ArticlePDF Available

Prediction of energy expenditure from heart rate monitoring during submaximal exercise

Taylor & Francis
Journal of Sports Sciences
Authors:

Abstract

The aims of this study were to quantify the effects of factors such as mode of exercise, body composition and training on the relationship between heart rate and physical activity energy expenditure (measured in kJ x min(-1)) and to develop prediction equations for energy expenditure from heart rate. Regularly exercising individuals (n = 115; age 18-45 years, body mass 47-120 kg) underwent a test for maximal oxygen uptake (VO2max test), using incremental protocols on either a cycle ergometer or treadmill; VO2max ranged from 27 to 81 ml x kg(-1) x min(-1). The participants then completed three steady-state exercise stages on either the treadmill (10 min) or the cycle ergometer (15 min) at 35%, 62% and 80% of VO2max, corresponding to 57%, 77% and 90% of maximal heart rate. Heart rate and respiratory exchange ratio data were collected during each stage. A mixed-model analysis identified gender, heart rate, weight, V2max and age as factors that best predicted the relationship between heart rate and energy expenditure. The model (with the highest likelihood ratio) was used to estimate energy expenditure. The correlation coefficient (r) between the measured and estimated energy expenditure was 0.913. The model therefore accounted for 83.3% (R2) of the variance in energy expenditure in this sample. Because a measure of fitness, such as VO2max, is not always available, a model without VO2max included was also fitted. The correlation coefficient between the measured energy expenditure and estimates from the mixed model without VO2max was 0.857. It follows that the model without a fitness measure accounted for 73.4% of the variance in energy expenditure in this sample. Based on these results, we conclude that it is possible to estimate physical activity energy expenditure from heart rate in a group of individuals with a great deal of accuracy, after adjusting for age, gender, body mass and fitness.
Prediction of energy expenditure from heart rate
monitoring during submaximal exercise.
Publication: Journal of Sports Sciences
Publication Date: 01-MAR-05
Author: Keytel, L.R.
; Goedecke, J.H. ;
Noakes, T.D. ;
Hiiloskorpi, H. ;
Laukkanen, R. ; van
der Merwe, L. ;
Lambert, E.V.
Prediction of energy expenditure from heart rate monitoring during submaximal
exercise.
Publication: Journal of Sports Sciences
Publication Date: 01-MAR-05
Author: Keytel, L.R. ; Goedecke, J.H. ; Noakes, T.D. ; Hiiloskorpi, H. ; Laukkanen, R. ;
van der Merwe, L. ; Lambert, E.V.
Abstract
The aims of this study were to quantify the effects of factors such as mode of
exercise, body composition and training on the relationship between heart rate
and physical activity energy expenditure (measured in kJ x [min.sup.-1]) and to
develop prediction equations for energy expenditure from heart rate. Regularly
exercising individuals (n = 115; age 18-45 years, body mass 47-120 kg)
underwent a test for maximal oxygen uptake (V[O.sub.2max] test), using
incremental protocols on either a cycle ergometer or treadmill; V[O.sub.2max]
ranged from 27 to 81 x ml [kg.sup.-1] x [min.sup.-1]. The participants then
completed three steady-state exercise stages on either the treadmill (10 min) or
the cycle ergometer (15 min) at 35%, 62% and 80% of V[O.sub.2max],
corresponding to 57%, 77% and 90% of maximal heart rate. Heart rate and
respiratory exchange ratio data were collected during each stage. A mixed-
model
analysis identified gender, heart rate, weight, V[O.sub.2max] and age as factors
that best predicted the relationship between heart rate and energy expenditure.
The model (with the highest likelihood ratio) was used to estimate energy
expenditure. The correlation coefficient (r) between the measured and estimated
energy expenditure was 0.913. The model therefore accounted for 83.3%
([R.sup.2]) of the variance in energy expenditure in this sample. Because a
measure of fitness, such as V[O.sub.2max], is not always available, a model
without V[O.sub.2max] included was also fitted. The correlation coefficient
between the measured energy expenditure and estimates from the mixed model
without V[O.sub.2max] was 0.857. It follows that the model without a fitness
measure accounted for 73.4% of the variance in energy expenditure in this
sample. Based on these results, we conclude that it is possible to estimate
physical activity energy expenditure from heart rate in a group of individuals with
a great deal of accuracy, after adjusting for age, gender, body mass and fitness.
Keywords: Energy expenditure, physical activity, prediction equations
Introduction
During moderate physical activity, there is a linear relationship between heart
rate and oxygen consumption. This heart rate--oxygen consumption relationship
is subject to both intra- and inter-individual variability. Heart rate may be partially
dissociated from energy expenditure by factors such as emotion, posture and
environmental conditions (Hebestreit & Bar-Or, 1998). The relationship between
heart rate and energy expenditure is linear only within a relatively narrow range
of approximately 90-150 beats x [min.sup.-1] (the so-called "flex heart rate")
during physical activity (Ceesay et al., 1989; Rennie, Hennings, Mitchell, &
Wareham, 2001; Spurr et al., 1988). During light activity or inactivity, there is
almost no slope to the relationship between heart rate and energy expenditure,
and for the purpose of measuring energy expenditure from heart rate it is
assumed that energy expenditure is equal to resting energy expenditure (Rennie
et al., 2001). A non-linear, discontinuous function has been found to be more
accurate than a linear relationship in predicting physical activity energy
expenditure from heart rate (Li, Deurenberg, & Hautvast, 1993).
Heart rate monitoring, for estimating free-living energy expenditure, has been
extensively validated using indirect calorimetry, doubly labelled water and whole-
room respirometry, and reported differences between measures range from -
20%
to +25% (Luke, Maki, Barkey, Cooper, & McGee, 1997). In large groups of
people, heart rate monitoring provides one of the most efficient and economical
means of estimating energy expenditure. In addition, heart rate monitoring
provides useful insights into the type of activity being undertaken over the
measurement period. Other assessment methods, such as doubly labelled water,
can only convey the total amount of physical activity measured, whereas heart
rate monitoring provides physiological information about the type of activities
being performed and describes the nature of day-to-day variability in energy
expenditure (Hebestreit & Bar-Or, 1998; Luke et al., 1997). While whole-room
respirometry and indirect calorimetry provide physiological information about the
nature of the activity being performed, these tools are not only costly to maintain,
but often take the participant out of his or her natural environment for the duration
of the measurement period (Luke et al., 1997).
In most previous studies investigating the use of heart rate in the prediction of
energy expenditure, individual calibration of the heart rate-energy expenditure
relationship was performed (Ceesay et al., 1989; Li
et al., 1993; Luke et al., 1997;
Spurr et al., 1988). Individual calibration requires that each participant complete a
progressive exercise test, during which time heart rate is simultaneously
measured, along with indirect calorimetry to estimate energy expenditure. Two
recent studies have investigated free-living energy expenditure with heart rate
monitoring utilizing prediction equations, generated on large samples of
individuals, instead of an individual calibration test (Hiilloskorpi et al., 1999;
Rennie et al., 2001). Hiilloskorpi et al. (1999) developed a prediction equation for
energy expenditure from heart rate, using multiple regression analysis, on a
sample of 87 healthy, active men and women. Factors found to have a significant
interaction with energy expenditure included age, weight and gender. Mode of
exercise (cycling versus running) did not contribute significantly to the model.
In a more recent study, Rennie et al. (2001) developed a prediction model using
a sample of 789 individuals. Factors found to have a significant effect on the
relationship between heart rate and energy expenditure included sitting heart rate
in addition to age, weight and gender. These variables were used to predict the
slope and the intercept of the regression line between energy expenditure and
heart rate. This energy expenditure equation was then further validated on an
independent sample of 97 individuals and found to have a correlation coefficient
(r) of 0.73. Rennie et al. (2001) demonstrated the utility of developing equations
for estimating physical activity energy expenditure, from the heart rate-energy
expenditure relationship in large, representative samples of individuals, with
reasonable accuracy and the potential for wide application in epidemiological
studies.
The main aim of the present study was to further characterize the factors that
influence the relationship between energy expenditure and heart rate during
moderate to vigorous activity in regularly exercising persons. A second aim was
to develop a
prediction equation for energy expenditure from heart rate, adjusting
for these factors.
Methods
Part 1: Developing the energy expenditure prediction equation
Participants. The participants were recruited from a local fitness centre, group-
based exercise programmes, running clubs and cycle races. Altogether, 127
regularly exercising men and women volunteered for the study (of which 115 had
complete data). The participants were familiar either with a cycle ergometer or
motor-driven treadmill, and ranged in age from 19 to 45 years. They were free
from any known cardiac or metabolic disorders and were not currently taking any
chronic medication. The physical characteristics of the participants are presented
in Table I. The participants were tested on two occasions, after self-selecting the
mode of exercise (cycle ergometer, n = 69; treadmill, n = 46). The Ethics and
Research Committee of the University of Cape Town, Faculty of Health Sciences,
approved the study and informed consent was obtained from all participants
before the trials began.
A second sample of regularly exercising individuals (n = 17) was subsequently
recruited, independent of the first sample, to test the validity of the prediction
model. The second sample was recruited from a local fitness centre, and
represented a wide range of ages (21-53 years), weights (51105 kg) and fitness
(V[O.sub.2max] = 34-74.3 ml x [kg.sup.-1] x [min.sup.1]).
Body composition. Body fatness was expressed as the sum of seven skinfolds
(biceps, triceps, subscapular, suprailiac, anterior thigh, abdominal and medial
calf). Percentage body fat was estimated using the equations of Dumin and
Womersley (1974).
Maximal oxygen consumption. During the first visit to the laboratory, maximal
oxygen consumption (V[O.sub.2ma
x), maximal heart rate, and peak power output
or peak treadmill running speed were measured. Maximal oxygen uptake was
measured during either a progressive treadmill or cycle test to exhaustion. During
the treadmill test, the starting treadmill speed was 12 km x [h.sup.-1] for the men
and 10 km x [h.sup.-1] for the women, and it was increased by 0.5 km x [h.sup.-
1]
every 30 s until volitional exhaustion, as described previously (Noakes, Myburgh,
& Schall, 1990). In the cycle test to exhaustion, participants were tested on an
electronically braked cycle ergometer (Lode, Gronigen, The Netherlands). Each
participant started cycling at an exercise intensity of 3.33 W x [kg.sup.-1] body
weight for 150 s, after which the work rate was increased by 50 W for a further
150 s. The exercise intensity was then increased by 25 W every 150 s up to the
point of exhaustion (Hawley & Noakes, 1992). Maximal heart rate was defined as
that heart rate achieved at the point of exhaustion. During both the treadmill and
cycle tests, the participants wore a facemask attached to an Oxycon Alpha
automated gas analyser (Oxycon, Jaeger, The Netherlands). Before each test,
the gas analyser was calibrated using a Hans Rudolph 5530 3-litre syringe and a
two-point calibration technique, using a 5% C[O.sub.2]/95% [N.sub.2] gas
mixture and fresh air. The rate of oxygen consumption (V[O.sub.2]), rate of
carbon dioxide production (VC[O.sub.2]) and the respiratory exchange ratio
(RER) were calculated using conventional equations (Weir, 1990). Peak power
output and peak treadmill running speed were defined as the workload at which
the participant could no longer maintain the pace of the treadmill or maintain a
cadence of 70 rev x [min.sup.-1].
Submaximal testing and estimation of energy expenditure. The participants
returned to the laboratory within a week and performed a submaximal test. The
cycle ergometer submaximal test protocol consisted of three consecutive
workloads, each lasting 15 min, during which the participants cycled at 25%, 55%
and 70
63% and 80% of V[O.sub.2max] respectively. The submaximal treadmill protocol
consisted of three consecutive workloads, each lasting 10 min, at 35%, 50% and
70% (corresponding to approximately 41%, 63% and 80% of V[O.sub.2max]
respectively) of previously determined peak treadmill running speed. Minute-to-
minute heart rate was recorded using the Polar Vantage heart rate monitor (Polar
Electro, Finland) and respiratory exchange measurements (V[O.sub.2] and
VC[O.sub.2]) were collected and used to estimate energy expenditure, based on
the equations of Weir (1990), during the last 5 min of each of the stages. The
submaximal heart rate data from the last 5 min of each stage were used to
subsequently calculate predicted energy expenditure on the basis of individual
regression equations. Factors that were significantly correlated with heart rate or
V[O.sub.2] were used in the model to predict energy expenditure.
Part 2: Validation of prediction model on an independent sample
For the purpose of validation, the energy expenditure values from a 20-min self-
selected cardiovascular session were predicted on an independent sample of
individuals, recruited from a local fitness centre. These participants were
instructed to choose either a single 20-min cardiovascular workout or two 10-min
exercise bouts.
Participants. The 17 participants (9 males, 8 females) were free from known
cardiovascular and metabolic disorders and took part in some form of
cardiovascular physical activity at least three times a week. The participants met
the inclusion criteria of the original study and their physical characteristics are
presented in Table II.
Body composition and maximal test to exhaustion. The participants reported to
the laboratory on two different occasions within 7 days. During their first visit, the
participants had their body composition measured using the near infrared
reactance technique (Futrex Inc., Gaithersburg, MD, USA). They then performed
a maximal test to exhaustion on an electronically braked cycle ergometer (Lode,
Gronigen, The Netherlands) as previously described (Hawley & Noakes, 1992).
During the test, oxygen consumption and carbon dioxide productions were
measured as described above.
Estimation of physical activity energy expenditure. During the second visit, the
participants reported to the laboratory in a 2-h post-prandial state. They were
instructed not to engage in any strenuous physical activity during the preceding
24 h. All participants completed a 20-min cardiovascular exercise session as part
of an independent study in progress. The cardiovascular component was
performed following a 5-min warm-up consisting of 23 min of walking and 2-
3 min
of light jogging. The participants then chose to either complete one 20min
continuous cardiovascular exercise session or two 10-min sessions on a self-
selected piece of fitness centre equipment. Throughout the exercise session, the
participants' heart rate, V[O.sub.2] and VC[O.sub.2] were monitored continuously
using the [K4b.sup.2] portable gas analyser (Cosmed, Italy). Minute-by-minute
energy expenditure (kJ x [min.sup.-1]) was then determined using the non-
protein
caloric equivalents for oxygen. Before each test, the portable gas analyser was
calibrated using a Hans Rudolph 5530 3-1itre syringe and a two-point calibration
technique, using a 5% C[O.sub.2].sub.1]/16% [0.sub.2] gas mixture and fresh air.
The analyser outputs were processed to calculate breath-by-breath ventilation,
V[O.sub.2], VC[O.sub.2] and the respiratory exchange ratio using conventional
equations (Weir, 1990).
Statistical analysis
Initially, 127 individuals volunteered to participate in the study. The final sample
size was 115 because of incomplete heart rate and V[O.sub.2max] for 12
participants. The initial exploratory data analyses to determine factors that may
have significantly contributed to the relationship between heart rate and energy
expenditure included Box plots and scatter plots for all variables (not shown).
Univariate (means, standard deviations) and bivariate (correlation coefficients)
summary statistics were then calculated for all variables.
Based on these analyses, we fitted a mixed model for predicting energy
expenditure. The factors gender, weight, age and V[O.sub.2max] were modelled
as fixed effects, and participants as random effects, with three repeated
measurements of energy expenditure (and fixed heart rate) for each participant.
In the model, the covariance matrix between the measurements for each
participant was unstructured and compound symmetry was assumed for the
covariances between participants.
A second mixed model was fitted under the rationale that, in certain settings, a
test of maximal oxygen consumption might be impractical or not available. The
second model included all the variables and assumptions in the original model
except V[O.sub.2max]. For inner validation, both models were tested on an
independent sample of participants (n = 17), who completed 20 min of
cardiovascular exercise.
The initial exploratory analyses were performed using the Statistica data analysis
software system (version 6.1, Statsoft, Southern Africa Inc., 2002). Statistical
modelling was done with SAS[R] Proprietary Software Release 8.2 (USA).
Results
Characteristics of sample used to develop the prediction equation
The characteristics of the participants are presented in Table I. The participants
represented a wide range of morphology and fitness: age 19-45 years of age,
body weight 47-116 kg, percentage body fat 4.8-37.8% and V[O.sub.2max] 27-
81
ml x [kg.sup.-1] x [min.sup.-1]. There were no differences in mean age, weight,
percentage body fat or V[O.sub.2max] between the participants who underwent
treadmill testing versus those that underwent cycle ergometer testing. There
were significant differences in weight, percentage body fat and V[O.sub.2max]
between the sexes (Table I, P < 0.00001).
Characteristics of sample used for inner validation
The characteristics of the participants are presented in Table II. The participants
in this sample were similar to those used in the original study and represented a
broad range in body composition. Percentage body fat ranged from 9.4 to 21.6%
in the men and from 21.6 to 30.6 % in the women. Similarly, there was a wide
range in the performance data, with V[O.sub.2max] ranging from 38.7 to 73.8 ml
x [kg-.sup.-1] x [min.sup.-1] in the men and from 34.3 to 49.6 ml x [kg.sup.-1] x
[min.sup.-1] in the women.
The participants in both samples were equally matched for age and weight. The
participants (males and females combined) in the original study were slightly fitter
(mean V[O.sub.2max] 53.5 [+ or -] 0.5 ml. [kg-.sup.-1]. [min.sup.-1]) than those
who took part in the validation study (mean V[O.sub.2max] 48.1 [+ or -] 0.5 ml x
[kg.sup.-1] x [min.sup.-1]); this difference was not statistically significant.
Prediction equations of energy expenditure from heart rate: Mixed-model analysis
A mixed model was used to derive the following equation for predicting physical
activity energy expenditure (EE):
EE = -59.3954 + gender x (-36.3781 + 0.271 x age + 0.394 x weight + 0.404
V[O.sub.2max] + 0.634x heart rate) + (1 - gender) x (0.274 x ag
e + 0.103x weight
+ 0.380x V[O.sub.2max] + 0.450 x heart rate)
where gender = 1 for males and 0 for females. Table III shows the above model
in a different format. The likelihood ratio test for goodness-of-fit [chi square] =
262.73 on five degrees of freedom with P < 0.0001. The results of type III tests
for the fixed effects in the mixed model are presented in Table IV. The degrees of
freedom for the F-tests were calculated using Satterthwaite's method.
In Figure 1, the measured energy expenditure is regressed against estimated
energy expenditure. The correlation coefficient (r) is 0.913, so [R.sup.2] = 83.3%
of the variation in measured energy expenditure in the sample is explained by the
model.
[FIGURE 1 OMITTED]
A second model, which contained no
measure of fitness, was also fitted. The final
prediction equation for energy expenditure using age, gender, weight and heart
rate was:
EE = gender x (-
55.0969 + 0.6309 x heart rate + 0.1988 x weight + 0.2017 x age)
+ (1 - gender) x (-20.4022 + 0.4472 x heartrate - 0.1263 x weight + 0.074 x age)
where gender = 1 for males and 0 for females. Table V shows the above model
in a different format. The likelihood ratio test for goodness-of-fit [chi square] =
360.68 on five degrees of freedom with P < 0.0001. The results of type III tests
for the fixed effects in the mixed model are given in Table VI. The degrees of
freedom for the F-tests were calculated using Satterthwaite's method.
In Figure 2, the measured energy expenditure is regressed against estimated
energy expenditure. The coefficient of correlation 0.857, so [R.sup.2] = 73.4% of
the variation in measured energy expenditure in the sample is explained by the
model.
[FIGURE 2 OMITTED]
Independent sample analysis for inner validation
Data from an independent sample of 17 participants (8 females, 9 males) were
used to validate both models. Predicted energy expenditure using the first model,
which included a measure of fitness (V[O.sub.2max]), correlated with measured
energy expenditure during self-selected cardiovascular fitness training (r = 0.836,
P < 0.0001; Figure 3). Using the second model for measuring energy
expenditure, with no measure of fitness, the correlation coefficient was 0.77 (P <
0.0001) (Figure 4).
[FIGURE 3-4 OMITTED]
Agreement
Because we used a mixed model (with random participant effects), we had to use
maximum likelihood estimation instead of least squares. The result is that even
the estimates for the initial sample which was used to develop the equations are
slightly biased. The bias of the estimates and their random variation for the four
sets of estimates are summarized in Table VII. The bias is the difference
between the predicted and the corresponding actual value of energy expenditure,
and the 95% limits of absolute agreement were calculated as described in
Atkinson and Nevill (1998). It is interesting to note that the bias in the initial
sample is on average in the opposite direction to that for the validation sample.
The fact that the agreement limits become wider down the table is completely
logical. We believe that these limits are narrow enough for the underlying models
to be of practical use.
Discussion
In this study, we demonstrated that physical activity energy expenditure during
moderate- to high-intensity exercise may be predicted with good accuracy in a
group of individuals varying widely in age, fitness and morphology, without the
need for individual calibration. This study denotes an improvement over existing
studies in the estimation of physical activity energy expenditure using heart rate
monitoring. The proposed model (using heart rate, age, weight, gender and level
of fitness V[O.sub.2max])) accounted for 70% of the variation in observed energy
expenditure in an independent sample of people completing a self-selected 20-
min cardiovascular exercise session.
Previous studies (Li et al., 1993; Rutgers, Klijn, & Deurenberg, 1997) have cited
poor agreement between energy expenditure estimated using heart rate
monitoring and measured energy expenditure. These prediction equations were
developed on small samples, not representative of the population to which the
equation was to be applied. Rutgers et al. (1997) developed a prediction equation
based on the heart rate and energy expenditure data acquired from 13 elderly
individuals. The authors concluded that the use of heart rate monitoring to
measure energy expenditure was inaccurate over 3 days of measurement, citing
large discrepancies between energy expenditure estimation using the individual
calibration
curve and a group curve. Li et al. (1993) also reported poor agreement
for the estimation of energy expenditure using heart rate monitoring between
group and individually derived estimates. Once again, this sample was relatively
small, consisting of only 40 persons.
The current study represents an improvement over existing studies (Hiilloskorpi
et al., 1999; Rennie et al., 2001) that used heart rate monitoring to estimate
physical activity energy expenditure, without individual calibration. Previously,
Rennie et al. (2001) used the variables that significantly interacted with energy
expenditure to predict the slope, intercept and the heart rate flex point for
measured versus predicted physical activity energy expenditure. In that study,
the variables of sitting heart rate, age, weight and gender were found to have a
significant impact on the slope, intercept and heart rate flex point. These
investigators were then able to use the slope and intercept of the linear model to
place 98% of the participants in their sample in either the same or adjacent
quartiles for the measured and estimated physical activity levels. Their model has
implications for physical activity classification in epidemiological models. In the
current study, we derived linear equations, based on mixed-model analyses.
These equations yield predictions that correlate significantly with the test sample
as well as the independent validation sample.
Previously, Hiilloskorpi et al. (1999) developed an equation to predict energy
expenditure using the variables of heart rate, age, weight and gender. They
showed that the mode of exercise, cycling versus running, did not significantly
affect the final prediction of energy expenditure. We also found that the mode of
exercise did not affect the estimation of energy expenditure, and therefore
suggest that the proposed equation may be used for both running and cycling
activities. During our inner validation study, we even found good agreement with
other models of continuous activity, such as stationary rowing ergometry and
stationary stair-climbing activities.
Hiilloskorpi et al. (1999) did not include any measure of physical fitness or
V[O.sub.2max] in their prediction equation, citing a need to produce an equation
for estimating energy expenditure independent of laboratory testing. We found
that when a measure of the level of cardiorespiratory fitness such as
V[O.sub.2max] is included, the accuracy of the prediction improved. The
correlation coefficients (r) of the study sample were 0.913 for the model
V[O.sub.2max] with and 0.857 for the model without V[O.sub.2max]. The
increase in variation explained by the model including V[O.sub.2max] is 83.4%-
73.4% = 10%. The correlation coefficients of the validation sample were 0.836 for
the model with V[O.sub.2max] and 0.77 for the model without V[O.sub.2max].
The increase in variation explained by the model including V[O.sub.2max] is
approximately 10%. It is well known that training results in adaptations in the
heart rate response to increasing workloads (Meijer, Westerterp, & Verstappen,
1999; Wilmore et al., 1996). Therefore, it is not surprising that an indirect
measure of cardiorespiratory fitness improves the accuracy of the prediction of
energy expenditure from heart rate. This finding is in line with the s
tudy of Rennie
et al. (2001), in which sitting heart rate was found to play a significant role in the
prediction of energy expenditure from heart rate monitoring. Rennie et al. (2001)
proposed that resting heart rate when sitting was a useful proxy measure
ment for
fitness, since previous studies have found an inverse association between resting
tachycardia and maximal exercise capacity (Blair, Kannel, Kohl, Goodyear, &
Wilson, 1989), as well as a positive relationship between regular participation in
physical activity and lower resting heart rate, independent of age (Steinhaus et
al., 1988).
Hiilloskorpi et al. (1999) found that including age in the regression model did not
significantly improve the variance. This is at odds with the current study, as we
found that age did contribute significantly to the final mixed model. This
difference may partly be explained by differences in sample characteristics.
Hiilloskorpi et al. (1999) acknowledges a relatively narrow age range, with few
participants older than 5
0 years or younger than 25 years. In our study, the mean
age of the 72 men was 31 years (range 19-50 years) and that of the 43 women
was 30 years (range 22-44 years). In the study of Hiilloskorpi et al. (1999), the
mean age of the 45 men was 40 years and that of the 43 women was 38 years.
Their participants were notably older than those in the current study. In addition
to the age discrepancies between the two studies, there were also discrepancies
between the fitness of the two samples. In the current study, the mean
V[O.sub.2max] for the men was 59.2 ml x [min.sup.-1] x [kg.sup.-1] and for the
women it was 45.7 59.2 ml x [min.sup.-1] x [kg.sup.-1]; in the study of Hiilloskorpi
et al. (1999), the mean values were 48.5 and 39.5 ml x [min.sup.-1] x [kg.sup.-1]
for the men and women respectively. These demographic differences may partly
account for the differences found between the two prediction models. Rennie et
al. (2001) also found that age impacted in the regression model of physical
activity energy expenditure from heart rate. It may be argued that the sample
used to generate the prediction equation comprised a well-trained group of
individuals, but we feel that they represented a typical fitness centre population.
Maximal oxygen uptake ranged from 27.0 to 64.1 ml x [min.sup.-1] x [kg.sup.-1]
in the women and from 38.0 to 81.4 ml x [min.sup.-1] x [kg.sup.-1] in the men. It
was the intention of the present study to apply this equation to the general
exercising population and, as a result, our recruitment focused on a local fitness
centre, amateur running clubs and cycling races. While our average fitness levels
were unlike those presented in both Hiilloskorpi et al. (1999) and Rennie et al.
(2001), we have demonstrated that the inclusion of V[O.sub.2max], as
a proxy for
fitness, improves the predictability of our group-based equation.
While many other studies (Hiilloskorpi et al., 1999; Li et al., 1993; Rennie et al.,
2001; Strath et al., 2000) have used similar approaches to develop prediction
equations without individual calibration, not all of them (Hiilloskorpi et al., 1999;
Strath et al., 2000) used an independent sample for inner validation of the
developed model and, in some cases, did not report inner validation of the
developed model (Strath et al., 2000) or used the same sample for which the
original prediction equation was developed (Hiilloskorpi et al., 1999). This may
lead to elevated levels of agreement between the prediction models and
measured estimates, due to the homogeneous nature of samples. For example,
Strath et al. (2000) estimated physical activity during moderate-intensity exercise
using heart rate monitoring and reported good agreement (r = 0.87) between
measured and estimated energy expenditure; however, in this study, no inner
validation was performed on an independent sample of participants. Conversely,
Rennie et al. (2001) validated a prediction equation for physical activity levels,
developed on a sample of 789 individuals, on a smaller subset of 97 individuals.
During this inner validation, 98% of the subset was placed in the same or
adjacent quartiles during comparison of measured and estimated physical activity
levels. In the current study, we found good agreement on an independent sample
of participants. The prediction equation explained 71% of the variance in
estimated energy expenditure in an independent sample, during self-selected
cardiovascular exercise training.
Finally, for practical application the proposed equations represent an
improvement in the estimation of energy
expenditure from heart rate over existing
equations. They may be used in large population-based studies for health
purposes. Further research is needed on the simultaneous measurement of
physical activity energy expenditure and heart rate. Predictive equations that
estimate energy expenditure for health research and promotion are required for a
wider variety of activities, particularly for intermittent activity or activity conducted
at lower intensities.
Table I. Characteristics of the sample used to
develop the prediction equation (mean [+ or -] s)
Treadmill
Men (n = 22) Women (n = 24)
Age (years) 30 [+ or -] 7 30 [+ or -] 6
Weight (kg) * 76 [+ or -] 10 66 [+ or -] 11
Percent body fat * 14.5 [+ or -] 4.8 26.8 [+ or -] 5.2
V[O.sub.2max] (ml x
[kg.sup.-1] x
[min.sup.-1]) * 65.0 [+ or -] 8.6 49.0 [+ or -] 9.7
Maximal heart
rate (beats x
[min.sup.-1]) 189 [+ or -] 9 184 [+ or -] 9
Cycle ergometer
Men (n = 50) Women (n = 19)
Age (years) 31 [+ or -] 6 31 [+ or -] 6
Weight (kg) * 81 [+ or -] 13 62 [+ or -] 6
Percent body fat * 16.6 [+ or -] 4.2 23.1 [+ or -] 5.0
V[O.sub.2max] (ml x
[kg.sup.-1] x
[min.sup.-1]) * 55.3 [+ or -] 8.3 48.3 [+ or -] 8.1
Maximal heart
rate (beats x
[min.sup.-1]) 187 [+ or -] 11 185 [+ or -] 9
* P < 0.00001, differences between the sexes.
Table II. Characteristics of the sample used
for inner validation (mean [+ or -] s)
Men (n = 9) Women (n = 8)
Age (years) 29 [+ or -] 8 34 [+ or -] 10
Weight (kg) * 81 [+ or -] 14 62 [+ or -] 9
Percent body fat * 14.8 [+ or -] 5.1 26.0 [+ or -] 3.9
V[O.sub.2max]
(ml x [kg.sup.-1]
x [min.sup.-1]) * 54.3 [+ or -] 11.4 42.4 [+ or -] 5.4
Maximal heart
rate (beats x
[min.sup.-1]) 190 [+ or -] 9 178 [+ or -] 15
* P < 0.00001, differences between the sexes.
Table III. The estimates and their standard
errors for the fixed effects of the model
including fitness
Men Women
Standard Standard
Effect Estimate error Estimate error
Intercept -95.7735 9.5734 -59.3954 17.1314
Heart rate 0.6344 0.0137 0.4498 0.0165
Weight 0.3942 0.0642 0.1032 0.1166
V[O.sup.2max] 0.4044 0.0837 0.3802 0.1575
Age 0.2713 0.1120 0.2735 0.2087
Table IV. Table with type III analysis for fixed
effects of model including fitness
Degrees of
Effect freedom F-value P-value
Gender 2, 109 56.05 < 0.0001
Heart rate x gender 2, 125 1444.98 < 0.0001
Weight x gender 2, 100 19.23 < 0.0001
V[O.sub.2max] x gender 2, 101 14.57 < 0.0001
Age x gender 2, 101 3.79 0.0258
Table V. The estimates and their standard errors for the
fixed effects of the mixed model without fitness
Men Women
Standard Standard
Effect Estimate error Estimate error
Intercept -55.0969 5.5780 -20.4022 7.2318
Heart rate 0.6309 0.0137 0.4472 0.0165
Weight 0.1988 0.0619 -0.1263 0.1061
Age 0.2017 0.1180 0.0740 0.1742
Table VI. Results of type III tests for fixed
effects of model excluding fitness.
Degrees of
Effect freedom F-value P-value
Gender 2, 109 14.43 0.0002
Heart rate x gender 2, 125 1428.63 < 0.0001
Weight x gender 2, 99.9 5.86 0.0039
Age x gender 2, 100 1.55 0.2170
Table VII. Summary of the bias of the energy
expenditure (in kJ x [min.sup.-1]) estimates
and their random variation for the four sets
of estimates
Standard
Sample Model Mean deviation
Initial Including fitness -1.06 7.83
Initial No fitness -5.79 9.85
Validation Including fitness 8.19 9.19
Validation No fitness 6.27 9.65
Sample 95% limits of agreement
Initial -16.41 14.28
Initial -25.10 13.52
Validation -9.83 26.21
Validation -12.65 25.19
Note: The bias is the difference between the
predicted and the corresponding actual value
of energy expenditure.
Acknowledgments
We thank all of the participants for their cooperation. This study was funded by Polar
Electro Oy, Kempele, Finland, the Medical Research Council of South Africa, and the
Nellie Atkinson and Harry Crossley Staff Research Funds of the University of Cape
Town.
References
Atkinson, G., & Nevill, A. M. (1998). A review of the statistical methods employed to
assess measurement error (reliability) in variables relevant in sports medicine. Sports
Medicine, 26, 217-238.
Blair, S. N., Kannel, W. B., Kohl, H. W., Goodyear, N., &
Wilson, P. W. 1989. Surrogate measures of physical activity and physical fitness:
Evidence for sedentary traits of resting tachycardia, obesity, and low vital capacity.
American Journal of Epidemiology, 129, 1145-1156.
Ceesay, S. M., Prentice, A. M., Day, K. C., Murgatroyd, P. R., Goldberg, G. R., Scott,
W., & Spurr, G. B. 1989. The use of heart rate monitoring in the estimation of energy
expenditure: A validation study using indirect whole-body calorimetry. British Journal of
Nutrition, 61, 175-186.
Dumin, J. V., & Womersley, J. 1974. Body fat assessed from total body density and its
estimation from skinfold thickness: Measurements on 481 men and women aged from 16
to 72 years. British Journal of Nutrition, 32, 77-97.
Hawley, J. A., & Noakes, T. D. 1992. Peak power output predicts maximal oxygen
uptake and performance time in trained cyclists. European Journal of Applied Physiology
and Occupational Physiology, 65, 79-83.
Hebestreit, H., & Bar-Or, O. 1998. Influence of climate on heart rate in children:
Comparison between intermittent and continuous exercise. European Journal of Applied
Physiology and Occupational Physiology, 78, 7-12.
Hiilloskorpi, H., Fogelholm, M., Laukkanen, R., Pasanen, M., Oja, P., Manttari, A., &
Natri, A. 1999. Factors affecting the relation between heart rate and energy expenditure
during exercise. International Journal of Sports Medicine, 20, 438-443.
Li, R., Deurenberg, P., & Hautvast, J. G. 1993. A critical evaluation of heart rate
monitoring to assess energy expenditure in individuals. American Journal of Clinical
Nutrition, 58, 602-607.
Luke, A., Maki, K. C., Barkey, N., Cooper, R., & McGee, D. 1997. Simultaneous
monitoring of heart rate and motion to assess energy expenditure. Medicine and Science
in Sports and Exercise, 29, 144-148.
Meijer, E. P., Westerterp, K. R., & Verstappen, F. T. 1999. Effect of exercise training on
total daily physical activity in elderly humans. European Journal of Applied Physiology
and Occupational Physiology, 80, 16-21.
Noakes, T. D., Myburgh, K. H., & Schall, R. 1990. Peak treadmill running velocity
during the V[O.sub.2max] test predicts running performance. Journal of Sports Sciences,
8, 35-45.
Rennie, K. L., Hennings, S. J., Mitchell, J., & Wareham, N. J. 2001. Estimating energy
expenditure by heart-rate monitoring without individual calibration. Medicine and
Science in Sports and Exercise, 33, 939-945.
Rutgers, C. J., Klijn, M. J., & Deurenberg, P. 1997. The assessment of 24-hour energy
expenditure in elderly women by minute-by-minute heart rate monitoring. Annals of
Nutrition and Metabolism, 41, 83-88.
Spurr, G. B., Prentice, A. M., Murgatroyd, P. R., Goldberg, G. R., Reina, J. C., &
Christman, N. T. 1988. Energy expenditure from minute-by-minute heart-rate recording:
Comparison with indirect calorimetry. American Journal of Clinical Nutrition, 48, 552-
559.
Steinhaus, L. A., Dustman, R. E., Ruhling, R. O., Emmerson, R. Y., Johnson, S. C.,
Shearer, D. E., Shigeoka, J. W., & Bonekat, W. H. 1988. Cardio-respiratory fitness of
young and older active and sedentary men. British Journal of Sports Medicine, 22, 163-
166.
Strath, S. J., Swartz, A. M., Bassett, D. R., Jr., O'Brien, W. L., King, G. A., & Ainsworth,
B. E. 2000. Evaluation of heart rate as a method f
or assessing moderate intensity physical
activity. Medicine and Science in Sports and Exercise, 32 (suppl.), S465-S470.
Weir, J. B. 1990. New methods for calculating metabolic rate with special reference to
protein metabolism. Nutrition, 6, 213-221.
Wilmore, J. H., Stanforth, P. R., Gagnon, J., Leon, A. S., Rao, D. C., Skinner, J. S., &
Bouchard, C. 1996. Endurance exercise training has a minimal effect on resting heart
rate: The HERITAGE Study. Medicine and Science in Sports and Exercise, 28, 829-835.
(Accepted 1 June 2004)
L. R. KEYTEL (1), J. H. GOEDECKE (1), T. D. NOAKES (1), H. HIILOSKORPI (2),
R. LAUKKANEN (3), L. VAN DER MERWE (4), & E. V. LAMBERT (1)
(1) MRC/UCT Exercise Science and Sports Medicine Unit, University of Cape Town
Medical School, Newlands, South Africa; (2) The Urho Kavela Kekkonen Institute for
Health Promotion Research, Tampere, Finland; (3) University of Oulu, Oulu, Finland;
and (4) Biostatistics Unit, Medical Research Council of South Africa, Tygerberg, South
Africa
Correspondence: L. R. Keytel, MRC/UCT Exercise Science and Sports Medicine Unit,
Department of Human Biology, University of Cape Town Medical School, PO Box 115,
Newlands 7725, South Africa. E-mail: lkeytel@sports.uct.ac.za
t Rate Check out this amazing resource for A Target Heart Rate Calculator Tool
Improve Aerobic Threshold Naturally. EnergyDropz (TM) Increases O2 Levels in
Body.
Email this article | Print this article
COPYRIGHT 2005 E & F N Spon
Abstract
The aims of this study were to quantify the effects of factors such as mode of
exercise, body composition and training on the relationship between heart rate
and physical activity energy expenditure (measured in kJ x [min.sup.-1]) and to
develop prediction equations for energy expenditure from heart rate. Regularly
exercising individuals (n = 115; age 18-45 years, body mass 47-120 kg)
underwent a test for maximal oxygen uptake (V[O.sub.2max] test), using
incremental protocols on either a cycle ergometer or treadmill; V[O.sub.2max]
ranged from 27 to 81 x ml [kg.sup.-1] x [min.sup.-1]. The participants then
completed three steady-state exercise stages on either the treadmill (10 min) or
the cycle ergometer (15 min) at 35%, 62% and 80% of V[O.sub.2max],
corresponding to 57%, 77% and 90% of maximal heart rate. Heart rate and
respiratory exchange ratio data were collected during each stage. A mixed-model
analysis identified gender, heart rate, weight, V[O.sub.2max] and age as factors
that best predicted the relationship between heart rate and energy expenditure.
The model (with the highest likelihood ratio) was used to estimate energy
expenditure. The correlation coefficient (r) between the measured and estimated
energy expenditure was 0.913. The model therefore accounted for 83.3%
([R.sup.2]) of the variance in energy expenditure in this sample. Because a
measure of fitness, such as V[O.sub.2max], is not always available, a model
without V[O.sub.2max] included was also fitted. The correlation coefficient
between the measured energy expenditure and estimates from the mixed model
without V[O.sub.2max] was 0.857. It follows that the model without a fitness
measure accounted for 73.4% of the variance in energy expenditure in this
sample. Based on these results, we conclude that it is possible to estimate
physical activity energy expenditure from heart rate in a group of individuals with
a great deal of accuracy, after adjusting for age, gender, body mass and fitness.
Keywords: Energy expenditure, physical activity, prediction equations
Introduction
During moderate physical activity, there is a linear relationship between heart
rate and oxygen consumption. This heart rate--oxygen consumption relationship
is subject to both intra- and inter-individual variability. Heart rate may be partially
dissociated from energy expenditure by factors such as emotion, posture and
environmental conditions (Hebestreit & Bar-Or, 1998). The relationship between
heart rate and energy expenditure is linear only within a relatively narrow range
of approximately 90-150 beats x [min.sup.-1] (the so-called "flex heart rate")
during physical activity (Ceesay et al., 1989; Rennie, Hennings, Mitchell, &
Wareham, 2001; Spurr et al., 1988). During light activity or inactivity, there is
almost no slope to the relationship between heart rate and energy expenditure,
and for the purpose of measuring energy expenditure from heart rate it is
assumed that energy expenditure is equal to resting energy expenditure (Rennie
et al., 2001). A non-linear, discontinuous function has been found to be more
accurate than a linear relationship in predicting physical activity energy
expenditure from heart rate (Li, Deurenberg, & Hautvast, 1993).
Heart rate monitoring, for estimating free-living energy expenditure, has been
extensively validated using indirect calorimetry, doubly labelled water and whole-
room respirometry, and reported differences between measures range from -20%
to +25% (Luke, Maki, Barkey, Cooper, & McGee, 1997). In large groups of
people, heart rate monitoring provides one of the most efficient and economical
means of estimating energy expenditure. In addition, heart rate monitoring
provides useful insights into the type of activity being undertaken over the
measurement period. Other assessment methods, such as doubly labelled water,
can only convey the total amount of physical activity measured, whereas heart
rate monitoring provides physiological information about the type of activities
being performed and describes the nature of day-to-day variability in energy
expenditure (Hebestreit & Bar-Or, 1998; Luke et al., 1997). While whole-room
respirometry and indirect calorimetry provide physiological information about the
nature of the activity being performed, these tools are not only costly to maintain,
but often take the participant out of his or her natural environment for the duration
of the measurement period (Luke et al., 1997).
In most previous studies investigating the use of heart rate in the prediction of
energy expenditure, individual calibration of the heart rate-energy expenditure
relationship was performed (Ceesay et al., 1989; Li et al., 1993; Luke et al.,
1997; Spurr et al., 1988). Individual calibration requires that each participant
complete a progressive exercise test, during which time heart rate is
simultaneously measured, along with indirect calorimetry to estimate energy
expenditure. Two recent studies have investigated free-living energy expenditure
with heart rate monitoring utilizing prediction equations, generated on large
samples of individuals, instead of an individual calibration test (Hiilloskorpi et al.,
1999; Rennie et al., 2001). Hiilloskorpi et al. (1999) developed a prediction
equation for energy expenditure from heart rate, using multiple regression
analysis, on a sample of 87 healthy, active men and women. Factors found to
have a significant interaction with energy expenditure included age, weight and
gender. Mode of exercise (cycling versus running) did not contribute significantly
to the model.
In a more recent study, Rennie et al. (2001) developed a prediction model using
a sample of 789 individuals. Factors found to have a significant effect on the
relationship between heart rate and energy expenditure included sitting heart rate
in addition to age, weight and gender. These variables were used to predict the
slope and the intercept of the regression line between energy expenditure and
heart rate. This energy expenditure equation was then further validated on an
independent sample of 97 individuals and found to have a correlation coefficient
(r) of 0.73. Rennie et al. (2001) demonstrated the utility of developing equations
for estimating physical activity energy expenditure, from the heart rate-energy
expenditure relationship in large, representative samples of individuals, with
reasonable accuracy and the potential for wide application in epidemiological
studies.
The main aim of the present study was to further characterize the factors that
influence the relationship between energy expenditure and heart rate during
moderate to vigorous activity in regularly exercising persons. A second aim was
to develop a prediction equation for energy expenditure from heart rate, adjusting
for these factors.
Methods
Part 1: Developing the energy expenditure prediction equation
Participants. The participants were recruited from a local fitness centre, group-
based exercise programmes, running clubs and cycle races. Altogether, 127
regularly exercising men and women volunteered for the study (of which 115 had
complete data). The participants were familiar either with a cycle ergometer or
motor-driven treadmill, and ranged in age from 19 to 45 years. They were free
from any known cardiac or metabolic disorders and were not currently taking any
chronic medication. The physical characteristics of the participants are presented
in Table I. The participants were tested on two occasions, after self-selecting the
mode of exercise (cycle ergometer, n = 69; treadmill, n = 46). The Ethics and
Research Committee of the University of Cape Town, Faculty of Health
Sciences, approved the study and informed consent was obtained from all
participants before the trials began.
A second sample of regularly exercising individuals (n = 17) was subsequently
recruited, independent of the first sample, to test the validity of the prediction
model. The second sample was recruited from a local fitness centre, and
represented a wide range of ages (21-53 years), weights (51105 kg) and fitness
(V[O.sub.2max] = 34-74.3 ml x [kg.sup.-1] x [min.sup.1]).
Body composition. Body fatness was expressed as the sum of seven skinfolds
(biceps, triceps, subscapular, suprailiac, anterior thigh, abdominal and medial
calf). Percentage body fat was estimated using the equations of Dumin and
Womersley (1974).
Maximal oxygen consumption. During the first visit to the laboratory, maximal
oxygen consumption (V[O.sub.2max), maximal heart rate, and peak power
output or peak treadmill running speed were measured. Maximal oxygen uptake
was measured during either a progressive treadmill or cycle test to exhaustion.
During the treadmill test, the starting treadmill speed was 12 km x [h.sup.-1] for
the men and 10 km x [h.sup.-1] for the women, and it was increased by 0.5 km x
[h.sup.-1] every 30 s until volitional exhaustion, as described previously (Noakes,
Myburgh, & Schall, 1990). In the cycle test to exhaustion, participants were
tested on an electronically braked cycle ergometer (Lode, Gronigen, The
Netherlands). Each participant started cycling at an exercise intensity of 3.33 W x
[kg.sup.-1] body weight for 150 s, after which the work rate was increased by 50
W for a further 150 s. The exercise intensity was then increased by 25 W every
150 s up to the point of exhaustion (Hawley & Noakes, 1992). Maximal heart rate
was defined as that heart rate achieved at the point of exhaustion. During both
the treadmill and cycle tests, the participants wore a facemask attached to an
Oxycon Alpha automated gas analyser (Oxycon, Jaeger, The Netherlands).
Before each test, the gas analyser was calibrated using a Hans Rudolph 5530 3-
litre syringe and a two-point calibration technique, using a 5% C[O.sub.2]/95%
[N.sub.2] gas mixture and fresh air. The rate of oxygen consumption
(V[O.sub.2]), rate of carbon dioxide production (VC[O.sub.2]) and the respiratory
exchange ratio (RER) were calculated using conventional equations (Weir,
1990). Peak power output and peak treadmill running speed were defined as the
workload at which the participant could no longer maintain the pace of the
treadmill or maintain a cadence of 70 rev x [min.sup.-1].
Submaximal testing and estimation of energy expenditure. The participants
returned to the laboratory within a week and performed a submaximal test. The
cycle ergometer submaximal test protocol consisted of three consecutive
workloads, each lasting 15 min, during which the participants cycled at 25%, 55%
and 70% of the previously determined peak power output, corresponding to 41%,
63% and 80% of V[O.sub.2max] respectively. The submaximal treadmill protocol
consisted of three consecutive workloads, each lasting 10 min, at 35%, 50% and
70% (corresponding to approximately 41%, 63% and 80% of V[O.sub.2max]
respectively) of previously determined peak treadmill running speed. Minute-to-
minute heart rate was recorded using the Polar Vantage heart rate monitor (Polar
Electro, Finland) and respiratory exchange measurements (V[O.sub.2] and
VC[O.sub.2]) were collected and used to estimate energy expenditure, based on
the equations of Weir (1990), during the last 5 min of each of the stages. The
submaximal heart rate data from the last 5 min of each stage were used to
subsequently calculate predicted energy expenditure on the basis of individual
regression equations. Factors that were significantly correlated with heart rate or
V[O.sub.2] were used in the model to predict energy expenditure.
Part 2: Validation of prediction model on an independent sample
For the purpose of validation, the energy expenditure values from a 20-min self-
selected cardiovascular session were predicted on an independent sample of
individuals, recruited from a local fitness centre. These participants were
instructed to choose either a single 20-min cardiovascular workout or two 10-min
exercise bouts.
Participants. The 17 participants (9 males, 8 females) were free from known
cardiovascular and metabolic disorders and took part in some form of
cardiovascular physical activity at least three times a week. The participants met
the inclusion criteria of the original study and their physical characteristics are
presented in Table II.
Body composition and maximal test to exhaustion. The participants reported to
the laboratory on two different occasions within 7 days. During their first visit, the
participants had their body composition measured using the near infrared
reactance technique (Futrex Inc., Gaithersburg, MD, USA). They then performed
a maximal test to exhaustion on an electronically braked cycle ergometer (Lode,
Gronigen, The Netherlands) as previously described (Hawley & Noakes, 1992).
During the test, oxygen consumption and carbon dioxide productions were
measured as described above.
Estimation of physical activity energy expenditure. During the second visit, the
participants reported to the laboratory in a 2-h post-prandial state. They were
instructed not to engage in any strenuous physical activity during the preceding
24 h. All participants completed a 20-min cardiovascular exercise session as part
of an independent study in progress. The cardiovascular component was
performed following a 5-min warm-up consisting of 23 min of walking and 2-3 min
of light jogging. The participants then chose to either complete one 20min
continuous cardiovascular exercise session or two 10-min sessions on a self-
selected piece of fitness centre equipment. Throughout the exercise session, the
participants' heart rate, V[O.sub.2] and VC[O.sub.2] were monitored continuously
using the [K4b.sup.2] portable gas analyser (Cosmed, Italy). Minute-by-minute
energy expenditure (kJ x [min.sup.-1]) was then determined using the non-
protein caloric equivalents for oxygen. Before each test, the portable gas
analyser was calibrated using a Hans Rudolph 5530 3-1itre syringe and a two-
point calibration technique, using a 5% C[O.sub.2].sub.1]/16% [0.sub.2] gas
mixture and fresh air. The analyser outputs were processed to calculate breath-
by-breath ventilation, V[O.sub.2], VC[O.sub.2] and the respiratory exchange ratio
using conventional equations (Weir, 1990).
Statistical analysis
Initially, 127 individuals volunteered to participate in the study. The final sample
size was 115 because of incomplete heart rate and V[O.sub.2max] for 12
participants. The initial exploratory data analyses to determine factors that may
have significantly contributed to the relationship between heart rate and energy
expenditure included Box plots and scatter plots for all variables (not shown).
Univariate (means, standard deviations) and bivariate (correlation coefficients)
summary statistics were then calculated for all variables.
Based on these analyses, we fitted a mixed model for predicting energy
expenditure. The factors gender, weight, age and V[O.sub.2max] were modelled
as fixed effects, and participants as random effects, with three repeated
measurements of energy expenditure (and fixed heart rate) for each participant.
In the model, the covariance matrix between the measurements for each
participant was unstructured and compound symmetry was assumed for the
covariances between participants.
A second mixed model was fitted under the rationale that, in certain settings, a
test of maximal oxygen consumption might be impractical or not available. The
second model included all the variables and assumptions in the original model
except V[O.sub.2max]. For inner validation, both models were tested on an
independent sample of participants (n = 17), who completed 20 min of
cardiovascular exercise.
The initial exploratory analyses were performed using the Statistica data analysis
software system (version 6.1, Statsoft, Southern Africa Inc., 2002). Statistical
modelling was done with SAS[R] Proprietary Software Release 8.2 (USA).
Results
Characteristics of sample used to develop the prediction equation
The characteristics of the participants are presented in Table I. The participants
represented a wide range of morphology and fitness: age 19-45 years of age,
body weight 47-116 kg, percentage body fat 4.8-37.8% and V[O.sub.2max] 27-81
ml x [kg.sup.-1] x [min.sup.-1]. There were no differences in mean age, weight,
percentage body fat or V[O.sub.2max] between the participants who underwent
treadmill testing versus those that underwent cycle ergometer testing. There
were significant differences in weight, percentage body fat and V[O.sub.2max]
between the sexes (Table I, P < 0.00001).
Characteristics of sample used for inner validation
The characteristics of the participants are presented in Table II. The participants
in this sample were similar to those used in the original study and represented a
broad range in body composition. Percentage body fat ranged from 9.4 to 21.6%
in the men and from 21.6 to 30.6 % in the women. Similarly, there was a wide
range in the performance data, with V[O.sub.2max] ranging from 38.7 to 73.8 ml
x [kg-.sup.-1] x [min.sup.-1] in the men and from 34.3 to 49.6 ml x [kg.sup.-1] x
[min.sup.-1] in the women.
The participants in both samples were equally matched for age and weight. The
participants (males and females combined) in the original study were slightly fitter
(mean V[O.sub.2max] 53.5 [+ or -] 0.5 ml. [kg-.sup.-1]. [min.sup.-1]) than those
who took part in the validation study (mean V[O.sub.2max] 48.1 [+ or -] 0.5 ml x
[kg.sup.-1] x [min.sup.-1]); this difference was not statistically significant.
Prediction equations of energy expenditure from heart rate: Mixed-model
analysis
A mixed model was used to derive the following equation for predicting physical
activity energy expenditure (EE):
EE = -59.3954 + gender x (-36.3781 + 0.271 x age + 0.394 x weight + 0.404
V[O.sub.2max] + 0.634x heart rate) + (1 - gender) x (0.274 x age + 0.103x weight
+ 0.380x V[O.sub.2max] + 0.450 x heart rate)
where gender = 1 for males and 0 for females. Table III shows the above model
in a different format. The likelihood ratio test for goodness-of-fit [chi square] =
262.73 on five degrees of freedom with P < 0.0001. The results of type III tests
for the fixed effects in the mixed model are presented in Table IV. The degrees of
freedom for the F-tests were calculated using Satterthwaite's method.
In Figure 1, the measured energy expenditure is regressed against estimated
energy expenditure. The correlation coefficient (r) is 0.913, so [R.sup.2] = 83.3%
of the variation in measured energy expenditure in the sample is explained by the
model.
[FIGURE 1 OMITTED]
A second model, which contained no measure of fitness, was also fitted. The
final prediction equation for energy expenditure using age, gender, weight and
heart rate was:
EE = gender x (-55.0969 + 0.6309 x heart rate + 0.1988 x weight + 0.2017 x age)
+ (1 - gender) x (-20.4022 + 0.4472 x heartrate - 0.1263 x weight + 0.074 x age)
where gender = 1 for males and 0 for females. Table V shows the above model
in a different format. The likelihood ratio test for goodness-of-fit [chi square] =
360.68 on five degrees of freedom with P < 0.0001. The results of type III tests
for the fixed effects in the mixed model are given in Table VI. The degrees of
freedom for the F-tests were calculated using Satterthwaite's method.
In Figure 2, the measured energy expenditure is regressed against estimated
energy expenditure. The coefficient of correlation 0.857, so [R.sup.2] = 73.4% of
the variation in measured energy expenditure in the sample is explained by the
model.
[FIGURE 2 OMITTED]
Independent sample analysis for inner validation
Data from an independent sample of 17 participants (8 females, 9 males) were
used to validate both models. Predicted energy expenditure using the first model,
which included a measure of fitness (V[O.sub.2max]), correlated with measured
energy expenditure during self-selected cardiovascular fitness training (r = 0.836,
P < 0.0001; Figure 3). Using the second model for measuring energy
expenditure, with no measure of fitness, the correlation coefficient was 0.77 (P <
0.0001) (Figure 4).
[FIGURE 3-4 OMITTED]
Agreement
Because we used a mixed model (with random participant effects), we had to use
maximum likelihood estimation instead of least squares. The result is that even
the estimates for the initial sample which was used to develop the equations are
slightly biased. The bias of the estimates and their random variation for the four
sets of estimates are summarized in Table VII. The bias is the difference
between the predicted and the corresponding actual value of energy expenditure,
and the 95% limits of absolute agreement were calculated as described in
Atkinson and Nevill (1998). It is interesting to note that the bias in the initial
sample is on average in the opposite direction to that for the validation sample.
The fact that the agreement limits become wider down the table is completely
logical. We believe that these limits are narrow enough for the underlying models
to be of practical use.
Discussion
In this study, we demonstrated that physical activity energy expenditure during
moderate- to high-intensity exercise may be predicted with good accuracy in a
group of individuals varying widely in age, fitness and morphology, without the
need for individual calibration. This study denotes an improvement over existing
studies in the estimation of physical activity energy expenditure using heart rate
monitoring. The proposed model (using heart rate, age, weight, gender and level
of fitness V[O.sub.2max])) accounted for 70% of the variation in observed energy
expenditure in an independent sample of people completing a self-selected 20-
min cardiovascular exercise session.
Previous studies (Li et al., 1993; Rutgers, Klijn, & Deurenberg, 1997) have cited
poor agreement between energy expenditure estimated using heart rate
monitoring and measured energy expenditure. These prediction equations were
developed on small samples, not representative of the population to which the
equation was to be applied. Rutgers et al. (1997) developed a prediction
equation based on the heart rate and energy expenditure data acquired from 13
elderly individuals. The authors concluded that the use of heart rate monitoring to
measure energy expenditure was inaccurate over 3 days of measurement, citing
large discrepancies between energy expenditure estimation using the individual
calibration curve and a group curve. Li et al. (1993) also reported poor
agreement for the estimation of energy expenditure using heart rate monitoring
between group and individually derived estimates. Once again, this sample was
relatively small, consisting of only 40 persons.
The current study represents an improvement over existing studies (Hiilloskorpi
et al., 1999; Rennie et al., 2001) that used heart rate monitoring to estimate
physical activity energy expenditure, without individual calibration. Previously,
Rennie et al. (2001) used the variables that significantly interacted with energy
expenditure to predict the slope, intercept and the heart rate flex point for
measured versus predicted physical activity energy expenditure. In that study,
the variables of sitting heart rate, age, weight and gender were found to have a
significant impact on the slope, intercept and heart rate flex point. These
investigators were then able to use the slope and intercept of the linear model to
place 98% of the participants in their sample in either the same or adjacent
quartiles for the measured and estimated physical activity levels. Their model has
implications for physical activity classification in epidemiological models. In the
current study, we derived linear equations, based on mixed-model analyses.
These equations yield predictions that correlate significantly with the test sample
as well as the independent validation sample.
Previously, Hiilloskorpi et al. (1999) developed an equation to predict energy
expenditure using the variables of heart rate, age, weight and gender. They
showed that the mode of exercise, cycling versus running, did not significantly
affect the final prediction of energy expenditure. We also found that the mode of
exercise did not affect the estimation of energy expenditure, and therefore
suggest that the proposed equation may be used for both running and cycling
activities. During our inner validation study, we even found good agreement with
other models of continuous activity, such as stationary rowing ergometry and
stationary stair-climbing activities.
Hiilloskorpi et al. (1999) did not include any measure of physical fitness or
V[O.sub.2max] in their prediction equation, citing a need to produce an equation
for estimating energy expenditure independent of laboratory testing. We found
that when a measure of the level of cardiorespiratory fitness such as
V[O.sub.2max] is included, the accuracy of the prediction improved. The
correlation coefficients (r) of the study sample were 0.913 for the model
V[O.sub.2max] with and 0.857 for the model without V[O.sub.2max]. The
increase in variation explained by the model including V[O.sub.2max] is 83.4%-
73.4% = 10%. The correlation coefficients of the validation sample were 0.836 for
the model with V[O.sub.2max] and 0.77 for the model without V[O.sub.2max].
The increase in variation explained by the model including V[O.sub.2max] is
approximately 10%. It is well known that training results in adaptations in the
heart rate response to increasing workloads (Meijer, Westerterp, & Verstappen,
1999; Wilmore et al., 1996). Therefore, it is not surprising that an indirect
measure of cardiorespiratory fitness improves the accuracy of the prediction of
energy expenditure from heart rate. This finding is in line with the study of Rennie
et al. (2001), in which sitting heart rate was found to play a significant role in the
prediction of energy expenditure from heart rate monitoring. Rennie et al. (2001)
proposed that resting heart rate when sitting was a useful proxy measurement for
fitness, since previous studies have found an inverse association between
resting tachycardia and maximal exercise capacity (Blair, Kannel, Kohl,
Goodyear, & Wilson, 1989), as well as a positive relationship between regular
participation in physical activity and lower resting heart rate, independent of age
(Steinhaus et al., 1988).
Hiilloskorpi et al. (1999) found that including age in the regression model did not
significantly improve the variance. This is at odds with the current study, as we
found that age did contribute significantly to the final mixed model. This
difference may partly be explained by differences in sample characteristics.
Hiilloskorpi et al. (1999) acknowledges a relatively narrow age range, with few
participants older than 50 years or younger than 25 years. In our study, the mean
age of the 72 men was 31 years (range 19-50 years) and that of the 43 women
was 30 years (range 22-44 years). In the study of Hiilloskorpi et al. (1999), the
mean age of the 45 men was 40 years and that of the 43 women was 38 years.
Their participants were notably older than those in the current study. In addition
to the age discrepancies between the two studies, there were also discrepancies
between the fitness of the two samples. In the current study, the mean
V[O.sub.2max] for the men was 59.2 ml x [min.sup.-1] x [kg.sup.-1] and for the
women it was 45.7 59.2 ml x [min.sup.-1] x [kg.sup.-1]; in the study of Hiilloskorpi
et al. (1999), the mean values were 48.5 and 39.5 ml x [min.sup.-1] x [kg.sup.-1]
for the men and women respectively. These demographic differences may partly
account for the differences found between the two prediction models. Rennie et
al. (2001) also found that age impacted in the regression model of physical
activity energy expenditure from heart rate. It may be argued that the sample
used to generate the prediction equation comprised a well-trained group of
individuals, but we feel that they represented a typical fitness centre population.
Maximal oxygen uptake ranged from 27.0 to 64.1 ml x [min.sup.-1] x [kg.sup.-1]
in the women and from 38.0 to 81.4 ml x [min.sup.-1] x [kg.sup.-1] in the men. It
was the intention of the present study to apply this equation to the general
exercising population and, as a result, our recruitment focused on a local fitness
centre, amateur running clubs and cycling races. While our average fitness levels
were unlike those presented in both Hiilloskorpi et al. (1999) and Rennie et al.
(2001), we have demonstrated that the inclusion of V[O.sub.2max], as a proxy for
fitness, improves the predictability of our group-based equation.
While many other studies (Hiilloskorpi et al., 1999; Li et al., 1993; Rennie et al.,
2001; Strath et al., 2000) have used similar approaches to develop prediction
equations without individual calibration, not all of them (Hiilloskorpi et al., 1999;
Strath et al., 2000) used an independent sample for inner validation of the
developed model and, in some cases, did not report inner validation of the
developed model (Strath et al., 2000) or used the same sample for which the
original prediction equation was developed (Hiilloskorpi et al., 1999). This may
lead to elevated levels of agreement between the prediction models and
measured estimates, due to the homogeneous nature of samples. For example,
Strath et al. (2000) estimated physical activity during moderate-intensity exercise
using heart rate monitoring and reported good agreement (r = 0.87) between
measured and estimated energy expenditure; however, in this study, no inner
validation was performed on an independent sample of participants. Conversely,
Rennie et al. (2001) validated a prediction equation for physical activity levels,
developed on a sample of 789 individuals, on a smaller subset of 97 individuals.
During this inner validation, 98% of the subset was placed in the same or
adjacent quartiles during comparison of measured and estimated physical activity
levels. In the current study, we found good agreement on an independent sample
of participants. The prediction equation explained 71% of the variance in
estimated energy expenditure in an independent sample, during self-selected
cardiovascular exercise training.
Finally, for practical application the proposed equations represent an
improvement in the estimation of energy expenditure from heart rate over
existing equations. They may be used in large population-based studies for
health purposes. Further research is needed on the simultaneous measurement
of physical activity energy expenditure and heart rate. Predictive equations that
estimate energy expenditure for health research and promotion are required for a
wider variety of activities, particularly for intermittent activity or activity conducted
at lower intensities.
Table I. Characteristics of the sample used to
develop the prediction equation (mean [+ or -] s)
Treadmill
Men (n = 22) Women (n = 24)
Age (years) 30 [+ or -] 7 30 [+ or -] 6
Weight (kg) * 76 [+ or -] 10 66 [+ or -] 11
Percent body fat * 14.5 [+ or -] 4.8 26.8 [+ or -] 5.2
V[O.sub.2max] (ml x
[kg.sup.-1] x
[min.sup.-1]) * 65.0 [+ or -] 8.6 49.0 [+ or -] 9.7
Maximal heart
rate (beats x
[min.sup.-1]) 189 [+ or -] 9 184 [+ or -] 9
Cycle ergometer
Men (n = 50) Women (n = 19)
Age (years) 31 [+ or -] 6 31 [+ or -] 6
Weight (kg) * 81 [+ or -] 13 62 [+ or -] 6
Percent body fat * 16.6 [+ or -] 4.2 23.1 [+ or -] 5.0
V[O.sub.2max] (ml x
[kg.sup.-1] x
[min.sup.-1]) * 55.3 [+ or -] 8.3 48.3 [+ or -] 8.1
Maximal heart
rate (beats x
[min.sup.-1]) 187 [+ or -] 11 185 [+ or -] 9
* P < 0.00001, differences between the sexes.
Table II. Characteristics of the sample used
for inner validation (mean [+ or -] s)
Men (n = 9) Women (n = 8)
Age (years) 29 [+ or -] 8 34 [+ or -] 10
Weight (kg) * 81 [+ or -] 14 62 [+ or -] 9
Percent body fat * 14.8 [+ or -] 5.1 26.0 [+ or -] 3.9
V[O.sub.2max]
(ml x [kg.sup.-1]
x [min.sup.-1]) * 54.3 [+ or -] 11.4 42.4 [+ or -] 5.4
Maximal heart
rate (beats x
[min.sup.-1]) 190 [+ or -] 9 178 [+ or -] 15
* P < 0.00001, differences between the sexes.
Table III. The estimates and their standard
errors for the fixed effects of the model
including fitness
Men Women
Standard Standard
Effect Estimate error Estimate error
Intercept -95.7735 9.5734 -59.3954 17.1314
Heart rate 0.6344 0.0137 0.4498 0.0165
Weight 0.3942 0.0642 0.1032 0.1166
V[O.sup.2max] 0.4044 0.0837 0.3802 0.1575
Age 0.2713 0.1120 0.2735 0.2087
Table IV. Table with type III analysis for fixed
effects of model including fitness
Degrees of
Effect freedom F-value P-value
Gender 2, 109 56.05 < 0.0001
Heart rate x gender 2, 125 1444.98 < 0.0001
Weight x gender 2, 100 19.23 < 0.0001
V[O.sub.2max] x gender 2, 101 14.57 < 0.0001
Age x gender 2, 101 3.79 0.0258
Table V. The estimates and their standard errors for the
fixed effects of the mixed model without fitness
Men Women
Standard Standard
Effect Estimate error Estimate error
Intercept -55.0969 5.5780 -20.4022 7.2318
Heart rate 0.6309 0.0137 0.4472 0.0165
Weight 0.1988 0.0619 -0.1263 0.1061
Age 0.2017 0.1180 0.0740 0.1742
Table VI. Results of type III tests for fixed
effects of model excluding fitness.
Degrees of
Effect freedom F-value P-value
Gender 2, 109 14.43 0.0002
Heart rate x gender 2, 125 1428.63 < 0.0001
Weight x gender 2, 99.9 5.86 0.0039
Age x gender 2, 100 1.55 0.2170
Table VII. Summary of the bias of the energy
expenditure (in kJ x [min.sup.-1]) estimates
and their random variation for the four sets
of estimates
Standard
Sample Model Mean deviation
Initial Including fitness -1.06 7.83
Initial No fitness -5.79 9.85
Validation Including fitness 8.19 9.19
Validation No fitness 6.27 9.65
Sample 95% limits of agreement
Initial -16.41 14.28
Initial -25.10 13.52
Validation -9.83 26.21
Validation -12.65 25.19
Note: The bias is the difference between the
predicted and the corresponding actual value
of energy expenditure.
Acknowledgments
We thank all of the participants for their cooperation. This study was funded by Polar
Electro Oy, Kempele, Finland, the Medical Research Council of South Africa, and the
Nellie Atkinson and Harry Crossley Staff Research Funds of the University of Cape
Town.
References
Atkinson, G., & Nevill, A. M. (1998). A review of the statistical methods employed to
assess measurement error (reliability) in variables relevant in sports medicine. Sports
Medicine, 26, 217-238.
Blair, S. N., Kannel, W. B., Kohl, H. W., Goodyear, N., &
Wilson, P. W. 1989. Surrogate measures of physical activity and physical fitness:
Evidence for sedentary traits of resting tachycardia, obesity, and low vital capacity.
American Journal of Epidemiology, 129, 1145-1156.
Ceesay, S. M., Prentice, A. M., Day, K. C., Murgatroyd, P. R., Goldberg, G. R., Scott,
W., & Spurr, G. B. 1989. The use of heart rate monitoring in the estimation of energy
expenditure: A validation study using indirect whole-body calorimetry. British Journal of
Nutrition, 61, 175-186.
Dumin, J. V., & Womersley, J. 1974. Body fat assessed from total body density and its
estimation from skinfold thickness: Measurements on 481 men and women aged from 16
to 72 years. British Journal of Nutrition, 32, 77-97.
Hawley, J. A., & Noakes, T. D. 1992. Peak power output predicts maximal oxygen
uptake and performance time in trained cyclists. European Journal of Applied Physiology
and Occupational Physiology, 65, 79-83.
Hebestreit, H., & Bar-Or, O. 1998. Influence of climate on heart rate in children:
Comparison between intermittent and continuous exercise. European Journal of Applied
Physiology and Occupational Physiology, 78, 7-12.
Hiilloskorpi, H., Fogelholm, M., Laukkanen, R., Pasanen, M., Oja, P., Manttari, A., &
Natri, A. 1999. Factors affecting the relation between heart rate and energy expenditure
during exercise. International Journal of Sports Medicine, 20, 438-443.
Li, R., Deurenberg, P., & Hautvast, J. G. 1993. A critical evaluation of heart rate
monitoring to assess energy expenditure in individuals. American Journal of Clinical
Nutrition, 58, 602-607.
Luke, A., Maki, K. C., Barkey, N., Cooper, R., & McGee, D. 1997. Simultaneous
monitoring of heart rate and motion to assess energy expenditure. Medicine and Science
in Sports and Exercise, 29, 144-148.
Meijer, E. P., Westerterp, K. R., & Verstappen, F. T. 1999. Effect of exercise training on
total daily physical activity in elderly humans. European Journal of Applied Physiology
and Occupational Physiology, 80, 16-21.
Noakes, T. D., Myburgh, K. H., & Schall, R. 1990. Peak treadmill running velocity
during the V[O.sub.2max] test predicts running performance. Journal of Sports Sciences,
8, 35-45.
Rennie, K. L., Hennings, S. J., Mitchell, J., & Wareham, N. J. 2001. Estimating energy
expenditure by heart-rate monitoring without individual calibration. Medicine and
Science in Sports and Exercise, 33, 939-945.
Rutgers, C. J., Klijn, M. J., & Deurenberg, P. 1997. The assessment of 24-hour energy
expenditure in elderly women by minute-by-minute heart rate monitoring. Annals of
Nutrition and Metabolism, 41, 83-88.
Spurr, G. B., Prentice, A. M., Murgatroyd, P. R., Goldberg, G. R., Reina, J. C., &
Christman, N. T. 1988. Energy expenditure from minute-by-minute heart-rate recording:
Comparison with indirect calorimetry. American Journal of Clinical Nutrition, 48, 552-
559.
Steinhaus, L. A., Dustman, R. E., Ruhling, R. O., Emmerson, R. Y., Johnson, S. C.,
Shearer, D. E., Shigeoka, J. W., & Bonekat, W. H. 1988. Cardio-respiratory fitness of
young and older active and sedentary men. British Journal of Sports Medicine, 22, 163-
166.
Strath, S. J., Swartz, A. M., Bassett, D. R., Jr., O'Brien, W. L., King, G. A., & Ainsworth,
B. E. 2000. Evaluation of heart rate as a method for assessing moderate intensity physical
activity. Medicine and Science in Sports and Exercise, 32 (suppl.), S465-S470.
Weir, J. B. 1990. New methods for calculating metabolic rate with special reference to
protein metabolism. Nutrition, 6, 213-221.
Wilmore, J. H., Stanforth, P. R., Gagnon, J., Leon, A. S., Rao, D. C., Skinner, J. S., &
Bouchard, C. 1996. Endurance exercise training has a minimal effect on resting heart
rate: The HERITAGE Study. Medicine and Science in Sports and Exercise, 28, 829-835.
(Accepted 1 June 2004)
L. R. KEYTEL (1), J. H. GOEDECKE (1), T. D. NOAKES (1), H. HIILOSKORPI (2),
R. LAUKKANEN (3), L. VAN DER MERWE (4), & E. V. LAMBERT (1)
(1) MRC/UCT Exercise Science and Sports Medicine Unit, University of Cape Town
Medical School, Newlands, South Africa; (2) The Urho Kavela Kekkonen Institute for
Health Promotion Research, Tampere, Finland; (3) University of Oulu, Oulu, Finland;
and (4) Biostatistics Unit, Medical Research Council of South Africa, Tygerberg, South
Africa
Correspondence: L. R. Keytel, MRC/UCT Exercise Science and Sports Medicine Unit,
Department of Human Biology, University of Cape Town Medical School, PO Box 115,
Newlands 7725, South Africa. E-mail: lkeytel@sports.uct.ac.za
... Several other caveats also exist when considering HR as a surrogate for effort intensity, namely intra-and inter-personal variability, sensitivity to other factors such as emotional arousal and environmental conditions, and the non-linearity of the relationship between HR and metabolic rate beyond the moderate HR range 26,37 . For example, the notable inter-individual variability observed in HR response times underscores the challenge of creating a one-size-fits-all controller and suggests that individualized settings could enhance effectiveness. ...
Article
Full-text available
Despite their potential, exoskeletons have not reached widespread adoption in daily life, partly due to the challenge of seamlessly adapting assistance across various tasks and environments. Task-specific designs, reliance on complex sensing and extensive data-driven training often limit the practicality of the existing control strategies. To address this challenge, we introduce an adaptive control strategy for hip exoskeletons, emphasizing minimal sensing and ease of implementation. Using only insole pressure and heart rate (HR) sensing, the controller modulates assistance across various locomotor tasks. We evaluated this strategy with twelve able-bodied participants in a real-world scenario including level walking, stairs, and inclines. The controller successfully adapted assistance timing and amplitude to different activities. This resulted in effort intensity reductions (measured by oxygen uptake) of up to 12.6% compared to walking with no exoskeleton, and up to 25.5% compared to walking with the exoskeleton in zero-torque mode. Cardiodynamic response of HR, although delayed, proved sufficient for adaptation in tasks lasting longer than around 45 s, and delay-induced limitations primarily affected brief bouts of abrupt change in intensity. However, we found discernible patterns in HR shortly after the onset of such changes that can be exploited to improve responsiveness. Our findings underscore the potential of HR as a promising measure of user effort intensity, encouraging future research to explore its integration into advanced adaptive algorithms.
... In parallel with the exploration of additional physiological signals, this research investigates alternative predictive models to elucidate further information embedded within HR or its combination with other physiological parameters. Historically, linear regression has been commonly employed to predict oxygen uptake [1,27,28]. However, this method has significant limitations in predicting . ...
Article
Full-text available
Oxygen uptake (V˙O2) is an essential metric for evaluating cardiopulmonary health and athletic performance, which can barely be directly measured. Heart rate (HR) is a prominent physiological indicator correlated with V˙O2 and is often used for indirect V˙O2 prediction. This study investigates the impact of HR placement on V˙O2 prediction accuracy by analyzing HR data combined with the respiratory rate (RESP) and minute ventilation (V˙E) from three anatomical locations: the chest; arm; and wrist. Twenty-eight healthy adults participated in incremental and constant workload cycling tests at various intensities. Data on V˙O2, RESP, V˙E, and HR were collected and used to develop a neural network model for V˙O2 prediction. The influence of HR position on prediction accuracy was assessed via Bland–Altman plots, and model performance was evaluated by mean absolute error (MAE), coefficient of determination (R2), and mean absolute percentage error (MAPE). Our findings indicate that HR combined with RESP and V˙E (V˙O2HR+RESP+V˙E) produces the most accurate V˙O2 predictions (MAE: 165 mL/min, R2: 0.87, MAPE: 15.91%). Notably, as exercise intensity increases, the accuracy of V˙O2 prediction decreases, particularly within high-intensity exercise. The substitution of HR with different anatomical sites significantly impacts V˙O2 prediction accuracy, with wrist placement showing a more profound effect compared to arm placement. In conclusion, this study underscores the importance of considering HR placement in V˙O2 prediction models, with RESP and V˙E serving as effective compensatory factors. These findings contribute to refining indirect V˙O2 estimation methods, enhancing their predictive capabilities across different exercise intensities and anatomical placements.
... Energy expenditure during the post-exercise recovery period was estimated from HR using a mixed model analysis that accounts for age, sex, body mass, and . VO 2 peak and which has previously been shown to estimate energy expenditure with a high degree of accuracy [34]. ...
Article
Full-text available
A healthy body mass contributes to a positive quality of life, and for overweight/obese individuals, weight loss of even modest proportions improves health-related outcomes. A novel approach to promoting body mass regulation is to pair exercise with high altitude, thereby upregulating metabolic processes and increasing caloric expenditure. As an added measure of body mass regulation, data suggest that high altitude stimulates the production of the appetite-suppressing hormone leptin. Issues arise, however, given that high altitude compromises aerobic exercise capacity. Whereas exercising at high altitude may compromise exercise intensity and duration, recovering at high altitude retains the integrity of exercise while still potentially conferring the benefits of the low oxygen environment on energy expenditure and energy intake. The purpose of this study was to perform a proof-of-concept pilot test on whether post-exercise exposure to a simulated high altitude influenced acute energy balance. Twelve healthy men and women ran for 30 min at a moderate intensity on two separate occasions. Following exercise, participants recovered for 30 min while breathing either sea level air or low oxygen air simulating high altitude (equivalent to 4500 m elevation). Blood samples and hunger ratings were collected pre-exercise and post-recovery. Heart rate was recorded throughout exercise and recovery and used to calculate caloric expenditure. Post-exercise energy expenditure was significantly higher (p = 0.03) following high altitude recovery (139 ± 15 kcal) compared to sea-level recovery (98 ± 11 kcal). Participants reported a lower desire to eat when they recovered in the high altitude environment (p = 0.01), though post-recovery leptin concentrations were similar between the two conditions. Post-exercise exposure to a simulated high altitude environment represents a promising method for increasing daily caloric expenditure and lowering appetite. Given the pilot nature of this study, future research is needed to address the question of high altitude recovery on a larger sample over a longer time period and with robust measures of caloric expenditure.
... A short calibration procedure has been included in the protocol, it consists of three minutes of rest, in order to acquire the resting heart rate and two minutes of walking (at the subject's natural speed) to acquire heart rate during normal activity, by combining heart rate signal with anthropometric data the HRmax of each subject is evaluated. An evaluation of energy expenditure is carried out as reported in the literature (Keytel et al., 2005). It is possible to calculate a heart rate threshold, the value of the "Flex Heart Rate" below which energy expenditure is assumed to be equal to resting energy expenditure. ...
Conference Paper
Healthcare workers experience physically, emotionally stressful situations, are exposed to human suffering, experience pressure from interactions with patients and family members, and are under constant threat of infection, injury and stress. Healthcare workers are at greater risk of developing stress-related mental disorders, such as depression or post-traumatic stress disorder (Braquehais et al., 2023, doi:10.1016/j.mcna.2022.04.004). The COVID-19 pandemic has highlighted the need for healthcare organizations to ensure the well-being of healthcare workers. Indeed, more stressful working hours, the fear of being infected and the need to ensure immediate decision-making have significantly increased the risk of burnout, depression, anxiety and insomnia. In the USA and Europe, a series of regulations have been issued to preserve the health of workers, specific to the workload linked to the various tasks and in the literature work-related stress indices have been evaluated in the healthcare sector, linked to muscle disorders skeletal injuries due to patient handling, for nurses and personal care/assistance workers. However, biomechanical overload and the risk of damage to the musculoskeletal system are only one aspect linked to the health of the worker: the dimension of work stress has a significant role in the general well-being of the worker during his activities and a methodology for the objective assessment of mental and physical workload in work environments.In this study we propose the use of a wearable system (WWS by Smartex) compatible with work activity, to monitor and extract significant information on the workload, due to the physical demand and the physiological response to stress, on a sample of physiotherapists. The system consists of a t-shirt for the continuous detection of a cluster of physiological parameters, that can be stored and processed during work. The system detects an ECG lead via integrated textile electrodes, the respiratory signal through the measurement of thoracic movement, posture, and physical activity recognition, via an Inertial and Magnetic Measurement Unit (IMMU), integrated into the RUSA device, a portable data logger dedicated to the acquisition, processing, storage and/or transmission of data.The RUSA is connected to the garment through a simple plug and can be easily unplugged when necessary. The T-shirt is absolutely similar to a common underwear. The base yarn is composed by antibacterial materials to guarantee a safe and prolonged use. The sensors are made of fibers that are directly woven during the production process to be fully integrated in the garment without discontinuity. The shirts come in male and female version and in several sizes to fit the largest number of users. The data acquired by the RUSA are processed on board to extract the following parameters: Hearth Rate (HR), HR Variability, RR interval, signal quality, Breathing rate, activity classification, activity intensity. The RUSA can save data on a Flash Memory (microSD), transmit data via Bluetooth® 2.1), save and transmit them simultaneously, without losing information in case of interruption of wireless transmission.A pilot study has been performed on 11 physiotherapists, engaged in XX sessions. During the study, the cluster of physiological data have been combined with a set of meta-data related to the work session such as the type of intervention (i.e. neurological rehabilitation, orthopaedic rehabilitation, etc), the level of physical impairment of the patient (according to modified Rankin score and Communicative disability scale), the working place etc.). As well as to the results of NASA questionnaire that has been administrated after each acquisition section to the physiotherapists. Preliminary results on the stress level will be presented, in parallel to evaluate the use of the IMMU platform for the overload of the musculoskeletal system, research on the posture evaluation in the rehabilitation workplace has been performed. The accuracy of a single IMMU to retrieve trunk angles was assessed by comparison with stereophotogrammetry. The results revealed that the IMMU is adequately effective in determining sagittal angles but has limitations in assessing lateral and transverse angles in a natural and uncontrolled environment.
... The purpose of this study was to investigate the effects of 8-week hiking bench training on the cardiorespiratory and muscular responses of highly trained single-handed dinghy sailors during hiking The VO 2 max is the gold standard for assessing aerobic capacity and is used as an indicator of superior performance in endurance athletes of mixed ability levels (Millet et al., 2003). A lower VO 2 or energy consumption during a given task implies heightened efficiency, making athletes more likely to achieve outstanding performance (Keytel et al., 2005). HR is an important parameter for evaluating an individual's response to training loads and is also one of the most cost-effective methods for assessing energy expenditure and physiological changes (Roos et al., 2013). ...
Article
Full-text available
To investigate the effects of 8‐week hiking bench training on cardiorespiratory and muscular responses of highly trained sailors during hiking emulation. Twenty‐four sailors were assigned into two groups: the hiking bench training group (HTG, n = 12) and the control group (CG, n = 12). Both groups maintained their regular training with the HTG performed two additional hiking bench training sessions per week for 8 weeks, while the CG performed an equivalent duration of on‐water sailing training. Physiological responses were assessed by performing four successive 3‐min hiking bouts on a sailing emulation ergometer before and after the 8‐week training period. Comparing the pretest, both groups exhibited a significant decrease (p < 0.05) in the percentage of maximal oxygen uptake (%VO2max) and maximal heart rate (%HRmax); the HTG experienced a greater decrease in %VO2max in bouts 2 and 3. The root mean square (RMS) of rectus femoris (RF), vastus lateralis (VL), rectus abdominis (RA), and external oblique decreased significantly (p < 0.05), whereas the mean power frequency (MPF) of RF, VL, and RA exhibited an increasing trend. The RMS of RF and RA in HTG were lower than those in CG in the initial three bouts; VL and EA in HTG were lower than those in CG in bouts 1 and 2 (p < 0.05). The MPF of RA in HTG was significantly increased in bouts 2, 3, and 4 (p < 0.05). Eight‐week hiking bench training could improve hiking economy and the activation of lower limb and trunk muscles delaying the onset of fatigue in sailors.
Article
Full-text available
Providing feedback on energy expenditure (EE) may be an important tool to support obesity prevention among manual wheelchair users (MWU). This paper presents a data-driven approach for estimating EE based on data collected from 40 participants (20 MWU and 20 controls without disability) across different activities (lying, sitting and wheelchair propulsion at different intensities). We extracted features from heart rate, inertial measurement units (IMU), and individual personal characteristics to develop activity classification and EE estimation algorithms and investigate the influence of personal characteristics on EE estimates. Support Vector Machines were selected as classifiers, while Support Vector Regressors, Gaussian Processes, Random Forest, XGBoost, and Neural Networks were selected as regression models. High classification accuracy was achieved with minor confusion between activities and EE estimation results showed high generalisation capabilities of the trained models on unseen participants. We explored the impact of changing the position of the IMU on the accuracy of EE estimations. We recommend the wrist as the primary location for sensor placement. It provides a good trade-off between accuracy, high wear compliance rates and the possibility of integrating our algorithms in already existing wearable devices. Our findings showed that including data collected from people without disabilities to develop EE estimation algorithms for MWU did not enhance the estimation accuracy. In conclusion, data-driven algorithms based on wearable sensors and personal characteristics are effective for activity classification and EE estimation in MWU, but need to be personalized and further developed for daily life settings to be ecologically valid.
Article
Full-text available
Bu araştırmanın amacı Catapult markasına ait farklı modellerdeki mikro elektro mekanik sistemlerin (i) ortalama ve (ii) maksimal kalp atım hızı ölçümündeki geçerlilik ve güvenirliliğini incelemektir. Araştırmaya 30 elit akademi futbol oyuncusu (Yaş: 22± 3.4 yıl; Boy: 175 ± 8.4cm; Kilo: 69 ± 4.5 kg) gönüllü olarak katılmıştır. Katılımcıların kalp atım hız değerlerini farklı atım aralıklarında ölçebilmek için yürüme, jog, koşu ve sprint gibi farklı aktiveleri içeren takım sporu simülasyon döngüsü uygulatılmıştır. Takım sporu simülasyon döngüsü esnasında sporcuların ortalama ve maksimum kalp atım hızları Catapult marka mikro elektro mekanik sistemleri aracılığıyla tüm test esnasında kayıt altına alınmıştır. Katılımcılara eş zamanlı olarak dört farklı Catapult mikro elektro mekanik sistem ekipmanı giydirilmiştir. İki tane Catapult S7 yeleğe gömülü ve iki tanede H 10 Polar bandı aracılığıyla ölçüm yapan Catapult X7 ünitesinden eş zamanlı olarak veri elde edilmiştir. Bu yöntemle, modeller arası ve modellerin birimleri arasındaki farklılıklar incelenmiştir. Modellerin ve ünite farklılıklarını tespit etmek için tek yönlü ANOVA analizi yapılmıştır. Aynı zamanda farklılık hangi model ve üniteden kaynaklandığı tespit etmek için bonferroni post hoc analizi yapılmıştır. Modeller ve üniteler arasındaki ilişkiyi tespit etmek için pearson korelasyon analizi yapılmıştır. Tek yönlü ANOVA ortalama kalp atım hızları ölçümünde (F= 0,203; p=0,894; η²= 0,002) ve maksimum kalp atım hızı ölçümünde (F= 0,262; p=0,852; η²= 0,002) model ve birimler arasında istatistiksel olarak anlamlı farklılık görülmemiştir. Pearson korelasyon analizinde tüm karşılaştırmalar arasında (r ≤ 0,9) neredeyse mükemmel ilişikliler tespit edilmiştir. Araştırmanın bulguları, farklı Catapult modelleri ve ünitelerinin kalp atım hızı ölçümlerinde tutarlı ve güvenilir olduğunu göstermektedir. Bu iki sistemin birbiri yerine kullanılabileceği düşünülmektedir.
Article
Full-text available
Porous soft bioelectronics have attracted significant attention due to their high breathability, long‐term biocompatibility, and other unique features inaccessible in nonporous counterparts. However, fabricating high‐quality multimodal bioelectronic components that operate stably under strain on porous substrates, along with integrating microfluidics for sweat management, remains challenging. In this study, cellulose nanofibrils (CNF) are explored, biomass‐derived sustainable biomaterials, as nanofibril interfaces with unprecedented interfacial robustness to enable high‐quality printing of strain‐resilient bioelectronics on porous substrates by reducing surface roughness and creating mechanical heterogeneity. Also, CNF‐based microfluidics can provide continuous sweat collection and refreshment, crucial for accurate biochemical sensing. Building upon these advancements, a multimodal porous wearable bioelectronic system is further developed capable of simultaneously detecting electrocardiograms and glucose and beta‐hydroxybutyrate in sweat for monitoring energy metabolism and consumption. This work introduces novel strategies for fabricating high‐quality, strain‐resilient porous bioelectronics with customizable multimodalities to meet arising personalized healthcare needs.
Article
Full-text available
1. A modified heart rate (HR) method for predicting total energy expenditure (TEE) was cross-validated against whole-body calorimetry (CAL). Minute-by-minute HR was converted to energy expenditure (EE) using individual calibration curves when HR exceeded a pre-determined ‘FLEX’ value designed to discriminate periods of activity. (‘FLEX’ HR was defined as the mean of the highest HR during rest and the lowest HR during the lightest imposed exercise.) Sedentary EE (below FLEX) was calculated as the mean EE during lying down, sitting and standing at rest. Sleeping EE was calculated as basal metabolic rate (BMR) predicted from standard equations. 2. Calibration curves of oxygen consumption v. HR for different postures at rest and during exercise were obtained for twenty healthy subjects (eleven male, nine female); mean r 0.941 (SD 0.04). The mean FLEX HR for men and women were 86 (sd 10) and 96 (SD 6) beats/min respectively. 3. Simultaneous measurements of HR and EE were made during 21 h continuous CAL, which included 4 x 30 min imposed exercise (cycling, rowing, stepping, jogging). HR exceeded FLEX for a mean of 98 (SD 41) min. Mean TEE by CAL (TEE. CAL) was 8063 (sd 1445) kJ. 4. The HR method yielded a mean non-significant underestimate in TEE (TEE. HR) of 1.2 (sd 6.2)% (range−11.4 to + 10.6 %). Regression of TEE. HR ( y ) v. TEE. CAL ( X ) yielded Y = 0.868 X +927 kJ, r 0.943, se of the estimate 458 kJ, n 20. 5. The satisfactory predictive power and low cost of the method makes it suitable for many field and epidemiological applications.
Article
Full-text available
Total daily energy expenditure (TDEE) and energy expended in activity (EAC) were estimated by the minute-by-minute heart-rate method in 22 (16 men, 6 women) individually calibrated subjects and compared with values obtained by whole-body indirect calorimetry. Subjects followed four activity protocols during the 22 h in the calorimeter; no exercise (n = 6) and 2 (n = 5), 4 (n = 4), and 6 (n = 6) 30-min bouts of exercise on a bicycle ergometer at varying intensities. There were no statistically significant differences between the two methods in TDEE or EAC in any of the sex or protocol groupings. The regression of TDEE by heart rate on TDEE in the calorimeter was y = 0.92x + 1.0 MJ; (r = 0.87, SEE = 0.91 MJ). The heart-rate method also follows the varying activity patterns of individuals and can be used to closely estimate the TDEE and EAC of even small (n = 4-6) groups of subjects. In the present measurements, it gave a maximum error of TDEE for individuals of +20% and -15%.
Article
Full-text available
The prediction of energy expenditure (EE) from minute-by-minute heart rate (HR) recording was evaluated in 40 female workers, with emphasis on the inter- and intraindividual variation of this method. The results show that the relationship between EE and HR varied greatly between and within subjects. The inter- and intraindividual CVs were 14-18% and 11-20%, respectively. The poor limits of agreement in EE (mean difference +/- 2 SD) between group and individual calibration curves indicate that estimated EE from group curves is inferior to estimated EE from individual curves. Therefore, it is concluded that to have the best estimates of individual EE, only individual calibration curves should be used. At different occasions these curves should be generated again. Because the limits of agreement in EE between individual curves based on 18 and 9 different activities were wide (-2399 to 1817 kJ/16 h), it is preferable to have a wide range of different activities in the individual calibration procedure.
Article
The purposes of this study were firstly to determine the relationship between the peak power output (W peak) and maximal oxygen uptake (VO2max) attained during a laboratory cycling test to exhaustion, and secondly to assess the relationship betweenW peak and times in a 20-km cycling trial. One hundred trained cyclists (54 men, 46 women) participated in the first part of this investigation. Each cyclist performed a minimum of one maximal test during whichW max andVO2max were determined. For the second part of the study 19 cyclists completed a maximal test for the determination ofW peak, and also a 20-km cycling time trial. Highly significant relationships were obtained betweenW peak andVO2max (r=0.97,P<0.0001) and betweenW peak and 20-km cycle time (r= –0.91,P<0.001). Thus,W peak explained 94% of the variance in measuredVO2max and 82% of the variability in cycle time over 20 km. We concluded that for trained cyclists, theVO2max can be accurately predicted fromW peak, and thatW peak is a valid predictor of 20-km cycle time.
Article
Studies on physical activity, physical fitness, and health have been hampered because of invalid, unreliable, or impractical measures of physical activity. This report examines the validity of sedentary traits (resting tachycardia, obesity, and low vital capacity) as predictors of physical fitness as assessed by a maximal treadmill exercise test. Study participants were women ( n = 3,94.3) and men ( n = 15,627) with at least one visit to the Cooper Clinic in Dallas, Texas. Association of the sedentary traits with physical fitness was examined by multiple regression analyses. Sedentary traits were associated with physical fitness in all age and sex groups, accounting for 12–40% of the variance in treadmill time. When smoking, a simple physical activity index, and sedentary traits were included in a model to predict physical fitness, R 2 values ranged from 0.20 to 0.53 in women and 0.45 to 0.61 in men and were significant at p < 0.0001. These models account for approximately twice as much variance in physical fitness as has been reported previously. The addition of sedentary traits measurements to a simple physical activity index provides a valid estimate of physical fitness in epidemiologic studies.
Article
Twenty specialist marathon runners and 23 specialist ultra-marathon runners underwent maximal exercise testing to determine the relative value of maximum oxygen consumption (VO2max), peak treadmill running velocity, running velocity at the lactate turnpoint, VO2 at 16 km h-1, % VO2max at 16 km h-1, and running time in other races, for predicting performance in races of 10-90 km. Race time at 10 or 21.1 km was the best predictor of performance at 42.2 km in specialist marathon runners and at 42.2 and 90 km in specialist ultra-marathon runners (r = 0.91-0.97). Peak treadmill running velocity was the best laboratory-measured predictor of performance (r = -0.88(-)-0.94) at all distances in ultra-marathon specialists and at all distances except 42.2 km in marathon specialists. Other predictive variables were running velocity at the lactate turnpoint (r = -0.80(-)-0.92); % VO2max at 16 km h-1 (r = 0.76-0.90) and VO2max (r = 0.55(-)-0.86). Peak blood lactate concentrations (r = 0.68-0.71) and VO2 at 16 km h-1 (r = 0.10-0.61) were less good predictors. These data indicate: (i) that in groups of trained long distance runners, the physiological factors that determine success in races of 10-90 km are the same; thus there may not be variables that predict success uniquely in either 10 km, marathon or ultra-marathon runners, and (ii) that peak treadmill running velocity is at least as good a predictor of running performance as is the lactate turnpoint. Factors that determine the peak treadmill running velocity are not known but are not likely to be related to maximum rates of muscle oxygen utilization.
Article
Physiological profiles are described for 30 healthy young (20-31 years) and 30 healthy older (50-62 years) men. Half of the individuals in each group reported that during the previous five years they participated frequently in strenuous physical exercises; the other half reported sedentary lifestyles. A treadmill exercise test was used to determine maximal aerobic power (VO2 max). Heart rate and blood pressure were measured during rest, maximal exercise and recovery. The active older men demonstrated significantly lower resting heart rates, lower resting systolic and diastolic blood pressures, higher VO2 max, lower maximal exercise diastolic blood pressure and lower recovery heart rates than the age-matched sedentary men. Compared with the young sedentary men, the older active men had lower resting heart rates and higher VO2 max, walked longer on the treadmill, had lower recovery heart rates and weighed less. Older active men also had higher VO2 max levels than young sedentary men. In summary, physiological profiles of the older active men more closely resembled profiles of active men who were 30 years younger than those of older sedentary men. These results emphasize the range of benefits associated with exercise.
Article
1. Skinfold thicknesses at four sites – biceps, triceps, subscapular and supra-iliac – and total body density (by underwater weighing) were measured on 209 males and 272 females aged from 16 to 72 years. The fat content varied from 5 to 50% of body-weight in the men and from 10 to 61% in the women. 2. When the results were plotted it was found necessary to use the logarithm of skinfold measurements in order to achieve a linear relationship with body density. 3. Linear regression equations were calculated for the estimation of body density, and hence body fat, using single skinfolds and all possible sums of two or more skinfolds. Separate equations for the different age-groupings are given. A table is derived where percentage body fat can be read off corresponding to differing values for the total of the four standard skinfolds. This table is subdivided for sex and for age. 4. The possible reasons for the altered position of the regression lines with sex and age, and the validation of the use of body density measurements, are discussed.
Article
This study determined the effects of a 20-wk endurance training program (The HERITAGE Family Study) on resting heart rate (HRrest). HRrest was obtained on a sample of 26 men and 21 women during sleep; during resting metabolic rate and resting blood pressure measurement periods in the early morning following a 12-h fast and 24-h post-exercise; and at rest prior to a maximal bout of exercise. Following training, the subjects exhibited a 16.0 +/- 9.4% (mean +/- SD) increase in VO2max (P < 0.05), but the HRrest for each of the resting conditions was decreased by only 1.9 to 3.4 bpm (P < 0.05), or an average across the three conditions of 2.7 bpm. In a larger sample of 253 HERITAGE subjects, HRrest obtained only at the time of the resting blood pressure measurement decreased by only 2.6 bpm, while VO2max increased 17.7 +/- 10.0%. It is concluded that there is a significant, but small, decrease in resting heart rate as a result of 20 wk of moderate- to high-intensity endurance training; which suggests a minimal alteration in either, or both, intrinsic heart rate and autonomic control of HRrest.