Journal of Clinical Monitoring and Computing (2019) 33:815–824

https://doi.org/10.1007/s10877-018-0235-z

ORIGINAL RESEARCH

Photoplethysmographic characterization of vascular tone mediated

changes in arterial pressure: an observational study
Gerardo Tusman1 · Cecilia M. Acosta1 · Sven Pulletz2 · Stephan H. Böhm3 · Adriana Scandurra4 ·
Jorge Martinez Arca4 · Matías Madorno5 · Fernando Suarez Sipmann6,7,8

Received: 28 July 2018 / Accepted: 11 December 2018 / Published online: 15 December 2018
© Springer Nature B.V. 2018

Abstract
To determine whether a classification based on the contour of the photoplethysmography signal (PPGc) can detect changes
in systolic arterial blood pressure (SAP) and vascular tone. Episodes of normotension (SAP 90–140 mmHg), hypertension
(SAP > 140 mmHg) and hypotension (SAP < 90 mmHg) were analyzed in 15 cardiac surgery patients. SAP and two surrogates
of the vascular tone, systemic vascular resistance (SVR) and vascular compliance (Cvasc = stroke volume/pulse pressure)
were compared with PPGc. Changes in PPG amplitude (foot-to-peak distance) and dicrotic notch position were used to define
6 classes taking class III as a normal vascular tone with a notch placed between 20 and 50% of the PPG amplitude. Class
I-to-II represented vasoconstriction with notch placed > 50% in a small PPG, while class IV-to-VI described vasodilation
with a notch placed < 20% in a tall PPG wave. 190 datasets were analyzed including 61 episodes of hypertension [SAP = 159
(151–170) mmHg (median 1st–3rd quartiles)], 84 of normotension, SAP = 124 (113–131) mmHg and 45 of hypotension
SAP = 85(80–87) mmHg. SAP were well correlated with SVR (r = 0.78, p < 0.0001) and Cvasc (r = 0.84, p < 0.0001). The
PPG-based classification correlated well with SAP (r = − 0.90, p < 0.0001), SVR (r = − 0.72, p < 0.0001) and Cvasc (r = 0.82,
p < 0.0001). The PPGc misclassified 7 out of the 190 episodes, presenting good accuracy (98.4% and 97.8%), sensitivity
(100% and 94.9%) and specificity (97.9% and 99.2%) for detecting episodes of hypotension and hypertension, respectively.
Changes in arterial pressure and vascular tone were closely related to the proposed classification based on PPG waveform.
Clinical Trial Registration NTC02854852.

Keywords Arterial pressure · Photoplethysmography · Vasodilation · Vasoconstriction · Vascular tone

* Gerardo Tusman Episodes of hypertension or hypotension during anesthesia

gtusman@hotmail.com are common and when important are associated to postoper-
ative complications and can thus have an impact on patient’s
1
Department of Anesthesiology, Hospital Privado de outcome [1–5]. A tight intraoperative control of arterial
Comunidad, 7600 Mar del Plata, Buenos Aires, Argentina
blood pressure is therefore mandatory for the conduction
2
Department of Anesthesiology and Intensive Care Medicine, of a safe anesthesia. The invasive monitoring of blood pres-
Klinikum Osnabrueck, Osnabrueck, Germany
sure through an indwelling arterial catheter is the reference
3
Department of Anesthesiology and Intensive Care Medicine, method in high-risk patients whereas automated cuff oscillo-
Rostock University Medical Center, Rostock, Germany
metric noninvasive blood pressure (NIBP) is the most widely
4
Bioengineering Laboratory, Electronic Department, School used method for monitoring normal or low-risk anesthetized
of Engineering, Mar del Plata University, Mar del Plata,
Argentina patients [6]. These monitoring systems do not provide infor-
5 mation regarding the mechanism/s behind hyper or hypoten-
Instituto Tecnológico Buenos Aires (ITBA), Buenos Aires,
Argentina sive episodes so that interpretation, and hence decisions on
6 potential therapeutic interventions, must be done according
Hedenstierna Laboratory, Department of Surgical Sciences,
Uppsala University, Uppsala, Sweden to the clinical context. As one of the principal mechanisms
7 of hyper and hypotension during anesthesia is related to
CIBERES, Madrid, Spain
changes in systemic arterial vascular resistance induced by
8
Department of Critical Care, Hospital Universitario de La
Princesa, Madrid, Spain

13
Vol.:(0123456789)
816 Journal of Clinical Monitoring and Computing (2019) 33:815–824

anesthetic drugs [7–9], the possibility of monitoring vascular the PPGc with invasive arterial blood pressure and two
tone would be of clinical interest. surrogates of vascular tone in anesthetized patients under-
The pulse oximetry photoplethysmographic (PPG) signal going cardiac surgery.
provides non-invasive beat-by-beat information related to
the characteristics of the vascular system [10]. The asso-
ciation between PPG, arterial blood pressure and systemic
vascular resistance (SVR) has been previously described 1 Methods
[11–13]. Such relationship is based on the fact that the PPG
waveform represents the change of blood volume in a tested This prospective observational pilot study was performed
tissue (commonly the finger) during one beat [14, 15]. This in the operating theater of a community hospital with the
pulse flow wave is strongly influenced by ventricular-vascu- corresponding local IRB approval (Clinical Trial Regis-
lar interactions in a similar way as the forward and backward tration NTC02854852). After obtaining signed informed
pulse pressure waves [16, 17]. As opposed to the positive consent, we studied mechanically ventilated patients
returning backward pressure wave, the backward flow wave aged ≥ 18 years undergoing cardiac surgery. We excluded
returns to the heart as an inverse negative wave after reflect- emergency surgery and patients with baseline arrhythmias.
ing back from the peripheral arterial tree bifurcations [18]. Patients who developed intraoperative hypothermia (naso-
This leads to marked changes in the contour of the PPG pharyngeal temperature ≤ 36 °C), arrhythmias and a total
waveform (PPGc) anytime the arterial compliance is affected bleeding ≥ 350 mL were also excluded to avoid any source
by modifications in the peripheral vascular tone. of changes in the PPG signal.
Based on the dynamic changes observed in the PPGc in Patients were anesthetized using standard proce-
response to alterations in vascular tone we recently proposed dures. Routine monitoring included ECG, time-based
a classification in which we identified six differentiated pat- capnography, pulse oximetry and naso-pharyngeal
terns (Fig. 1) [19]. This classification stems from close clini- temperature (S5, Datex-Ohmeda, Helsinki, Finland).
cal observations and is supported by some published evi- Anesthesia was induced with propofol 1–1.5 mg kg−1,
dences [20–23] but has not properly been tested or validated vecuronium 0.08 mg kg −1 and fentanyl 10 µ kg −1 and
in anesthetized mechanically ventilated patients before. maintained with propofol 80 µ kg −1 min−1 and remifen-
We believe that the changes observed in PPGc recorded tanil 0.5 µ kg−1 min−1. Lungs were ventilated in volume-
at the finger level are related to modifications on the arte- controlled ventilation using the Advance workstation (GE
rial blood pressure induced by changes in vascular tone. Healthcare, Madison, WI, US) with a tidal volume of
We used the previously described PPGc classification sys- 8 mL kg−1 of predicted body weight, respiratory rate 15
tem to test this hypothesis. For this purpose we compared breaths min−1, PEEP 5 cm ­H2O, I:E ratio 1:2 and F
­ IO2 0.5.

Fig. 1  Classification of the vascular tone based on the visual inspec- class IV the notch reaches baseline although the backward wave is
tion of the photo-plethysmography waveform shape. The classifica- still evident. Class V shows a flat phase without notch and in the class
tion is based on the photoplethysmography (PPG) amplitude and on VI the notch becomes negative. Vasoconstriction shows less PPG
the positioning of the dicrotic notch. Normal PPG shape (class III) amplitude than normal meaning that blood flow decreases (less infra-
presented the dicrotic notch between 20 and 50% of the total PPG red light absorbance). The notch ascends and fuses with the systolic
amplitude. Vasodilation increases PPG amplitude because the finger pulse peak (classes II and I, respectively)
receives more blood flow (more infrared light absorbance). In PPG
Journal of Clinical Monitoring and Computing (2019) 33:815–824 817

1.1 Hemodynamic monitoring at the index finger of the tested hand. Data was recorded
at 100 Hz via a laptop using a customized data collecting
A femoral artery 5F catheter (Pulsion Medical Systems, system. The PPG signal is the amount of light absorbed by
Munich, Germany) and an internal jugular 8F catheter were the finger. In order to improve the visualization and have a
inserted before anesthesia induction. Pressure transduc- comprehensive range, the dimensionless PPG signal is pre-
ers were placed and zeroed at the phlebostatic level before sented in an arbitrary scale 0–100% by the Fluxmed device.
data recording. Invasive systolic (SAP), mean (MAP), dias- Any improvement in blood flow at the finger increases PPG
tolic (DAP) and pulse (PP = SAP − DAP) arterial pressures amplitude because more light is absorbed by a crescent
were obtained beat-by-beat. Normotension was defined amount of blood. The opposite is true when finger’s blood
as a SAP between 90 and 140 mmHg, hypertension as a flow decreases, making PPG amplitude small.
SAP > 140 mmHg and hypotension as a SAP < 90 mmHg The contour of the PPG was classified according to six
[4, 24]. patterns related to different vascular tone conditions as pre-
Cardiac output (CO) was continuously monitored by viously described [19]. This classification is based on (1)
pulse wave contour analysis after calibration with 3 stable the position of the dicrotic notch and (2) the amplitude of
transpulmonary thermodilutions every 20 min (PICCO Sci- the PPG waveform defined as the foot-to-peak PPG distance
ence, Pulsion Medical Systems, Munich, Germany). Stroke [14, 16–18]. The classification is then described as follows
volume (SV) was calculated as CO/heart rate. Central venous (Fig. 1):
pressure (CVP) was continuously measured by the central
venous catheter connected to the PICCO monitor. We thus • Class I: decreased PPG amplitude and dicrotic notch
calculated SVR and arterial vascular compliance (Cvasc) fusion with the systolic peak. Meaning: decrease in blood
as surrogates of systemic vascular tone in a beat basis [25]. flow due to a significant increase in vascular tone second-
SVR was derived from the standard formula: ary to severe vasoconstriction.
• Class II: decreased PPG amplitude and notch position-
SVR = (MAP − CVP)∕CO × 80
ing at the upper 50% of the PPG waveform maximal
Cvasc was calculated as SV/PP as described previously amplitude. Meaning: decrease in local blood flow due to
[26]. a moderate increase in vascular tone.
Pulse pressure variation (PPV) was calculated as [27]. • Class III: the PPG amplitude of this class was arbitrarily
[( ) (
PPV(%) = 100 × PPmax − PPmin ∕ PPmax + PPmin ∕2
) ] taken as the reference amplitude because in our experi-
ence class III is related to normal blood pressure. The
were ­PPmax and P
­ Pmin are the maximal and minimal pulse dicrotic notch is commonly located between 20 and 50%
pressure values determined over a single respiratory cycle. of PPG’s maximum amplitude in this class
The stroke volume variation (SVV) was calculated as [27]: • Class IV: increased PPG amplitude and notch position-
ing at 20% of the PPG’s waveform maximal amplitude.
[( ) ( ) ]
SVV(%) = 100 × SVmax − SVmin ∕ SVmax + SVmin ∕2
Meaning: increase in blood flow due to a reduced vascu-
The global end-diastolic volume index (GEDI) is a vol-
lar tone caused by mild vasodilation.
umetric preload parameter related to the volume of blood
• Class V: increased PPG amplitude where notch and dias-
within the heart, derived from the transpulmonary thermodi-
tolic wave are flat, located at the foot of the PPG. Mean-
lution (TPTD) curve as [28]:
ing: increased blood flow due to moderate vasodilation.
GEDI = Cardiac index × (mean transit time − downslope time) • Class VI: increased PPG amplitude and the notch becom-
Extravascular lung water index (ELWI) was calculated ing negative. Meaning: increase in blood flow caused by
as [29]: severe vasodilation.

EVLWI = (ITTV − ITBV)∕predicted body weight

where the intrathoracic thermal volume (ITTV = CO × mean 1.3 Protocol and data manipulation
transit time × 1000/60) and intrathoracic blood volume
(ITBV = GEDV × 1.25). Data was continuously recorded 45 min before sternotomy
and within 1 h after closing the chest with patients placed in
supine position during the whole protocol. These moments
1.2 Photoplethysmography were chosen to avoid any interference from surgical manip-
ulations and to minimize the risk of surgical bleeding.
Core temperature was maintained ≥ 36 °C using a warming Patients received a bolus of 3 mL kg−1 of crystalloids dur-
mattress. PPG was obtained by a conventional pulse oxime- ing the induction of anesthesia and a maintenance infusion
ter (FluxMed, MBMED, Buenos Aires, Argentina) placed of 4 mL kg−1 ­h−1 for the remaining surgical time.

13
818 Journal of Clinical Monitoring and Computing (2019) 33:815–824

Intraoperative hemodynamic management was based

on standard care guidelines used at our institution. Arte-
rial blood hypotension related to a pulse pressure variation
(PPV) > 12% was treated by a 2 mL kg−1 bolus of a crystal-
loid solution [25]. Any bleeding < 350 mL was immediately
compensated by a bolus of crystalloid at a 1:3 ratio. Arterial
blood hypotension associated to vasodilation according to
low SVR (SVR values below 15% of the one observed dur-
ing normal arterial blood pressure) was treated by norepi-
nephrine (NE) infusion between 0.01 and 0.15 µ kg−1 ­min−1
until reach normal arterial blood values. Arterial blood
hypertension was treated increasing the doses of anesthetic
drugs (propofol bolus of 20 mg together an increment in its
infusion to 90 µ kg−1 min−1 + fentanyl bolus of 150 µ + incre-
ment in remifentanyl infusion to 0.6–0.8 µ kg−1 min−1). If
SAP remained high after adjusting anesthetics doses, a
nitroglycerine infusion of 0.01–0.10 µ kg−1 ­min−1 was then
administered until its normalization.
The clocks from S5, PICCO and Fluxmed devices were
synchronized and data simultaneously recorded. Analysis
of the recorded files was performed off-line. We selected
episodes of arterial blood normotension, hypertension and/
or hypotension according to their definition based on SAP
values [4, 24]. We arbitrarily choose those episodes in a
sequential way, from the beginning to the end of the file, as
depicted with arrows in Fig. 2. We selected 20 beats at each
moment (arrows) and obtained mean values for each of the
studied variables. A mean value of each studied variable
build the final database. As no automated detection routines
exist, one of the investigators allocated the PPGc to the cor-
responding class by visual inspection (Fig. 1).
Radial arterial blood pressure was recorded in two addi-
tional patients using the S5 device, which simultaneously
collected the coordinated arterial blood pressure and PPG Fig. 2  Example of data analysis. Systolic (SAP), mean (MAP) and
diastolic (DAP) arterial blood pressures, photoplethysmography
raw data. This information was not provided by the PICCO
(PPG), systemic vascular resistance (SVR) and vascular compliance
monitor and was only used to build Figures showing exam- (Cvasc) are depicted in one patient. Horizontal broken lines in arte-
ples of our results. rial blood pressure showed the cut off value of SAP for the diagno-
sis of arterial blood hypertension (> 140 mmHg) and hypotension
(< 90 mmHg). The arrows indicate the moment selected arbitrarily in
1.4 Statistical analysis this particular patient to build the database according to the changes
in SAP. Note the simultaneous modifications observed in PPG ampli-
Statistical analysis was performed using IBM SPSS Statistic tude and in the surrogates of vascular tone (SVR and Cvasc) during
19.0.0 (IBM Company, USA) and Matlab® (Mathworks, data collection
Natick, MA, USA). Lilliefors test described a no normal dis-
tribution of data. Wilcoxon rank sum test and Spearman rank calculated for each patient. Thus, normalized value lies in
correlations coefficient were used for comparisons between an interval between 0, 1 for proper global correlation analy-
studied variables. Due to the high inter-individual variability sis. Results are expressed as mean ± standard deviation
in studied parameters, correlations were done after normal- (SD) or median and 1st–3rd quartiles as appropriated. A p
izing them by the following formula: value < 0.05 was considered statistically significant.
x − xmin A diagnostic test 2 × 2 table was used for the assess-
xnorm = ment of sensitivity = [TP/(TP + FN)] × 100; specific-
xmax − xmin
ity = [TN/(TN + FP)] × 100; positive predicted value = [TP/
where x is the absolute value of one parameter and xmin − xmax (TP + FP)] × 100; negative predicted value = [TN/
the minimum and maximum value of this parameter, (TN + FN)] × 100 and diagnostic accuracy = [(TP + TN)/

13
Journal of Clinical Monitoring and Computing (2019) 33:815–824 819

(TP + TN + FP + FN)] × 100; where TP is true positive, TN Table 2 because they were not continuously recorded only
is true negative, FP is false positive and FN is false negative. during thermodilutions.
The test was applied to determine the ability of our classifi- The SVR increased and Cvasc decreased during arte-
cation to detect arterial blood hypertension or hypotension rial hypertension and changed in opposite direction during
caused by changes in vascular tone; where the “disease” hypotensive episodes (p < 0.0001 both compared to nor-
defined abnormal blood pressure according to known cut-off motension). Changes in SAP were correlated with these
values [4, 24] and “healthy” is represented by normotension. vascular tone related variables: SAP versus SVR (r = 0.78,
p < 0.0001) and SAP versus Cvasc (r = − 0.84, p < 0.0001).
PPGc class allocation adequately discriminated hyperten-
sive and hypotensive from normotensive episodes (Table 2).
2 Results We found a good correlation between PPGc class and SAP
(r = − 0.90, p < 0.0001), MAP (r = − 0.88, p < 0.0001), DAP
In this analysis we studied 16 patients submitted to coro- (r = − 0.85, p < 0.0001), and PP (r = − 0.87, p < 0.0001).
nary artery bypass graft (CABG) surgery (Table 1). No PPGc class was also well correlated with SVR and Cvasc
patient developed new arrhythmias nor needed pacemaker [− 0.72 (p < 0.0001) and 0.77 (p < 0.0001) respectively] the
when collecting the data. One patient was excluded due to chose surrogates of systemic arterial impedance.
acute bleeding of > 350 mL and data from the remaining As expected PPGc amplitude decreased during hyper-
15 patients were analyzed. This patient presented a medi- tension and increased during hypotensive episodes when
cal coagulopathy after cardiopulmonary bypass that needed compared to normotension (Table 2, all p < 0.0001). This
more i.v. fluids, platelets and red cells transfusion. parameter was also well correlated with SAP (r = − 0.79,
According to the pre-defined blood pressure groups we p < 0.0001), MAP (r = − 0.76, p < 0.0001), DAP (r = − 0.75,
studied 84 normotensive, 61 hypertensive and 45 hypoten- p < 0.0001), PP (r = − 0.77, p < 0.0001), SVR (r = − 0.66,
sive episodes. A total of 190 complete dataset was analyzed p < 0.0001) and Cvasc (r = 0.82, p < 0.0001).
with an average of 13 ± 5 measurements per patient. SAP Figure 3 presents examples of the performance of the
ranged from 65 to 197 mmHg. PPGc classification in one patient during changes in arterial
Table 2 presents the recorded parameters of the differ- blood pressure. Figure 4 illustrates how PPG dynamically
ent arterial blood pressure groups. CVP, PPV and SVV changes in response to treatment of hypertensive and hypo-
remained at normal values during the protocol. Mean GEDI tensive episodes.
was 742 ± 88 mL m−2 and EVLWI 8.2 ± 1.4 mL kg−1 during The PPG-based classification system failed in 7 out of the
the protocol. These last two variables are not included in 190 measurements. Three episodes of normal arterial blood

Table 1  Patient’s data

Cardiac surgery patients (n = 15)

Age (years) 67 ± 19
Gender (% females) 33
Weight (kg) 85 ± 24
Height (cm) 169 ± 43
BMI (kg/cm2) 29 ± 8
Type of surgery (number of patients) Aorto-coronary bypass (13)
Aortic valvular replacement (stenosis) + bypass (2)
EF (%) 58 ± 17
LV function (number of patients) Normal systolic (10)–slight systolic dysfunction (5)–normal diastolic (9)–slight diastolic
dysfunction (6)
RV function -Tapse (mm) 25 ± 3
Chronic diseases (number of patients) Arterial hypertension (12)
Smokers (8)–diabetes (3)–chronic renal failure (1)–obesity (3)–hypotiroidism (2)–stroke (1)
Preoperatory drugs (number of patients) B-blockers (10)
Vasodilators (7)
Diuretics (1)–T4 (2)–statins (3)
Oral hypoglycemic agents (3)

Data is presented as mean ± SD

BMI body mass index, EF ejection fraction left ventricle (LV), RV right ventricle

13
820 Journal of Clinical Monitoring and Computing (2019) 33:815–824

Table 2  Main studied Parameter Hypertension p value Normotension p value Hypotension

parameters (n = 61) (n = 84) (n = 45)

SAP (mmHg) 159 (151–170) < 0.001 124 (113–131) < 0.001 85 (80–87)
MAP (mmHg) 100 (94–109) < 0.001 77 (72–85) < 0.001 53 (47–59)
DAP (mmHg) 67 (61–74) < 0.001 53 (49–59) < 0.001 37 (33–42)
PP (mmHg) 93 (86–104) < 0.001 70 (60–77) < 0.001 47 (40–51)
CO (L min−1) 4.6 (3.4–6.4) 0.030 4.2 (3.7–6.1) 0.026 3.8(2.9–5.4)
HR (bpm) 52 (47–60) 0.354 51 (48–62) 0.407 54 (48–65)
PPV (%) 4 (3–6) 0.406 6 (4–8) < 0.001 9 (5–12)
SVV (%) 3 (3–4) 0.512 5 (3–7) 0.066 7 (6–10)
CVP (mmHg) 10 (6–12) 0.004 8 (4–11) 0.016 7 (3–9)
SVR (dyn s cm−5) 1766 (1165–2105) < 0.001 1381 (962–1711) < 0.001 1075 (750–1252)
Cvasc (mL mmHg−1) 0.97 (0.64–1.25) < 0.001 1.17 (0.80–1.59) < 0.001 1.39 (1.10-2.00)
PPG class 2 (1–2) < 0.001 3 (3–3) < 0.001 5 (5–6)
PPG amplitude (%) 43 (38–52) < 0.001 56 (44–66) < 0.001 71(61–79)

SAP systolic arterial pressure, MAP mean arterial pressure, DAP diastolic arterial pressure, PP pulse pres-
sure, CO cardiac output, HR heart rate, PPV pulse pressure variation, SVV stroke volume variation, CVP
central venous pressure, SVR systemic vascular resistance, PPG photoplethysmography. p value = Wil-
coxon rank sum test compared with normotension. Data is presented as median and 1st–3rd quartiles

Fig. 3  Example of the PPG

classification in one patient.
Arterial blood pressure (ABP)
and photoplethysmography
(PPG) at different hemodynamic
states in patient #9. Note the
inverse relationship between
PPG and pulse pressure ampli-
tude

pressure were misclassified as arterial blood hypotension and 3 Discussion

one as hypertension (false positives). The other 3 episodes of
arterial blood hypertension were misclassified as normoten- In this observational pilot study we have found that a clas-
sion because patients presented a class III pattern at SAP above sification system based on the simple visual inspection of
140 mmHg (false negatives). The corresponding analysis of the photoplethysmography waveform’s morphology was
accuracy, sensibility and specificity of the PPGc-based classifi- closely related to episodes of arterial blood hypertension
cation to detect hypo and hypertensive episodes are depicted in or hypotension mediated by changes in vascular tone.
Table 3. The diagnostic test 2 × 2 table showed high accuracy During surgery patients go through different phases in
of PPGc to diagnose arterial blood hypotension (98.4%) and a dynamic way. Intraoperative hemodynamic alterations,
hypertension (97.8%).
Journal of Clinical Monitoring and Computing (2019) 33:815–824 821

Fig. 4  Example of Arterial

blood pressure and photop-
lethysmography recordings in
two patients. Arterial blood
pressure (ABP) and photo-ple-
thysmography (PPG) continu-
ous beat-by-beat recordings.
Black-dotted, black and gray
lines indicate the moments
when PPG was analyzed with
the corresponding waves
presented at the bottom of the
figure. Left: arterial blood
hypertension episode treated by
increasing anesthetic-drug doses
as described in methods. As a
result the initial low PPG ampli-
tude increases and the position
of the dicrotic notch descends
progressively as blood pressure
decreases. Right: a hypotensive
episode corresponded to a PPG
class V with an almost flat
diastolic wave (black dotted-line
wave). Noradrenaline infusion
was administered until PPGc
changed from class V to class
III. Noradrenaline was discon-
tinued once the normal PPGc
target was reached. The direc-
tion of change in PPG ampli-
tude and dicrotic notch position
correspond to the long and short
arrows marks, respectively

Table 3  Performance of the PPGc-based classification for diagnosing hypertensive or hypotensive episodes, manipulations in
arterial blood hypotension and hypertension the surgical field affecting the monitored arm (i.e. exter-
ABP Sensitivity Specificity Accuracy PPV NPV nal cuff compression) or in morbid obese patients among
others [30–33]. Early detection of changes in the vascu-
Hypotension 1 0.979 0.984 0.938 1 lar tone and its effect on blood pressure by simple PPGc
Hypertension 0.949 0.992 0.978 0.982 0.977 analysis could therefore facilitate the early diagnosis of
Diagnostic test 2 × 2 table for the assessment of sensitivity, specific- circulatory instability and allow for an early proactive cor-
ity, accuracy, positive predictive value (PPV) and negative predicted rective therapeutic intervention when needed.
value (NPV) of the PPG classification of hypotensive and hyperten- The PPGc classification would also help to interpret the
sive episodes
pathophysiologic mechanisms behind arterial blood pressure
changes during anesthesia and intensive care management.
in the form of hyper or hypotensive episodes, are com- Changes in arterial blood pressure induced by alterations
mon and caused by surgical manipulations, bleeding, the in vascular tone are common for most anesthetic and many
interaction of the anesthetic agents and previous patient’s sedative drugs [7–9]. Thus, medical therapies for intraopera-
chronic diseases. The standard blood pressure routine tive blood pressure changes could be better targeted to one of
monitoring is based on NIBP systems that intermittently its main pathophysiologic causative mechanisms.
asses blood pressure at regular preselected periods. This The clinical use of the arterial pulse pressure waveform
implies that many of such hemodynamic episodes that for monitoring hemodynamics has been well described
appear suddenly may go undetected [6]. Depending on [16–18]. The same is true for the PPG contour analysis,
the measurement rate set, routine NIBP monitoring leaves which has been used for the assessment of fluid respon-
“blind” periods between cuff compressions of variable siveness [34, 35] and for studying the effect of aging and
duration in which, sometimes important changes in the diseases on the vascular system [36, 37]. Previous publica-
hemodynamic status occur. Furthermore, the perfor- tions have highlighted the relationship between PPG and
mance of NIBP is often impaired in situations of severe arterial blood pressure. PPG has been used to estimate blood

13
822 Journal of Clinical Monitoring and Computing (2019) 33:815–824

pressure using different approaches like the pulse wave the changes in arterial blood pressure, guiding the treatment
velocity analysis [38], the artery clamp method [39, 40], as observed in Fig. 4. We did not test the potential role of
reappearance of the PPG waveform during cuff deflation [41, the proposed PPG based classification in the detection of
42] or the analysis of different PPG features among others other hemodynamic changes like hypovolemia in the studied
[43, 44]. The association between the PPGc and arterial wall patients. Therefore, whether such a classification can dif-
compliance has also been described. Lopez-Beltrán et al. and ferentiate vasoplegic from hypovolemic hypotension must
Jagomägi et al. calculated beat-by-beat vascular compliance be addressed in future studies.
in a non-invasive way using PPG and finger blood pressure
(Peñaz’s method) [45, 46]. These authors demonstrated that 3.1 Limitations
vascular compliance derived from PPG quickly changes with
the cold-pressure test and during arm elevation, respectively. Despite the protocol was design to avoid occult hypovolemia
Awad et al. and Middleton et al. have already studied the or hypervolemia, we cannot confirm or discard these con-
correlation between PPGc derived-parameters and SVR [11, ditions using PPV. The changes observed in arterial blood
13]. Dawber et al. focused on the upward position of the pressure could be affected not only by changes in the vas-
dicrotic notch of the PPGc for the screening of hypertensive cular tone but also for some degree of hypo/hypervolemia.
patients [20]. This notch’s position shift reflects the early Thus, the performance of our PPG based classification of
return of backward waves during cardiac systole and was the vascular tone could be influenced by a potential change
related to arterial blood hypertension. However, they did in patient’s volemia. We assume that the main changes in
not describe any PPGc changes to characterize arterial blood arterial blood pressure were caused by changes in the vas-
hypotension. cular tone in our patients. This assumption is because the
Our results are in line with these studies. We observed a fast changes in blood pressure were reversible and related to
good correlation between the PPGc class and SAP at blood changes in the surrogates of vascular tone (SVR and Cvasc)
pressure levels ranging from 65 to 197 mmHg. We also and not related to bleeding, fluid deficit or overload. Next
found good correlations between SAP and surrogates of studies must be done in controlled experimental environ-
vascular tone such as SVR and Cvasc. Taking all this infor- ments to test different hemodynamic scenarios.
mation together, our findings support and raise the interest of We only investigated those PPG parameters related to the
the use of PPGc as a non-invasive hemodynamic monitoring proposed classification—the amplitude and the position of
option [47]. the dicrotic notch. Other PPG derived variables related to
The effect on the PPGc of vasodilation induced by alco- the hemodynamic status like the plethysmographic variation
hol, nitrites or heat has been previously studied by several index, the stiffness index, the diastolic-to-systolic ratio (B/A
authors [21–23, 34]. In our daily practice we also observed ratio), the PPG second derivative or the PPG width were not
that arterial hypotension induced by anesthetic drug induced tested in this study. The addition of those PPG parameters
vasodilation had a clear effect on the PPGc; effect that is could improve the sensitivity and specificity of the PPGc
reversed by an intravenous infusion of a vasoconstrictor like method as described by Lee et al. [12].
noradrenaline [48–50]. Similar to arterial hypertension, the We did not analyze inter-observer agreement in this study
changes on PPGc during vasodilation-induced arterial hypo- and certainly the automatization of this PPGc-based estima-
tension could be explained by ventricular-vascular interac- tion of the vascular tone, including many other PPG-derived
tion. During hypotension the vascular tree is more compliant parameters, would eliminate any physician’s subjectivity and
and the backward wave returns slowly and late in diastole, improve the diagnostic performance of PPGc. Our primary
delaying and moving the dicrotic notch downwards. If vaso- intention at this stage was a preliminary description of a
dilation is moderate-to-severe we observed that the notch method that can provide clinicians with an easy, fast and
became flat and then negative. simple assessment of vascular tone and its association with
We believe our findings are of clinically interest because arterial blood pressure changes just by observing the stand-
most of severe postoperative complications, like acute myo- ard pulse oximeter’s waveform display and not depending
cardial infarction, intracranial hemorrhage or kidney failure on any more complex computational waveform analysis.
are related not only with the degree of intraoperative arte- All detected episodes of arterial blood normo-hyper-hypo-
rial blood hyper/hypotension but also with the duration and tension were selected for analysis but, as the investigator
number of such episodes [1–5]. The proposed classification was not blinded for the PiCCO tracings (as they were used
can help to quickly suspect an arterial hyper/hypotensive to detect the episodes), certain subjectivity in the selection
episode in patients where invasive arterial blood pressure could not be ruled out and could induce a bias in our final
or advanced hemodynamic monitoring is not indicated. results.
Besides, the classification could be potentially helpful to The studied cohort of cardiac surgery patients is an
differentiate the pathophysiological mechanisms that caused unstable population with vascular disease and endothelial
Journal of Clinical Monitoring and Computing (2019) 33:815–824 823

dysfunction, which results in common and frequent fluc- 9. Holte K, Foss NB, Svensen C, Lund C, Madsen JL. Epidural
tuations in blood pressure. Different cohorts like elderly or anesthesia, hypotension, and changes in intravascular volume.
Anesthesiology. 2004;100:281–6.
children or patients with different hemodynamic conditions 10. Allen J. Photopletysmography and its application in clinical
such as heart failure and/or acute bleeding could affect the physiological measurement. Physiol Meas. 2007;28:R1–39.
performance of a PPGc-based estimation of the vascular 11. Awad AA, Haddadin AS, Tantawy H, Badr TM, Stout RG, Sil-
tone. Future studies will be needed to test the accuracy of verman DG, Shelley KH. The relationship between the photop-
lethysmographic waveform and systemic vascular resistance. J
the proposed method in a broader population of patients. Clin Monit Comput. 2007;21:365–72.
12. Lee QY, Chan GSH, Redmond SJ, Middleton PM, Steel E,
Malouf P, Critoph C, Flynn G, O’Lone E, Lovell NH. Mul-
3.2 Conclusions tivariate classification of systemic vascular resistance using
photoplethysmography. Physiol Meas. 2011;32:1117–32.
13. Middleton PM, Chan GSH, Steel E, Malouf P, Critoph C, Flynn
Changes in arterial blood pressure mainly mediated by G, O’Lone E, Celler BC, Lovell NH. Fingertip photoplethys-
alterations in the vascular tone were closely related to a mographic waveform variability and systemic vascular resist-
classification based on the shape of the photo-plethysmo- ance in intensive care unit patients. Med Biol Eng Comput.
graphic waveform. The proposed non-invasive classifica- 2011;49:859–66.
14. De Trafford J, Lafferty K. What does photoplethysmography
tion has potential monitoring, diagnostic and therapeutic measure? Med Biol Eng Comput. 1984;22:479–80.
applications. 15. Wisely NA, Cook LB. Arterial flow waveforms from pulse
oximetry compared with measured Doppler flow waveforms.
Compliance with ethical standards Anaesthesia. 2001;56:556–61.
16. O’Rourke MF, Yaginuma T, Avolio AP. Physiological and
pathophysiological implications of ventricular/vascular cou-
Conflict of interest No potential conflicts of interest exist except for pling. Ann Biomed Eng. 1984;12:119–34.
Matías Madorno who is partner and manager of MBMed S.A; a com- 17. Korpas D, Hâlek J, Dolezal L. Parameters describing the pulse
pany that produce respiratory monitoring equipments. wave. Physiol Res. 2009;58:473–9.
18. Van den Bos GC, Westerhof N, Randall OS. Pulse-wave reflec-
Informed consent Informed consent was obtained from all individual tion: can it explain the differences between systemic and pul-
participant included in the study. monary pressure and flow waves? Circ Res. 1982;51:479–85.
19. Tusman G, Bohm SH, Suarez Sipmann F. Advanced uses of
pulse oximetry for monitoring mechanically ventilated patients.
Anesth Analg. 2017;201:62–71.
20. Dawber TR, Thomas HE Jr, McNamara PM. Characteristics of
References the dicrotic notch of the arterial pulse wave in coronary heart
disease. Angiology. 1973;24:244–55.
1. Walsh M, Devereaux PJ, Garg AX, Kurz A, Turan A, Rodseth 21. Morikawa Y. Characteristic pulse wave caused by organic
RN, Cvwinski J, Thabane L, Sessler DI. Relationship between nitrates. Nature. 1967;213:841–2.
intraoperative mean arterial pressure and clinical outcomes after 22. Morikawa Y, Matsuzaka J, Kuratsune M, Tsukamoto S, Maki-
noncardiac surgery. Toward an empirical definition of hypoten- sumi S. Plethysmographic study of effects of alcohol. Nature.
sion. Anesthesiology. 2013;119:507–15. 1968;220:186–7.
2. Basali A, Masch EJ, Kalfas I, Schubert A. Relation between peri- 23. Dillon JB, Hertzman AB. The form of the volume pulse in the
operative hypertension and intracranial hemorrhage after crani- finger pad in health, arteriosclerosis, and hypertension. Am
otomy. Anesthesiology. 2000;93:48–53. Heart J. 1941;21:172–90.
3. Reich DL, Bennet-Guerrero E, Bodin C, Hossain S, Winfree W, 24. Mancia G, Fagard R, Narkiewicz K. ESH/ESC guidelines
Krol M. Intraoperative tachycardia and hypertension are indepen- for the management of arterial hypertension. Eur Heart J.
dently associated wih adverse outcome in noncardiac surgery of 2013;34:2159–219.
long duration. Anesth Analg. 2002;95:273–7. 25. Milnor WR. Arterial impedance as ventricular afterload. Circ
4. Bijker JB, van Klei WA, Kappen TH, van Wolfswinkel L, Res. 1975;36:565–70.
Moons KGM, Kalkman CJ. Incidence of intraoperative hypo- 26. Randall OS, Westerhof N, van den Bos GC, Alexander B. Reli-
tension as a function of the chosen definition. Anesthesiology. ability of stroke volume to pulse pressure ratio for estimating
2007;107:213–20. and detecting changes in arterial compliance. J Hypertens.
5. van Waes JAR, van Klei WA, Wijeysundera DN, van Wolfswinkel 1986;4:293–6.
L, Lindsay TF, Beattie WS. Association between intraoperative 27. Marik PE, Monnet X, Teboul JL. Hemodynamic parameters to
hypotension and myocardial injury after vascular surgery. Anes- guide fluid theraphy. Ann Intensive Care. 2011;1:1.
thesiology. 2016;124:35–44. 28. Broch O, Renner J, Gruenewald M, Meybohm P, Höcker J,
6. Bartels K, Esper SA, Thiele RH. Blood pressure monitor- Schöttler J, Steinfath M, Bein B. Variation of left ventricular
ing for the anesthesiologist: a practical review. Anesth Analg. outflow tract velocity and global end-diatolic volume index reli-
2016;122:1866–79. ability predict fluid responsiveness in cardiac surgery patients.
7. Akata T. General anesthetics and vascular smooth muscle. Direct J Crit Care. 2012;27:325e7–14.
actions of general anesthetics on cellular mechanisms regulating 29. Sakka SG, Rühl CC, Pfeiffer UJ, Beale R, McLuckie A, Rein-
vascular tone. Anesthesiology. 2007;106:365–91. hart K, Meier-Hellmann A. Assessment of cardiac preload and
8. Lindorf HH. Investigation of the vascular effect of newer extravascular lung water by single transpulmonary thermodilu-
local anesthetics and vasoconstrictors. Oral Surg Oral Med. tion. Intensive Care Med. 2000;26:180–7.
1979;46:292–7.

13
824 Journal of Clinical Monitoring and Computing (2019) 33:815–824

30. Bur A, Herkner H, Vlcek M, Woisetschläger C, Derhasching U, 42. Talke P, Nichols RJ, Traber DL. Does measurement of systolic
Delle KG, Lagger A, Hirschi MM. Factors influencing the accu- blood pressure with a pulse oximeter correlate with conventional
racy of oscillometric blood pressure measurement in critically ill methods? J Clin Monit. 1990;6:5–9.
patients. Crit Care Med. 2003;31:793–9. 43. Bortolotto LA, Blacher J, Kondo T, Takazawa K, Safar ME.
31. Liu J, Hahn JO, Mukkamala R. Error mechanism of the oscil- Assessment of vascular aging and atherosclerosis in hypertensive
lometric fixed-ratio blood pressre measurement method. Ann subjects: second derivative of photoplethysmogram versus pulse
Biomed Eng. 2013;41:587–97. wave velocity. Am J Hypertens. 2000;13:165–71.
32. Pytte M, Dybwik K, Sexton J, Straume B, Nielsen W. Oscillo- 44. Awad AA, Ghobashy AM, Stout RG, Silverman DG, Shelley KH.
metric brachial mean artery pressures are higher than intra-radial How does plethysmogram derived from the pulse oximeter relate
mean artery pressures in intensive care unit patients receiving to arterial blood pressure in coronary artery bypass graft patients?
norepinephrine. Acta Anaesthesiol Scand. 2006;50:718–21. Anesth Analg. 2001;93:1466–71.
33. Araghi A, Bander JJ, Guzman JA. Arterial blood pressure moni- 45. Lopez-Beltrán EA, Blackshear PL, Finkelstein SM, Cohn JN.
toring in overweight critically ill patients: invasive or noninvasive. Non-invasive studies of peripheral vascular compliance using a
Crit Care. 2006;10:R64. non-occluding photoplethysmographic method. Med Biol Eng
34. Natalini G, Rosano A, Taranto M, Faggian B, Vittorelli E, Ber- Comput. 1998;36:748–53.
nardini A. Arterial versus plethysmographic dynamic indices to 46. Jagomägi K, Raamat R, Talts J, Ragun U, Länsimies E, Jurvelin
test responsiveness for testing fluid administration in hypotensive J. Recording of dynamic arterial compliance changes during hand
patients: a clinical trial. Anesth Analg. 2006;103:1478–84. elevation. Clin Physiol Funct Imaging. 2005;25:350–6.
35. Cannesson M, Besnard C, Durand PG, Bohé J, Jacques D. Rela- 47. Bendjelid K. The pulse oximetry plethysmographic curve revis-
tion between respiratory variations in pulse oximetry plethysmo- ited. Curr Opin Crit Care. 2008;14:348–53.
graphic waveform amplitude and arterial pulse pressure in venti- 48. Fischer GW, Levin MA. Vasoplegia during cardiac sur-
lated patients. Crit Care. 2005;9:R562–8. gery: current concepts and management. Semin Thorac Surg.
36. Millasseau SC, Kelly RP, Ritter JM, Chowienczyk PJ. Determi- 2010;22:140–4.
nation of age-related increases in large artery stiffness by digital 49. Overgaard CB, Dzavík V. Inotropes and vasopressors. Review of
pulse contour analysis. Clin Sci (Lond). 2002;103:371–7. physiology and clinical use in cardiovascular disease. Circulation.
37. Allen J, Murray A. Age-related changes in peripheral pulse 2008;118:1047–56.
shape characteristics at various body sites. Physiol Meas. 50. Morimatsu H, Uchino S, Chung J, Bellomo R, Raman J, Bux-
2003;24:297–307. ton B. Norephinephrine for hypotensive vasodilation after car-
38. Hashimoto J, Chonan K, Aoki Y, Nishimura T, Ohkubo T, Hozawa diac surgery: impact on renal function. Intensive Care Med.
A, Suzuki M, Matsubara M, Michimata M, Araki T, Imai Y. Pulse 2003;29:1106–12.
wave velocity and the second derivative of the finger photop- 51. Wajima Z, Shiga T, Imanaga K, Inoue T. Do induce hypertension
lethysmogram in treated hypertensive patients: their relationship and hypotension affect stroke volume variation in man? J Clin
and associating factors. J Hypertens. 2002;20:2415–22. Anaesth. 2012;24:207–11.
39. Peñaz J. Criteria for set point estimation in the volume clamp 52. Bouchcourt JP, Riva JA, Grignola JC. The increase of vasomo-
method of blood pressure measurement. Physiol Meas. tor tone avoids the ability of the dynamic preload indicators to
1992;41:5–10. estimate fluid responsiveness. BMC Anesthesiol. 2013;13:41.
40. Pinna GD, Maestri R, Mortara A. Estimation of arterial 53. Bagshaw SM, Brophy PD, Ronco CD. Fluid balance as a bio-
blood pressure variability by spectral analysis: comparison marker: impact of fluid overload on outcome in critically ill
between Finapres and invasive measurements. Physiol Meas. patients with acute kidney injury. Crit Care. 2008;12:169.
1996;17:147–9. 54. Morin JF, Mistry B, Langllois Y, Ma F, Chaumoun P, Holcroft
41. Wallace C, Barker D, Apert C, Tankersley SJ, Conroy JM, Kerns C. Fluid overload after coronary artery bypass grafting surgery
RE. Comparison of blood pressure measurement by Doppler and increases the incidence of postoperative complications. World J
by pulse oximetery techniques. Anesth Analg. 1987;66:1018–9. Cardiovasc Surg. 2011;1:18–23.