditure than walking on both the
treadmill and the track (P⬍0.001) for both genders (Fig.
1a). On the treadmill, the males expended 520.6 ⫾27.6 kJ
for 1600 m; this was significantly higher (P⬍0.05) than the
energy expenditure by the females (441.1 ⫾25.6 kJ). For
the walk, the males expended 370.4 ⫾17.7 kJ, and the
females expended 309.6 ⫾17.2 kJ for 1600 m (P⬍0.05
between genders). When energy expenditure was adjusted
for fat-free mass, the gender effect disappeared, but running
FIGURE 1—Total energy expenditures for 1600 m of walking and
running in males and females on the track and treadmill, expressed in
total expenditure (a), and normalized to fat-free mass (b). * P<0.05
versus walking; † P<0.05 versus females.
2130
Official Journal of the American College of Sports Medicine http://www.acsm-msse.org

exercise still required more energy than walking (P⬍0.01;
Fig. 1b).
In the previous literature, sitting metabolic rate was sub-
tracted from the total energy expenditure to acquire energy
expenditures for locomotion. Sitting energy expenditure
was 8.5 ⫾0.4 and 7.8 ⫾0.4 kJ for men and women,
respectively. After subtracting sitting energy expenditure
from the total energy expenditure, running energy expendi-
ture was still significantly greater than walking energy ex-
penditure (P⬍0.001) in both the men and women (tread-
mill: running males 437 ⫾27 kJ, females 378 ⫾23 kJ;
walking males 216 ⫾14 kJ, females 177 ⫾10 kJ; track:
running males 469 ⫾24 kJ, females 347 ⫾28 kJ; walking
males 196 ⫾20 kJ, females 164 ⫾14 kJ).
Of the 24 subjects, a subgroup of 17 subjects (10 females
and 7 males) performed the run and walk on the track. Total
energy expenditure during steady-state track running was
significantly greater than track walking (P⬍0.05), but no
significant differences were found between the track and
treadmill total energy expenditure. Adjusting energy expen-
diture for the sitting energy expenditure did not alter these
findings. Gender differences occurred on the track as well as
on the treadmill; these differences disappeared after adjust-
ing for fat-free mass.
The predictions using the ACSM or Le´ ger equations for
running did not differ significantly from actual energy ex-
penditures (Fig. 2). For walking, the actual energy expen-
diture was not significantly different than the ACSM or
Pandolf prediction. However, both McArdle’s table and van
der Walt’s prediction significantly overestimated (P⬍
0.001) energy expenditures for running and walking 1600 m
for both genders (Fig. 2), whereas the Epstein equation
underestimated the metabolic cost of running (P⬍0.01). To
ensure that we did not make a Type II error, power calcu-
lations comparing actual energy expenditures and the pre-
diction equations were performed. When we found signifi-
cant differences, the power approached 1.0. Where there
were no significant differences, the power ranged from
0.11– 0.35, indicating that considerably more subjects
would have been required to produce a significant differ-
ence between actual and predicted energy expenditures.
Hence, we are confident that the differences that we found
are real.
Figures 3–5 show the modified Bland–Altman plots for
the actual energy expenditure versus prediction models
available in the literature. The mean (solid line) ⫾2SD
ranges (dashed lines) for the entire group are shown in each
figure. When the ACSM’s prediction model was compared
FIGURE 2—Actual total energy expenditure (solid) compared with
energy expenditure predictions by ACSM (diagonal lines), McArdle
(M) (dotted), van der Walt (horizontal lines), Le´ger (checkered), and
Pandolf (vertical lines) for 1600 m. Values reported in means ⴞSE.
*P<0.05 between actual expenditures and predicted.
FIGURE 3—Comparison of difference be-
tween measured energy expenditure (kJ) and
ACSM’s prediction for energy expenditure
for running (a) and walking (b) and McAr-
dle’s table for energy expenditure for run-
ning (c) and walking (d) 1600 m using the
modified Bland–Altman technique. Males
represented by open diamonds, and females
by solid squares. Solid line is total error (TE)
from zero, with ⴞ2 SD (dashed lines). Figures
4 and 5 follow the same format.
MEASURED VERSUS PREDICTED ENERGY EXPENDITURE Medicine & Science in Sports & Exercise姞
2131

with the measured value (Fig. 3a), a TE of ⫺20.0 kJ was
observed with 95% confidence intervals (⫾2SD)of
⫾92.1 kJ for running, and a TE of 14.4 kJ was observed
with 95% confidence intervals of ⫾68.7 kJ while walking
(Fig. 3b). The McArdle tables, compared with actual
energy expenditure for running and walking, had a TE of
⫺53.5 kJ with 95% confidence intervals of ⫾95.3 kJ for
running (Fig. 3c) and a TE of ⫺98.3 kJ with 95% con-
fidence intervals of ⫾81.1 kJ for walking (Fig. 3d). When
comparing actual energy expenditures against van der
Walt’s prediction model, a TE of ⫺46.9 kJ with 95%
confidence intervals of ⫾100.4 kJ was observed for run-
ning (Fig. 4a), and a TE of ⫺67.1 kJ with 95% confidence
intervals of ⫾74.5 kJ was found for walking (Fig. 4b). In
Figure 4c, the comparison with Le´ger’s prediction for
running energy expenditures resulted in a TE of ⫺10.1 kJ
with 95% confidence intervals at ⫾92.1 kJ. Pandolf’s
prediction model for walking energy expenditures noted
aTEof⫺10.0 kJ with a ⫾69.3 kJ 95% confidence
interval (Fig. 4d). Lastly, when the Epstein prediction
model was compared with the measured value (Fig. 5), a
TE of 135.5 kJ was observed with 95% confidence in-
tervals of ⫾167 kJ for running.
A significant (P⬍0.001) inverse relationship between
V
˙O
2peak
and percent fat was observed (r ⫽⫺0.75), as well
as between V
˙O
2peak
and total energy expenditures adjusted
for fat-free mass (P⬍0.01) for running and walking (track:
r⫽⫺0.65, r ⫽⫺0.64; treadmill: r ⫽⫺0.57, r ⫽⫺0.69,
respectively). Thus, the more fit individuals used less en-
ergy per unit of fat-free mass for 1600 m than the less fit.
Moderately strong correlations ranging from 0.71 to 0.83
(P⬍0.01) were found between percent body fat and energy
expenditures normalized for body weight for both the track
and treadmill.
DISCUSSION
In the development of exercise prescription, knowing the
actual energy expenditure of the prescribed activity is es-
sential, particularly in weight loss programs. The ACSM
guidelines (1), as well as numerous well-published equa-
tions, exist. These guidelines and equations allow for the
prediction of walking or running, but how they compare
with actual measures has seldom been analyzed. Our find-
ings show that the total energy expenditure for running
1600 m was ⬃30% higher than walking the same distance,
regardless of gender. When resting energy expenditure was
subtracted from the total energy expenditure, as has been
done in earlier studies (21), the cost of locomotion was
⬃55% lower for males during walking than running, and
⬃52% lower for females. Similarly Farley et al. (13) ob-
served greater energy expenditures for running compared
with walking, and this occurred whether they were in nor-
mal or reduced gravity conditions. Our findings disagree
with those of Kram et al. (19), who reported that running
FIGURE 5—Comparison of difference between measured energy ex-
penditures (kJ) and Epstein’s energy expenditure prediction for run-
ning 1600 m using the modified Bland–Altman technique.
FIGURE 4 —Comparison of difference
between measured energy expenditures
(kJ) and van der Walt’s energy expendi-
ture prediction for running (a) and walk-
ing (b) and Le´ger’s energy expenditure
prediction for running 1600 m (c) and
Pandolf’s energy expenditure prediction
for walking 1600 m (d) using the modified
Bland–Altman technique.
2132
Official Journal of the American College of Sports Medicine http://www.acsm-msse.org

and walking 1600 m require the same amount of energy;
such a finding would agree with the theoretical physics
calculations.
Frequently, it is observed that women have a greater
difficulty than men in losing weight when in a walking or
running program. This difficulty could possibly be due to
the lower energy expenditure for walking and running than
we observed in women. In this study, the gender differences
observed between walking and running 1600 m were attrib-
uted to the significantly larger metabolically active tissue
(fat-free mass) found in the male population. By adjusting
for metabolically active tissue, gender differences disap-
peared, and the energy cost of running was 8.4 kJ per unit
of fat-free mass, and for walking was 5.9 kJ per unit of
fat-free mass for both genders.
We also hypothesized that an energy expenditure differ-
ence between the track and treadmill when walking or
running would not be observed when factors such as wind
influence were controlled. Our findings agree with this
hypothesis; as we noted, there was no difference between
the energy expenditure on the track and the treadmill. This
is consistent with McMiken et al. (24), who noted that
submaximal exercise on the track and treadmill elicited the
same physiological responses when running at low to mod-
erate speeds. Although we found no difference at 2.82
m·s
⫺1
, their research indicates that running at higher speeds
might potentially cause wind resistance to influence energy
expenditures (24). Pugh et al. (29,30) have reported that
running at speeds above 4.0 m·s
⫺1
(8.44 miles·h
⫺1
) can
result in an increase in energy expenditure up to 16%
because of the energy cost of overcoming wind resistance.
In the present study we specifically chose a speed below 4.0
m·s
⫺1
to minimize or eliminate potential influences of wind
resistance. Furthermore, this is one of the first studies com-
paring track and treadmill exercise at submaximal intensi-
ties using the portable unit (Cosmed K4 b
2
), and not the
more cumbersome Douglas bags (24,29,30). More recent
work (8) has used the portable K4 b
2
, investigating maximal
physiological responses on the track and treadmill, and
reporting no significant differences between the two sur-
faces. We have now also demonstrated that no differences
occur between track and treadmill measurements during
submaximal exercise.
One of the primary purposes of this study was to compare
the actual energy expenditures of walking and running with
the energy expenditure calculated through predictions equa-
tions. In many research and clinical settings, actual energy
expenditure methods cannot be utilized, resulting in reliance
on prediction equations. To address this issue, we compared
our actual measures against formulas or tables frequently
cited in the literature or used in clinical practice (the ACSM
prediction formula (1), McArdle tables (22), van der Walt
prediction formula (33), Epstein (12) and Le´ ger (20) pre-
diction models for running, and Pandolf prediction formula
for walking (27)). We observed that the ACSM formula
overestimated the average energy expenditure in steady-
state running of 1600 m by only 4.3% (21kJ), and under-
estimated energy expenditure by only 3.8% (13 kJ) for
1600 m. The total error of this overestimation was ⫺20.0 kJ
for running energy expenditures, and there was a 14.4-kJ
underestimation of walking energy expenditures. Similarly,
the equation by Le´ ger overestimated by 2% (⫺10.1 kJ), and
Pandolf overestimated by 2.8% (⫺10.0 kJ); these differ-
ences were minimal.
Unlike the above equations, the McArdle tables revealed
an 11% overestimation (⫺54.5 kJ) for running and a 30%
overestimation (⫺98.3 kJ) for walking. Similarly, van der
Walt’s prediction model overestimated running energy ex-
penditures by 10.4% (⫺46.9 kJ) and 19.7% (⫺67.1 kJ) for
walking. Again, this overestimation was evenly distributed
across the ranges of energy expenditures for both genders
for the tables and prediction model. The Epstein equation
was the only equation that substantially underestimated the
energy expenditure of running, and did so by the largest
margin of error. In fact, this underestimation occurred in a
linear fashion, such that at higher energy expenditures, a
greater underestimation occurred. The reason for the larger
discrepancy with this equation is unclear to us. Thus, cau-
tion is recommended when using these tables and prediction
models to estimate walking and running energy expenditure.
The error may not seem substantial for a single exercise
bout, but if these equations were used as part of a weight
loss program, this error would be magnified with daily
exercise, such that weekly energy expenditure values would
be significantly in error.
Closer examination of the Bland–Altman plots reveals
that there are larger limits of agreement for running than for
walking. Although total errors vary considerably between
prediction models, during running it is much more difficult
to predict energy expenditures for a 1600-m run than a
1600-m walk.
It was initially hypothesized that a significant inverse
correlation between V
˙O
2peak
and the total energy expen-
diture difference between walking and running 1600 m
would be found. Our results did not support this hypoth-
esis; however, the finding of a significant Pearson cor-
relation between V
˙O
2peak
and energy expenditures nor-
malized to fat-free mass may have two separate
explanations. First, it can be hypothesized that the more
fit individuals are simply more economical than unfit
individuals. This would be consistent with previous stud-
ies suggesting that trained individuals optimize locomo-
tion to minimize energy expenditure from the biochem-
istry of muscle tissue to gait economy (4 – 6,9,10,25).
Second, the metabolically active tissue working during
the run/walk has to carry nonmetabolically active tissue
(i.e., adipose tissue). Hence, the individual with a lower
percent body fat is carrying less excess weight, and
therefore consumes less energy per unit of fat-free mass;
thus, they are more economical. When expressing this
finding in terms of carrying external workloads, the ad-
dition of weight to walking or running has been shown to
increase energy expenditure in the literature (13,27).
From this, one could propose that body composition
might help explain why males have been found to have
better economies than females (10,17). When normalized
MEASURED VERSUS PREDICTED ENERGY EXPENDITURE Medicine & Science in Sports & Exercise姞
2133

to fat-free mass, the gender difference disappeared.
Hence, the less economical running energy expenditures
found in females might be explained by the fact that
females are traditionally less lean than males, and must
carry more adipose tissue than males.
Although theoretically walking and running a mile should
require the same work (5), running required more energy than
walking for 1600 m, regardless of whether the subject was on
the track or the treadmill. Comparison of the actual energy
expenditure with prediction equations reveals that the ACSM
and Le´ger’s prediction model for horizontal running are more
accurate in a young healthy population. For horizontal walking,
ACSM and Pandolf’s prediction models also appear more
accurate than the other equations evaluated.
The authors of this study would like to thank and acknowledge all
the Human Performance Lab personnel at Syracuse University for
their assistance with this research project.
REFERENCES
1. AMERICAN COLLEGE OF SPORTS MEDICINE.Guidelines for Exercise
Testing and Prescription, 6th Edition. Baltimore, MD: Lippincott
Williams & Wilkins, 2000, pp.163–176.
2. BLAND, J., and D. ALTMAN. Statistical methods for assessing agree-
ment between two methods of clinical measurement. Lancet
1:307–310, 1986.
3. BUNC, V., and R. DLOUHA. Energy cost of treadmill walking.
J. Sports Med. Phys. Fitness 37:103–109, 1997.
4. CAVAGNA, G., F. SAIBENE, and R. MARGARIA. Mechanical work in
running. J. Appl. Physiol. 19:249 –56, 1964.
5. CAVANAGH, P., and R. KRAM. The efficiency of human move-
ment—a statement of the problem. Med. Sci. Sports Exerc. 17:
304 –308, 1985.
6. CAVANAGH, P., and R. KRAM. Mechanical and muscular factors
affecting the efficiency of human movement. Med. Sci. Sports
Exerc. 17:326 –331, 1985.
7. CHANG, Y., and R. KRAM. Metabolic cost of generating horizontal
forces during human running. J. Appl. Physiol. 86:1657–1662,
1999.
8. CROUTER, S., C. FOSTER,P.ESTEN,G.BRICE, and J. PORCARI.
Comparison of incremental treadmill exercise and free range run-
ning. Med. Sci. Sports Exerc. 33:644 – 47, 2001.
9. DANIELS, J. A physiologist’s view of running economy. Med. Sci.
Sports Exerc. 17:332–338, 1985.
10. DANIELS, J., and N. DANIELS. Running economy of elite male and
elite female runners. Med. Sci. Sports Exerc. 24:483– 489, 1992.
11. ELIA, M., and G. LIVESEY. Energy expenditure and fuel selection in
biological systems: the theory and practice of calculations based
on indirect calorimetry and tracer methods. In: Metabolic Control
of Eating, Energy Expenditure and the Bioenergetics of Obesity,
A. P. Simopoulos (Ed.). Basel, Switzerland: Karger, 1992, pp.
69 –130.
12. EPSTEIN, Y., L. A. STROSCHEIN, and K. B. PANDOLF. Predicting
metabolic cost of running with and without backpack loads. Eur.
J. Appl. Physiol. Occup. Physiol. 56:495–500, 1987.
13. FARLEY, C., and T. MCMAHON. Energetics of walking and running:
insights from simulated reduced gravity experiments. J. Appl.
Physiol. 73:2709 –2712, 1992.
14. FREDERICK, E., T. CLARKE,J.LARSEN, and L. COOPER. The effects of
shoe cushioning on the oxygen demands of running. In: Biome-
chanical Aspects of Sports Shoes and Playing Surfaces, B. Nigg
and B. Kerr (Eds.). Calgary: University of Calgary Printing Ser-
vices, 1983, pp. 107–114.
15. GOTTSCHALL, J., and R. KRAM. Energy cost and muscular activity
required for propulsion during walking. J. Appl. Physiol. 94:
1766 –1772, 2003.
16. HAUSSWIRTH, C., A. X. BIGARD, and J. M. LECHEVALIER. The
Cosmed K4 telemetry system as an accurate device for oxygen
uptake measurements during exercise. Int. J. Sports Med. 18:449 –
453, 1997.
17. HEGLUND, J., F. INGJIER, and S. STROMME. Sex differences in per-
formance-matched marathon runners. Eur. J. Appl. Physiol. Oc-
cup. Physiol. 61:433– 439, 1990.
18. HOLT, K., J. HAMILL, and R. ANDERS. Predicting the minimal
energy cost of human walking. Med. Sci. Sports Exerc. 23:491–
498, 1991.
19. KRAM, R., and C. TAYLOR. Energetics of running: a new perspec-
tive. Nature 346:265–267, 1990.
20. Le´GER L., and D. MERCIER. Gross energy cost of horizontal tread-
mill and track running. Sports Med. 1:270 –277, 1984.
21. LEJEUNE, T., P. WILLEMS, and N. HEGLUND. Mechanics and ener-
getics of human locomotion on sand. J. Exp. Bio. 201:2071–2080,
1998.
22. MCARDLE, W., F. KATCH, and V. KATCH.Exercise Physiology:
Energy Nutrition, and Human Performance. Baltimore, MD: Wil-
liams & Wilkins, 2001, pp. 1103–1115.
23. MCLAUGHLIN, J., G. A. KING,E.T.HOWLEY,D.R.J.BASSETT, and
B. E. AINSWORTH. Validation of the COSMED K4 b2 portable
metabolic system. Int. J. Sports Med. 22:280 –284, 2001.
24. MCMIKEN, D., and J. DANIELS. Aerobic requirements and maxi-
mum aerobic power in treadmill and track running. Med. Sci.
Sports. 8:14 –17, 1976.
25. MORGAN, D., D. BRANSFORD,D.COSTILL,J.DANIELS,E.HOWLEY,
and G. KRAHENBUHL. Variation in the aerobic demand of running
among trained and untrained subjects. Med. Sci. Sports Exerc.
27:404 – 409, 1995.
26. MOSTARDI, R., R. KUBICA,A.VEICSTEINAS, and R. MARGARIA. The
effect of increased body temperature due to exercise on the heart
rate and on the maximal aerobic power. Eur. J. Appl. Physiol.
Occup. Physiol. 33:237–245, 1974.
27. PANDOLF, K., B. GIVONI, and R. GOLDMAN. Predicting energy ex-
penditure with loads while standing or walking very slowly.
J. Appl. Physiol. 43:577–581, 1978.
28. PINNINGTON, H. C., P. WONG,J.TAY,D.GREEN, and B. DAWSON.
The level of accuracy and agreement in measures of FEO
2
, FECO
2
and VE between Cosmed K4b2 portable, respiratory gas analysis
system and a metabolic cart. J. Sci. Med. Sport 4:324 –335, 2001.
29. PUGH, L. The influence of wind resistance in running and walking
and the mechanical efficiency of work against horizontal or ver-
tical forces. J. Physiol. 213:255–276, 1971.
30. PUGH, L. Oxygen intake in track and treadmill running with
observations on the effect of air resistance. J. Physiol. 207:823–
835, 1970.
31. ROBERTS, T., R. KRAM,P.WEYAND, and C. TAYLOR. Energetics of
bipedal running: metabolic cost of generating force. J. Exp. Bio.
201:2745–2751, 1998.
32. SERWAY, R., and J. JEWETT.Principles of Physics. Saunders Col-
lege Publishing, 1998, pp. 25–50.
33. VAN DER WALT, W., and C. WYNDHAM. An equation for prediction
of energy expenditure of walking and running. J. Appl. Physiol.
34:559 –563, 1973.
34. VESCOVI, J. D., S. L. ZIMMERMAN,W.C.MILLER, and B. FERNHALL.
Effects of clothing on accuracy and reliability of air displacement
plethysmography. Med. Sci. Sports Exerc. 34:282–285, 2002.
2134
Official Journal of the American College of Sports Medicine http://www.acsm-msse.org