Beruflich Dokumente
Kultur Dokumente
DOI 10.1007/s00421-009-1112-8
1 3
ORI GI NAL ARTI CLE
The acute eVects of running on blood pressure estimation
using pulse transit time in normotensive subjects
Mico Yee-Man Wong Emma Pickwell-MacPherson
Yuan-Ting Zhang
Accepted: 5 June 2009 / Published online: 20 June 2009
Springer-Verlag 2009
Abstract Pulse transit time (PTT) is a potential parameter
for cuZess blood pressure (BP) estimation. Since exercise
induces changes in arterial properties that can inXuence the
relationship between BP and PTT, we investigate whether
PTT can be used to estimate BP after successive bouts of
exercise. PTT-foot, PTT-peak (time intervals from the peak
of electrocardiogram R-wave to the foot and peak of pho-
toplethysmogram, respectively) and BP of 41 normotensive
subjects (aged 25 4 years) were measured in the Wrst test.
A repeatability test was then conducted on 14 subjects after
6 months. Each test included two periods of running on the
treadmill at 10 and 8 km/h (with a rest in between). In both
tests, systolic BP (SBP) was closely correlated with PTT-
foot and PTT-peak. For each subject, the best Wt linear rela-
tionships between SBP and PTTs were determined over all
phases of each test. The diVerences between the linear Wts
and measured data were greater after the second period of
running for all subjects in both tests. This implied that the
relationships started to change after the second period of
running. When SBP in the repeatability test was predicted
using the linear regression coeYcients from the Wrst test,
the linear Wt after the Wrst period of exercise was still better
than after the second. The repeated observations in both
tests suggest that PTT is a potential parameter for cuZess
BP estimation after one period of exercise, but would need
re-calibration (relationship between BP and PTTs) for mea-
surements after successive phases of exercise.
Keywords Arterial blood pressure Exercise
Pulse transit time Acute eVects
Abbreviations
BP Blood pressure
DBP Diastolic blood pressure
ECG Electrocardiogram
HR Heart rate
PPG Photoplethysmogram
PTT Pulse transit time
PWV Pulse wave velocity
SBP Systolic blood pressure
Introduction
Pulse transit time (PTT) or pulse wave velocity (PWV) has
the potential for blood pressure (BP) prediction (Allen et al.
1981; Pollak and Obrist 1983; Steptoe et al. 1976; Wong
and Zhang 2004). When the heart contracts, a pressure
pulse wave is generated and propagates along the arterial
tree with PWV. PWV is formulated with the Moens
Korteweg equation (McDonald 1974), which relates PWV to
the geometric properties of the arterial wall such as radius,
wall thickness, stiVness and blood density. An inverse of
PWV, PTT, is a parameter that refers to the time required
for the pressure pulse wave to travel through a segment of
M. Y.-M. Wong E. Pickwell-MacPherson Y.-T. Zhang (&)
Joint Research Centre for Biomedical Engineering,
Department of Electronic Engineering,
The Chinese University of Hong Kong,
Rm. 404, Ho Sin Hang Eng. Bldg., Shatin, Hong Kong
e-mail: ytzhang@ee.cuhk.edu.hk
Y.-T. Zhang
SIAT Institute of Biomedical and Health Engineering,
Chinese Academy of Sciences, Shenzhen, China
Y.-T. Zhang
Key Laboratory for Biomedical Informatics
and Health Engineering, Chinese Academy of Sciences,
Shenzhen, China
170 Eur J Appl Physiol (2009) 107:169175
1 3
arterial tree (McDonald 1974). When BP increases, the
arteries become stiVer and PTT decreases.
The relationship between BP and PTT during exercise
has been examined. Marie et al. (1984) found that PTT was
correlated to systolic BP (SBP) (r = 0.81) and diastolic
BP (DBP) (r = 0.52) during bicycle and handgrip exer-
cise. The beat-to-beat inverse relationship between SBP
and PTT has been supported by Lass et al. (2004) in normo-
tensive and hypertensive subjects. Teng et al. (2005) mea-
sured the variation of PTT and BP of 12 normotensive
subjects 1 h after treadmill exercise and found that although
PTT did not change signiWcantly, both Wnger and brachial
BP were diVerent after the exercise, indicating that the rela-
tionship between BP and PTT changes after exercise.
The geometric and elastic properties of arteries can be
aVected by exercise, so that blood viscosity increases (Naka
et al. 2003), stiVness decreases (Kingwell et al. 1997) and
arterial diameter increases (Studinger et al. 2003). These
exercise-induced changes in the properties of the arterial tree
can aVect the relationship between BP and PTT, which, in
turn, aVects BP estimation using the PTT-technique. There-
fore, we investigated the relationship between PTT and BP
after subjects had been running on a treadmill at two speeds.
We also evaluated whether PTT could be used to estimate
BP when subjects were recovering after exercise.
Methods
Experiments were conducted on normotensive subjects,
including 22 males and 19 females (aged 25 4 years). A
repeatability test was carried out after 6 months on 14 sub-
jects aged 26 4 years, including 8 females, with the labo-
ratory temperature kept between 22 and 24C. The subjects
were given verbal and written instructions before they gave
their informed consent prior to the experiment. No subjects
had cardiovascular problems and they did not take medica-
tion on the day of the experiment. Participants were asked
to abstain from planned exercise for at least 24 h before the
experiment.
The exercise was carried out on a treadmill (C956, Precor,
USA). An in-house-designed circuitry with three electrodes,
a light emitting diode (IR91-21C/F7, Taiwan) and a photode-
tector (PT91-21B/F7, Taiwan) were used to detect ECG and
PPG from a subjects Wngertips (Wong and Zhang 2005,
2006). The Wltering passbands for analog ECG and PPG
were 248 and 0.334 Hz, respectively. Similar Wltering
passbands have been used for ECG and PPG detection (Park
et al. 2005; Teng and Zhang 2006). ECG and PPG were sam-
pled at 1 kHz and recorded by a data acquisition system (DI-
220, DATAQ Instruments WinDaq, USA) for 45 s. When
clear ECG and PPG waveforms were detected, the recording
was started. After exercise, subjects immediately put their
Wngertips on the sensors and it usually took less than 5 s to
obtain the typical ECG and PPG waveforms. Brachial SBP
and DBP were measured on the arm by an electronic BP
monitor (HEM-907, OMRON, Japan) (Assaad et al. 2002;
White and Anwar 2001). The subjects were seated and their
forearms placed on the desk, while a cuV was wrapped
around the right arm kept at the level of the heart. Every sub-
ject contributed 13 data sets taken during the four pre-deter-
mined phases: Pre-exercise (data sets 13), Post-exercise1
(data sets 46, 5 min of rest after the Wrst period of exercise),
Post-exercise2 (data sets 79, 5 min of rest after the second
period of exercise) and Recovery (data sets 1013, while the
subject had rested for a further 40 min).
Subjects rested for 3 min before conducting the experi-
ment (Fig. 1). In the Pre-exercise phase, BP was obtained
four times and ECG and PPG were recorded simulta-
neously between BP measurements. Subjects rested and
were then directed to run on the treadmill at 10 km/h for
3 min, followed by another 5-min period of rest (Post-exer-
cise1). Then subjects ran on the treadmill at 8 km/h for
3 min, followed by another 5-min period of rest (Post-
exercise2). Four BP measurements and three recordings of
ECG and PPG were conducted in each Post-exercise1 and
Fig. 1 Outline of the experimental procedures. 113 indicate the data
sets of simultaneous measurements of electrocardiogram (ECG) and
photoplethysmogram (PPG) (duration 45 s). In the Pre-exercise, Post-
exercise1 and Post-exercise2 phases, blood pressure (BP) measure-
ments were taken four times and ECG and PPG were recorded simul-
taneously between BP measurements. In the Recovery phase,
simultaneous recordings of ECG and PPG were followed by BP
measurements
represents BP measurement by automatic BP machine denotes simultaneous recordings of ECG and PPG
15min.
20min.
25min.
40min.
Post
exercise1
Post
exercise2
Recovery
1 2 3 4 5 6 7 8 9 10 11 12 13
Pre
exercise
5min. 5min. 5min.
3min.
10km/h
running
3min.
8km/h
running
3min.
rest
Eur J Appl Physiol (2009) 107:169175 171
1 3
Post-exercise2 phases. In the recovery period, measure-
ments were acquired at 15, 20, 25 and 40 min after the end
of Post-exercise2. The averages of electronic BP measure-
ments before and after the recording of ECG and PPG in
the phases, Pre-exercise, Post-exercise1 and Post-exercise2,
were considered as references to provide a reliable baseline
and to compensate for the fast BP changes due to exercise.
Signal processing
The signals were analyzed oZine using Matlab (Math-
Works, USA). Figure 2 shows the deWnitions of the PTTs.
The PTT-peak and PTT-foot were the time intervals from
the peak of the ECG R-wave to the peak and the foot of the
PPG, respectively. A program was used to detect the char-
acteristic points on ECG and PPG to measure PTTs (Wong
and Zhang 2005, 2006). The ECG and PPG were Wltered by
lowpass Wlters with cutoV frequencies at 25 and 18 Hz,
respectively (Foo and Wilson 2006). The Wltered ECG was
diVerentiated and the peaks of the ECG derivative were
detected. The peaks of the ECG R-wave were measured
from 40 ms before the peak positions of the ECG derivative
until 40 ms thereafter. The foot and peak of the Wltered PPG
were measured by Wnding the minimum and maximum
points, respectively, in the range deWned from the times
corresponding to the peaks of the R-wave to 350 ms after
the peaks. The averages of PTT-peak and PTT-foot were
calculated for each data set. Heart rate (HR) was measured
as the reciprocal of average successive time intervals
between the peaks of the R-wave (RRI).
Statistical analysis
A students t test was used to evaluate whether there was a
diVerence between the sample means in the Wrst and repeat-
ability tests (p < 0.05). The relationship between PTT and
BP was evaluated by the correlation coeYcient r. r > 0.8 or
r < 0.8 was considered as close correlation between the
pair (Salkind 2005).
The BP for each individual was modeled as a linear
function of PTT: BP = mPTT + c. The coeYcients m and c
were the slope and Y-intercept, respectively, such that they
were determined by least squares Wt to all BP and PTT pairs
for each subject. The diVerence between measured and
predicted BP using the least squares method was calculated
in the Wrst and repeatability tests. Additionally, BP in the
repeatability test was estimated using the regression coeY-
cients (m and c) obtained from the Wrst test. The estimated
diVerences, deWned as the subtraction of the measurement
from its corresponding prediction, are presented as
mean SD. The means of the absolute values of the esti-
mated diVerences are also given.
Results
The Wrst test
Variation of the parameters
Figure 3 illustrates the variation of BP, HR, PTT-peak and
PTT-foot. All parameters showed changes after the Wrst and
second periods of running: SBP, DBP and HR increased
and PTT-foot and PTT-peak decreased. In the Recovery
phase, only SBP and PTT-peak returned to their baseline
values.
In the Post-exercise1 and Post-exercise2 phases, SBP,
PTT-peak and PTT-foot diVered after the two exercise
phases with diVerent speeds, and good correlation was
found between SBP and PTTs (Table 1). DBP, however,
remained unchanged in these post-exercise periods. Thus,
DBP was not well correlated with either PTT-foot or PTT-
peak.
We found that SBP was more closely related to PTT-foot
(r = 0.85) than to PTT-peak (r = 0.77). Furthermore,
PTT-foot was inversely related to SBP in all subjects
(0.98 < r < 0.32). Figure 4 depicts the scatter plots of
SBP versus PTTs with linear regression lines of the sub-
jects, who show the strongest and weakest correlations. The
plots demonstrate that large variations of PTT-peak and
PTT-foot were provoked. In subjects whose SBP was
poorly correlated with PTTs, the range of SBP was usually
not as large as that observed in subjects with close correla-
tion. The BP range induced by exercise aVected the likeli-
hood of occurrence of high correlation between BP and
PTTs. Therefore, weak correlations between DBP and
PTTs were mostly observed in subjects who had a small
range of DBP [60 mmHg < mean DBP < 65 mmHg in the
Wrst test (Fig. 3b)].
BP estimation after the Wrst and second periods of running
Table 2 summarizes the estimated diVerences between the
predicted and measured BP after the Wrst and second
Fig. 2 PTT-peak and PTT-foot were the time intervals measured from
the peak of the electrocardiogram (ECG) R-wave to the peak and the
foot of the photoplethysmogram (PPG), respectively. RRI represented
the time interval between the successive peaks of the ECG R-wave
172 Eur J Appl Physiol (2009) 107:169175
1 3
periods of running. When BP was predicted using a least
squares Wt in all BP and PTT pairs for each subject, all esti-
mated diVerences in the Post-exercise2 phase were diVerent
from those in the Post-exercise1. The diVerences of SBP
and DBP predictions using PTT-foot in Post-exercise2
phase were larger than in Post-exercise1; and the diVer-
ences of SBP and DBP estimations using PTT-peak were
smaller in Post-exercise2 than in Post-exercise1. PTT-foot
always gave smaller mean absolute diVerences for SBP
estimation in both periods compared to the PTT-peak.
Repeatability test
Variation of the parameters
Figure 3 summarizes the mean values of parameters
throughout the experiment in the repeatability test. All
parameters except DBP showed changes in the two post-
exercise periods. PTT-foot and DBP over all phases were
also consistent with those in the Wrst test and PTT-foot
again diVered (p = 0.013) in two post-exercise periods.
Correlation between SBP and PTTs (r < 0.80) and the
inverse correlation between SBP and PTT-foot for all sub-
jects (0.97 < r < 0.53) were also found in the repeatabil-
ity test (Table 1).
BP estimation after the Wrst and second periods of running
BP was predicted from a second set of PTT measurements
using least squares Wts. In all BP and PTT pairs for each
subject (Table 2), using PTT-foot to calculate SBP and
DBP, the measurements matched those from the Post-
exercise1 phase more closely than those from the Post-
exercise2 phase, suggesting that the second period of exer-
cise aVected the calibration. Also SBP estimation using
PTT-foot matched the measured data more closely than if
PTT-peak had been used. The discrepancy between BP esti-
mation diVerences in the two post-exercise periods revealed
that successive bouts of exercise had a pronounced eVect on
BP estimation using PTT-technique.
Table 2 further shows the diVerences between the pre-
dicted and measured BP in two post-exercise periods
when BP in the repeatability test is predicted using regres-
sion coeYcients from the Wrst test for each individual.
Compared to the estimation diVerences obtained using a
least squares Wt, the mean absolute diVerences increased
when BP was predicted using regression coeYcients
60
64 65
62
59
62
63
61
20
30
40
50
60
70
80
90
100
Pre-
exercise
Post-
exercise1
Post-
exercise2
Recovery
Pre-
exercise
Post-
exercise1
Post-
exercise2
Recovery
Pre-
exercise
Post-
exercise1
Post-
exercise2
Recovery
Pre-
exercise
Post-
exercise1
Post-
exercise2
Recovery
Pre-
exercise
Post-
exercise1
Post-
exercise2
Recovery
D
B
P
(
m
m
H
g
)
* *
*
107
124
118
105
103
123
118
101
60
80
100
120
140
160
180
S
B
P
(
m
m
H
g
)
*
*
^
*
+
*
72
95 96
86
69
92 91
82
40
50
60
70
80
90
100
110
120
130
140
H
R
(
b
p
m
)
* *
*
^
* *
+
*
286
254
263
289
298
253
262
296
200
220
240
260
280
300
320
340
360
P
T
T
-
p
e
a
k
(
m
s
)
*
*
^ ^
*
*
+
157
132
144
165
163
129
143
165
100
120
140
160
180
200
220
P
T
T
-
f
o
o
t
(
m
s
)
*
*
*
*
*
++
(b)
(a)
(c)
(d)
(e)
Fig. 3 Mean SD of a systolic blood pressure (SBP), b diastolic
blood pressure (DBP), c heart rate (HR) and pulse transit time (d PTT-
peak and e PTT-foot) for four phases in the Wrst test (black bars) and
repeatability test (white bars). The mean values of the parameters are
shown on top of the bars. Asterisks (*) indicate diVerences (p < 0.05)
from the baselines (Pre-exercise phase). Pluses (+) represent that the
parameters in the Post-exercise2 phase were diVerent (p < 0.05) from
those in the Post-exercise1 phase. Hats (^) indicate that the parameters
in the Wrst and repeatability tests are signiWcantly diVerent (p < 0.05)