Pulse Oximeter Plethysmographic Waveform Changes
in Awake, Spontaneously Breathing,
Hypovolemic Volunteers
Susan P. McGrath, PhD,* Kathy L. Ryan, PhD,† Suzanne M. Wendelken, MS,*
Caroline A. Rickards, PhD,† and Victor A. Convertino, PhD†
BACKGROUND: The primary objective of this study was to determine whether alterations in the
pulse oximeter waveform characteristics would track progressive reductions in central blood
volume. We also assessed whether changes in the pulse oximeter waveform provide an
indication of blood loss in the hemorrhaging patient before changes in standard vital signs.
METHODS: Pulse oximeter data from finger, forehead, and ear pulse oximeter sensors were
collected from 18 healthy subjects undergoing progressive reduction in central blood volume
induced by lower body negative pressure (LBNP). Stroke volume measurements were simulta-
neously recorded using impedance cardiography. The study was conducted in a research
laboratory setting where no interventions were performed. Pulse amplitude, width, and area
under the curve (AUC) features were calculated from each pulse wave recording. Amalgamated
correlation coefficients were calculated to determine the relationship between the changes in
pulse oximeter waveform features and changes in stroke volume with LBNP.
RESULTS: For pulse oximeter sensors on the ear and forehead, reductions in pulse amplitude,
width, and area were strongly correlated with progressive reductions in stroke volume during
LBNP (R2 � 0.59 for all features). Changes in pulse oximeter waveform features were observed
before profound decreases in arterial blood pressure. The best correlations between pulse
features and stroke volume were obtained from the forehead sensor area (R2 � 0.97). Pulse
oximeter waveform features returned to baseline levels when central blood volume was restored.
CONCLUSIONS: These results support the use of pulse oximeter waveform analysis as a
potential diagnostic tool to detect clinically significant hypovolemia before the onset of
cardiovascular decompensation in spontaneously breathing patients. (Anesth Analg 2010;X:
●●●–●●●)
Hemorrhage is a leading cause of death in bothcivilian and military trauma.1,2 Consequently, it isof paramount importance that the hypovolemia
associated with hemorrhage and its severity be detected
early to enable treatment to begin as soon as possible in
both prehospital and hospital settings; this is underscored
by reports that early intervention before the development
of hemorrhagic shock is associated with a lower mortality
rate.3 Unfortunately, compensatory mechanisms that act to
defend arterial blood pressure during hemorrhage can
mask the progressive loss of blood until seconds before
cardiovascular collapse when rapid deterioration of
blood pressure and loss of consciousness occur.4 Thus,
the use of blood pressure as an indicator of blood loss
can often prove to be too late and does not allow for early
intervention.
Traditionally, pulse oximeters have only been utilized
for the indirect assessment of arterial oxygen saturation
(Spo2) and heart rate in the clinical setting. However,
previous studies have shown that the photoplethysmogram
(PPG), frequently obtained from pulse oximeters, may
provide a reliable indication of hemodynamic changes,
including hypovolemia.5–9 For the most part, these inves-
tigations involved analysis of respiration-induced oscilla-
tions of the PPG in either anesthetized surgical patients
undergoing positive pressure ventilation or healthy pa-
tients who underwent only moderate blood loss (e.g., 450
mL). The question of whether features of the PPG may
provide diagnostic insight into the progression of central
hypovolemia in spontaneously breathing trauma patients
during the prehospital emergency phase of care has not
been studied. To address this gap, an assessment is needed
in awake, nonventilated patients to identify the relation-
ship, if any, between changes in the PPG and reductions in
central blood volume that mimic hemorrhage. In this study,
we had the unique opportunity to investigate the pulse
shape features of the PPG obtained from finger, forehead,
and ear sensors during an experimental model of central
hypovolemia and to determine which features best serve as
early indicators of reduction in central blood volume and
impending cardiovascular collapse. The hemodynamic
challenge was created using lower body negative pressure
(LBNP), which safely and progressively reduces central
blood volume by sequestering blood in the lower extremi-
ties through application of negative pressure around the
legs and abdomen. LBNP has been shown to be an effective
model of the early phases of hemorrhage in conscious
humans4,10 and allowed us to test the hypothesis that
reduced pulse amplitude, width, and/or area under the
From the *Thayer School of Engineering, Dartmouth College, Hanover, New
Hampshire; and †US Army Institute of Surgical Research, San Antonio,
Texas.
Accepted for publication November 11, 2009.
Study funding: Funding information is provided at the end of the article.
Disclosure: The authors report no conflicts of interest.
Address correspondence and reprint requests to Susan McGrath, PhD,
Thayer School of Engineering, Hanover, NH 03755. Address e-mail to
susan.p.mcgrath@dartmouth.edu.
Copyright © 2010 International Anesthesia Research Society
DOI: 10.1213/ANE.0b013e3181cb3f4a
XXX 2010 • Volume X • Number XX www.anesthesia-analgesia.org 1
curve (AUC) calculated from recordings of the PPG pulse
wave would be associated with progressive reductions in
central blood volume and precede changes in standard vital
signs. Our approach is unique in that experiments were
conducted in awake, healthy subjects undergoing extreme
hemodynamic challenge in a controlled laboratory setting
and in the absence of the complicating effects of positive
pressure ventilation or drugs.
METHODS
Experimental procedures and protocols were reviewed and
approved by the IRB for the use of human subjects at the
Brooke Army Medical Center at Fort Sam Houston, Texas.
Each subject gave written informed voluntary consent to
participate in the experiments.
Study Design
LBNP was used in the present investigation as an experi-
mental tool to simulate loss of central blood volume (e.g.,
hemorrhage) in conscious, healthy humans.10 With the use of
a neoprene skirt designed to form an airtight seal between the
subject and the chamber, the application of negative pressure
to the lower body with the subject in a supine position
results in a redistribution of blood away from the upper
body (head and heart) to the lower extremities and abdo-
men. Thus, this model provides a unique method of
investigating conditions of controlled, progressive, experi-
mentally induced hypovolemic hypotension.
Each subject underwent exposure to an LBNP protocol
designed to test his or her tolerance to experimentally
induced hypotensive hypovolemia. The LBNP protocol con-
sisted of a 5-minute baseline period (0 mm Hg) followed by
5 minutes of chamber decompression to �15, �30, �45,
and �60 mm Hg and additional increments of �10 mm Hg
every 5 minutes until the onset of cardiovascular collapse
or the completion of 5 minutes at �100 mm Hg. Cardio-
vascular collapse was identified in real time by the attend-
ing investigator by a precipitous decrease in systolic blood
pressure (SBP) �15 mm Hg concurrent with the onset of
presyncopal symptoms such as bradycardia, gray out
(loss of color vision), tunnel vision, sweating, nausea, or
dizziness.
Selection of Participants
Eighteen healthy nonsmoking subjects (8 men, 10 women)
with mean � sd age of 23 � 4 years, body weight of 69 �
11 kg, and height of 171 � 10 cm were recruited to
participate in the study. Female subjects underwent an
initial urine test before experimentation to ensure that they
were not pregnant. Subjects maintained their normal sleep
and dietary patterns, refrained from exercise, and abstained
from caffeine and other autonomic stimulants such as
prescription or nonprescription drugs for at least 24 hours
before each experimental protocol. All subjects received a
verbal briefing and a written description of all procedures
and risks associated with the experiments and were famil-
iarized with the laboratory, the protocol, and procedures.
Methods of Measurement
All subjects were instrumented with 3 Nonin� pulse oximeter
sensors (Nonin Medical, Plymouth, MN; OEM III module,
16-bit data format) placed on the forehead, finger, and ear.
The sensor on the forehead was securely taped to the skin,
whereas the finger sensor was placed on the ring finger of the
left hand. Unlike standard pulse oximeters that have
autocalibration capability, the Nonin pulse oximeter did
not alter the raw waveform signal with an autogain mecha-
nism. In addition, an infrared finger PPG blood pressure
monitor (Finometer� Blood Pressure Monitor, TNO-TPD
Biomedical Instrumentation, Amsterdam, The Nether-
lands) and an electrocardiogram were used to record
beat-by-beat arterial pressures and heart rate. The Finom-
eter blood pressure cuff was placed on the middle finger of
the left hand, which in turn was held at heart level.
Excellent estimates of directly measured intraarterial blood
pressures during various physiologic maneuvers have been
demonstrated with this device.11–14 Beat-to-beat stroke vol-
ume was measured noninvasively using thoracic electrical
bioimpedance with an HIC-2000 Bio-Electric Impedance Car-
diograph (Bio-Impedance Technology, Chapel Hill, NC). The
thoracic electrical bioimpedance technique is based on the
resistance changes in the thorax to a low-intensity (4-mA),
high-frequency (70-kHz) alternating current applied to the
thorax by 2 surface electrodes placed at the root of the neck
and 2 surface electrodes placed at the xiphoid process at the
midaxillary line. Ventricular stroke volume was calculated
according to the technique described by Kubicek et al.15
Correlation coefficients of 0.70 to 0.93 have been reported in
stroke volume measurements simultaneously made with
thoracic electrical bioimpedance and thermodilution tech-
niques.16 Despite this close association, stroke volume
measurements made with noninvasive techniques are esti-
mations of the actual stroke volume and are therefore
presented as percentage change from baseline levels.
The PPG and Spo2 values from the Nonin pulse oxime-
ter sensors were continuously and simultaneously recorded
to a laptop PC using Java-based data collection software at
a sample frequency of 75 Hz.
Analysis
Subjects reached cardiovascular collapse (i.e., maximal
LBNP tolerance) at different absolute LBNP levels based on
their individual physiologic responses. Because these re-
sponses are the same at cardiovascular collapse indepen-
dent of the LBNP level at which an individual subject
reaches this point,17 we chose to normalize each individu-
al’s data by reapportioning their responses to equal frac-
tions between 0% LBNP tolerance (baseline) and 100%
LBNP tolerance, the level at which the LBNP protocol was
terminated as a result of impending cardiovascular collapse
(presyncope). This approach allowed us to consider the
data from all subjects relative to their maximal capacity for
LBNP tolerance.18–20
Pulse amplitude, width, and AUC for each cardiac cycle
were automatically extracted from the PPG data using a
feature extractor implemented in Matlab� (The Math-
Works, Natick, MA). Figure 1 depicts these PPG features
extracted by the analysis algorithms. Pulse amplitude was
calculated as the difference between the PPG values of the
cardiac peak and the preceding valley. Pulse width was
calculated as the time width at half the cardiac peak
amplitude. The AUC was calculated as a standard line
Photoplethysmogram Analysis for Tracking Hypovolemia
2 www.anesthesia-analgesia.org ANESTHESIA & ANALGESIA
integral minus the DC area. The amplitude, width, and area
calculated from the PPG were normalized with respect to
their baseline values to study the trends while accounting
for individual physiologic variability. Baseline values were
calculated by averaging each feature over the last 3 minutes
of the pre-LBNP baseline period. Similarly, values for all
variables were averaged over the last 3 minutes of each
LBNP level and recovery and for 1 minute before presyn-
cope. A 1-way (percentage LBNP level) randomized block
(subjects) analysis of variance for repeated measures was
used for comparison of outcome variables. If statistical
differences were found, Bonferroni-corrected comparisons
with baseline measurements were performed to determine
the first percentage level of LBNP that could be distin-
guished statistically from baseline. Amalgamated correla-
tion coefficients (i.e., coefficients generated by correlations
between mean values of stroke volume changes and PPG
pulse shape features at each LBNP tolerance level) were
calculated to determine the relationship between the
changes in PPG features and changes in stroke volume
during LBNP and subsequent recovery. All data are pre-
sented as mean � se, and exact P values are presented for
all comparisons.
RESULTS
A representative tracing of the beat-to-beat pulse oximeter
PPG waveform characteristics recorded from the ear sensor
of 1 subject before, during, and after the progressive LBNP
protocol is presented in Figure 2. PPG pulse amplitude,
width, and area decreased with increasing LBNP in all
pulse oximeter sensor sites. Immediately after the termina-
tion of LBNP with accompanying cardiovascular collapse,
most subjects demonstrated a sudden rebound (increase) in
pulse amplitude and area to levels greater than baseline,
whereas the recovery of pulse width was more gradual.
Figure 3 illustrates group means for PPG pulse ampli-
tude width and area, as well as alterations in hemodynamic
variables during LBNP. LBNP induced a progressive re-
duction in central blood volume, indicated by changes in
stroke volume (Fig. 3C) and a reduction in SBP at 60% of
LBNP tolerance (Fig. 3A). Spo2 did not change from
baseline values (99.2% � 0.2%) during the LBNP exposure
(99.3% � 0.3% at 100% of LBNP tolerance). LBNP caused a
reduction in PPG pulse wave amplitude until the final
point (100% of LBNP tolerance), at which it increased
slightly (Fig. 3D). PPG pulse width did not change from
baseline at early stages of LBNP but decreased at 80% and
100% of LBNP tolerance (Fig. 3E). These changes in pulse
amplitude and width resulted in a progressive decrease in
PPG area that became statistically significant (P � 0.001) at
60% of LBNP tolerance (Fig. 3F). Upon cessation of LBNP,
all hemodynamic and PPG pulse features returned to
pre-LBNP baseline values (Fig. 3).
Table 1 presents amalgamated correlation coefficients
between the changes in stroke volume and changes in PPG
waveform characteristics. We found correlation coefficients
(R2) �0.59 between stroke volume and PPG pulse ampli-
tude, pulse width, and area during LBNP and recovery in
the ear and forehead sensors, with extremely high correla-
tion coefficients (0.91 from ear sensor and 0.97 from
forehead sensor) between stroke volume and AUC. Corre-
lations with stroke volume were not as strong for pulse
amplitude and AUC in the finger sensor.
DISCUSSION
Strong correlations between the reduction in stroke volume
and pulse amplitude, width, and area during LBNP sup-
port our hypothesis that PPG waveform characteristics
obtained from a standard pulse oximeter may prove to be
sensitive and specific as early indicators of blood loss in
Figure 1. Illustration of pulse amplitude, pulse width, and area under the curve (AUC) extracted from the photoplethysmogram (PPG) waveform.
XXX 2010 • Volume X • Number XX www.anesthesia-analgesia.org 3
trauma patients. Our data are consistent with previous
studies that demonstrated that PPG pulse shape features
change significantly and predictably during hypovolemia in
anesthetized surgical patients.5–8 To our knowledge, how-
ever, this investigation is the first to demonstrate, in con-
scious, spontaneously breathing humans, the ability of PPG
pulse wave features to provide early and continuous indi-
cations of central hypovolemia during preshock hemor-
rhage when standard vital signs such as arterial blood
pressure and Spo2 remain within clinically normal levels
(Fig. 3).
Hemorrhagic shock remains the leading cause of death
on the battlefield. Data from current military conflicts
demonstrate that the mortality rate of combat casualties is
dramatically increased when treatment commences after,
rather than before, the onset of circulatory shock.3 It is clear
from these data that early recognition of progressive blood
loss and the need for immediate intervention before the
development of shock could prove critical in the attempt to
save lives. In this regard, the measurement of physiologic
responses that are altered before changes in standard vital
signs such as SBP could improve survival from hemor-
rhagic trauma. Results from the present investigation
support the notion that the measurement of PPG pulse
characteristics can represent a significant adjunct to current
monitoring capabilities by providing information about the
degree of blood volume reduction before any profound
change in SBP. Although SBP changes statistically at 60%
LBNP, it is still well above any clinical alert threshold (e.g.,
110 mmHg21) until 100% LBNP. The decrease in SBP at this
point indicates that, by Advanced Trauma Life Support
definition, subjects were progressing into class III of shock,
which is associated with loss of effective blood volume
between 30% and 40%.22 Because pulse oximetry is ubiq-
uitously used for monitoring patient status, it is likely that
embedding new algorithms for measurement of PPG pulse
characteristics into current oximeters would be logistically
simple and readily accepted by the medical community.
The correlations between the PPG pulse features and
stroke volume measurements were typically higher when
measured from the ear and forehead sensors than the finger
sensor, particularly for pulse amplitude and, consequently,
pulse area. Progressive reduction in central blood volume is
associated with both increased sympathetic nerve activity and
concomitant peripheral vasoconstriction.10 The amplitude of
finger PPG is reduced by increased sympathetic activation8,23
and vasoconstriction.24 In addition, blood flow to the
forehead seems to be less influenced by sympathetically
mediated reflex peripheral vasoconstriction,8 suggesting
that PPG waveforms measured at the forehead may there-
fore more closely reflect alterations in central hemodynam-
ics associated with blood loss (e.g., stroke volume) than
those in the finger. Our results support this postulate and
indicate that the forehead region may be a better location
than a peripheral digit for predicting early blood volume loss
using the PPG.8 Operationally, the use of a head sensor site is
also preferred because the finger sensor is more prone to
Figure 2. Representative tracings of the ear photoplethysmogram (PPG) waveform, pulse amplitude, pulse width, and area from a single
subject. Vertical lines denote pressure changes of the lower body negative pressure (LBNP) device. The final vertical line denotes the onset
of presyncope.
Photoplethysmogram Analysis for Tracking Hypovolemia
4 www.anesthesia-analgesia.org ANESTHESIA & ANALGESIA
motion artifact25 and extremity injury is very common, par-
ticularly on the battlefield,26 which would potentially elimi-
nate any possibility of using the finger site.
We observed an immediate restoration of PPG pulse wave
amplitude, width, and area in all subjects upon termination of
LBNP (Figs. 2 and 3). Recovery of the PPG pulse wave is not
unexpected because the immediate restoration of central
blood volume with the cessation of LBNP is related to the
complete recovery of stroke volume, sympathetic nerve activ-
ity, and peripheral blo