Change in pulse transit time and pre eje

The present study 18 investigated the feasibility of using pulse transit time PTT to 19 track variations in pre-ejection period PEP during progressive 20 central hypovolaemia induced by

PROOF

1

6

7 Gregory S H Chan1,2, Paul M Middleton1,3, Branko

8 G Celler, PhD1, Lu Wang1 and Nigel H Lovell,

9 PhD1,2,4

10

11

Chan GSH, Middleton PM, Celler BG, Wang L, Lovell NH Change

12

in pulse transit time and pre-ejection period during head-up

tilt-in-13

duced progressive central hypovolaemia.

14

J Clin Monit Comput 2007

15 ABSTRACT Objective Traditional vital signs such as heart rate

16

(HR) and blood pressure (BP) are often regarded as insensitive

17

markers of mild to moderate blood loss The present study

18

investigated the feasibility of using pulse transit time (PTT) to

19

track variations in pre-ejection period (PEP) during progressive

20

central hypovolaemia induced by head-up tilt and evaluated the

21

potential of PTT as an early non-invasive indicator of blood

22

loss Methods About 11 healthy subjects underwent graded

23

head-up tilt from 0 to 80° PTT and PEP were computed from

24

the simultaneous measurement of electrocardiogram (ECG),

25

finger photoplethysmographic pulse oximetry waveform

(PPG-26

POW) and thoracic impedance plethysmogram (IPG) The

27

response of PTT and PEP to tilt was compared with that of

28

interbeat heart interval (RR) and BP Least-squares linear

29

regression analysis was carried out on an intra-subject basis

30

between PTT and PEP and between various physiological

31

variables and sine of the tilt angle (which is associated with the

32

decrease in central blood volume) and the correlation

33

coefficients (r) were computed Results During graded tilt,

34

PEP and PTT were strongly correlated in 10 out of 11 subjects

35

(median r = 0.964) and had strong positive linear correlations

36

with sine of the tilt angle (median r = 0.966 and 0.938

37

respectively) At a mild hypovolaemic state (20–30°), there

38

was a significant increase in PTT and PEP compared with

39

baseline (0°) but without a significant change in RR and BP

40

Gradient analysis showed that PTT was more responsive to

41

central volume loss than RR during mild hypovolaemia (0–20°)

42

but not moderate hypovolaemia (50–80°) Conclusion PTT

43

may reflect variation in PEP and central blood volume, and is

44

potentially useful for early detection of non-hypotensive

45

progressive central hypovolaemia Joint interpretation of PTT

46

and RR trends or responses may help to characterize the extent

47

of blood volume loss in critical care patients

48

KEY WORDS pulse transit time (PTT), pulse transmission time,

49

pre-ejection period, head-up tilt, hypovolaemia, blood loss.

50 51

INTRODUCTION

53 Early detection of internal bleeding has often been a

dif-54 ficult task for clinicians Vital sign monitors that are

cur-55 rently in use in emergency department (ED) or in

56 emergency transport vehicles measure a range of

physio-57 logical variables including heart rate (HR) and blood

58 pressure (BP) but these variables are often regarded as

59 insensitive markers of mild to moderate blood loss [1, 2]

60 The decrease in central blood volume during early stage of

61 blood loss typically triggers a baroreflex response that acts

62

to maintain a perfusing BP despite a decline in stroke

63 volume BP may not decrease considerably until about

From the 1Biomedical Systems Laboratory, School of Electrical

Engineering and Telecommunications, University of New South

Wales, Sydney, NSW 2052, Australia;2Graduate School of

Bio-medical Engineering, University of New South Wales, Sydney,

NSW 2052, Australia;3Prince of Wales Clinical School, University

of New South Wales, Sydney, NSW 2031, Australia; 4National

Information and Communications Technology Australia (NICTA),

Eveleigh, NSW 1430, Australia.

Received 10 May 2007 Accepted for publication 9 July 2007.

Address correspondence to N H Lovell, Graduate School of

Biomedical Engineering, University of New South Wales, Sydney,

NSW 2052, Australia.

E-mail: N.Lovell@unsw.edu.au

PROOF

64 30% of total blood volume has been lost, by which time

65 patients are at high risk of cardiovascular collapse as a

66 result of haemorrhagic shock [1, 3–6] Delayed control of

67 haemorrhage has been recognized as a major contributor

68 to preventable trauma deaths and has often been related to

69 delays in the assessment or diagnosis of haemorrhage [7,

70 8] There are potentially large benefits to the critical care

71 clinician if small volume losses could be diagnosed early,

72 accurately and reproducibly simply by the assessment of a

73 physiological variable that can be conveniently derived

74 from existing patient monitoring equipment

75 Recently, a significant amount of research effort has

76 been devoted to the pulse transit time or the pulse

77 transmission time (PTT) [9, 10] PTT is typically

mea-78 sured as the time interval from the R-wave of the

elec-79 trocardiogram (ECG) to a reference point on the systolic

80 upstroke of a subsequent peripheral pulse wave It consists

81 of two components: the pre-ejection period (PEP), which

82 corresponds to the timing from the onset of ventricular

83 depolarisation to the onset of ventricular ejection, and the

84 vascular transit time (VTT), which defines the period for

85 the arterial pulse wave to travel from the aortic valve to

86 the peripheral arteries In particular, the PEP component

87 of PTT is known to vary with cardiac preload [11–13]

88 Recent studies have shown that respiratory variation in

89 PTT/PEP could predict fluid responsiveness in patients

90 [14, 15] From the perspective of clinical monitoring,

91 PTT has the potential to become widely applied in patient

92 care since its derivation only requires ECG and a

93 peripheral pulse waveform, such as the finger

photople-94 thysmographic pulse oximetry waveform (PPG-POW)

95 which has been commonly used for the monitoring of

96 arterial oxygen saturation (SpO2) Both ECG and

PPG-97 POW are routinely measured by existing vital sign

98 monitors, and their measurement is totally noninvasive

99 and causes minimal discomfort to the patients By

moni-100 toring PTT in a continuous beat-by-beat manner, it may

101 be possible to identify subtle change in the patientÕs

car-102 diovascular status caused by small amounts of progressive

103 blood loss over time

104 Previous studies have identified the potential value of

105 PTT in the detection of hypotension caused by central

106 hypovolaemia [16, 17] Ahlstrom et al showed that PTT

107 could track changes in systolic BP during simulated

108 hypovolaemia with lower body negative pressure (LBNP)

109 [16] A study of actual haemorrhage in dogs by Ochiai

110 et al demonstrated that hypotension caused by acute

111 blood loss could be potentially identified as a prolongation

112 in PTT [17] However, these studies involved a high

113 degree of central hypovolaemia which subsequently led to

114 hypotension It is not clear whether PTT is also useful for

115 detecting mild hypovolaemia in the absence of significant

116 BP reduction

117 The purpose of the present study was to identify the

118 change in PTT associated with the decline in central

119 blood volume, similar to that which occurs during mild to

120 moderate blood loss Graded head-up tilt was used as a

121 model to simulate progressive central hypovolaemia [3–5,

122

18, 19] The sine of the tilt angle (sinh) is proportional to

123 the hydrostatic effect of head-up tilting [20, 21], and a

124 linear relationship has been found between sinh and the

125 decrease in thoracic fluid content [22] Although graded

126 head-up tilt is not truly equivalent to actual blood loss

127 since the blood volume is merely re-distributed to the

128 lower body under gravitational influence rather than

129 actually lost from the circulatory system, it may at least

130 simulate most of the cardiovascular responses to a

pro-131 gressive decline in central blood volume similar to that

132 occurring during haemorrhage

133

In the current study, the change in PTT and PEP at

134 different levels of central blood volume induced by graded

135 tilt was examined along with corresponding responses in

136 interbeat heart interval (RR) and BP Intra-subject

137 regression analysis was carried out (1) between PTT and

138 PEP to determine how much PEP contributed to the

139 PTT variations associated with change in central blood

140 volume, and (2) between the different physiological

141 variables and sinh to determine the association of the

142 variables with central blood volume Moreover, the

gra-143 dient of the variables with respect to tilt angle increment

144 was computed to provide a measure of the directional

145 change in the variable in response to a further decrease in

146 central blood volume at a given volume status represented

147

by the tilt angle A positive/negative gradient would

148 indicate an increasing/decreasing trend in the variable as

149 volume loss progressed

150 METHODS AND MATERIALS

151

Subject

152 About 11 healthy subjects (10 males and 1 female, aged

153 18–44 years, mean age 30 years) were studied Prior to

154 the experiment, subjects were requested to provide

155 information about their physical condition and none

re-156 ported any history of cardiovascular or respiratory disease

157 Written informed consent was obtained from all

partici-158 pants, and the study was approved by the Human

Re-159 search Ethics Advisory (HREA) Panel of the University of

160 New South Wales

161

Measurement devices and systems

162 PPG-POW was measured from the tip of the right index

163 finger using a reflection mode infrared finger probe

PROOF

164 (ADInstruments, Sydney, Australia) ECG was acquired

165 from the lead I configuration and amplified with a

bioam-166 plifier (ADInstruments, Sydney, Australia) The thoracic

167 impedance plethysmogram (IPG) was acquired using the

168 Tetrapolar High-Resolution Impedance Monitor, also

169 known as the THRIM (UFI, Morro Bay, USA) The two

170 Ag/AgCl electrodes for current injection were placed on

171 the right clavicle and on the left leg respectively A constant

172 sinusoidal alternating current of 50 kHz and 0.1 mA rms,

173 which was not perceivable by the subjects, was applied

174 between the two current injecting electrodes The two

175 Ag/AgCl electrodes for the measurement of thoracic IPG

176 were positioned over the sternum: one at the top of the

177 sternum and another superior to the xiphoid process From

178 the anatomical perspective, this electrode arrangement

179 should produce a thoracic IPG which reflects the change in

180 blood volume predominantly in the aorta and in the

tho-181 racic vessels [23] The signals were recorded and digitised at

182 a sampling rate of 1000 Hz using the Powerlab data

183 acquisition system (ADInstruments, Sydney, Australia) BP

184 measurements, including systolic blood pressure (SBP),

185 diastolic blood pressure (DBP), mean arterial pressure

186 (MAP), and pulse pressure (PP), were obtained using a

187 clinically approved oscillometric BP device (Colin Co.,

188 Japan) from a cuff placed around the left arm over the

189 brachial artery

191 The subjects were advised not to eat for at least 2 h prior

192 to the study, with any meal to be free of alcohol and

193 caffeine beverages The subjects were also asked not to

194 undertake any intensive exercise within 12 h before the

195 study All measurements were made in a quiet dimly lit

196 room at an ambient temperature of approximately 24°C

197 The subject initially rested in a supine position on the tilt

198 table for a period of 20 minutes The subjectÕs feet were

199 supported by a footboard, and straps were applied at the

200 levels of waist and knees to stabilize the body during

head-201 up tilting Measurements were made at each of the

fol-202 lowing tilt angles in incremental order: 0, 10, 20, 30, 40,

203 50, 60 and 80° At each tilt angle, PPG-POW, ECG and

204 thoracic IPG were simultaneously recorded for a period of

205 15 s, followed by a measurement of BP A 15 s

mea-206 surement period is considered sufficient to encompass at

207 least one respiratory cycle, allowing the influence of

208 respiratory phase on the measurements to be minimized

209 by averaging Once measurements at the current tilt angle

210 were completed, the subject was tilted to the next angle

211 After each tilt, and before the next phase of measurement

212 commenced, a 1.5 min adaptation period allowed the

213 measured cardiovascular variables to settle to a stable level,

214 which generally takes up to 30 s [24] Measurements were

215 made with the subject breathing spontaneously BP

216 measurements and finger PPG-POW signals were

ac-217 quired with the subjectÕs forearms supported by armrests

218 maintained at close to the heart level

219

Signal processing and parameter extraction

220 All signal processing and feature extraction were

imple-221 mented in Matlab (the MathWorks Inc., Natick, USA)

222 The R-wave peaks were detected from the ECG signal

223 using a set of automatic programming routines involving

224 lowpass filtering, differentiation, and threshold peak

225 detection The processing of the PPG-POW and the AC

226 component of the thoracic IPG involved four main stages:

227 (1) Lowpass filtering—An 8th order Butterworth lowpass

228 filter with a 3-dB point at 18 Hz was designed to remove

229 high frequency noise Zero-phase filtering was

imple-230 mented, which involved filtering the signal in both forward

231 and backward directions, to eliminate phase distortion (2)

232 Baseline removal—The baseline of the two signals was

233 approximated by moving averaging For the PPG-POW, a

234

2 s window (3-dB point at 0.23 Hz) was used, whereas for

235 the thoracic IPG, a 1.5 s window (3-dB point at 0.3 Hz)

236 was used The baseline component was subsequently

sub-237 tracted from the respective signals (3) Differentiation—

238

A 5-point digital differentiator was designed to differentiate

239 the two signals to obtain the first derivative (d1), the second

240 derivative (d2), and the third derivative (d3), namely

241 d1POW, d2POW and d3POW for

PPG-242 POW, and dZ/dt, d2Z/dt2and d3Z/dt3for thoracic IPG

243 The high order derivatives were generally noisy and

244 therefore were smoothed by moving averaging with a

245 31.3 ms window (3-dB point at 14 Hz) (4) Pulse

detec-246 tion—A threshold detection algorithm was implemented

247 for detecting the systolic peaks from the derivatives of the

248 two signals All the data traces were free of artefact and

249 therefore artefact rejection was not necessary

250 Several timing parameters were derived from ECG,

251 PPG-POW, and thoracic IPG, including RR, PTT, PEP

252 and VTT RR was computed as the time interval

be-253 tween successive R-wave peaks PTT was computed as

254 the time interval between the R-wave peak and the

ar-255 rival of the subsequent pulse in finger d1PPG-POW (see

256 Figure 1) In the present study, d1PPG-POW was taken

257

as the reference pulse signal for PTT measurement due to

258 its close association with arterial inflow [25] The

refer-259 ence point chosen for PTT computation was the onset or

260 foot of d1PPG-POW, which could be reliably detected

261 from the systolic peaks in d3PPG-POW based on the

262 second derivative method [26]

263 PEP was computed as the time interval between

R-264 wave peak and the onset of ventricular ejection detected

265 from the subsequent thoracic dZ/dt pulse (see Figure1)

PROOF

266 R-wave peak was used as the reference point to represent

267 ventricular depolarisation due to the reliability of its

268 detection [16, 17] The onset of ventricular ejection was

269 identified from thoracic dZ/dt by locating the so-called

270 B-point, which appeared as an incisura at the base of the

271 rising edge of the large systolic wave in dZ/dt [27] and

272 was detected using the derivatives [28]

274 The RR, PTT, PEP and VTT of a subject at a given tilt

275 angle were averaged over the 15 s recording period The

276 mean and standard error (SE) of all subject measurements

277

at each tilt angle were calculated, and the mean ± SE was

278 plotted against sinh The range and coefficient of variation

279

of PTT and PEP were computed for 0 and 80° Moreover,

280 the gradient of the variable was calculated as the difference

281

in the variable divided by the difference in sinh between

282 successive tilt angles, to measure the directional change of

283 the variable in response to a unit decrement in central

284 blood volume The average gradients of the variable in

285 three stages were computed: (1) 0–20° (mild

hypovola-286 emia), (2) 20–50° (mild-to-moderate hypovolaemia), (3)

287 50–80° (moderate hypovolaemia) Nonparametric

Fried-288 manÕs ANOVA test for repeated measures was used to

289 determine whether any significant change occurred in the

290 variable and its gradient during sequential tilting, and when

291 significant change was detected, Wilcoxon signed rank test

292 was performed post hoc with Bonferroni correction to test

293 whether there was significant increase/decrease in the

294 variable from baseline (0°) or in its gradient between the

295 three stages Wilcoxon signed rank test with Bonferroni

296 correction was also used to test whether there was a

sig-297 nificant positive/negative gradient in each stage For all

298 statistical tests, p < 0.05 was considered significant

Least-299 squares linear regression analysis was carried out between

300 PTT and PEP, and between each variable and sinh The

301 correlation coefficient (r) was computed The regression

302 relationship was considered significant if p < 0.05

303 RESULTS

304 The results are expressed as mean ± SE Overall, there was

305 significant change in RR (p < 0.001), PTT (p < 0.001),

306 PEP (p < 0.001), VTT (p < 0.001), DBP (p < 0.05), MAP

307 (p < 0.05) and PP (p < 0.01) during tilting but no

signifi-308 cant change in SBP (p > 0.05) Table1 shows the values

309

of RR, PTT, PEP, VTT, SBP, DBP, MAP and PP at

Time (s)

PEP

PTT

ECG

Thoracic dZ/dt

d1PPG−POW VTT

Fig 1 Detection of PEP, PTT and VTT from ECG (top), thoracic dZ/

dt (middle), and d1PPG-POW (bottom) PEP corresponds to the time

interval from the R-wave peak (square) to the onset of ventricular ejection

detected from thoracic dZ/dt (triangle), PTT corresponds to the time interval

from the R-wave peak to the foot of the d1PPG-POW pulse (circle), and

VTT corresponds to the time interval between the onset of ventricular

ejection and the foot of the d1PPG-POW pulse.

Table1 Physiological variables at different tilt angles

h (°) RR (ms) PTT (ms) PEP (ms) VTT (ms) SBP (mmHg) DBP (mmHg) MAP (mmHg) PP (mmHg)

0 1031 ± 17 190 ± 6 109 ± 5 81 ± 3 104 ± 3 60 ± 1 75 ± 2 44 ± 2

10 1031 ± 30 195 ± 6 117 ± 6* 78 ± 3 107 ± 2 60 ± 1 76 ± 2 47 ± 2

20 998 ± 25 202 ± 7* 122 ± 7* 80 ± 3 106 ± 3 58 ± 2 75 ± 3 47 ± 2

30 966 ± 32 204 ± 7* 126 ± 6* 78 ± 3 107 ± 3 59 ± 2 75 ± 2 48 ± 3

40 929 ± 36* 208 ± 7* 132 ± 7* 77 ± 3 106 ± 3 58 ± 2 75 ± 2 47 ± 2

50 881 ± 45* 212 ± 7* 137 ± 6** 75 ± 3 105 ± 3 61 ± 2 76 ± 3 44 ± 2

60 847 ± 40** 211 ± 8* 139 ± 7** 72 ± 3** 105 ± 3 61 ± 2 78 ± 2 43 ± 2

80 797 ± 45** 215 ± 7** 143 ± 7** 72 ± 3 105 ± 3 63 ± 2 79 ± 2 42 ± 2

Results are presented as mean ± SE ** p < 0.01, * p < 0.05: significant increase/decrease from 0° (with Bonferroni correction made) Abbreviations: h = tilt angle.

PROOF

310 different tilt angles and any significant changes from

311 baseline There was no significant change in RR from

312 baseline at 10–30° but there was significant decrease at 40°

313 and above PTT was significantly higher than baseline at

314 20° and above whereas PEP was significantly higher than

315 baseline at 10° and above At 0°, PTT ranged from 152 to

316 228 ms with CV of 11%, whereas PEP ranged from 75 to

317 134 ms with CV of 16% At 80°, PTT ranged from 160 to

318 255 ms with CV of 11%, whereas PEP ranged from 98 to

319 174 ms with CV of 16% The percentage change in mean

320 PTT and PEP between 0 and 80° were 13 and 31%

321 respectively There was no significant change in VTT

322 from baseline except for a significant decrease at 60° No

323 significant change from baseline was observed in SBP,

324 DBP, MAP and PP In Figures 2–4, the mean ± SE of

325 each variable is plotted against sinh As sinh increased, RR

326 decreased with the rate of decrease tending to be greater at

327 higher tilt angles PTT and PEP, on the other hand,

in-328 creased linearly with sinh, while VTT decreased slightly

329 The BP variables did not appear to change with sinh at

330 low tilt angles, although there was a tendency for MAP

331 and DBP to increase and for PP to decrease at high tilt

332 angles

333 Table 2 shows the gradients of RR, PTT, PEP, VTT,

334 SBP, DBP, MAP and PP at the three stages (0–20°, 20–

335 50° and 50–80°) and any significantly positive gradient

336 (rising trend) or negative gradient (falling trend) A

sig-337 nificantly negative gradient was identified in RR at 20–

338 50° and 50–80° and in PP at 20–50° A significantly

339 positive gradient was identified in PTT at 0–20° and 20–

340 50°, in PEP at all three stages and in MAP at 50–80°

341 Overall, there was a significant change in gradient in RR

342 (p < 0.05) and DBP (p < 0.05) but not in other variables A

348 significant decrease in gradient compared with 0–20° was

349 identified in RR at both 20–50° and 50–80°

350 The results of intra-subject regression analysis of PEP

351 against PTT, and RR, PTT, PEP, VTT, SBP, DBP,

352 MAP and PP against sinh are shown in Table 3 The

353 correlation between PEP and PTT was generally strong

354 (median r = 0.964, range of r from 0.626 to 0.988), and

355

10 out of 11 subjects had positive and significant

regres-356 sion relationships (p < 0.05) The regression slope of PEP

357 against PTT was significantly positive (1.18 ± 0.13) PEP

358 had the strongest correlation with sinh (median

359

r = 0.966), and the regression relationships were positive

700

800

900

1000

1100

sinΘ

RR

Fig 2 RR against sinh As sinh increases, RR decreases with the rate of

decrease tending to be greater at higher tilt angles.

50 100 150 200 250

sinΘ

PTT

PEP

VTT

Fig 3 PTT, PEP and VTT against sinh As sinh increases, PEP and PTT increase in a linear manner, while VTT decreases slightly.

40 50 60 70 80 90 100 110

sinΘ

SBP

MAP

DBP

PP

Fig 4 SBP, MAP, DBP and PP against sinh The BP variables do not appear to change with sinh at low tilt angles although there is a tendency for MAP and DBP to increase and for PP to decrease at high tilt angles.

PROOF

360 and significant in 10 out of 11 subjects The subject who

361 had poor correlation between PEP and sinh (r = 0.417,

362 p > 0.05) also had poor correlation between PEP and PTT

363 (r = 0.626, p > 0.05) The regression slope of PEP against

364 sinh was significantly positive (33.9 ± 4.4 ms) PTT

365 showed a strong correlation with sinh (median r = 0.938),

366 and the regression relationships were positive and

signif-367 icant in 8 out of 11 subjects The regression slope of PTT

368 against sinh was significantly positive (25.0 ± 3.0 ms) RR

369 also showed a strong correlation with sinh (median

370 r = )0.927), and the regression relationships were

nega-371 tive and significant in 9 out of 11 subjects The regression

372 slope of RR against sinh was significantly negative

373 ()245 ± 47 ms) VTT showed a moderate negative

cor-374 relation with sinh (median r = )0.665), but the regression

375 relationships were negative and significant in only 5 out of

376 11 subjects The regression slope of VTT against sinh was

377 significantly negative ()8.8 ± 2.7 ms) The regression

378 relationships between the BP variables and sinh varied

379 considerably between subjects and did not reach statistical

380 significance for most subjects, and the regression slopes

381 were not significant

382 DISCUSSION

383 The present study highlights the potential value of PTT as

384

a sensitive early non-invasive marker of falling central

385 blood volume Graded head-up tilt from 0 to 80° has been

386 used as a model to simulate the transition from mild to

387 moderate central hypovolaemia, similar to that occurs in

388 progressive blood loss An important new finding of the

389 present study is that PTT can signal a drop in central

390 blood volume relative to the normovolaemic state (0°) at a

Table2 Gradients of physiological variables at the three stages

h (°) RR (ms) PTT (ms) PEP (ms) VTT (ms) SBP (mmHg) DBP (mmHg) MAP (mmHg) PP (mmHg) 0–20 )98 + 49 35 + 8** 38 + 9** )3 + 6 5 + 5 )5 + 4 )2 + 5 10 + 6

20–50 )285 + 77**## 23 + 6** 35 + 4** )11 + 5 )2 + 4 6 + 3 2 + 3 )8 + 2*

50–80 )380 + 69**## 14 + 13 28 + 9* )14 + 12 1 + 8 10 + 4 16 + 6* )9 + 6

Results are presented as mean ± SE ** p < 0.01, * p < 0.05: significantly positive/negative gradient (with Bonferroni correction made).

##

p < 0.01,#p < 0.05: significant increase/decrease from 0–20° (with Bonferroni correction made) Abbreviations: h = tilt angle.

Table3 Correlation coefficients from intra-subject regression analysis

Subject PEP-PTT RR-h PTT-h PEP-h VTT-h SBP-h DBP-h MAP-h PP-h

1 0.969* )0.952* 0.829* 0.922* )0.937* 0.186 )0.181 0.306 0.275

2 0.964* )0.514 0.982* 0.978* )0.584 )0.015 )0.249 )0.176 0.155

3 0.931* )0.618 0.957* 0.966* )0.790* 0.521 0.912* 0.741* )0.101

4 0.981* )0.941* 0.980* 0.966* )0.612 )0.819* )0.716* )0.682 )0.244

5 0.982* )0.960* 0.996* 0.978* )0.753* )0.030 0.257 )0.126 )0.161

6 0.717* )0.855* 0.890* 0.899* )0.365 0.708* 0.559 0.708* 0.128

7 0.965* )0.927* 0.938* 0.987* 0.065 )0.263 0.201 0.308 )0.461

8 0.988* )0.988* 0.988* 0.990* )0.958* )0.051 0.846* 0.703 )0.570

9 0.813* )0.944* 0.627 0.843* )0.804* )0.708* )0.455 0.338 )0.295

10 0.626 )0.927* 0.515 0.417 0.473 )0.213 )0.273 0.272 )0.081

11 0.867* )0.774* 0.622 0.899* )0.665 0.730* 0.947* 0.622 )0.898* Med r 0.964 )0.927 0.938 0.966 )0.665 )0.030 0.201 0.308 )0.161

Max r 0.988 )0.514 0.996 0.990 0.473 0.730 0.947 0.741 0.275

Min r 0.626 )0.988 0.515 0.417 )0.958 )0.819 )0.716 )0.682 )0.898

Mean m 1.18* )245* 25.0* 33.9* )8.8* )0.16 2.49 2.92 )2.65

±SE m ±0.13 ±47 ±3.5 ±4.4 ±2.7 ±2.12 ±2.32 ±1.66 ±1.26

* p < 0.05: significant Abbreviations: PEP-PTT = PEP against PTT, RR-h = RR against sine of the tilt angle etc The table displays the correlation coefficients of the intra-subject regressions, with the last five rows corresponding to the median, maximum and minimum of the subjects Õ correlation coefficients, and the mean and standard error of the regression slopes.

PROOF

391 much early stage than RR and BP A significant rise in

392 PTT occurred at 20° tilt whereas a significant fall in RR

393 only occurred at 40° tilt and above, while BP did not

394 show any significant change from baseline at all tilt angles

395 The change in PTT during mild to moderate central

396 hypovolaemia has been shown to reflect predominantly

397 the change in PEP, and this finding is in agreement with

398 the study by Newlin which revealed considerable

con-399 tribution of PEP to PTT variation [29] A significantly

400 positive linear correlation between PTT and PEP has

401 been observed in 10 out of 11 subjects The

non-signif-402 icant correlation observed in one subject was partly

403 attributed to the lack of PEP response to tilting as

indi-404 cated by the poor correlation between PEP and sinh, but

405 generally, the correlation coefficient between PTT and

406 PEP was high (median r = 0.964), justifying the potential

407 use of PTT to monitor PEP variations during mild change

408 in volume status Progressive prolongation of PEP/PTT

409 during graded head-up tilt is believed to reflect a decline

410 in stroke volume caused by the reduction of cardiac

411 preload (or end diastolic volume) as a result of orthostatic

412 volume shift from the central venous pool to the lower

413 body [11] During head-up tilt, the hydrostatic effect of

414 tilting is proportional to sinh which reflects the body axis

415 component of gravitational pull exerted on the blood

416 volume inside the body [20, 21] As demonstrated by the

417 present study, PEP and PTT had strong positive

correla-418 tion (r > 0.8) with sinh in most subjects and the overall

419 regression slopes were significantly positive These

find-420 ings are consistent with the observed linear relationship

421 between sinh and the decrease in thoracic fluid content

422 during graded head-up tilt [22] and suggest that PEP and

423 PTT may reflect proportional change in central blood

424 volume or preload

425 A new way of studying the haemodynamic effect of

426 progressive hypovolaemia using gradient/trend analysis of

427 PTT and RR has been presented in this study that may

428 permit better characterization of the different stages of

429 blood loss From the perspective of clinical application,

430 we propose that the changing trends in PTT and RR may

431 be more useful than their absolute values for identifying

432 and distinguishing between different phases of progressive

433 blood loss It is well recognized that trends in

physio-434 logical variables, both with evolving pathology and with

435 resuscitative measures, are very useful diagnostically and

436 prognostically in acute illness Although PTT may be

437 augmented in hypovolaemia compared with

normovola-438 emia, in a real life situation critical care clinicians often

439 need to diagnose blood loss without prior knowledge of

440 the patientsÕ pre-haemorrhage physiological variables

441 It seems not possible to use absolute values of PTT to

442 identify patients with low central blood volume because

443 of the high degree of inter-subject variability, as

444 demonstrated by the considerable overlap in the ranges of

445 PTT comparing the normovolaemic state (0°) with the

446 most hypovolaemic state (80°) and the high inter-subject

447

CV (11%) relative to the percentage difference (13%)

448 between the two states Alternatively, the trend or

gra-449 dient of PTT may be useful for identifying patients who

450 are progressively losing blood, since dynamic volume

451 decrease may result in a rising trend in PTT over time In

452 the current study, three different stages of physiological

453 response to central volume loss have been identified:

454 Stage 1 (0–20°): This stage simulated mild central

455 hypovolaemia A rising trend was observed in PTT/PEP

456

as preload decreased No significant falling trend was

457 observed in RR, probably because small decrement in

458 central blood volume at a mild hypovolaemic state was

459 not sufficient to trigger noticeable baroreflex response

460 Stage 2 (20–50°): This stage simulated

mild-to-mod-461 erate central hypovolaemia PTT/PEP continued to show

462

a rising trend as preload decreased A falling trend was also

463 observed in RR, which could be attributed mostly to

464 vagal withdrawal and also to sympathetic activation The

465 more negative RR gradient in this stage compared with

466 stage 1 might result from augmented baroreflex

respon-467 siveness as central blood volume decreased [30]

468 Stage 3 (50–80°): This stage simulated moderate central

469 hypovolaemia A significant rising trend was not observed

470

in PTT even though PEP was increasing with preload

471 reduction, and this suggested that the decline in VTT

472 might have offset the rise in PEP The decline in VTT

473 indicated an increase in pulse wave velocity (PWV)

in-474 duced by sympathetic activation, which might result from

475

a rise in MAP/DBP causing a passive increase in arterial

476 stiffness or from an increase in myocardial contractility

477 [31, 32] PEP, despite continuing to increase, showed a

478 weaker rising trend in this stage, possibly because the

479 lengthening effect of preload reduction was opposed

480

by the shortening effect of sympathetic activation [12, 13]

481

A falling trend in RR continued in this stage as a result of

482 the combined influence of vagal withdrawal and

sympa-483 thetic activation with further reduction in central blood

484 volume

485 Based on the observed physiological response to the three

486 stages of simulated hypovolaemia, it is clear that rising trend

487

in PTT can be a useful marker for progressive volume loss in

488 stages 1 and 2 (mild and mild-to-moderate hypovolaemia),

489 but not when the patient has entered stage 3 (moderate

490 hypovolaemia) In stage 3, sympathetic activation is

be-491 lieved to cause variation in VTT which reduces the ability

492

of PTT to follow changes in PEP The shortcoming of PTT

493 can be mitigated, however, by also considering RR, which

494 tends to fall sharply as a result of an enhanced sympathetic

495 tone The present study has demonstrated that the joint

496 interpretation of PTT and RR trends may offer promising

PROOF

497 possibility of not only detecting the presence, but also

498 estimating the extent of progressive blood loss For

exam-499 ple, a change in the patientÕs status from rising PTT and

500 unchanged RR to unchanged/falling PTT and falling RR

501 may indicate the transition from mild hypovolaemia to

502 moderate hypovolaemia

504 Although an increase in afterload (represented by MAP/

505 DBP) due to peripheral vasoconstriction may also lead to

506 an increase in PEP [33], it was unlikely to be the major

507 cause of the tilt-induced change in PEP and PTT because

508 MAP/DBP did not have a significant linear relationship

509 with sinh The lack of concomitant BP change with mild

510 decrease in central blood volume induced by head-up tilt

511 has been reported elsewhere [34–37], and this finding

512 supports the concept of insensitivity of BP to small

vol-513 ume loss [1] It is known that in phase I of haemorrhage

514 (loss of up to 750 ml or 15% of total blood volume),

515 sympathetic activation would help to maintain a stable BP

516 despite a drop in stroke volume, and only until blood loss

517 reaches a critical level (30–40% of total blood volume), a

518 decompensatory phase II commences during which BP

519 and HR fall dramatically [3–5] In contrast to previous

520 studies which utilized PTT as a surrogate marker of BP

521 change for detecting hypovolaemia-induced hypotension

522 [16, 17], the present study demonstrates that PTT may in

523 fact be a more robust indicator of mild volume loss than

524 BP itself and may signal an early stage of hypovolaemia

525 well before the phase II hypotension occurs The

theo-526 retical basis for the use of PTT as a surrogate marker of BP

527 was initially suggested to be the potential negative

cor-528 relation between VTT/PWV and BP [38] but the

rela-529 tionship between PTT and BP can be significantly

530 influenced by the variation in PEP which may oppose the

531 change in VTT [17, 29, 39] The results of this study have

532 provided further evidence that PTT and PEP can change

533 in a disassociated way from VTT and BP during mild to

534 moderate central hypovolaemia

536 In this study, progressive central hypovolaemia was

in-537 duced in healthy awake subjects by incremental head-up

538 tilt from 0 to 80° The use of head-up tilt as a model to

539 simulate the major haemodynamic response to

haemor-540 rhage in humans has been documented elsewhere [3–5,

541 18, 19] Although tilt-induced central hypovolaemia is

542 not identical to actual blood loss since the blood volume is

543 merely re-distributed to the lower body rather than

544 actually lost from the circulatory system, the initial

car-545 diovascular response to haemorrhage is essentially the

546 same as that elicited by a reduction in central blood

vol-547 ume, e.g by head-up tilt or by lower body negative

548 pressure (LBNP) [4–6, 19] About 24° head-up tilt

pro-549 duces a similar cardiovascular response to 15 mmHg

550 LBNP [36], which approximates mild haemorrhage (loss

551

of 400–550 ml or 10% of total blood volume) [6], while

552 60° head-up tilt produces a similar central cardiovascular

553 response to 20–40 mmHg LBNP [37], which

approxi-554 mates moderate haemorrhage (loss of 550–1000 ml or

555

10–20% of total blood volume) [6] However, a

limi-556 tation of using head-up tilt as a model of blood loss is that

557 the regional blood volume changes and the associated

558 vascular responses induced by gravitational fluid shift to

559 the lower body can be different from that in actual

560 haemorrhage [37, 40] Nonetheless, head-up tilt may still

561

be regarded as an acceptable model to simulate most of the

562 cardiovascular effect of falling central blood volume that

563 occurs in blood loss

564

Comparison with actual haemorrhage in anaesthetized dogs

565 Since the present findings are based on a simulated model of

566 haemorrhage, whether the results are applicable to an actual

567 blood loss situation remains to be investigated Kubitz et al

568 studied variation in PEP and cardiac preload during acute

569 haemorrhage in pigs, but concluded that PEP was not

570 sensitive to the change in intravascular volume status [41]

571 Ochiai et al showed that acute blood loss led to significant

572 prolongation in VTT and PTT, yet with no significant

573 change in PEP [17] We suggest that one reason for the lack

574

of PEP change in these haemorrhage studies may be due to

575 the magnitude of blood loss being severe given the presence

576

of hypotension It was noted in the present study that PEP

577 showed a weaker rising trend in stage 3 (50)80°) compared

578 with stage 1 and 2 (0–50°), which suggests that PEP may be

579 less sensitive to volume change as the degree of

hypovola-580 emia becomes more severe, most likely due to the opposite

581 effect of preload reduction and sympathetic activation

582 Another possible reason for the difference may be the effect

583

of experimental procedure on the physiologic response of

584 haemorrhaged animals, including the method of blood

585 withdrawal and the induction of anaesthesia which may

586 have confounding effects on the cardiovascular response to

587 volume loss [42, 43] For example, the use of isoflurane in

588 the study by Ochiai et al could lead to vasodilatation and

589 might subsequently influence the PEP/PTT response to

590 haemorrhage

591

Technical aspects of PTT/PEP derivation

592 The ability of PTT to monitor PEP variation may depend

593

on which part of the peripheral pulse waveform is used as a

594 reference point for PTT measurement In the present study,

PROOF

595 the first derivative of the finger PPG-POW

(d1PPG-596 POW) was used as the reference pulse waveform since it is

597 considered to be closely related to peripheral arterial flow

598 [25] The foot of d1PPG-POW was used as the reference

599 point for pulse arrival, in order to eliminate the potential

600 contribution of the rising time of systolic upstroke on PTT,

601 so that the PTT measurement would more closely reflect

602 the variation in PEP In fact, it is possible that sympathetic

603 activation during haemorrhage may induce a change in the

604 systolic rising time that opposes the prolongation of PEP

605 caused by preload reduction

606 For PEP measurements, the present study used thoracic

607 IPG to identify the onset of ventricular ejection The

608 thoracic dZ/dt pulse waveform has been regarded as a

609 measure of intrathoracic blood volume change and

610 experimental evidence tended to suggest a major role

611 played by systolic blood volume expansion in the

612 ascending aorta [23, 44], although the precise anatomic

613 site of its onset (B-point) remains speculative

Neverthe-614 less, it has been demonstrated that B-point occurred

615 synchronously with the first heart sound which marks the

616 onset of ventricular contraction [27] and the use of

B-617 point to estimate PEP has been validated by comparison

618 with the standard technique based on carotid pulse and

619 phonocardiogram [45] Comparing PEP measurements in

620 our study with Stafford et al [11], the differences between

621 the mean PEP at equivalent tilt angles were actually quite

622 small (the differences were 1 ms at 0°, 3 ms at 10°, )2 ms

623 at 20 and 30°, and 5 ms at 60°), despite the difference in

624 methodology used for PEP computation The close

625 agreement between the PEP measurements in our study

626 and in Stafford et al provides us with further reassurance

627 that the thoracic IPG-based technique is reliable

629 The ability of PTT to identity early stages of hypovolaemia

630 has a potentially enormous benefit to clinical practice, in

631 particular for those cases associated with covert

haemor-632 rhage into body cavities that are not easily recognizable at

633 the beginning Delayed control of abdominal, pelvic or

634 intrathoracic haemorrhage has been recognized as a major

635 contributor of preventable trauma deaths and is often caused

636 by delays in the assessment or diagnosis of haemorrhage [7,

637 8] Notably, it would be of great interest if such events could

638 be detected as early as possible based on information that

639 could be obtained from existing patient monitoring devices

640 Although PTT may not be as good as PEP for detecting

641 preload variation due to the confounding effect of VTT, it

642 can be easily computed from simultaneous measurements of

643 ECG and finger pulse oximetry, both of which have been

644 routine patient monitoring techniques for some years The

645 measurements of ECG and finger PPG-POW are totally

646 noninvasive, cause minimal discomfort to the patients, and

647 can be obtained continuously in a beat-to-beat manner

648 which may permit the early detection of small physiological

649 perturbations It would certainly be advantageous to critical

650 care clinicians if these two routine measurements can

pro-651 vide information relevant to the diagnosis of blood loss in

652 addition to their conventional use for HR and SpO2

653 monitoring

654 Apart from detecting progressive blood loss, the

re-655 sponse pattern of PTT/PEP to graded tilt at different

656 volume status may also have direct relevance towards the

657 use of tilt test in the clinical assessment of hypovolaemia

658 and fluid responsiveness Due to a lack of response to

659 volume challenge in some patients who are suspected to

660

be hypovolaemic, the test for fluid responsiveness has

661 often been considered an important initial therapeutic

662 question [46] In clinical practice, one method to test for

663 fluid responsiveness is to measure the haemodynamic

664 change by first tilting the patient to the reverse

Trendel-665 enburg position (30° head-up tilt) to induce relative

666 depletion of central blood volume then to the

Trendel-667 enburg position (30° head-down tilt) to simulate volume

668 expansion [47] Ideally, the change in stroke volume or

669 cardiac output during the manoeuvre would define fluid

670 responsiveness, but in cases where stroke volume or

car-671 diac output measurements are not available or not

pre-672 ferred due to their invasive nature, non-invasive indices

673 such as PTT/PEP may be useful alternatives Previous

674 studies have demonstrated the potential value of

respira-675 tory fluctuation in PTT/PEP in predicting fluid

respon-676 siveness [14, 15] The current study has further

677 demonstrated the possibility of using PTT/PEP for

678 assessing fluid responsiveness by studying their dynamic

679 change during tilt manoeuvres, although more clinical

680 studies are required to validate this

681 For the PTT technique to be applied clinically, several

682 issues need to be addressed in future investigations; firstly,

683

it is unclear what the optimal duration is for the reliable

684 detection of an increasing/decreasing trend of PTT/RR

685 associated with blood loss Certainly, the analysis period

686 has to be sufficiently long since both PTT and RR exhibit

687 respiratory fluctuations as well as other spontaneous low

688 frequency oscillations [14, 39, 48] that may confound the

689 genuine trend related to physiological perturbations

690 Secondly, there is a need to identify patient groups whose

691 PEP/PTT may have limited responsiveness to a change in

692 preload, such as those who suffer from heart failure [11]

693 Thirdly, the PTT measurement may be influenced by the

694 contact force with the sensor [49], the peripheral

tem-695 perature [50] and the limb position [51] Whether these

696 factors would affect the applicability of PTT in the

697 monitoring of critical care patients remains to be

inves-698 tigated

PROOF

700 In conclusion, this study has shown that PTT may reflect

701 variation in PEP and is potentially useful for early detection

702 of non-hypotensive progressive central hypovolaemia

703 Joint interpretation of PTT and RR trends or responses

704 may help to characterize the extent of blood volume loss

705 Further work is required to evaluate the applicability of

706 PTT in the examination of critical care patients who may be

707 suffering from haemorrhage

708 We would like to thank Dr Ross Odell for his valuable advice on

709 data analysis.

710

711 REFERENCES

712 1 American College of Surgeons Shock In: ATLS Instructors

713 Manual Chicago: First Impressions; 1993 p 75–94.

714 2 McGee S, Abernethy WB III, Simel DL The rational clinical

715 examination Is this patient hypovolemic? JAMA 1999;

716 281(11): 1022–9.

717 3 Evans RG, Ventura S, Dampney RA, Ludbrook J Neural

718 mechanisms in the cardiovascular responses to acute central

719 hypovolaemia Clin Exp Pharmacol Physiol 2001; 28(5–6):

721 4 Hainsworth R, Drinkhill MJ Regulation of blood volume In:

722 Jordan D, Marshall J, eds Cardiovascular regulation London:

723 Portland, 1995: p 77–91.

724 5 Secher NH, Van Lieshout JJ Normovolaemia defined by

725 central blood volume and venous oxygen saturation Clin Exp

726 Pharmacol Physiol 2005; 32(11): 901–10.

727 6 Cooke WH, Ryan KL, Convertino VA Lower body negative

728 pressure as a model to study progression to acute hemorrhagic

729 shock in humans J Appl Physiol 2004; 96(4): 1249–61.

730 7 Gruen RL, Jurkovich GJ, McIntyre LK, Foy HM, Maier RV.

731 Patterns of errors contributing to trauma mortality: lessons

732 learned from 2,594 deaths Ann Surg 2006; 244(3): 371–80.

733 8 Anderson ID, Woodford M, de Dombal FT, Irving M.

734 Retrospective study of 1000 deaths from injury in England

735 and Wales Br Med J (Clin Res Ed) 1988; 296(6632):

737 9 Foo JY, Lim CS Pulse transit time as an indirect marker for

738 variations in cardiovascular related reactivity Technol Health

739 Care 2006; 14(2): 97–108.

740 10 Naschitz JE, Bezobchuk S, Mussafia-Priselac R, Sundick S,

741 Dreyfuss D, Khorshidi I, Karidis A, Manor H, Nagar M, Peck

742 ER, et al Pulse transit time by R-wave-gated infrared

pho-743 toplethysmography: review of the literature and personal

744 experience J Clin Monit Comput 2004; 18(5–6): 333–42.

745 11 Stafford RW, Harris WS, Weissler AM Left ventricular

sys-746 tolic time intervals as indices of postural circulatory stress in

747 man Circulation 1970; 41(3): 485–92.

748 12 Boudoulas H Systolic time intervals Eur Heart J 1990;

749 11(Suppl I): 93–104.

750

13 Lewis RP, Rittogers SE, Froester WF, Boudoulas H A critical

751

review of the systolic time intervals Circulation 1977; 56(2):

752

146–58.

753

14 Bendjelid K, Suter PM, Romand JA The respiratory change in

754

preejection period: a new method to predict fluid

responsive-755

ness J Appl Physiol 2004; 96(1): 337–42.

756

15 Feissel M, Badie J, Merlani PG, Faller JP, Bendjelid K

Pre-757

ejection period variations predict the fluid responsiveness of

758

septic ventilated patients Crit Care Med 2005; 33(11): 2534–9.

759

16 Ahlstrom C, Johansson A, Uhlin F, Lanne T, Ask P

Nonin-760

vasive investigation of blood pressure changes using the pulse

761

wave transit time: a novel approach in the monitoring of

he-762

modialysis patients J Artif Organs 2005; 8(3): 192–7.

763

17 Ochiai R, Takeda J, Hosaka H, Sugo Y, Tanaka R, Soma T.

764

The relationship between modified pulse wave transit time and

765

cardiovascular changes in isoflurane anesthetized dogs J Clin

766

Monit Comput 1999; 15(7–8): 493–501.

767

18 Matzen S, Perko G, Groth S, Friedman DB, Secher NH Blood

768

volume distribution during head-up tilt induced central

hyp-769

ovolaemia in man Clin Physiol 1991; 11(5): 411–22.

770

19 Pawelczyk JA, Matzen S, Friedman DB, Secher NH

Cardio-771

vascular and hormonal responses to central hypovolaemia in

772

humans In: Secher, NH, Pawelczyk, JA and Ludbrook, J, eds,

773

Blood loss and shock Edward Arnold, London, 1994: pp 25–

774

36.

775

20 Blomqvist CG, Stone HL Cardiovascular adjustments to

776

gravitational stress In: Shepherd JT, Abboud FM, eds

Hand-777

book of physiology, The cardiovascular system, Peripheral

778

circulation and organ blood flow Sect 2, vol 3 Bethesda:

779

American Physiological Society, 1983: p 1025–63.

780

21 Iwase S, Mano T, Saito M Effects of graded head-up tilting on

781

muscle sympathetic activities in man Physiologist 1987; 30(1

782

Suppl): S62–3.

783

22 Laszlo Z, Rossler A, Hinghofer-Szalkay HG Cardiovascular

784

and hormonal changes with different angles of head-up tilt in

785

men Physiol Res 2001; 50(1): 71–82.

786

23 Patterson RP, Wang L, Raza B, Wood K Mapping the

car-787

diogenic impedance signal on the thoracic surface Med Biol

788

Eng Comput 1990; 28(3): 212–6.

789

24 Toska K, Walloe L Dynamic time course of hemodynamic

790

responses after passive head-up tilt and tilt back to supine

po-791

sition J Appl Physiol 2002; 92(4): 1671–6.

792

25 Cook LB Extracting arterial flow waveforms from pulse

oxi-793

meter waveforms apparatus Anaesthesia 2001; 56(6): 551–5.

794

26 Chiu YC, Arand PW, Shroff SG, Feldman T, Carroll JD.

795

Determination of pulse wave velocities with computerized

796

algorithms Am Heart J 1991; 121(5): 1460–70.

797

27 Lababidi Z, Ehmke DA, Durnin RE, Leaverton PE, Lauer

798

RM The first derivative thoracic impedance cardiogram.

799

Circulation 1970; 41(4): 651–8.

800

28 DeMarzo AP, Lang RM A new algorithm for improved

801

detection of aortic valve opening by impedance cardiography.

802

Comput Cardiol 1996;373–76.

803

29 Newlin DB Relationships of pulse transmission times to

pre-804

ejection period and blood pressure Psychophysiology 1981;

805

18(3): 316–21.

806

30 Pawelczyk JA, Raven PB Reductions in central venous

807

pressure improve carotid baroreflex responses in conscious

808

men Am J Physiol 1989; 257(5 Pt 2): H1389–95.