0 ratings0% found this document useful (0 votes) 256 views9 pagesCycle Ergometry
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content,
claim it here.
Available Formats
Download as PDF or read online on Scribd
0195:9131/2072205.070482 000
MEDGGNE AND SCIENCE IN SPORTS AND EXERCISE
Copright © Te60 by tne Anca Coupe Sports Wedcine
vt 22, 80.5
Pantin USA,
Accurate prediction of VOsmax in cycle
ergometry
THOMAS W. STORER, JAMES A. DAVIS, and
VINCENT J. CAIOZZO
Laboratory of Exercise Science,
El Camino College,
Torrance, CA 90506 (T.W.S.):
Laboratory of Applied Physiology,
California State University,
Long Beach, CA 90840 (J.A.D.); and
Neuromuscular Research Laboratory,
Division of Orthopedics,
Department of Surgery,
University of California,
Irvine, CA 92717 (V.I.C.)
ABSTRACT
STORER, T. W., J. A. DAVIS, and V. J. CAIOZZO. Accurate
prediction of VOsex in cycle ergometry. Med. Sci. Sports Exerc,
Vol, 22, No. 5, pp. 704-712, 1990, Numerous equations exist for
predicting VOsms from the duration (an analog of maximal work
ate, Wow) of a treadmill graded exercise test (GXT). Since a similar
equation for cycle ergometry (CE) was not available, we saw the need
to develop such an equation, hypothesizing that CE VOmy, could be
accurately predicted due to its more direct relationship with W. Thus,
healthy, sedentary males (N'= 115) and females (= 116), aged 20-
70 yr, were given a 15 W-min” CE GXT, The following multiple
linear regression equations which predict VOtaw, (ml-min") from
the independent variables of Wana (W), body weight (kg), and age (yr)
were derived from our subjects
‘Males: Y = 10.51 (W) + 6.35 (kg) ~ 10.49 (yr) + 519.3 ml-min“
939, SEE = 212 ml-min™
Females: ¥ = 9.39 (W)-+ 7.7 (kg) ~ 5.88 (yr) + 136.7 ml-mi
R =0.932, SEE= 147 ml min”
Using the 95% confidence limits as examples of worst case errors,
‘our equations predict VOsnw, to within 10% ofits true value. Internal
(double cross-validation) and external cross-validation analyses
Yylelded r values ranging between 0.920 and 0.950 for the male and
female regression equations. These results indicate that use of the
‘equations generated in this study fora 15 W-min™' CE GXT provides
accurate estimates of VO3n
WORK RATE, FUNCTIONAL CAPACITY, PREDICTION
EQUATIONS
The evaluation of exercise tolerance is important in
assessing cardiopulmonary health status or in identify-
ing the potential for endurance performance. An im-
portant index of exercise tolerance is the amount of
‘Accepted fox pblestonSetebey,1989,
704
oxygen that can be taken up by the working muscles
during maximal exercise (31). Typically, a graded ex-
ercise test (GXT) performed on a treadmill or cycle
ergometer is used to determine this important param-
eter of aerobic function. Ideally, the GXT includes
measurements of ventilation and gas exchange so that
maximal oxygen uptake (VOzmax) can be determined
directly. However, the assessment of VO2ma: during a
GXT is technically difficult and time consuming and
involves expensive equipment. In 1973, Bruce et al. (8)
introduced an equation to predict VOzmax based on
duration of exercise on the Bruce treadmill protocol.
The rationale for this approach is that, in normal
subjects, VO» is linearly related to work rate and work
rate is linearly related to test duration. Hence, VO, is
linearly related to test duration. This equation, with its
relatively low standard error of estimate, is widely used
today. Other investigators (6,15,16,26,27) have also
developed equations which predict VOzms from tread-
mill test duration.
From a physiological perspective, it is preferable to
predict VO from a maximal work rate instead of
test duration since VOs is directly linked to work rate
but is only indirectly linked to test duration through
work rate. However, work rates can neither easily nor
accurately be determined for treadmill exercise. This
may explain the widespread use of test duration instead
of maximal work rate to predict VOzma. from this
testing mode. Unlike the treadmill, cycle ergometers
provide easily and accurately quantifiable work rates
(21,33) and yield VO, values at submaximal work rates
that are less variable than those obtained from treadmill
exercise (5, p. 338). Both of these factors suggest thatACCURATE PREDICTION OF VOzmax
the estimation of VO2max from maximal work rate may
be improved for cycle ergometer exercise when com-
pared to treadmill exercise,
At present, estimates of VOzmax using cycle ergometry
are made from submaximal test data which usually
employ the linear relationship between a series of sub-
maximal work rates and the corresponding heart rates
(4.29). This requires an extrapolation of the heart rate
vs work rate regression line to the maximal heart rate
(usually age-predicted). This heart rate is then used to
predict a maximal work rate. Finally, the predicted
miaximal work rate is used to predict VOoma. Correc-
tion factors, some of which were derived empirically
(5, p. 349), may be used to adjust the predicted VOamux
value for the effects of age, gender, and/or body size
(2,4,5). Although these tests may be attractive in terms
of their ease of administration and lack of required
maximal effort, they have been shown to have relatively
poor predictive accuracy (10-27% error) with high
standard errors of estimate (10,28).
While there are numerous cycle ergometer GXT
protocols in the literature, a frequently used protocol
consists of the work rate increasing 15 W- min"! (23,
34). However, an equation for estimating VO2na, does
not exist for this cycle ergometer GXT protocol. Hence,
it was the purpose of this study to develop an equation
which would predict VO2ma, using the maximal work
rate achieved during a standard cycle ergometer graded
exercise test as the principal prediction variable. These
equations could then be used to assess cardiopulmonary
functional capacity when direct measurement of VO,
is not feasible and/or 10 provide normal standards
against which a measured VO2max could be evaluated,
METHODS
In order to characterize the cardiopulmonary re-
sponses to cycle ergometer exercise, 115 males and 116
females, aged 20-70 yr, underwent graded exercise
testing. These sedentary, nonsmoking, and apparently
healthy subjects were recruited in a manner that re-
sulted in approximately 20 subjects per gender in the
following age decades: 20-29, 30-39, 40-49, 50-59,
and 60-69 yr. A subject was considered sedentary if he
or she did not exercise more than once per week in any
activity designed 10 develop cardiopulmonary fitness
(1). To be considered a nonsmoker, subjects must not
have smoked for the past 15 yr. Subjects were taking
no medications and were free from cardiovascular and/
or pulmonary disease at the time of the study. Subject
characteristics are displayed in Table 1.
All subjects were recruited and tested at two institu-
tions: El Camino College (ECC) in Torrance, CA, and
the University of California at Irvine (UCD. A third
institution, Harbor-UCLA Medical Center (H-UCLA),
705
‘TABLE 1, Subject characterises and stout by ests.
Wales Rm
‘go Weight Height’ Age Weight” Wott
TestSio Nh eS) (em) a)
0p
K 38 42s G04 78 SH a7 he tsa
Pe a
argo 2047 4082 soi-iod abr as 10-175
vot
x
81 425 834 1798 62 486 G27 1660
280 48 107 70 132 108 65
Range 20-69 $8-120 169-195 20-70 41-94 144-108
P OOS >005 <005 3005 >005 <005
Tota
X15 425 925 1788 16 4437 t68
49 48 2973 143 9 66
Range 20-69 49-192 161-196 20-70 1-101 144-108
EGC = E\ Camino Cooge; UCI = Univers of Calfmia, iv.
served as a reference site in order to ensure that the
cardiopulmonary data collected at ECC and UCI were
not different. The H-UCLA data collection system has
been previously described and validated (7). Prior to
the start of the present investigation, 15 subjects (nine
from ECC and six from UCI) performed incremental
cycle ergometer exercise tests both at their own insti-
tution and at the reference laboratory. There was good
agreement between values for maximal oxygen uptake
(VOzmox) at each test site and the reference site. Paired
tests revealed no significant differences (P > 0.05)
between mean VO2ma values obtained at H-UCLA
(2.60 I-min™) and at ECC (2.67 I-min"!); ¢ = 2.44.
Similarly, no significant differences were observed be-
tween the mean VO2max Values measured at H-UCLA,
(2.89 1- min“) and at UCI (2.95 |-min7!); ¢= 1.93. The
correlation coefficients for VOzmax between H-UCLA
and the two test sites were r= 0.97 (ECC) and r = 0.96
(uch.
After giving their informed consent and completing
routine screening tests (e.g,, resting 12-lead ECG and
blood pressure), subjects cycled for 4 min at 0 W.
‘Thereafter, the cycle ergometer work rate increased in
15-W-min"' increments until the subject reached his
or her limit of tolerance. The pedal rate was maintained
at 60 rpm throughout the test as confirmed by a pedal
revolution counter. Subjects were verbally encouraged
by test administrators to provide a true maximal effort.
A Monark ergometer was used at each test site to
provide the work rates. The ergometers were calibrated
at L-wk intervals.
Pulmonary ventilation and gas exchange were meas-
ured either breath-by-breath with an on-line data ac-
quisition system (Exertrend, Alpha Technologies, La-
guna Hills, CA) or at 30-5 intervals using a semi-
automated, mixing chamber system (9). The Exertrend
system exployed a turbine volume transducer (13) for
the measurement of expired minute ventilation (Ve). A
continuous sample of expired gas, drawn from the706
mouthpiece, was transported via heated sampling lines
to electronic gas analyzers for the measurement of
oxygen (O;) and carbon dioxide (CO;) concentrations
(Applied Electrochemistry $3-A, Sunnyvale, CA, and
Beckman LB-2, Fullerton, CA, respectively). The com-
puter of this system sampled each breath for its volume
and Oz and CO» concentrations at the rate of 50 Hz.
‘After time alignment, the ventilation and gas exchange
data were cross-multiplied to yield oxygen uptake
(VOz), carbon dioxide output (VCO;), and the res
tory exchange ratio (R). The gas analyzers were cali-
brated before each test with gases of known concentra-
tion. The turbine volume transducer was calibrated
each day with a calibration syringe. The breath-by-
breath data were averaged and displayed over 30-s
intervals to correspond with the data format of the
semi-automated system.
With the semi-automated system, inspired minute
ventilation (V;) was measured with a Parkinson-Cowan
dry gas spirometer fitted with an optical encoder. A
digital panel meter displayed these ventilation measure-
ments. Expired ventilation was determined from V;
using the Haldane transformation (12). The expired air
was directed into a 5-1 mixing chamber. Samples of this
expired air were analyzed for their fractional concentra
tions of O; (FEo,) and CO: (FEco,). Gas analyzers,
identical to those used at ECC, were used for these
analyses and were calibrated before each test with gases
of known concentrations. Inspired ventilation, FEo,,
and FEco, were entered manually every 30 $ into a
personal computer. Standard equations were used to
calculate VO;, VCO2, and R. In both equipment con-
figurations, VOsmex and maximal work rate (Wins) Were,
determined for a complete 30-s Collection interval a
maximal exercise.
‘Throughout each test, the ECG was monitored con-
tinuously by use of the CMS lead placement. Electro-
cardiograms were recorded during the last 10 s of each
minute, Blood pressures were taken every 3 min
throughout the test.
‘While test duration (an analog of W,,,.) is known to
be well correlated with VOzmay, the age (3), body weight
(5, p. 321; 33), and gender (2,3) of subjects are also
associated with VOzmax. Hence, multiple linear regres-
sion, using VO2ma (ml-min“!) as the dependent vari-
able and Wrua (W), body weight (ke), and age (yr) as
independent variables, was used to generate gender-
specific equations for prediction of VO2max- Similarly,
a gender-independent equation, utilizing the entire
sample of 231 male and female subjects, was developed
for the purpose of comparison with other equations in
the literature. Each independent variable was tested for
its significance as a predictor of the dependent variable
(24, pp. 66-72).
The regression equations were validated using both
internal and external (independent) cross-validation,
MEDICINE AND SCIENCE IN SPORTS AND EXERCISE
procedures. The internal cross-validation was per-
formed using the double cross-validation procedure
described by Kerlinger and Pedhazur (24, p. 284).
Briefly, this method required the original sample to be
divided randomly into two groups. eg., A and B. A
prediction equation was then derived from each of the
subgroups. Following this, the equation developed from
the A subgroup was applied to the B subgroup in order
to predict the dependent variable (VOzmax in this case)
for the B subgroup; the predicted value was then com-
pared to the measured value obtained on the B
subgroup using simple regression analysis. This proce-
dure was repeated by applying the prediction equation
developed from the B subgroup to the A subgroup and
once again correlating the predicted value with the
measured value. The equations are valid if the two
simple correlation coefficients (r) are similar (24, p.
284), A second comparison was made between the
multiple correlation coefficients (R) of the two entire
samples (males and females) and the R values of their
respective subgroups. In this latter analysis, a shrinkage
formula was applied to the calculated R in order to
correct for sampling bias (30, p. 11), This correction
was small (about 0.04 of an R unit) for our data because
of our large sample size and the small number of
independent variables. Again, if the R values are simi-
lar, the equations developed from the entire samples
are considered valid (24, p. 284).
The external cross-validation consisted of applying
the male and female regression equations found in this
study to an independent sample of 36 subjects (26
males, 10 females). The mean + SD age (yt), weight
(kg), VOzmax (mI-min™), and Wimax (W) for the 26 male
subjects were 37.4 + 9.3 yr, 85.6 + 17.4 ke, 2762 + 643
ml-min™, and 200.5 + 41.9 W, respectively, and were
26.1 + 10.4 yr, 60.5 + 9.8 kg, 2153 + 493 ml-min™,
and 169.1 + 35.5 W, respectively, for the ten female
subjects. Fifteen of the cross-validation subjects were
studied 2 yr after the main study was completed using
a calibrated Monark cycle ergometer (model 868) fol-
lowing the same test protocol and using the Exertrend
breath-by-breath system previously described. The re-
maining 21 subjects had been tested 9 yr earlier, in
duplicate, using this same protocol on a Godart electri-
cally braked cycle ergometer. Oxygen uptake, VCO2,
and R were determined from measurements of minute
ventilation and O; and CO; concentrations in the ex-
pired air using a breath-by-breath system previously
described (11). Nine of the 21 subjects tested 9 yr earlier
were re-evaluated, in duplicate, after 9 wk of endurance
training,
As with the subjects of the present study, the cross-
validation subjects were sedentary nonsmokers and
were apparently healthy. They gave their informed
consent and completed the same screening tests.
Stepwise multiple linear regression equations wereACCURATE PREDICTION OF VOzmax
707
TABLE 2. Values at maximal exercise by tes ite fo the males, females and males pls fale.
ales Females
Wow VO Hie We VOrn, Haw
Test Site wy {enl-min™") opm) Res - (w) {rnl-min) (bpm) Prawn
coc «X= ato ~~=CreS=—~S—CGO SSCS 18 «16884 S~S«TGOSC«
0 369 5926 40008 2 3579 165 OND
Neo MINS 44706960620 1.08 580 mo 4004.00
max= 7700 gag m0 = D201
uc! x 206.8 27704 1740 121 126.6 1563.4 1686 126
435, ste 19307 220 3867 m4 00
nao woo Hinze 102 Wms 80 ago 320
20 atez0 1001 m0 Daa
2005 00530053005 <0 3005005 > 05
Toss armas AT 1.28 wor tetra
“12 as 180009 31 8 vo a0
Nets MN 10301128088 mmo 201.00
mao aie20 10018 i070 20D 182
Tota, vee 21003732128
les 21 74a 7 008
and 580 m0 801.00
Females NAX= 2904162010
ECC = El Camino College; UCI = University of California, Invine. ~
generated using the Statistical Programs for the Social TABLE. Equator resto of VO i
Sciences (SPSS-X). The ability of the equations devel- Equation Independent SEE (a
oped from the present study to predict VO2may in the pewter ia Ste em varieties ex Coeicets|Conet8 a maleic)
cross-validation samples was assessed by Pearson Prod- 1 Miles Nese W) pee aceon!
a " N= 115 Body weight (kg) 6.35
uct-Moment correlation coefficients and by standard Age hy) 1048
errors of estimate (SEE). Possible significant differences 2 Fama Woah ase nus asses
between the means of measured vs predicted VOamax in Wm ti6 Boayweghttg) 771
the external cross-validation study were examined by a Age tt 5.86
dependent ¢-test. The P < 0.05 level was chosen to 3 ce a m ed 4034 097 1870
i i i lmdes Body wosht
indicate statistical significance. ees a
Gender 2502
RESULTS
Values at maximal exercise for W (W), VO; (ml-
min“), HR (bpm), and Rem (the respiratory exchange
ratio, ie., VCO;/VO;, at termination of exercise) are
presented in Table 2. Evidence for the achievement of
VOrmay included a) Riera in excess of 1,00 (ef. 25) and
b) attainment of age-predicted HRnax- None of the 231
subjects had an Rien Jess than 1,00; the mean + SD
value was 1.25 + 0.09. The mean difference between
measured HRmax and that predicted as 220 minus age
for the 231 subjects was 2.8 bpm. This is well within
the £10 bpm SEE reported for the relationship between
maximal heart rate and age (18).
Stepwise multiple linear regression, with VO2max (m+
min”) as the dependent variable vs the independent
variables of Wax (W), body weight (kg), and age (yt),
‘was used to generate equations specific for each gender
and one equation independent of gender. These equa-
tions, along with their multiple correlation coefficients
(R)and SEE, are presented in Table 3. Each independ-
ent variable in all three equations was a significant
predictor of the dependent variable. The F-test for the
increment in R? resulting from the addition of each
* Gender cosct sO for males and 1 for females,
independent variable to the equation was significant at
P< 0.05 (24). Thus, while Wax was selected as the
first predictor variable in these equations (accounting
for 84.1%, 81.7%, and 91.6% of the common variance
in the male, female, and gender-independent equations,
respectively), statistically significant contributions to
the prediction of VOznax were made with the addition,
of body weight and age.
The correlation coefficients and the corresponding
SEE for the measured vs predicted VOzmax Values were
0.94 and 196 ml-min”' for the males, 0.93 and 134
ml-min for the females, and 0.97 and 194 ml-min~!
for the gender-independent equations, respectively. Dis-
played in Figure 1 are the individual data points for
predicted VOzma from the gender-specific equations
(males in Panel a and females in Panel b) and the
gender-independent equation (males plus females in
Panel c), with the corresponding values for measured
WOema
The results of the internal, double cross-validation
analysis of the male gender-specific equation yielded r708
5000
4 4000-
i
$3000:
i
2000
1000
oi
9 ¥e0o 700030004000 $000
zd VO, rox (mina "
Moore, mos (nia)
4000. Ponel
€ 3000.
£
2 2000.
2 000 wae
° tooo 2000 ©3000-4000
Weasured Vy mex (mimia™*)
000, Pane! ¢
‘4000
i
z
‘y 3000.
i
$2000 \
$1000 2 eaee*
o a
a” 1doo "ado sd” 4do0 S000
eesred 0 moe i”
Figure 1—Invidual datapoints for comparison of predicted VOrna
to measured VOsmu. Panel a: data points for all 115 male subjects.
Panel b; data points for all 116 female subjects, Panel c: data points
for all 231 male plus female subjects, Equations used to predict
VOtne: ia Panels a, b, and ¢ correspond to equations 1, 2, and 3,
respectively, as reported in Table 3.
values between 0.937 and 0.942, with SEE between 186
ml-min“! and 209 ml-min”'. The multiple correlation
coefficients generated in this analysis were 0.943 and
0.938; after correction for shrinkage (30), they were
0.939 and 0.934. For the female gender-specific equa-
tion, the r values were between 0.920 and 0.937, with
SEE between 122 ml-min“' and 159 ml-min“', The
multiple correlation coefficients were 0.928 and 0.936;
after correction for shrinkage, they were 0.924 and
0.943.
Application of the gender-specific equations devel-
oped in this study to the 36 external cross-validation,
MEDICINE AND SCIENCE IN SPORTS AND EXERCISE
subjects is depicted in Figure 2. The correlation coeffi-
cient between the predicted VO2max and the measured
VOrma, was 0.95, with a SEE of 176 ml-min™, The
ference between the mean values for the measured
VOrma, and the predicted VO2ma (20 ml-min™,
0.8%) was not statistically significant.
To further evaluate the utility of these equations
(generated and cross-validated on sedentary subjects),
wwe applied the male gender-specific equation to nine of
the male external cross-validation subjects. Although
previously sedentary, these nine men had undergone 9
wk of cycle ergometer endurance exercise training, im-
proving VOzma, on average, by 25% (11). Figure 3
illustrates that these nine subjects remained on essen
tially the same regression line (similar slopes and y-
intercepts) but, predictably, moved further up that same
line as a result of their improved functional capacity.
Statistical comparison of the two slopes (14) revealed a
5000
'g 4000.
i
z
3000.
i
$* 2000
3 1000
i
0 =
7doo 2000-3000” «doo Sabo
Measured dy mx (mein)
Fare 2—Indvius datapoints for 36 cross-validation subjects (26
‘males and 10 females) for comps
tured VOrnn- The gender-sp
3) were used to estimate VOs,
ison of predicted VOnnax to meas-
‘equations (equation 1 or 2 in Table
PreTaiing mt
4000} eos
Post Teining —
rose
Predicted Voy mex (rin?)
4
® 1600” 2000” 3000 4000 ” 5000
Measured 0p max (mtn?)
Figure 3—Comparison of predicted vs measured VOseu in
slopes of the two regression lines were not si
suggesting that the male gender-specific equation is
improved ca
jorespiratory fitness consequent to endurance exercise
training.ACCURATE PREDICTION OF VO2max
DISCUSSION
. The principal finding of the present study is that
‘VOsmux can be accurately predicted from a cycle ergom-
eter graded exercise test. This is indicated by the high
correlations and low standard errors of estimate of our
equations. The accuracy of these equations can be
examined by computing the worst-case errors, i.e. the
errors that would be obtained if we assumed that the
true value was at the lower or upper 95% confidence
limit (CL), For example, our male gender-specific equa-
tion predicts a VO2max Of 2774 ml-min~! for our average
male subject who had a Winax of 207.1 W, weighed 82.5
kg, and was 42.5 yr of age. As the SEE for this equation
is 212 ml-min”", the lower and upper 95% confidence
limits of predicted VOzma. for our male subjects are
2354 ml-min™! and 3194 ml-min“', respectively. The
accuracy of this prediction is then defined as
s SEE
Tower 95% CL vatue * 8° vpper 955 CL value * '°%
val
For our sample of male subjects, the accuracy is 9.0%
and 6.6% at the lower and upper 95% CL, respectively.
For our sample of female subjects, the accuracy is
10.9% and 7.6% at the lower and upper 95% CL,
respectively. Our gender-independent equation yields
accuracies of prediction at the lower and upper 95%
confidence limits of 7.8% and 5.9%, respectively, for
the males and 15.1% and 9.4% for the females. Hence,
our equations, on average, predict VOomax to within
10% of its true value for 95 out of every 100 subjects.
As a comparison, we computed the accuracy of the
Bruce et al. (8) general equation using this technique
and obtained values of 10.4% and 7.3% for the males
and 14.7% and 9.3% for the females at the lower and
upper 95% CL, respectively. Hence, our gender-specific
equations and the gender-independent equation of
Bruce et al. (8) yield similar results,
The demonstrated accuracy of our equations is due
primarily to the high correlation between VO2max and
Wax. Bruce et al. (8) demonstrated that maximal time
on the treadmill (analogous to Wms) was the first
variable selected in their stepwise multiple regression
analysis for prediction of VO2mas (in units of ml-kg™!
min~), Time on the treadmill accounted for 82% of
the common variance in their study. Similarly, Wax
(W) was the first variable chosen to predict VOrmax for
both the males and the females of the present study
(Table 4). To facilitate a more direct comparison of our
data to that of Bruce et al. (8), we pooled our male and
female subjects to generate a gender-independent equa-
tion. Once again, Wimax was the first variable chosen in
the stepwise multiple regression procedure, accounting
for 92% of the common variance (Table 4). The larger
coefficient of determination (R?) demonstrated in the
present study suggests that VOzmax is more closely re-
709
TABLE 4, Comparison of mute egresson statis for equations generated in
{he present study and for he Bruce et (8) gende-ndependen quatn,
P ‘nerementin
‘Equation Number and st
‘Sample ble Entered ROR
1 Waa) 1 O97 oBt
Males, present study Ago) 2 0881 0888 aes
Weghttig) «3099 0882 0.016
2 Woe 1 0904 0817
Females, present Aga(y) 2 0919 08s 0.028
study Weight (ig) «90832 0888 0.003,
3 Wa) 1 0957 0916
Maes and females, Ago(y) 2 0969 0927 ott
present study Weight ic) «30985 0839 0.06
Gender 4 0871 o983 0015
Bruceat al, males and Duraton(min) 10907 0422
females Gender 2 0800 0846 0.003
Age) 3 0923 0852 0.008
Weighttig) 40925 0.855 0.008
lated to Winax in cycle ergometry than in treadmill test
duration, which itself is highly correlated to treadmill
work rate. The problem with treadmill test duration is,
that two people can accomplish the same maximal time
on the treadmill, but one may have a higher measured
‘VOsms than the other because the formers less efficient
in performing maximal treadmill exercise (19). Because
the efficiency of cycle ergometry is virtually the same
for everyone, two people that reach the same Wrox
should have the same VOzma, assuming that their
capacity to perform anaerobic work during maximal
graded exercise testing is similar.
We chose to develop separate equations for males
and females rather than pool their data and thus obtain
agender-independent multiple regression equation with
coefficient for gender. It has been well established that
‘WO is lower in females if expressed in absolute terms
(ml-min™). While Bruce et al. (8) reported gender as
the second variable selected for inclusion in their pre-
diction of VOzmax. analysis of our pooled data for men
and women (Table 4) revealed that gender was the last
variable added to the equation, entering after both
weight and age. Gender did, however, significantly in-
‘crease the magnitude of R® as displayed in Table 4.
While the absolute VO, (I-min“') for treadmill ex-
ercise at a given work rate is substantially affected by
body weight, such is not the case in cycle ergometer
exercise since the body weight is supported by the seat.
However, Wasserman and Whipp (33) have reported
that VO; is influenced by the subject's body weight
even in this weight-supported exercise due to the differ-
ences in the O; cost of moving the legs; for any given
work rate, they found that the VO; was 3.8 ml-min"!
higher for each additional kilogram of body weight. In
our male and female subjects, body weight was a sig-
nificant predictor of VOzm., entering the regression
equation at step 2 for the females and at step 3 for the
males. When the male and female data were pooled,
‘body weight was the second independent variable which
entered the equation, significantly increasing R?.710
‘The maximal VO; declines with age due to a number
of potential factors including decreases in lean body
mass and reductions in maximal exercise values for
cardiac output, pulmonary ventilation, and pulmonary
diffusing capacity (5, pp. 385-386). While these changes
should have been accounted for by decreases in maxi-
mal work rate, age remained a significant predictor of
VOrmx in our equations, entering the equation at the
third (last) step for the women and at the second step,
for men. Age entered the Bruce et al. equation at the
third step and was categorized with weight as an insig-
nificant predictor of VOams (8). The small increment
in R? resulting from the addition of age to our gender-
independent equation was, however, statistically signif-
icant,
Table 5 contains a comparison of predictor variables,
correlation coefficients, and SEE between the present
study and other investigations using maximal treadmill
or cycle ergometer protocols. When several equations
were presented by other studies, the equation which
yielded the highest R and the lowest SEE was selected
for comparison. It is apparent from Table 5 that the
equations of the present study yield high multiple cor-
relation coefficients and low SEE.
The double cross-validation procedure is the most
rigorous method for validating regression equations (24,
p. 284), Validity is established when this procedure
yields high simple correlation coefficients between the
‘rue value and the predicted value for the dependent
variable in subgroups of the original sample and when
these high correlations are similar. The correlation coef
ficients (1) obtained from this procedure were 0.937
and 0.942 for the males and 0.920 and 0.937 for the
females, thus verifying that the original regression equa-
tions derived from the entire male and the entire female
samples are valid. Furthermore, the R values generated
in the double cross-validation procedure were also high
and quite similar to the R values obtained from the
regression analyses using all the male and all the female
subjects. Thus, our gender-specific equations are gen-
eralizable and should provide valid estimates of VO2mox
in other subject samples which have characteristics
similar to our sample of subjects.
MEDICINE AND SCIENCE IN SPORTS AND EXERCISE
Application of the gender-specific equations devel-
oped in this study to the 26 male and ten female
independent (i¢., external) cross-validation subjects
produced a validity coefficient of r = 0.95, with a SEE
of 176 ml-min™, Figure 2 illustrates this relationship.
‘An added feature of the present validation sample is
that 21 of the male subjects were tested 9 yr earlier, in
duplicate, using an entirely different equipment config-
uration but with the same cycle ergometer protocol. It
would seem apparent, therefore, that experimenter or
equipment specificity has little impact on the utility of
these equations.
Nine of the 21 sedentary male subjects tested 9 yr
earlier underwent 9 wk of endurance exercise training,
which improved their VOzmax, on average, by 25%. As
revealed in Figure 3, the accuracy of prediction was not
affected by improved cardiorespiratory fitness; the
slopes, y-intercepts, and SEE are nearly identical. The
mean difference between the actual and predicted
VOrmax before training was —0.5% (-43 ml-min™').
After training, the mean difference was ~2.4% (—110
ml-min™), Neither difference was statistically signifi-
cant, although there was a trend toward underpredic-
tion. These results suggest that the equations generated
in this study may also be used to estimate VO2ma, in a
more active population. This observation was also
made by Bruce et al. (8) and Froelicher et al. (20). Both
groups of investigators were unable to demonstrate
significant differences in predicting VOzmax from tread-
mill duration as a function of habitual level of activity.
‘While there are many positive aspects in utilizing the
prediction equations generated in this study, this in
ect approach does result in the loss of diagnostic data
‘that would have been available had ventilation and gas
exchange been measured. Failure to make these meas-
urements throughout an exercise test reduces the pos-
sibility of detecting other important parameters of aero-
bic function, e.g., the anaerobic threshold (34). Ad
tionally, Hansen et al. (22) have recently demonstrated
that patients with cardiovascular disease have a VOs vs
work rate slope during incremental exercise which is
significantly lower than predicted. Thus, use of predic-
tion equations which include maximal work rate as a
TABLE 5. Comparison of preditonvaabie,corelaton coeticint (Ror), and SEE betwen regression equations found inthe erature and those foundin the present
study forthe predtion of VO
Wark
Study Device Voviabes* ort SEE (mi-min~") SEE (l-kg"-minY)
Present study, 1969 cE Wi. WT, Aa 0999 22 257
15 228
Bruce et a, 1979 6) ™ D,G,Age, HT, Actity 0828 245 312
Jones a, 1985 (23) ce Age, WT, HT set an 563
Town and Galdng 1977 (82) ™ ‘Age,HA, WT, %6Grade 0.898 330 408
Froelcher ta, 1975 (20) ™ D 0870 368 an
Fox, 1973(17)| cE HR at 1500. 0760 286 839
= W = work rat; WT = body weigh: G = gender,
= Duraon of test Acty =
itu evel of physical aciviy; HT = height eraACCURATE PREDICTION OF VO2max
predictor variable would overestimate the true VO2max
of these patients. It is apparent, therefore, that one must,
consider the relative advantages and disadvantages in
using indirect vs direct methods for the assessment of
VOrmax- When the constraints of equipment, time, ex-
pense, and technical assistance prohibit the direct meas-
urement of VO;, use of accurate prediction equations
such as the ones generated in the present study provides
an attractive alternative. However, caution is advised if
these equations are used in subjects with known or
suspected cardiovascular disease.
CONCLUSIONS
‘We have generated equations to predict VOzmax that
have high correlation coefficients, low standard errors
of estimate, and levels of accuracy which allow predic-
tion of VOzmax to within 10% of its true value in 95 out
of every 100 subjects. These equations have been vali-
dated with rigorous internal and external validation
procedures, including an independent sample of sub-
jects, some of whom were tested 9 yr earlier with an
equipment configuration different from that used to
test the study sample. Results of the internal cross-
validation study suggest that our equations are gener-
alizable. Further proof of the generalizability of our
REFERENCES
1. AMERICAN COLLEGE OF SORTS MEDICINE. The recommended
{quanttiy and quality of exercise for developing and maintaining
fitness in healthy adults. Med. Sci. Sports. Exerc. 10:ii-x, 1978.
2. AsrRAND, I. Aerobic work capacity in men and women with
special reference to age. Acta Physiol. Scand. 49(Suppl. 169):
92, 1960.
43. ASTRAND, P-O. Experimental studies of physical working capac-
ity in relation to sex and age. Copenhagen: Munksexard, 1952,
pp. [-171.
4. AstRaND, PO. and I. RuvaiNG. A nomogram for caleulatc
of aerobic capacity (physical fitness) from pulse rate during
submaximal work. J. Appl. Physiol. 7:218-221, 1954.
5. AsTRAND, P-O, and K, RODAHL. Textbook of Work Physiology,
2nd Ed, New York: McGraw-Hill Book Company, 1977.
6. BaLke, B.and R. A. WARE, An experimental study of “physical
fitness" of airforce personnel. U.S. Armed Forces Med. J. X:675—
688, 1959,
7. Beaver, W. L, K, WASSERMAN, and B. J. Winer, Ondine
‘computer analysis and breath-by-breath graphical display of ex-
ercise function tests, J. Appl. Physiol, 34:128-132, 1973,
8. Bruce, R. A, F. KusuMl, and D. Hoswer, Maximal oxygen
intake and nomographic assessment of functional aerobic im-
pairment in cardiovascular disease. Am. Heart J. 85:546-562,
1973.
9. Catozz0, V.J.,J.A. Davis, . J. BERRIMAN, RB. VANDAGRIFF,
and C. A, Paierro. Effect of high intensity exercise on the Ver
Veco: relationship. J. Appl. Physiol. 62:1460-1464, 1987,
10, COLEMAN, A. E, Validation of a submaximal test of maximal
‘oxygen intake. J, Sports Med. Phys. Fitness 16:106-111, 1976.
11, Davis, J. A., M. H. FRANK, B. J. Wate, and K. WASSERMAN.
‘Anaerobic threshold alterations caused by endurance training in
‘middleaged men. J. Appl. Physiol. Respirat. Environ. Exerc
Physiol. 86:1039-1046, 1979.
12, Davis, JA V. J. Calozz0, N. LaMARRA, et al. Does the gas
mL
equations was demonstrated in our external cross-vali-
dation study. Additionally, our male gender-specific
equation appears to be insensitive to the effects of
endurance exercise training. For those practitioners
who prefer cycle ergometry as the exercise test mode,
use of the equations generated in this study provides
accurate estimates of VO2max for the assessment of
cardiopulmonary functional capacity and/or normal
standards against which a measured VO2mx may be
evaluated, Our equations may be specific, however, to
the GXT protocol we used, i i
in work rate, For practitioners who test young and fit
subjects, the 15 W-min"' GXT protocol results in
lengthy test durations. These subjects should be evalu-
ated using protocols employing larger work rate incre-
ments, e.g., 30 W-min“'. It will then be necessary to
validate the present equations for these different pro-
tocols.
“The authors wish to thank Mr. Richard B. Vandaghif, Mr. Matthew
J. Berenda, and Mr. Michael Fowler for their assistance with this
Project.
“This research was supported in part by a grant trom the American
Lung Association of Los Angeles County awarded to James A, Davi.
“The validation phase of this study was performed at the Exercise
Physiology Laboratory in the Division of Respiratory and Critical Care
Medicine and Physiology at Harbor-UCLA Medical Center, Torrance,
‘CA 90509,
‘Address for correspondence: T. W. Storer, Ph.D., Laboratory of
Exercise Science, EI Camino College, Torrance, CA 90506.
‘exchange anaerobic threshold occur at a fixed blood lactate
concentration of 2 or 4 mM? Int. J. Sports Med. 4:89-93, 1983,
1, Davis, J. A. and N. Lawarea. A turbine device for accurate
volume measurement during exercise. Aviat. Space Environ
Med. 55:455, 1984.
14, Dixon, W. J. and F. J. Massey, Introduction o Statistical Analy
sis, 3rd Ed, New York: McGraw-Hill, 1969, p, 200,
15, ELlestap, M. H,, C. G, Buongvisr, and J. P. Naucuton.
Standards for adult exercise testing laboratories. Circulation
59421-8304, 1979,
16, Foster, C., A.'S. JACKSON, M. L, PoLLock, et al. Generalized
‘equations for predicting functional capacity from treadmill per-
ormance. Am. Heart J. 107:1229-1234, 198,
17. Fox, E. L.A simple, accurate technique for predicting maximal
aerobic power. J. Appl. Physiol. 35:914-916, 1973.
18. Fox, S. M,, J.P. NauGHTON, and W. L.’ HASKELL. Physical
activity and the prevention of coronary heart disease. nn. Clin
Res, 3:404-432, 1971
19, FRoscicwer, V. F., H. Bramment, G. Davis, 1. NOGUERA, A.
STEWART, and MC, LaNcasteR. A’ comparison of the reproduc-
ibility and physiologic response to three maximal treadmill ex-
ecise protocols. Chest 65:512-517, 1974.
20. FRosLicneR, V. F., A. J. THOMPSON, I. NOGUERA, G. Davis, A.
J, StewakT, and I, H. TRIEBWASSER. Prediction of maximal
‘oxygen consumption: Comparison of the Bruce and Balke tread-
mill protocols. Chest 68:331~336, 1975
21. HaNseN, J. E. Respiratory abnormalities: exercise evaluation of
the dyspnie patient. In: Cardiopulmonary Exercise Testing, A.
R, Leff (ed.). Orlando, FL: Grune and Stratton, Inc, 1986, pp.
69-88.
22, HANSEN, J.B. D. Y. Sue, A. One, and K, WASSERMAN. Relation
‘of oxygen uptake to work rate in normal men and men with
circulatory disorders. Am. J. Cardiol. $9:669-674, 1987,
23, JONES, N.L., L. MAKRIDES, C. Hircwack, T. CHYPCHAR, and N.2
m4,
25,
26,
2,
;. RoweLt, L. BH. L, Tavtor, and Y. WaNG.
McCartney, Norman standards for an ineremental progressive
‘yele egometer test. Am. Rev. Respir, Dis. 131:700—708, 1985,
KeRuinaer, F. N. and E. J. Pepuazur, Multiple Regression in
Behavioral Research. New York: Holt, Reinhart, and Winston,
Inc., 1973, pp. 66-72, p. 284.
McAgpuz, W. D,, F. I. Kate, and V. L. Katou, Exercise
Physiology. Energy, Nutrition, and Human Performance, 2nd Ed,
Philadelphia: Lea and Febiger, 1987, p. 176,
Nauaitton, J. P. and R. HAiDeR. Methods of exercise testing
In: Exercise Testing and Exercise Training in Coronary Heart
Disease, J. P. Naughton, H. K. Hellerstein, and L. C. Mohler
(Eas.). New York: Academic Press, 1973, pp. 79-81
Poutock, M. L., R. L, BOHANNON, K. H. Cooper, et al. A
‘comparative analysis of four protocols for maximal’ treadmill
stress testing, Am. Heart J. 92:39~46, 1976.
imitations to
prediction of maximal oxygen intake, J. Appl. Physiol. 19:919~
29,
30,
31
32,
3
34,
MEDICINE AND SCIENCE IN SPORTS AND EXERCISE
927, 1964.
ShosrRAND, T. Changes in respiratory organs of workmen at an
‘ore melting works, Acta Med. Scand. (Suppl) 196:687-695, 1947
‘TaTSUoKA, M. M. Validation studies: the use of multiple regres-
sion equations. In: Selected Topies in Advanced Statistics, M. M.
‘Tatsuoka (Ed,). Champaign, IL: Institute and Ability Testing,
1976, p.L
TAYLOR, H, Ly E. BUSKIRK, and A. HeNscueL. Maximal oxygen
intake as an objective measure of cardio-tespiratory performance.
J. Appl. Physiol. 8:73-80, 1955.
Town, G. P. and L. A. GouDinc. Treadmill test to predict
maximum aerobic capacity. J. Phys. Educ. 15:6-7, 197
WassenMian, K. and B. J. Winpp. Exercise physiology in health
‘and disease. Am. Rev. Respir. Dis. 112:219-249, 1975.
Wassceman, K., B.J. War, 8, N. Kovat, and W. L. Beaver,
Anaerobic threshold and respiratory gas exchange during exer-
cise. J. Appl. Physiol. 35:236-243, 1973.