ORIGINAL PAPER

Pulse Wave Velocity and Digital Volume Pulse as IndirectEstimators of Blood Pressure: Pilot Study on Healthy Volunteers

Juan M. Padilla Æ Enrique J. Berjano ÆJavier Saiz Æ Rafael Rodriguez Æ Lorenzo Facila

� Springer Science+Business Media, LLC 2009

Abstract The purpose of the study was to asses the

potential use of pulse wave velocity (PWV) and digital

volume pulse (DVP) as estimators of systolic (SBP) and

diastolic (DPB) blood pressure. Single and multiple cor-

relation studies were conducted, including biometric

parameters and risk factors. Brachial-ankle PWV (baPWV)

and DVP signals were obtained from a Pulse Trace PWV

and Pulse Trace PCA (pulse contour analysis), respec-

tively. The DVP (obtained by photoplethysmography),

allowed stiffness (SI) and reflection indexes (RI) to be

derived. The first study on 47 healthy volunteers showed

that both SBP and DPB correlated significantly both with

baPWV and SI. Multiple regression models of the baPWV

and the waist-to-hip ratio (WHR) allowed SBP and DBP to

be modeled with r = 0.838 and r = 0.673, respectively. SI

results also employed WHR and modeled SBP and DBP

with r = 0.852 and r = 0.663, respectively. RI did not

correlate either with SBP or DBP. In order to avoid the use

of ultrasound techniques to measure PWV, we then

developed a custom-built system to measure PWV by

photoplethysmography and validated it against the Pulse

Trace. With the same equipment we conducted a second

pilot study with ten healthy volunteers. The best SBP

multiple regression model for SBP achieved r = 0.997 by

considering the heart-finger PWV (hfPWV measured

between R-wave and index finger), WHR and heart rate.

Only WHR was significant in the DBP model. Our findings

suggest that the hfPWV photoplethysmography signal

could be a reliable estimator of approximate SBP and could

be used, for example, to monitor cardiac patients during

physical exercise sessions in cardiac rehabilitation.

Keywords Blood pressure � Digital volume pulse �Pulse wave velocity � Photoplethysmography �Stiffness index � Reflection index

Introduction

Noninvasive monitoring of blood pressure (BP) is usually

performed by traditional sphygmomanometry. However,

this technique is unable to monitor short-term changes, so

that a noninvasive method of measuring beat-to-beat BP

would be extremely valuable. This is especially crucial in

monitoring cardiac patients undergoing treadmill exercise

tests during cardiac rehabilitation. It has been suggested

that rate-pressure product (i.e., heart rate 9 systolic blood

pressure) is the key monitoring parameter during this

procedure (Pierson et al. 2004; Nieminen et al. 2008).

Our objective was therefore to assess the potential use of

certain cardiovascular signals (obtained non invasively)

as estimators of systolic (SBP) and diastolic (DPB) blood

pressure. Since a near linear correlation exists between

transit time TT (measured from the R wave) and BP, this

parameter could be used as a surrogate marker of

J. M. Padilla � E. J. Berjano � J. Saiz � R. Rodriguez

Institute for Research and Innovation on Bioengineering (I3BH),

Universidad Politecnica de Valencia, Valencia, Spain

J. M. Padilla

Instituto Tecnologico de Morelia, Morelia, Michoacan, Mexico

E. J. Berjano (&)

Departamento de Ingenierıa Electronica, Universidad Politecnica

de Valencia, Camino de Vera, 46022 Valencia, Spain

e-mail: eberjano@eln.upv.es

L. Facila

Departamento de Cardiologıa, Hospital Provincial de Castellon,

Valencia, Spain

123

Cardiovasc Eng

DOI 10.1007/s10558-009-9080-5

pressure (Geddes et al. 1981; Chen et al. 2000). However,

in order to obtain more general and reproducible results,

we initially chose two cardiovascular signals, the pulse

wave velocity (PWV) and the digital volume pulse

(DVP), obtained from commercially available equipment.

PWV is known to be an indicator of arterial stiffness

(Lehmann 1999) and is largely determined by BP. A

regional estimation of the properties of the arterial wall can

be made by measuring local PWV, for instance in the aorta,

where an invasive recording technique is used to record the

aortic pulse wave velocity (aPWV) (Liu et al. 2004).

Finally, brachial-ankle pulse wave velocity (baPWV) is a

non invasive PWV measurement system which shows a

good correlation with aPWV obtained by invasive record-

ing (Yamashina et al. 2002).

Digital Volume Pulse (DVP) is an alternative method of

measuring arterial stiffness (Chowienczyk et al. 1999) that

provides valid information on the pressure pulse waveform

(Millasseau et al. 2002). DVP can be recorded in the finger

by photoplethysmography (PPG) using an infrared optical

transducer. It usually exhibits an early systolic peak and a

later peak or point of inflection that occurs a short time

(DTDVP) after the first peak in early diastole (see Fig. 1).

It can provide two indexes: stiffness index (SI), which

relates with large artery stiffness, and the reflection index

(RI), which relates with vascular tone (Millasseau et al.

2002). The DVP waveform is determined mainly by the

characteristics of systemic circulation, including pressure

wave reflection and PWV of pressure in the aorta and large

arteries (Chowienczyk et al. 1999). To date, there have

been no previous studies employing either baPWV or DVP

to estimate BP.

The aim of this work was to asses the potential use of

pulse wave velocity (PWV) and digital volume pulse

(DVP) as estimators of systolic (SBP) and diastolic (DPB)

blood pressure in healthy subjects.

Materials and Methods

This study was divided into two phases. In the first, the

potential use of brachial-ankle PWV (baPWV) and digital

volume pulse (DVP) were studied as estimators of systolic

(SBP) and diastolic (DPB) blood pressure. Both signals

were obtained from commercially available equipment. In

the second phase, we studied the potential use of other

signals related with PWV and obtained by means of custom

built-equipment based on photoplethysmography.

First Phase

The first phase involved 47 healthy volunteers

(33.78 ± 9.04 years). The subjects were examined in the

supine position, with electrocardiogram electrodes placed

on both wrists after a 15 min rest. Blood pressure, baPWV

and DVP were measured three times for averaging. baPWV

was recorded by Pulse Trace PWV (Micro Medical, Kent,

UK). This equipment uses a 4 MHz Doppler probe to

identify the arrival of the arterial pulse. The arterial pulse

waveform was sequentially measured in two locations of

the arterial tree (right brachial and right ankle) and was

timed using the R wave of the ECG (Lead I) as shown in

Fig. 2. The Pulse Trace PWV automatically computed the

transit time (DTPWV) from the onset time difference

between brachial and ankle waveforms. baPWV was then

calculated by dividing the externally measured distance

between brachial-ankle locations (Lba) by the DTPWV:

baPWV ¼ Lba =DTPWV ð1ÞThe distance Lba between the measuring sites (i.e., the

brachial and posterior tibial arteries) was measured with a

tape measure. To reduce the influence of body contours on

the distance, the tape measure was held above the surface of

the body parallel to the plane of the examination table. Five

Fig. 1 a Digital Volume Pulse

(DVP) waveform obtained by

photoplethysmography. It

consists of a systolic peak (a)

and diastolic peak (b) separated

by a transit time delay

(DTDVP). b Arterial system

diagram highlighting the aortic

segment: the systolic peak of

DVP waveform corresponds to

the blood forward wave (Fw)

while the diastolic peak

characterizes the blood reflected

wave(s) (Rw) in the arterial

system

Cardiovasc Eng

123

partial distances were measured, as shown in Fig. 3: (1)

from suprasternal notch to inferior edge of the umbilicus;

(2) from the inferior edge of the umbilicus to the iliac crest;

(3) from to the iliac crest to the medial malleolus in sam-

pling site on the right leg; (4) from suprasternal notch to

head of humerus; and (5) from head of humerus to the

forearm. Lba was then calculated by subtracting the distance

between the suprasternal notch to medial malleolus site on

the posterior tibial artery from the distance between the

suprasternal notch to the forearm site on the brachial artery.

DVP waveform was obtained by using a Pulse Trace

PCA (Micro Medical, Kent, UK) which employs a pho-

toplethysmography (PPG) transducer. This techniques fol-

lows the changes in the reflectance of infrared light

(k = 940 nm) through the finger pulp. The PPG transducer

was placed on the index finger of the right hand. The Pulse

Trace PCA provides two indexes from the DVP: the

Stiffness Index (SI) and Reflection Index (RI). SI (mea-

sured in m/s) is defined as the ratio between subject’s

height (h) and DTDVP (see Fig. 1a):

SI ¼ h =DTDVP ð2Þ

The reflection index (RI) (measured as %) is defined as

the ratio between the amplitudes of the second (b) and first

(a) peaks of the DVP waveform (see Fig. 1a):

RI ¼ b=a ð3Þ

Regarding the risk factors, the waist to hip ratio (WHR)

was calculated as the minimal abdominal circumference

between the xiphoid process and iliac crests (i.e., waist)

divided by the circumference over the femoral heads (i.e.,

hips). The body mass index (BMI) was calculated as the

ratio of body weight (kg) to the square of height (m). BP

was measured using an automatic sphygmomanometer

Omrom M6 (Omrom Healthcare, Milton Keynes, UK),

following the guidelines of the European Society of

Cardiology and Hypertension (2003). Age and current

smoking status were determined by questionnaire.

Preliminary tests (in order to design the experimental

set-up) showed the marked influence of ambient tempera-

ture on the hemodynamic parameters (blood pressure,

PWV and DVP). This parameter was therefore carefully

controlled during the tests and kept at 21 ± 1�C. The

acquisition of the DVP signal was an automatic process,

i.e., the Pulse Trace PCA (Micro Medical, Kent, UK) only

showed DVP values when the recording was of sufficient

quality, which was achieved after pre-warming the sub-

ject’s finger.

Fig. 2 Transit time (DTPWV) was computed from the onset time

difference between the signals detected in two sites of measurement:

the brachial and ankle arteries. The waveforms were obtained by

means of an ultrasound sensor and are shown as brachial and ankle

waveforms. The R wave of ECG was used by the Pulse Trace PWV

(Micro Medical, Kent, UK) as timing reference to calculate DTPWV

automatically, since only one ultrasound sensor was employed. TS and

TD are systolic and diastolic period, respectively. The figure shows an

actual register obtained in our study

Fig. 3 Partial distances employed to calculate the right brachial–

ankle distance (Lba). Lba was calculated as the sum of the partial

distances 1, 2 and 3, minus the sum of the partial distances 4 and 5

(Lba = 1 ? 2 ? 3 - 4 - 5). The partial distances were (1) from

suprasternal notch to inferior edge of the umbilicus; (2) from the

inferior edge of the umbilicus to the iliac crest; (3) from the iliac crest

to the medial malleolus in the sampling site on the right leg; (4) from

suprasternal notch to head of humerus; and (5) from head of humerus

to the forearm

Cardiovasc Eng

123

Second Phase

In order to avoid the use of ultrasound techniques to

measure PWV, in the second part of the study we designed

a system to measure PWV by photoplethysmography

instead of using a Doppler probe. The equipment (see

Fig. 4) consisted of a signal conditioning module con-

nected to a data acquisition card controlled by LabView

software (National Instruments, Austin, TX, USA). The

module processed an ECG signal (Lead I), and two pho-

toplethysmography signals which registered the changes in

the transmittance of infrared light through the index finger

(index finger pulse wave) and the big toe (big toe pulse

wave). With these three signals, we defined three new

parameters: ftPWV (PWV measured between index finger

and big toe), hfPWV (PWV measured between heart -R

wave of ECG- and index finger), and htPWV (PWV

measured between heart and big toe). In each case PWV

was calculated by dividing the externally measured dis-

tance between each location by the recorded pulse transit

time (PTT). In the case of photoplethysmography signals,

PTT was measured by the onset of each pulse. The Fig. 5

shows typical waveforms obtained from the specially

designed equipment, and the PTT measured for each sig-

nal. These signals provided an additional stiffness index

and an additional reflection index, which were compared to

those obtained from the Pulse Trace PCA (Micro Medical,

Kent, UK) for validation. After validating the system, we

conducted the second phase of the study on ten healthy

volunteers (30.22 ± 7.06 years). The measuring protocol

was similar to the first phase, but included measuring the

new parameters.

Statistical Analysis

All statistical analyses were performed using SPSS 13

(SPSS, Chicago, IL, USA). Data were expressed as

mean ± standard deviation. Pearson0s correlation coeffi-

cients (r) were calculated to compare the relationships

between the variables under study and risk factors. Values

of P \ 0.05 were considered to indicate statistical signifi-

cance. Multiple regression analyses were automatically

conducted by a stepwise regression method (backward

elimination). Once the models had been obtained, they

were tested by means of the Durbin-Watson statistic,

Fisher-Snedecor statistic, variance inflation factor (VIF),

and the condition index.

Results

First Phase Overview

The first phase of the study evaluated the relationships

between cardiovascular parameters, cardiovascular mea-

surements (baPWV, SI and RI), and risk factors. Table 1

shows the basic characteristics of the 47 healthy subjects in

this group. The cardiovascular parameters included SBP,

DBP, mean blood pressure (MBP), pulse pressure (PP) and

heart rate (HR). The values of the variables under study

obtained were: baPWV of 8.4 ± 1.4 m/s, SI of 6.28 ±

0.83 m/s, and RI of 68.53 ± 8.07%. Table 2 shows the

correlation coefficients and significance between baPWV,

SI, RI, cardiovascular parameters and risk factors.

Assessment of baPWV

baPWV was positively and significantly correlated with all

the cardiovascular parameters and presented moderately

high correlation coefficients with SBP (r = 0.57,

P = 0.0001), DBP (r = 0.55, P = 0.0001) and MBP

(r = 0.60, P = 0.0001). It was also positively and signif-

icantly correlated with some risk factors, such as BMI

(r = 0.48, P = 0.001), WHR (r = 0.58, P = 0.0001) and

Fig. 4 Custom-built equipment to measure PWV by means of photoplethysmography. a The equipment consisted of a computer with a data

acquisition card and a signal conditioning module for ECG and photoplethysmography signals. b Inner view of the signal conditioning module

Cardiovasc Eng

123

waist (r = 0.60, P = 0.0001). The lack of correlation

between age and baPWV was unexpected, since baPWV

assesses arterial stiffness and this should correlate with age.

We therefore looked for confusion variables in the ana-

lyzed sample (i.e., variables which distort the relationship

between baPWV and age) and we repeated the correlation

analysis but excluding volunteers who smoked more than

Fig. 5 Typical waveforms

measured by the custom-built

equipment: ECG and two

photoplethysmography signals

at index finger and big toe. The

pulse transit time (PTT)

measured between index finger

and big toe was used to

calculate ftPWV (a), PTT

between heart (wave R of ECG)

and index finger to calculate

hfPWV (b), and PTT between

heart and big toe to calculate

htPWV (c)

Table 1 Basic characteristics of the healthy subjects who took part in

the first phase of the study (n = 47)

Characteristic Mean SD

Cardiovascular parameters SBP (mmHg) 114.37 11.70

DBP (mmHg) 67.93 6.83

MBP (mmHg) 82.86 9.00

PP (mmHg) 46.37 8.06

HR (bpm) 62.00 8.38

Risk factors Age (years) 33.78 9.04

BMI (kg/m2) 24.14 4.12

WHR (cm/cm) 0.86 0.10

Waist (cm) 85.38 13.36

SBP systolic blood pressure, DBP diastolic blood pressure, MBPmean blood pressure, PP pulse pressure, HR heart rate, BMI body

mass index, and WHR waist to hip ratio

Table 2 Correlation coefficients (r) and statistical significance (P)

between cardiovascular parameters, risk factors, brachial-ankle pulse

wave velocity (baPWV), stiffness index (SI) and reflection index (RI)

in the first phase of the study (47 healthy volunteers) (significant

values are highlighted in bold)

baPWV SI RI

r P r P r P

SBP 0.57 0.0001 0.41 0.004 0.07 [0.05

DBP 0.55 0.0001 0.40 0.005 0.16 [0.05

MBP 0.60 0.0001 0.44 0.002 0.13 [0.05

PP 0.35 0.018 0.24 [0.05 -0.04 [0.05

HR 0.36 0.014 0.29 0.046 -0.47 0.001

RI 0.10 [0.05 0.40 0.005

SI 0.55 0.0001 0.40 0.005

Age 0.28* [0.05* 0.56 0.0001 0.13 [0.05

BMI 0.48 0.001 0.19 [0.05 0.16 [0.05

WHR 0.58 0.0001 0.33 0.026 0.14 [0.05

Waist 0.60 0.0001 0.35 0.017 0.22 [0.05

SBP systolic blood pressure, DBP diastolic blood pressure, MBPmean blood pressure, PP pulse pressure, HR heart rate, BMI body

mass index, and WHR waist to hip ratio

* Correlation between age and baPWV became significant (r = 0.37,

P \ 0.05) when 11 smokers and BMI [ 30 kg/m2 volunteers were

excluded from the analysis

Cardiovasc Eng

123

three cigarettes a day and 11 subjects with BMI [ 30 kg/

m2. In this case, we did find a significant relationship

between baPWV and age, with a correlation coefficient

r = 0.37 (P \ 0.05).

In the multiple regression analyses for SBP and DBP,

using baPWV, the best models included baPWV and the

waist to hip ratio (WHR) as independent variables. The

regression models were:

SBP ¼ 25:43 þ 83:10�WHR þ 2:1� baPWV

DBP ¼ 26:33 þ 32:70�WHR þ 1:68� baPWV

with correlation coefficients of r = 0.838 and r = 0.673,

respectively.

Assessment of DVP (SI and RI)

The stiffness index derived from DVP (SI) was positively

and significantly correlated with moderately high correla-

tion coefficients with SBP (r = 0.41, P = 0.004), DBP

(r = 0.40, P = 0.005), MBP (r = 0.44, P = 0.002) and

with a low correlation coefficient with HR (r = 0.29,

P \ 0.05). SI was also positively and significantly corre-

lated with the risk factors WHR (r = 0.33, P \ 0.03),

waist (r = 0.35, P \ 0.02) and age (r = 0.56, P \ 0.001),

but did with BMI. In contrast, RI did not correlate either

with blood pressure parameters (SBP, MBP DBP) or risk

factors (BMI, WHR, waist and age) (see Table 2). RI was

negatively and significantly correlated only with HR

(r = -0.47, P = 0.001). The results also showed that

baPWV was positively and significantly correlated with SI

(r = 0.55, P = 0.001), but not with RI. On the other hand,

SI was positively and significantly correlated with RI

(r = 0.40, P = 0.005).

In the multiple regression analyses for SBP and DBP,

using SI and RI, the best models included SI and the waist

to hip ratio (WHR) as independent variables. The regres-

sion models were:

SBP ¼ 17:84 þ 72:23�WHR þ 2:7� SI

DBP ¼ 19:17 þ 42:03�WHR þ 2:1� SI

with correlation coefficients of r = 0.852 and r = 0.663,

respectively.

Second Phase

For the second phase of the study, we designed special

equipment to measure PWV by photoplethysmography.

We conducted a validation test against the Pulse Trace

PCA and PWV by measuring the parameters SI and transit

time (TT), respectively. The correlation coefficients

between the measurements from both systems were 0.88

(P = 0.004) for SI, and 0.744 (P = 0.034) for TT. Table 3

shows the basic characteristics of the 10 healthy volunteers

who took part in the second part of the study. Table 4

shows the values of the signals obtained in the study

measured by Pulse Trace PCA and PWV (Micro Medical,

Kent, UK) and by the custom-built equipment.

In the multiple regression analyses for SBP, using the

variables measured by the custom-built equipment, the best

model included hfPWV, the parameter waist to hip ratio

(WHR), and heart rate (HR) as independent variables. The

regression model was:

SBP ¼ 26:93 þ 51:49�WHR þ 15:82� hfPWV

� 0:375� HR

with a correlation coefficient r = 0.997. In contrast, the

best model for DBP only included the waist to hip ratio

(WHR) as an independent variable:

DBP ¼ 36:89 þ 38:90�WHR

with a correlation coefficient r = 0.758.

Discussion

Assessment of the Brachial-Ankle Pulse Wave Velocity

(baPWV)

The baPWV value obtained in our study (8.40 ± 1.42 m/s)

was lower than those previously reported: 11.08 ± 1.87 m/

s (Shiotani et al. 2005), 11.15 ± 1.04 m/s (Naidu et al.

2005), 11.04 ± 1.26 m/s (Liu et al. 2005), and

16.23 ± 0.93 m/s (Nakamura et al. 2003). In these studies,

baPWV was measured by techniques completely different

to ours: while Shiotani et al. (2005), Liu et al. (2005), and

Nakamura et al. (2003) used a volume-plethysmographic

system, and Naidu et al. (2005) employed a custom-made

Table 3 Basic characteristics of the healthy volunteers considered in

the second phase of the study (n = 10)

Characteristic Mean SD

Cardiovascular parameters SBP (mmHg) 111.25 12.45

DBP (mmHg) 69.37 6.80

MBP (mmHg) 78.87 16.59

PP (mmHg) 41.80 8.02

HR (bpm) 62.00 8.38

Risk factors Age (yrs) 30.22 7.06

BMI (kg/m2) 23.05 4.08

WHR (cm/cm) 0.83 0.12

Waist (cm) 81.67 16.58

SBP systolic blood pressure, DBP diastolic blood pressure, MBP,

mean blood pressure, PP pulse pressure, HR heart rate, BMI body

mass index, and WHR waist to hip ratio

Cardiovasc Eng

123

device based on the oscillometric method. The discrepancy

between our absolute baPWV value and those from the

previous studies could be due to the different measuring

techniques.

In the relationship between baPWV and the cardiovas-

cular parameters (see Table 2), we observed a moderately

high correlation between baPWV and BP (r = 0.57 with

SBP, and r = 0.55 with DBP). These values are in general

inside the range of correlation coefficients previously

reported for MBP by Shiotani et al. (2005) (r = 0.65,

P \ 0.001), and for SBP by Nakamura et al. (2003)

(r = 0.643, P \ 0.0001) and Liu et al. (2005) (r = 0.32,

P \ 0.05).

The correlations found between baPWV and risk factors

in the healthy volunteers in our study were partially in

disagreement with previous studies. We observed a mod-

erate correlation between baPWV and BMI (r = 0.48),

which is in agreement with Liu et al. (2005) (r = 0.46,

P \ 0.05), but not with Shiotani et al. (2005), who did not

find any correlation. We also found a significant correlation

with WHR (r = 0.58), which had not been reported pre-

viously. Initially, we did not find a significant correlation

between baPWV and age. However, since previous studies

on healthy volunteers had found high correlations between

baPWV and age (Nakamura et al. 2003; Liu et al. 2005),

we repeated the correlation analysis but excluding 11

smokers and BMI [ 30 kg/m2 volunteers. In this case, we

did find a significant relationship between baPWV and age,

with a correlation coefficient r = 0.34. Since our popula-

tion was composed of 70% males, our results are in good

agreement with those published by Im et al. (2007), who

found r = 0.30 in men and r = 0.11 in women. Our results

also confirm that smoker and BMI [ 30 kg/m2 status could

possibly be confusion variables in this type of analysis, as

has previously been suggested (McVeigh et al. 1997; Yufu

et al. 2007).

baPWV as an Estimator of Blood Pressure

To our knowledge, this is the first study where the baPWV

signal was employed to estimate SBP and DBP. However,

our findings are comparable to those obtained in similar

studies. For instance, Millasseau et al. (2002) reported a

regression model to infer the carotid-femoral PWV

(cfPWV) by using age and mean blood pressure (MBP), and

found a correlation coefficient of r = 0.71 (P = 0.0001).

Asmar et al. (1995) also reported a regression model with

healthy volunteers in which cfPWV was explained by age

and SBP (r = 0.685, P = 0.001). On the other hand, Kubo

et al. (2002) reported different models to explain baPWV by

using age and MBP in cardiac patients. Our results are

partially in agreement with those obtained by Millasseau

et al. (2002) and Kubo et al. (2002) since the highest cor-

relation coefficient found in our study was also for MBP

(r = 0.60) (see Table 2). However, our interest is focused

on models that explain SBP and DBP by using other car-

diovascular parameters, and in this respect, our findings are

hence new.

Assessment of the Digital Volume Pulse (DVP)

The SI value found in our study (6.28 ± 0.83 m/s) was

lower than those previously reported by Millasseau et al.

(2002) (8.4 m/s) and by Alty et al. (2007) (9.4 m/s), both

on healthy subjects. Since age is known to affect SI, this

disagreement could be due to age differences between our

subjects (33.78 ± 9.04 years) and those studied by

Millasseau et al. (2002) (47 ± 13.8 years) and Alty et al.

(2007) (50 ± 13.6 years).

We observed a moderate correlation between SI and

blood pressure (r = 0.41 for SBP, r = 0.40 for DBP and

r = 0.44 for MBP), which is in agreement with that found

by Millasseau et al. (2003) (r = 0.32 for SBP, r = 0.48 for

DBP and r = 0.45 for MBP) in a group of healthy subjects.

Concerning the relationship between SI and risk factors

in healthy volunteers (see Table 2), our results showed a

moderately high correlation between SI and age (r = 0.56),

which was similar to that previously reported by Millas-

seau et al. (2003) (r = 0.63). We also observed that SI

presented a low correlation with WHR (r = 0.33), and was

slightly higher with waist measurement (r = 0.35). Finally,

we did not find any correlation between SI and BMI. These

relationships had not been studied previously.

Mean RI value found by us in healthy volunteers was

68.53 ± 8.07%, which is in agreement with that reported

by Chowienczyk et al. (1999), also in healthy volunteers

(60.00 ± 5.50%). In our study RI did not correlate either

Table 4 Values of the signals obtained in the second phase of the

study on 10 healthy volunteers measured by both Pulse Trace PWV

(Micro Medical, Kent, UK) and custom-built equipment

Measuring equipment Parameter Mean SD

Pulse trace PCA and PWV

(Micro Medical, Kent, UK)

baPWV (m/s) 7.91 0.89

SI (m/s) 6.13 0.76

RI (%) 67.73 10.52

Custom-built equipment ftPWV (m/s) 6.94 1.20

hfPWV (m/s) 4.12 0.49

htPWV (m/s) 4.97 0.59

SI* (m/s) 6.15 1.20

RI* (%) 67.32 9.06

BaPWV brachial-ankle pulse wave velocity, SI stiffness index, RIreflection index (these two parameters measured from Pulse Trace

PWV of Micro Medical, Kent, UK), ftPWV index finger-big toe pulse

wave velocity, hfPWV heart-index finger pulse wave velocity, htPWVheart- big toe pulse wave velocity, SI* stiffness index; RI* reflection

index (both parameters measured from the custom-built equipment)

Cardiovasc Eng

123

with SBP, DBP, MBP or with PP (see Table 2), which is

also in agreement with Millasseau et al. (2003). We also

found a significant negative correlation between RI and HR

(r = -0.473, P = 0.001), which has also been previously

reported by Chowienczyk et al. (1999) and Millasseau et al.

(2003).

Finally, even though the relationship between baPWV

and SI had not been studied previously, the correlation we

found (r = 0.55, P = 0.0001) was similar to that found by

Millasseau et al. (2002) between carotid-femoral PWV

(cfPWV) and SI (r [ 0.65, P \ 0.01). The relationship

between PWV and arterial stiffness has been previously

discussed (Hamilton et al. 2007). In addition, by using the

Moens–Kortweg formula, it is know that PWV is propor-

tional to the square root of the elastic modulus of the

arterial wall. However, it has been also pointed out that the

square root relationship also means that a change in PWV

is not a particularly sensitive measure of change in physical

arterial properties (Hughes et al. 2004).

DVP as an Estimator of Blood Pressure

Only the SI parameter derived from DVP was found to be

useful in the multiple regression models. Once more, and

although there are no previous studies with which to

compare our SBP and DBP models, we should point out

that Millasseau et al. (2002) reported a model with a high

correlation coefficient (r = 0.69, P = 0.0001) in which SI

was explained both by age and by MBP. This is in close

agreement with our experimental results, in which the

highest correlation coefficients were between SI and age,

and between SI and MBP.

hfPWV as an Estimator of Blood Pressure

First of all, we found a high correlation coefficient

(r = 0.88) between the SI measured by the custom-built

equipment and that measured by the Pulse Trace PCA

(Micro Medical, Kent, UK), which suggests that both the

transmittance photoplethysmography technique (in the

custom-built equipment) and the reflectance technique

(Pulse Trace PCA) offer similar results. Moreover, the

mean ftPWV value measured in our study (6.94 ± 1.20 m/

s) was similar to those reported in healthy volunteers by

Tsai et al. (2005) (6.39 ± 0.93 m/s) and Chen et al. (2004)

(6.49 ± 0.92 m/s).

Secondly, in the multiple regression analyses for SBP

the best model included hfPWV, waist to hip ratio (WHR),

and heart rate (HR) as independent variables. The high

correlation coefficient (r = 0.997) given by this model

suggests that hfPWV could be a good estimator of the beat-

to-beat SBP. hfPWV could therefore be useful, for

instance, for monitoring cardiac patients during physical

exercise in cardiac rehabilitation. Unfortunately, since

DBP was only explained by the WHR, this means that we

did not find any signal that could potentially be used to

estimate DBP. Finally, future work should be conducted

with subjects undergoing physical exercise to assess the

reliability of our current results.

Conclusions

The findings of this study suggest that:

(1) Both the baPWV and SI signals (measured from

DVP) are good estimators of SBP and DBP. In

contrast, RI (measured from DVP) does not signifi-

cantly correlate with any blood pressure parameter.

(2) Estimation of SBP and DBP in multiple regression

analysis is improved by including the waist to hip

ratio (WHR). In this case, similar correlation coeffi-

cients are obtained from baPWV and SI.

(3) Results are improved when new variables measured

by the custom-built equipment are considered in the

multiple regression analyses. The best model to

explain the SBP (r = 0.997) includes the hfPWV

(PWV measured by photoplethysmography between

R-wave of ECG and index finger), along with the

waist to hip ratio (WHR) and heart rate (HR) as

independent variables. DBP could not be explained

by any signal from the custom-built equipment.

(4) These results suggest that hfPWV could be a good

estimator of the beat-to-beat SBP, which could be

very useful for monitoring cardiac patients during

physical exercise in cardiac rehabilitation.

Acknowledgments The translation of this paper was funded by the

Universidad Politecnica de Valencia, Spain. This study was partially

supported by Merce V. Electromedicina (Valencia, Spain) and by

‘‘IMIDT-Programa de Investigacion y Desarrollo Tecnologico’’ of the

‘‘Instituto de la Mediana y Pequena Industria de la Comunidad Val-

enciana (IMPIVA)’’.

References

Alty SR, Angarita-Jaimes N, Millasseau SC, Chowienczyk PJ.

Predicting arterial stiffness from the digital volume pulse

waveform. IEEE Trans Biomed Eng. 2007;54:2268–75.

Asmar R, Benetos A, Topouchian J, Laurent P, Pannier B, Brisac AM,

et al. Assessment of arterial distensibility by automatic pulse

wave velocity measurement. Validation and clinical application

studies. Hypertension. 1995;26:485–90.

Chen W, Kobayashi T, Ichikawa S, Takeuchi Y, Togawa T.

Continuous estimation of systolic blood pressure using the pulse

arrival time and intermittent calibration. Med Biol Eng Comput.

2000;38:569–74.

Chen YK, Wu HT, Chi CK, Tsai WC, Chen JY, Wang MC. A new dual

channel pulse wave velocity measurement system. Bioinformatics

Cardiovasc Eng

123

and Bioengineering, 2004. BIBE 2004. Proceedings. Fourth IEEE

symposium on volume, 19–21 May 2004: 17–21.

Chowienczyk PJ, Kelly RP, MacCallum H, Millasseau SC, Andersson

TL, Gosling RG, et al. Photoplethysmographic assessment of

pulse wave reflection: blunted response to endothelium-depen-

dent beta2-adrenergic vasodilation in type II diabetes mellitus. J

Am Coll Cardiol. 1999;34:2007–14.

European Society of Cardiology and Hypertension. European Society

of Hypertension-European Society of Cardiology guidelines for

the management of arterial hypertension. J Hypertens.

2003;21:1011–53.

Geddes LA, Voelz MH, Babbs CF, Bourland JD, Tacker WA. Pulse

transit time as an indicator of arterial blood pressure. Psycho-

physiology. 1981;18:71–4.

Hamilton PK, Lockhart CJ, Quinn CE, McVeigh GE. Arterial

stiffness: clinical relevance, measurement and treatment. Clin

Sci (Lond). 2007;113:157–70.

Hughes SM, Dixon LJ, McVeigh GE. Arterial stiffness and pulse

wave velocity: problems with terminology. Circulation.

2004;109:e3. (author reply e3).

Im JA, Lee JW, Shim JY, Lee HR, Lee DC. Association between

brachial-ankle pulse wave velocity and cardiovascular risk

factors in healthy adolescents. J Pediatr. 2007;150:247–51.

Kubo T, Miyata M, Minagoe S, Setoyama S, Maruyama I, Tei C. A

simple oscillometric technique for determining new indices of

arterial distensibility. Hypertens Res. 2002;25:351–8.

Lehmann ED. Clinical value of aortic pulse-wave velocity measure-

ment. Lancet. 1999;354:528–9.

Liu PY, Tsai WC, Lin CC, Hsu CH, Haung YY, Chen JH. Invasive

measurements of pulse wave velocity correlate with the degree

of aortic valve calcification and severity associated with matrix

metalloproteinases in elderly patients with aortic valve stenosis.

Clin Sci (Lond). 2004;107:415–22.

Liu H, Yambe T, Zhang X, Saijo Y, Shiraishi Y, Sekine K, et al.

Milyagin VA and Nitta S Comparison of brachial-ankle pulse

wave velocity in Japanese and Russians. Tohoku J Exp Med.

2005;207:263–70.

McVeigh GE, Morgan DJ, Finkelstein SM, Lemay LA, Cohn JN.

Vascular abnormalities associated with long-term cigarette

smoking identified by arterial waveform analysis. Am J Med.

1997;102:227–31.

Millasseau SC, Kelly RP, Ritter JM, Chowienczyk PJ. Determination

of age-related increases in large artery stiffness by digital pulse

contour analysis. Clin Sci (Lond). 2002;103:371–7.

Millasseau SC, Kelly RP, Ritter JM, Chowienczyk PJ. The vascular

impact of aging and vasoactive drugs: comparison of two digital

volume pulse measurements. Am J Hypertens. 2003;16:467–72.

Naidu MU, Reddy BM, Yashmaina S, Patnaik AN, Rani PU. Validity

and reproducibility of arterial pulse wave velocity measurement

using new device with oscillometric technique: a pilot study.

Biomed Eng. 2005;4:49. (Online Aug 23, 2005).

Nakamura U, Iwase M, Nohara S, Kanai H, Ichikawa K, Iida M.

Usefulness of brachial-ankle pulse wave velocity measurement:

correlation with abdominal aortic calcification. Hypertens Res.

2003;26:163–7.

Nieminen T, Leino J, Maanoja J, Nikus K, Viik J, Lehtimaki T, et al.

The prognostic value of haemodynamic parameters in the

recovery phase of an exercise test. The finnish cardiovascular

study. J Hum Hypertens. 2008;22:537–43.

Pierson LM, Miller LE, Herbert WG. Predicting exercise training

outcome from cardiac rehabilitation. J Cardiopulm Rehabil.

2004;24:113–8.

Shiotani A, Motoyama M, Matsuda T, Miyanishi T. Brachial-ankle

pulse wave velocity in Japanese university students. Intern Med.

2005;44:696–701.

Tsai WC, Chen JY, Wang MC, Wu HT, Chi CK, Chen YK, et al.

Association of risk factors with increased pulse wave velocity

detected by a novel method using dual-channel photoplethys-

mography. Am J Hypertens. 2005;18:1118–20.

Yamashina A, Tomiyama H, Takeda K, Tsuda H, Arai T, Hirose K,

et al. Validity, reproducibility, and clinical significance of

noninvasive brachial-ankle pulse wave velocity measurement.

Hypertens Res. 2002;25:359–64.

Yufu K, Takahashi N, Hara M, Saikawa T, Yoshimatsu H.

Measurement of the brachial-ankle pulse wave velocity and

flow-mediated dilatation in young, healthy smokers. Hypertens

Res. 2007;30:607–12.

Cardiovasc Eng

123