Heart-Lung Interactions: Implications for Non-Invasive Evaluation of Changes in Blood Volume

Heart-lung interactions: Implications for non-invasive evaluation of changes in blood volume

Thesis for the degree of Philosophiae Doctor (PhD)

Cand. Med. Ingrid Elise Hoff

Institute of Clinical Medicine, Faculty of Medicine

University of Oslo

Oslo, Norway and

The Norwegian Air Ambulance Foundation

Oslo, Norway and

Department of Anaesthesiology

Division of Emergencies and Critical Care

Oslo University Hospital

Oslo, Norway

2019 2

Series of dissertations submitted to the Faculty of Medicine, University of Oslo

ISBN 978-82-8377-539-6

Cover: Hanne Baadsgaard Utigard. Print production: Reprosentralen, University of Oslo. 3

To Ingvild and Benedikte

Table of Contents

1 ACKNOWLEDGEMENTS ...... 9 2 ABBREVIATIONS AND ACRONYMS ...... 11 3 LIST OF PAPERS ...... 13 4 SYNOPSIS ...... 14 5 INTRODUCTION ...... 17 5.1 Why normovolaemia? Oxygen delivery and consumption ...... 17 5.2 Cardiac output and cardiac output measurement ...... 18 5.3 Venous return and the Frank-Starling-curve ...... 20 5.4 Haemodynamic effects of respiration ...... 24 5.5 Dynamic variables and the concept of fluid responsiveness...... 26 5.6 Pulse pressure variations (∆PP) ...... 27 5.6.1 Considerations for the use ∆PP for haemodynamic evaluation ...... 29 5.6.1.1 Ventilation ...... 29 5.6.1.2 Intraabdominal hypertension ...... 30 5.6.1.3 Cardiac dysfunction ...... 31 5.7 The photoplethysmographic waveform amplitude variation (∆POP), perfusion index (PI) and pleth variability index (PVI) ...... 31 5.7.1 Limitations to the use of ∆POP, PVI and PI for haemodynamic evaluation ..... 33 5.7.1.1 Vascular tone ...... 33 5.7.1.2 Processing algorithms ...... 34

5.8 End-tidal (EtCO2) and volumetric (VtCO2) carbon dioxide ...... 35

5.8.1 Limitations of EtCO2 and VtCO2 for haemodynamic evaluation ...... 38 6 AIM AND RESEARCH QUESTIONS ...... 39 7 MATERIAL AND METHODS ...... 40 7.1 Study populations ...... 40 7.2 Lower Body Negative Pressure (LBNP) ...... 40 7.3 Right ventricular pacing (RVP) ...... 42 7.4 Passive leg raise (PLR) ...... 43 7.5 Modification of respiration: Non-invasive positive pressure ventilation (NPPV), continuous positive airway pressure (CPAP) and positive expiratory pressure (PEP) ..... 44 7.6 Cardiac output measurements ...... 45 6

7.7 The volume-clamp method for blood pressure measurement ...... 46 7.8 Sampling of haemodynamic data ...... 47 7.9 Calculation of respiratory variations in pulse pressure (∆PP) and the photoplethysmographic waveform amplitude (∆POP) ...... 47

7.10 Calculation of end-tidal (EtCO2) and volumetric (VtCO2) carbon dioxide ...... 49 7.11 Statistical methods ...... 50 7.11.1 Linear mixed models ...... 50 7.11.2 Correlation ...... 51 7.11.3 Receiver operating characteristics plot (ROC-plot) ...... 52 8 RESULTS ...... 54 8.1 Study I ...... 54 8.2 Study II ...... 54 8.3 Study III ...... 55 9 DISCUSSION ...... 56 9.1 Main results ...... 56 9.1.1 The ability of ∆PP, ∆POP, PVI and PI to reflect hypovolaemia during spontaneous breathing and non-invasive positive pressure ventilation (NPPV) ...... 56

9.1.2 The ability of EtCO2 and VtCO2 to reflect hypovolaemia ...... 59 9.1.3 The ability of ∆PP and ∆POP to detect hypovolaemia during positive expiratory pressure (PEP) and continuous positive airway pressure (CPAP) ...... 63 9.2 Methodological considerations ...... 66 9.2.1 Study populations ...... 66 9.2.1.1 Choice of study populations ...... 66 9.2.1.2 Sample size calculation ...... 67 9.2.1.3 Randomisation ...... 69 9.2.2 Models ...... 69 9.2.2.1 Lower body negative pressure (LBNP) as a model of central hypovolaemia 69 9.2.2.2 Non-invasive positive pressure ventilation (NPPV) to induce respiratory variations in stroke volume ...... 70 9.2.2.3 Right ventricular pacing for the reduction of cardiac output ...... 71 9.2.3 Measurements and calculations: Sources of potential errors ...... 71 9.2.4 Statistical methods ...... 74 9.2.4.1 Linear mixed models ...... 74 9.2.4.2 Correlation ...... 75 7

9.2.4.3 Receiver operating characteristics (ROC) plots ...... 75 9.3 Ethical considerations ...... 76 9.4 Future perspectives ...... 77 10 CONCLUSIONS ...... 79 11 REFERENCES ...... 80 12 PAPERS ...... 89

1 ACKNOWLEDGEMENTS

First and foremost, I want to thank my supervisors Knut Arvid Kirkebøen, Svein Aslak

Landsverk and Lars Øivind Høiseth for giving me the opportunity to work within the field of research. This group of supervisors is small, but highly efficient; consisting of different members who complement each other perfectly and thus cater to a PhD-student`s every need.

This work required custom-made technology that demanded some unusual skills, and practical as well as theoretical advice has been provided swiftly and at all hours, for which I am very grateful.

Many thanks also to The Norwegian Air Ambulance Foundation and its donors who have made this work possible through educational, practical and financial support. It has been a privilege being one of your PhD-students, and I`m very grateful to Hans Morten Lossius who originally invited me into this community of researchers.

This work would not have been possible without the trust and cooperation of the patients and volunteers who participated in the different studies. I`m impressed and humbled by the will to contribute even under difficult circumstances.

I`m also very grateful for the friendship, flexibility and support granted by the leaders of my clinical department; Anne Bøen and Kristin Sem Thagaard, and more recently co-pilot Anders

Nordby. Like my supervisors, they have patiently watched me try to juggle tasks within quite different areas over several years without complaining.

Many thanks also to Jonny Hisdal for never ending positivity and cooperation on Papers I and

III. I`m also very grateful to Jo Røislien for educational statistical guidance, and to Jon Bach for programming services for Papers I and II. 10

I would also like to thank all my colleagues at the Department of Anaesthesiology, Ullevål, for creating an atmosphere where discussions around research, clinical work and pure nonsense are equally welcome. Special thanks to Kirsti Myre and Baard Ingvaldsen for teaching me a lot in my early years, and to the anaesthesiologists and nurses of the cardiothoracic recovery unit for their assistance during data collection for Paper II.

And last, but not least, I want to thank my family for providing constant and unconditional support and inspiration, and making it all possible.

Oslo, November 2019

2 ABBREVIATIONS AND ACRONYMS

ARDS Acute respiratory distress syndrome

AUC Area under the curve

CO Cardiac output

CO2 Carbon dioxide

CPAP Continuous positive airway pressure

CVP Central venous pressure

∆PP Pulse pressure variations

∆POP Photoplethysmographic waveform amplitude variations

DO2 Oxygen delivery

EtCO2 End-tidal carbon dioxide concentration

Hgb Haemoglobin

ICU Intensive care unit

IPPV Intermittent positive pressure ventilation

LBNP Lower body negative pressure

MAP Mean arterial pressure

NPPV Non-invasive positive pressure ventilation

PaO2 The partial pressure of oxygen 12

PEP Positive expiratory pressure

PI Perfusion index

PLR Passive leg raise

Pmsf Mean systemic filling pressure

PP Pulse pressure

PVI Pleth variability index

RAP Right atrial pressure

ROC Receiver operating characteristics

RVP Right ventricular pacing

RVR Resistance to venous return

SaO2 Oxygen saturation

SV Stroke volume

SVV Stroke volume variation

TAVI Transcatheter aortic valve implantation

VCO2 Eliminated volume of carbon dioxide per minute

VR Venous return

VtCO2 Eliminated volume of carbon dioxide per tidal volume

VTI Velocity time integral 13

3 LIST OF PAPERS

Paper I

Hoff IE, Høiseth LØ, Hisdal J, Røislien J, Landsverk SA, Kirkebøen KA. Respiratory

Variations in Pulse Pressure Reflect Central Hypovolaemia during Noninvasive Positive

Pressure Ventilation.

Critical Care Research and Practice, Volume 2014, Article ID 712728, https://doi.org:10.1155/2014/712728.

Paper II

Hoff IE, Høiseth LØ, Kirkebøen KA and Landsverk, SA. Volumetric and end-tidal capnography for the detection of cardiac output changes in mechanically ventilated patients early after open heart surgery.

Critical Care Research and Practice, Volume 2019, Article ID 6393649, https://doi.org/10.1155/2019/6393649.

Paper III

Hoff IE, Hisdal J, Landsverk SA, Røislien J, Kirkebøen KA, Høiseth LØ. Respiratory variations in pulse pressure and photoplethysmographic waveform amplitude during positive expiratory pressure and continuous positive airway pressure in a model of progressive hypovolaemia.

Submitted.

4 SYNOPSIS

Correct and timely evaluation of haemodynamic status is essential in critically ill patients and during surgery. Early detection of reduced blood volume due to capillary leak, intestinal losses or haemorrhage may prevent deterioration into organ failure and shock. But even mild to moderate hypovolaemia may cause hypoperfusion. Hypovolaemia may be difficult to detect based on clinical signs such as blood pressure and heart rate.1 Too much fluid may also be detrimental, as it may induce pulmonary oedema and respiratory failure, or increase bleeding from injuries. Arterial blood pressure is frequently used to guide haemodynamic decisions. However, arterial pressure also depends on vascular resistance, and reflects hypovolaemia only after substantial blood loss. Vasomotor tone may be altered by pain, temperature and medication; factors that are frequently present in patients during injury, surgery or critical illness. Haemodynamic evaluation in these situations remains a challenge.

Thus, monitoring of cardiac output, the amount of blood ejected by the heart each minute, is recommended in critically ill patients.2 However, methods for cardiac output monitoring are often invasive and skill-dependent, or available only to specific patient groups. We have explored some less invasive methods based on heart-lung interactions in new experimental and clinical settings.

Heart-lung interactions may be recognisable as changes in stroke volume (SV) and pulse pressure (PP) or as changes in exhaled carbon dioxide (CO2). Pulse pressure variations (∆PP) arise from respiratory induced changes in venous return and SV. Together with the corresponding changes in the photoplethysmographic waveform amplitude (∆POP) and the pleth variability index (PVI), they are referred to as dynamic variables. This term indicates a dynamic measure of function rather than a static image of volume or pressure. The magnitude 15

of the respiratory induced changes in SV reflects preload dependency, and may thus be a measure of volume status. The dynamic variables are validated in mechanically ventilated patients, but not during spontaneous breathing activity. The amount of exhaled CO2 depends on the production of CO2 in the tissues, the transport ability as in cardiac output or pulmonary flow, and ventilation. If metabolism and ventilation are stable, any changes in exhaled CO2 will be due to changes in pulmonary flow. The monitoring of end-tidal CO2 (EtCO2) as an indicator of spontaneous circulation after cardiac arrest is well established. However, exhaled

CO2 is infrequently used for haemodynamic evaluation within more normal ranges of cardiac output, for instance perioperatively or in the intensive care unit (ICU). Like SV, pulmonary flow also depends on venous return.

The subject of this thesis is how heart-lung interactions, measured by circulation or ventilation, may contribute to the haemodynamic evaluation of both spontaneously breathing and mechanically ventilated patients perioperatively and during critical care.

In study I, we investigated to what degree dynamic variables reflected progressive hypovolaemia during spontaneous breathing and non-invasive positive pressure ventilation

(NPPV). Central hypovolaemia was induced by lower body negative pressure (LBNP), where the lower body is placed in a vacuum chamber, and blood is sequestered into the lower extremities. NPPV was applied to un-medicated healthy volunteers. We found that ∆PP reflected hypovolaemia both during spontaneous breathing and during NPPV, and that the effect of hypovolaemia on ∆PP almost doubled when NPPV was applied. However, the application of NPPV alone did not impact ∆PP, meaning that the subjects had to be hypovolaemic in order for NPPV to have an effect on ∆PP. Neither ∆POP nor PVI reflected hypovolaemia, nor were they affected by the application of NPPV. 16

In study II, we evaluated the ability of EtCO2 and eliminated CO2-volume per breath, VtCO2, to reflect moderate changes in cardiac output in mechanically ventilated patients. Cardiac output was transiently reduced by right ventricular pacing (RVP), and increased by passive leg raise (PLR). The reductions in EtCO2 and VtCO2 correlated with the reduction in cardiac output, but there was no correlation between the changes during PLR. The changes in VtCO2 induced by changes in cardiac output were twice as large as the changes in EtCO2, which may make them easier to detect. However, only EtCO2 demonstrated a discriminative ability according to ROC plots.

In paper III, we continued to evaluate how different modes of respiration affect the ability of the dynamic variables to reflect hypovolaemia. Continuous positive airway pressure (CPAP) and positive expiratory pressure (PEP) is frequently applied to spontaneously breathing patients postoperatively and in the ICU. We applied different levels of CPAP and PEP to healthy volunteers during progressive hypovolaemia induced by LBNP. ∆PP and ∆POP both reflected hypovolaemia during nearly all levels of PEP and CPAP. Whereas high PEP augmented the increase in ∆PP, change in CPAP-level did not affect any of the variables significantly.

Thus, in papers I and III, the impact of respiratory changes on the heart and circulation was investigated. In paper II, the effects of circulatory changes on respiratory-derived variables were explored. Our goal was to explore the potential of these variables in contexts where they are not normally used, such as ∆PP and ∆POP during spontaneous breathing, and exhaled

CO2 for the detection of moderate, sudden changes in cardiac output.

5 INTRODUCTION

5.1 Why normovolaemia? Oxygen delivery and consumption

Adequate intravascular volume is necessary to ensure adequate perfusion and oxygen delivery

(DO2) to the tissue. Detection of haemorrhage before the development of shock, where perfusion and oxygen supply are insufficient for the metabolic demands of the cell3, improves

4 outcome. DO2 is determined by the formula

DO2 = (1.3 x Hgb x SaO2 + 0.0225 x PaO2) x 10 x CO

where Hgb = haemoglobin, SaO2 = arterial oxygen saturation, PaO2 = partial pressure of oxygen, and CO = cardiac output. As the fraction of dissolved oxygen in blood is insignificant under normal circumstances, the formula may be simplified:

DO2 ≈1.3 x Hgb x SaO2 x 10 x CO

Increased cardiac output can compensate for a decrease in SaO2 or Hgb, but not vice versa, as

SaO2 can only be increased to 100%, and supra-normal Hgb-values increase viscosity and reduce perfusion.5 Thus, cardiac output is a major determinant for oxygen delivery to the tissues.

In most clinical circumstances, oxygen consumption is independent of DO2, because the degree of oxygen extraction adapts to different levels of oxygen supply. This mechanism provides the tissues with stable supplies of oxygen. However, when oxygen supply falls below a critical level, oxygen consumption decreases.5 Metabolism shifts from aerobic to anaerobic, and the production of lactate increases.3

During haemorrhage heart rate and vascular resistance initially increase in order to maintain cardiac output. Especially young, otherwise healthy adults compensate with increased 18

sympathetic tone during hypovolaemia, maintaining their blood pressure and masking volume loss.3 With progressive hypovolaemia, the compensatory mechanisms become exhausted, and de-compensation begins. When appx. 70% of total blood volume is lost, adequate perfusion of the brain and heart is no longer maintained.6,7

CO2-production is proportional to oxygen consumption. CO2 is brought to the lungs by pulmonary flow and can be measured in the expired air as eliminated CO2 -volume; VCO2.

Thus, VCO2 becomes a measure of both oxygen consumption and pulmonary flow. In patients without major shunting, pulmonary flow equals cardiac output. In adequately oxygenated patients with normal haemoglobin concentration, cardiac output determines oxygen delivery.

5.2 Cardiac output and cardiac output measurement

Cardiac output is the product of SV, the amount of blood ejected per heartbeat, and heart rate.

SV depends on preload, contractility and afterload. Preload is defined by the degree of stretching of the cardiac muscle fibres, or diastolic wall stress. It depends on ventricular filling, which is therefore often simplistically referred to as preload. Up until a certain point, increased distension of the muscle fibres increases the contractile force of the heart.

Contractility refers to the strength of the ventricular muscle contraction from a specific fibre length, and is altered by neuro-humoral factors.8 Afterload describes the force the ventricle has to overcome to eject blood, or systolic wall stress.9 It depends on the pressure and thus resistance in the aorta, or, for the right ventricle, in the pulmonary artery, but also on the radius and wall thickness of the ventricle.8

Cardiac output can be cumbersome to measure directly. Therefore, estimates of cardiac output are often based on measurements of surrogates. Traditionally, cardiac output has been estimated by pulmonary thermodilution. This technically challenging and highly invasive 19

procedure carries risks for complications and is thus not warranted during low-risk procedures.

It has also been found inaccurate in low flow states.10 Thus, the attention has been drawn towards the development of minimally invasive methods, however, pulmonary artery thermodilution remains the reference technique that new methods are evaluated against.

Minimally invasive methods based on catheters in the femoral or even radial arteries have replaced the pulmonary artery catheter in some settings. Pulse contour analysis refers to the estimation of SV from the arterial pressure waveform. It is based on the close relationship between left ventricular SV, arterial impedance and arterial pulse pressure; referred to as ventriculo-arterial coupling.11 Pulse contour systems may be calibrated by transpulmonary dilution techniques. Calibrated systems are generally less prone to measurement errors than the uncalibrated systems, which to a larger extent are affected by changes in vascular resistance.12-14 Calibrated systems may give information on extravascular lung water and vascular resistance as well as cardiac output. The uncalibrated systems are based on demographic data with statistical corrections which may be inaccurate for the individual patient, but these systems have been found reliable in haemodynamically stable patients during routine surgery.15 Minimally-invasive methods that are validated for perioperative use also include oesophageal Doppler, which continuously calculates SV based on flow in the descending aorta. Pulse contour analysis and oesophageal Doppler are suitable for tracking transient changes in SV during major surgery.11,13

Transthoracic echocardiography is frequently used for non-invasive cardiac output measurement. SV is calculated by multiplying the velocity time integral (VTI) of the left ventricular outflow tract, measured by Doppler ultrasound, with the cross-sectional area.16

Non-invasive methods also include pulmonary capnotracking and partial rebreathing; two continuous measurement methods based on the differential Fick`s principle. Ventilation is 20

manipulated by hypo-and hyperventilation or by increasing dead space, and pulmonary blood flow is calculated from the changes in eliminated CO2. Electrical bioimpedance and bioreactance are other non-invasive methods for cardiac output measurements.17

A reduction in complication rate following goal-directed fluid therapy based on cardiac output monitoring has been demonstrated.18, 19,20 All methods have limitations associated with technology or patient physiology, and must be chosen according to the patient and procedure.

The trend or a change in cardiac output is often of higher clinical interest than the absolute values. Rather than giving continuous estimates, the variables we investigated in this thesis may be used to track or detect changes in cardiac output.

5.3 Venous return and the Frank-Starling-curve

Cardiac output depends on preload, afterload, contractility and heart rate. The Frank-Starling mechanism describes the relationship between preload and SV, and the heart`s ability to increase SV according to venous return.21 A volume load will increase SV significantly when preload is low, but not if preload is already normal or high. With progressive hypovolaemia, the heart`s operating point moves to the steeper part of the Frank-Starling curve, where a small increase in preload leads to a large increase in SV (Figure 1). Preload dependency describes a heart where both ventricles operate on the steep part of the curve, and consequently will increase SV with preload increase.22 This mechanism enables the body to increase SV during times of stress, when venous reserves are mobilised by neuro-humoral mediated contraction of small veins.23

Figure 1: The Frank-Starling curve. The Frank-Starling or cardiac function curve illustrates how SV increases with increasing preload, given the same contractility and afterload. In states of reduced preload, the operating point of the Frank-Starling-curve is shifted to the steep part of the curve, where a small increase in preload leads to larger increase in SV. After volume expansion, the operating point shifts to the right, where an additional increase in preload only leads to a small increase in SV. In patients with right ventricular failure, SV may decrease with an additional increase in preload (blue line). The slope of the Frank-Starling curve is affected by ventricular contractility and afterload.21

The veins contain approximately 75% of the total blood volume.24 The pressure in the venous reservoir under flow conditions equals mean systemic filling pressure (Pmsf ). Experimentally,

Pmsf is measured as the equilibrated pressure in the vascular bed between aorta and the right

21 atrium at zero flow. Pmsf depends on stressed intravascular blood volume; the volume that creates distending pressure to the vascular walls, and on vascular compliance.25 Venous return

(VR), and thus right ventricular filling, is determined by the pressure gradient between Pmsf and right atrial pressure (RAP), divided by the resistance to venous return (RVR).21,26 22

Thus, the venous return curve reflects both blood volume and vasomotor tone (Figure 2).

RVR is calculated as the gradient for venous return divided by cardiac output.

Figure 2: The curve of venous return. Venous return is determined by mean systemic filling pressure (Pmsf) and the pressure opposing venous return; right atrial pressure (RAP). The slope of the venous return curve represents resistance to venous return.21 Venous return decreases as RAP increases, as the gradient for venous return decreases. At zero flow, RAP equals Pmsf.

In steady state, venous return equals cardiac output. The interaction between venous return and cardiac function is reflected in RAP, which normally equals central venous pressure

(CVP).27,28 RAP represents the backpressure to venous return as well as the filling pressure of the right ventricle, which is a major determinant for SV (Figure 3).27

Figure 3: The Frank-Starling curve and the curve for venous return.

The Frank-Starling curve combined with the curve for venous return, as suggested by Guyton29, illustrates the relationship between venous return and cardiac output. The intersection of the two curves determines cardiac output. A volume load shifts the blue venous return curve upwards, as mean systemic filling pressure (Pmsf) and thus the gradient for venous return increases. In a non-responsive heart, the gradient may remain unchanged, because the heart is unable to translate the increase in preload into an increase in SV, and RAP increases

25 instead. Haemorrhage shifts the venous return curve downwards, as Pmsf and the gradient for venous return decreases. The red cardiac function curve is shifted upwards by increased sympathetic stimulation, and downwards with myocardial damage.21

This relationship between Pmsf, RAP and cardiac output was originally suggested by Guyton, based on experiments on dogs.29 Guyton`s model has been criticised for being too theoretical, as it doesn`t consider compensatory mechanisms that may occur in the regulation of human circulation. However, his theories have also largely been confirmed in later studies on humans.25,30

5.4 Haemodynamic effects of respiration

Air flow to the lungs is driven by changes in transpulmonary pressure; the pressure gradient between intra-alveolar and pleural pressure. During spontaneous respiration this gradient increases, as pleural pressure decreases due to expansion of the chest wall. During mechanical ventilation, gas insufflation leads to an increase in alveolar pressure which outweighs the concomitant increase in pleural pressure. The respiratory induced changes in intrathoracic pressure are smaller during spontaneous breathing than during mechanical ventilation.31

Respiration causes changes in SV by influencing ventricular afterload as well as preload

(Table 1). Spontaneous inspiration induces a reduction in RAP. The pressure gradient between Pmsf and RAP increases, which in turn increases venous return and right ventricular preload. In a healthy heart this leads to an increase in SV, which is visible as maximal pulse pressure after a pulmonary transit time of 2-3 s. During mechanical ventilation, cyclic swings in SV are caused by several mechanisms.32 In the inspiratory phase, the gradient for venous return is reduced as RAP increases and vena cava is compressed by increased pleural pressure.

Right ventricular afterload increases due to increased intra-alveolar pressure, and left ventricular preload increases as blood is squeezed from the lung towards the left ventricle.

Left ventricular afterload is reduced due to the increase in pleural pressure which reduces the transmural pressure of the left ventricle.32,31 The reduction in right ventricular preload is visible on the left side as reduced SV and minimum pulse pressure a few heart beats later.

According to the Frank-Starling curve, these respiratory induced variations in SV and pulse pressure are larger in a preload responsive heart. This is the basis for the prediction of fluid responsiveness. 25

Both mechanical and spontaneous inspiration leads to an increase in right ventricular afterload due to lung inflation.27,32 Left ventricular afterload is higher during spontaneous inspiration because the heart, surrounded by negative pleural pressure, pumps against a larger pressure gradient. Mechanical ventilation reduces this transmural pressure, and thus left ventricular afterload. This effect does not outweigh the reduction in preload on SV, but may have clinical benefits in a failing heart. Thus, spontaneous and mechanical ventilation induce opposite effects on the ventricular in- and outflow, except for right ventricular afterload, which always increases during inspiration.27 During normal spontaneous breathing, the inspiratory increase in preload is offset by the increase in RV afterload, leading to smaller variations in venous return and SV. During mechanical ventilation, regular cyclic variations in intrathoracic pressure induce large variations in venous return and SV, which are augmented by the increase in RV afterload and the decrease in LV afterload.

Mechanical ventilation Spontaneous ventilation

Right ventricle Left ventricle Right ventricle Left ventricle

Preload ↓ ↓ ↑ ↑

Afterload ↑ ↓ ↑ ↑

Table 1: Pre- and afterload during mechanical and spontaneous respiration. During the mechanical inspiratory phase, left ventricular preload is reduced after the transient increase following insufflation of air which squeezes blood towards the left ventricle. Modified after Alian.33

5.5 Dynamic variables and the concept of fluid responsiveness

SVV and ∆PP are referred to as dynamic variables. The dynamic variables also include ∆POP, which refers to the respiratory induced variations in the photoplethysmographic waveform amplitude, and the automated pleth variability index, PVI. The dynamic variables have been thoroughly investigated in mechanically ventilated patients.34,35 While not traditionally referred to as dynamic variables, EtCO2 and VtCO2 are other functional measures of cardiac output changes that are based on heart-lung interactions.

The dynamic variables are used to predict or track fluid responsiveness. Only 50 % of patients receiving fluid loads in the ICU increase their cardiac output after filling.36 For the rest, excess fluid may lead to pulmonary and peripheral oedema and impaired gas exchange; thus increasing morbidity and mortality.37,38 To avoid harmful fluid loads, fluid responsiveness, i.e. preload dependency, is tested. A “fluid responder” is usually defined as a patient that increases SV ≥ 10 -15% after a fluid load of 250 - 500 ml.39,40 The changes in SV are measured directly, or estimated using surrogates such as dynamic variables and exhaled CO2.

Traditionally, the optimal baseline threshold value for the prediction is calculated based on the sum of maximal sensitivity and specificity, frequently between 11-13%.40,41 A “grey zone approach” with values between 9 and 13%, which allows for the uncertainty related to the measurements, has been suggested.42 In order for a variable to predict fluid responsiveness the increase in cardiac output should be associated with the baseline value of the variable, and the change should be associated with the change in cardiac output.43 The latter describes a variable`s ability to track SV changes, which is especially relevant for the detection of hypovolaemia. Preload dependency may also be revealed by other methods that shift the heart`s position on the Frank-Starling curve, such as an increase in positive end-expiratory pressure (PEEP). The PEEP-challenge has been used to demonstrate the predictive ability of 27

44 45 ∆PP in ARDS-patients , of changes in VCO2 in patients scheduled for cardiac surgery , and of respiratory variations in CVP (∆CVP) in patients during cardiac surgery.46

Dynamic variables reflect preload dependency rather than absolute preload, and have been found superior to static variables such as ventricular filling pressures or end-diastolic volumes for the prediction of fluid responsiveness.36,39,41 Preload is not equal to preload dependency due to two factors: the Frank-Starling-relationship depends on the contractility of the heart, and the relationship between preload and SV is not linear, but curvilinear.47 Hence, a specific preload value does not equal a certain SV, as its effect on SV is determined by the heart`s position on the Frank-Starling curve. The difference between static and dynamic variables may be illustrated by the CVP, which will often be low in responders and high in nonresponders, but shows no discriminative ability for values in between. By contrast, respiratory variations in CVP have been shown to predict fluid responsiveness with good sensitivity and specificity.46,48

Fluid responsiveness does not equal a need for fluid. For instance, healthy upright subjects are normally fluid responders, but do generally not require fluid expansion. Hence, a fluid

49 challenge should be triggered by other signs of inadequate perfusion or DO2.

5.6 Pulse pressure variations (∆PP)

Pulse pressure, the difference between systolic and diastolic blood pressure, is proportional to left ventricular SV. Thus, respiratory changes in SV are reflected in pulse pressure, and pulse pressure variations may act as a surrogate for SVV (Figure 4). ∆PP expresses the magnitude of the respiratory variations over one respiratory cycle, and is calculated as

41 %

where PPmax is the maximal and PPmin the minimal pulse pressure within one respiratory cycle.

∆PP is thoroughly investigated and validated during mechanical ventilation, provided certain conditions which are listed below.35 Pooled sensitivity of 88% and specificity of 89% for

∆PP`s ability to predict fluid responsiveness was found in a meta-analysis in 2014.35 Some studies have shown that ∆PP predicts fluid responsiveness also during one-lung ventilation50 and prone positioning51, but with other threshold values than during normal conditions (lower and higher, respectively).

∆PP depends on aortic compliance as well as left ventricular SV. When arterial tone is increased, ∆PP is increased relative to SVV.52 The ∆PP/SVV-ratio may thus be indicative of vascular tone, and has been shown to predict the need for norepinephrine.53 Arterial tone may be assumed to stay constant over one respiratory cycle, but increases with progressive hypovolaemia.

The performance of ∆PP in patients with spontaneous breathing activity has been disappointing in clinical studies, and it is generally assumed that pleural pressure swings during spontaneous respiration are insufficient to reveal preload dependency.34,54,55 However, some smaller studies show that ∆PP may reveal preload dependency also during spontaneous breathing, given certain conditions. Cyclic variations in venous return may be amplified by augmenting the increase in preload and SV with deep inspiration56, or the decrease in venous return and SV by forced expiration during the Valsalva manoeuvre.57 Slow patterned breathing has also been shown to improve the dynamic variables` ability to reveal preload dependency in experimental studies.58,59 29

Figure 4: Pulse pressure variations in a fluid responsive (A) and a fluid non-responsive (B) heart. A ventricle that operates on the flat part of the Frank-Starling curve is less sensitive to swings in intrathoracic pressure.

5.6.1 Considerations for the use ∆PP for haemodynamic evaluation

The use of ∆PP depends on certain conditions related to physiology and treatment that affect

SV and SVV. These also apply to other dynamic variables that reflect SVV, such as ∆POP.

Some of these conditions are challenged when dynamic variables are applied during spontaneous breathing.

5.6.1.1 Ventilation

Adequate swings in intrathoracic pressure are required to produce SVV, and ∆PP is validated for the prediction of fluid responsiveness in mechanically ventilated patients with tidal volumes ≥ 8ml/kg.34,35 ∆PP has been found to increase with increasing tidal volumes.60,61

When tidal volumes increase, right ventricular preload is reduced, shifting the heart`s position on the Frank-Starling curve to the left, where the heart is more sensitive to preload changes.32

However, protective ventilation strategies with tidal volumes ≤ 8 ml/kg have become the 30

norm.34,62 In patients ventilated with lower tidal volumes, low ∆PP-values may reflect inadequate changes in intrathoracic pressure rather than lack of fluid responsiveness.49,63 In these patients, procedures have been proposed which for a short period of time increase intrathoracic pressure swings, such as temporary increase in tidal volume62 or PEEP.46 The predictive properties of ∆PP during low tidal volume have been found to improve when ∆PP cut-off is set below the conventional 12%.34,63

It has been suggested that changes in venous return are induced by changes in intrathoracic pressure, rather than by tidal volumes as such.61,64 Hence, chest compliance affects dynamic variables. Decreased lung compliance reduces the transmission of airway pressure to pleural and pericardial pressures, and ∆PP becomes unreliable in this setting.65,66 However, ∆PP has been found to predict fluid responsiveness with excellent specificity but moderate sensitivity in ARDS-patients ventilated with low tidal volumes and high PEEP, as the swings induced by high PEEP compensate for the smaller swings induced by ventilation with smaller tidal volumes.47 Increased chest compliance reduces the inspiratory decrease in venous return and thus ∆PP. Thus, ∆PP is not validated during open chest conditions.67,68

High respiratory rate decreases SVV and ∆PP. Usually, the pulmonary transit time ensures that the maximum and minimum pulse pressure of both ventricles coincide to produce pulse pressure variations, with the minimum pulse pressure 2 -3 seconds after inspiration. An increase in respiratory rate with constant HR and pulmonary transit time leads to opposite effects on right and left ventricular SV, and SVV decreases. Thus, SVV and ∆PP are validated for HR/RR-ratio > 3.6.69

5.6.1.2 Intraabdominal hypertension

In patients with intra-abdominal hypertension, intra-abdominal pressure reaches almost the same level as intra-thoracic pressure during positive pressure inspiration. Venous return is reduced, and ∆PP may be low even if the patient is fluid responsive.37,39

5.6.1.3 Cardiac dysfunction

Right ventricular dysfunction and increased pulmonary artery pressure reduce the predictive ability of dynamic variables.39 High baseline ∆PP may be due to right ventricular dysfunction as well as preload responsiveness.70 Respiratory variations in right ventricular SV may be due to respiratory variations in pulmonary vascular resistance and not right ventricular preload, as the right ventricle is sensitive to changes in afterload. Preload increase in a dysfunctional right heart may lead to right heart failure.

Dynamic variables are also affected by irregular cardiac rhythm such as atrial fibrillation, because differences in filling time affect SV, and gives variations in SV and ∆PP that are not due to changes in venous return or afterload. ∆PP also has limited predictive value in patients with aortic stenosis.71

5.7 The photoplethysmographic waveform amplitude variation (∆POP), perfusion index (PI) and pleth variability index (PVI)

Most frequently used for oxygen saturation measurement, the photoplethysmographic signal also provides information about circulation.72-74 For transmission plethysmography, infrared light (940 m) is emitted from a light source and sent through the tissue of the finger or ear.75

The light detected on the other side is processed into a signal. The variability of this signal depends on the absorption of light from arterial blood, but also from tissue, bone and skin. As blood absorbs infrared light, more light is absorbed during the systole which in the original 32

signal represents the trough, whereas the peak represents the diastole. The signal is then flipped so that the systole is reflected in the peaks, analogue to the arterial blood pressure. In addition to the cardiac component which is believed to reflect pulse synchronous arterial blood and thus left ventricular SV, the photoplethysmographic signal consists of a baseline component which reflects slower respiratory induced changes in venous blood.76

The respiratory variation in the photoplethysmographic amplitude within one respiratory cycle is described as ∆POP. Analogue to ∆PP, ∆POP is calculated by the formula

where POPmax is the maximum and POPmin the minimum amplitude within one respiratory cycle.

Good ability to predict fluid responsiveness and good correlations with ∆PP have been found for ∆POP in mechanically ventilated patients.77,78 The use of ∆POP to reveal preload dependency is based on its ability to reflect SVV. As for SVV and ∆PP, sufficient intrathoracic pressure swings are required. Thus, in spontaneously breathing subjects primarily other aspects of the signal have been explored. The cardiac amplitude is related to

SV, and reductions in amplitude height, width and AUC have been demonstrated in mild hypovolaemia.33,79-81

The perfusion index expresses the ratio between the pulse synchronous (alternating current;

AC) and constant (direct current; DC) light absorption, or the relationship between pulsatile and non-pulsatile blood volume in the photoplethysmographic signal. It is calculated as

PI =

The PI has been shown to decrease with hypovolaemia82 and increase during a passive leg raise manoeuvre.83 Hence, PI reflects volume changes, but not based on respiratory variations like ∆PP, ∆POP and PVI. The pleth variability index, a commercially available and automated algorithm for the detection of respiratory changes in the perfusion index84, is calculated using the formula

where PImax is the maximal and PImin the minimal perfusion index.

Like ∆POP, PVI has mostly been investigated in mechanically ventilated patients. PVI predicts fluid responsiveness reasonably well, but appears inferior to ∆PP in tracking volume changes during mechanical ventilation.53,85 In spontaneously breathing subjects, ∆POP and

PVI have been found to track cardiac output changes during a PLR-manoeuver, but unable to reliably predict an increase in cardiac output.86-88 PVI was also found to reflect hypovolaemia induced by LBNP on a group level, but not in individual subjects, due to considerable overlap of PVI –values between LBNP-levels.89 PVI has also been found to predict changes in blood pressure following administration of dexmedetomidine90 and spinal anesthesia.91

5.7.1 Limitations to the use of ∆POP, PVI and PI for haemodynamic evaluation

As ∆POP mainly reflects changes in SV, it shares the same considerations as SVV and ∆PP regarding mechanical ventilation, tidal volume, HR/RR-ratio and closed chest conditions.

However, the photoplethysmographic signal is more complex, and influenced by other factors in addition to these.

5.7.1.1 Vascular tone

∆POP is affected by changes in SV, but also by changes in vascular tone and venous blood volume.92 Vasoactive medications and increased sympathetic nerve activity induced by pain or stress may lead to a reduction in the photoplethysmographic amplitude, especially in the finger signal, and the finger plethysmogram has thus been proposed as a monitor of vasomotor tone.75,93 Although a decrease of the amplitude due to increased sympathetic activity may be the consequence of reduced cardiac output, it is unspecific, and thus regarded as disturbance when the signal is used for the evaluation of SV changes.

5.7.1.2 Processing algorithms

Processing algorithms may alter qualities of the signal that are relevant for haemodynamic evaluation. The “auto-centering” function (high-pass filter) eliminates parts of the signal resulting from light absorption by venous blood and surrounding tissue (low frequency waves).94 The autogain or autoscale function adjusts the size of the signal for clinical use, and must be deactivated in order to observe changes in the amplitude on the screen.84 The signal processing algorithms may differ between manufacturers, and the variability of ∆POP has been found to depend on the pulse oximetry device.95 Hence, several authors have concluded that the pulse oximeter can only be a reliable tool for research purposes when the algorithms are open, the raw signal provided, and the data collected over a sufficient amount of time and presented in a standardised manner.96-98

The correlation between ∆POP and ∆PP has been shown to improve with advanced signal processing.99 Analysis of several different features of the photoplethysmographic amplitude by an automated algorithm was able to discriminate between euvolaemia and hypovolaemia with good sensitivity and specificity in a population of spontaneously breathing and mechanically ventilated trauma patients.7 35

5.8 End-tidal (EtCO2) and volumetric (VtCO2) carbon dioxide

CO2 is produced in the tissues, transported with the blood and eliminated in the lungs. Thus,

100,101 exhaled CO2 depends on metabolism, pulmonary flow and ventilation. If metabolism and ventilation are constant, changes in exhaled CO2 reflect changes in pulmonary flow.

Minor shunting may be regarded as negligible in this context, so that pulmonary flow may be considered equal to cardiac output.

Measurement of CO2 in the expired air is referred to as capnography. Both EtCO2, which measures end-expiratory CO2-concentration, and VCO2 or VtCO2, which measure eliminated

CO2 in ml per minute or per tidal volume, may reflect changes in cardiac output. Whereas

∆PP and ∆POP reflect variability in SV, EtCO2 and VtCO2 reflect reductions or increases in pulmonary flow. Analogue to ∆PP, a given value does not equal a given amount of cardiac output, as the relationship between cardiac output and expiratory CO2 differs between patients and within patients over time.

EtCO2 has most frequently been studied in relation to cardiopulmonary resuscitation, where it marks the return of spontaneous circulation after cardiac arrest102 and reflects the quality of

103 104,105 106,107 heart compressions. Both animal and human studies show that EtCO2 reflect changes in cardiac output. In recent years, EtCO2 has been investigated as a surrogate for SV when testing for fluid responsiveness. An increase in EtCO2 > 5% or 2 mmHg during PLR or a fluid load has been shown to accurately predict fluid responsiveness.101,108-110

EtCO2 is easily acquired by side-stream or mainstream capnography, whereas VCO2/VtCO2- calculation requires real-time integration of flow and CO2-curves. Whereas eliminated CO2 based on the method of partial rebreathing or pulmonary capnotracking for the continuous 36

evaluation of cardiac output has been extensively studied, only a few studies have investigated VCO2/VtCO2 with regard to its ability to reflect changes in pulmonary flow, and

106,111,112 even fewer have directly compared VtCO2 and EtCO2 in this context. VtCO2 is calculated from the area under the curve of the volumetric capnogram (Figure 5)113, and

VCO2 is VtCO2 multiplied with respiratory rate. VCO2 has been shown to reflect cardiac output changes and predict fluid responsiveness, and some studies indicate that VCO2 may

45,112 reveal preload dependency with higher sensitivity and specificity than EtCO2. The volumetric capnogram further allows estimation of dead space and the effects of recruitment, and may be used to optimise PEEP.113 When pulmonary flow and ventilation are constant,

114 VCO2 equals CO2-production, and may thus be a measure of metabolism. These features may be of particular interest in the ICU.

Figure 5: The volumetric capnogram, where CO2 is plotted against tidal volume. The blue area indicates eliminated CO2 pr breath; VtCO2. End-tidal CO2-concentration is measured at the highest point of the slope. 37

A reduction in cardiac output leads to a reduction in expired CO2 by two mechanisms: firstly, less CO2 is brought to the lungs. Secondly, a reduction in cardiac output will lead to under-

105,111 perfused lung areas and increased dead space, from which no CO2 will be excreted. The reduction in VCO2 and EtCO2 is transient. A reduction in transport capacity leads to storage of CO2 in the tissues. Eventually, increased venous CO2-content will lead to an increase in

106 CO2-elimination from the perfused alveoli, and a new steady state is reached. With an increase in cardiac output, previously under-perfused tissues are circulated and the CO2- transport to the lungs is improved, leading to a transient “overshoot” of exhaled CO2 before a new steady state is reached. Increased cardiac output may also increase exhaled CO2 by recruiting underperfused lung areas and improve the ventilation/perfusion-ratio. This effect is of high significance in hypovolaemic conditions, where ventilation/perfusion-ratio will improve significantly. In states of normovolaemia, the effect on exhaled CO2 is probably limited.

There is a close physiological relationship between pulmonary flow and exhaled CO2. With

VCO2, VtCO2 and EtCO2 readily available on modern ventilators (Figure 6), their value as non-invasive tools for haemodynamic monitoring in mechanically ventilated patients is of interest. 38

Figure 6: EtCO2, VCO2 and VtCO2 continuously displayed on a standard ventilator (Drӓger Infinity C500).

VCO2 is the product of VtCO2 and respiratory rate.

5.8.1 Limitations of EtCO2 and VtCO2 for haemodynamic evaluation

The detection of cardiac output changes based on exhaled CO2 relies on stable metabolism and ventilation. It has so far only been validated in mechanically ventilated patients. A recent study concluded that EtCO2 did not reliably predict fluid responsiveness in spontaneously breathing healthy volunteers undergoing PLR, although sensitivity was high.115

Pulmonary disease, shunting and increased dead space represent limitations to the method, although reports differ.45,112 A change in ventilation/perfusion-ratio alters the relationship between cardiac output and exhaled CO2. Obstructive diseases affect the slope of the capnogram, and thus EtCO2 in particular, as EtCO2 corresponds to the end of the plateau of the curve. 39

6 AIM AND RESEARCH QUESTIONS

The research questions in this thesis aimed to illuminate whether heart-lung-interactions may be non-invasive means of detecting blood volume changes in patients with and without mechanical ventilation. The specific research questions were as follows:

- Do ∆PP, ∆POP, PVI and PI reflect hypovolaemia during spontaneous

breathing and non-invasive positive pressure ventilation (NPPV)?

- Do VtCO2 and EtCO2 track changes in cardiac output in mechanically

ventilated patients, and do the changes in VtCO2 and EtCO2 correspond with

the changes in cardiac output?

- Do ∆PP and ∆POP reflect hypovolaemia in spontaneously breathing subjects

during positive expiratory pressure (PEP) and continuous positive airway

pressure (CPAP)?

7 MATERIAL AND METHODS

7.1 Study populations

The studies for paper I and III were conducted on healthy young adult volunteers (≥18 years) in a clinical circulation and research laboratory at Oslo University Hospital, Aker. These study populations consisted mainly of physically active students; recruited with a non- probability sampling method.116 For paper I, 16 patients were included, and 14 studied. Two subjects were unable to comply with NPPV, and had to be excluded. For paper III, 20 subjects were included and studied. The subjects practised breathing through the facemask and the resistors before start. During both studies, several of the subjects showed signs of discomfort and impending circulatory collapse, and the protocol had to be interrupted before completion of all LBNP-levels. This has been described previously, and the level and time at which the individual subject experiences cardiovascular collapse has been referred to as “maximal

LBNP-tolerance”.79

Study population for paper II consisted of adult patients (≥18 years) scheduled for open coronary artery bypass surgery or aortic valve replacement at Oslo University Hospital,

Ullevål. 40 patients were included, 33 studied. Seven patients were included, but not studied, due to changes in operative schedule or complications during or after surgery (bleeding, pacemaker dependency). The subjects were included consecutively according to the presence of the first author (non-probability convenience sampling).116

7.2 Lower Body Negative Pressure (LBNP)

LBNP induces central hypovolaemia and may be used to investigate cardiovascular and autonomic responses in conscious humans. LBNP is applied by placing the lower body in a vacuum chamber which is sealed around the subject`s waist (Figure 7).117 Negative pressure is 41

induced via suction at the lower end of the chamber, inducing incremental levels of vacuum in the chamber which in turn leads to a sequestering of blood in the lower abdomen and extremities.118 The method is validated in a number of studies as a reliable method of reducing preload.118-120 LBNP has been found to induce cardiovascular changes comparable to those that occur during haemorrhage. A study comparing primates during both haemorrhage and LBNP with humans during LBNP showed that in a 70 kg person, LBNP-levels of 30, 60 and 90 mmHg equalled blood losses of 450, 1000 and 1600 ml, respectively.121 The LBNP- method is considered safe in healthy volunteers because the changes may be reversed immediately by relieving the negative pressure, thus restoring central blood volume. This is also an indirect test of fluid responsiveness, as termination of the negative pressure leads to normalisation of cardiac output.

The response to LBNP varies, and may depend on gender and age among other factors. The

LBNP-model has been shown to induce different changes in peripheral vascular resistance and redistribution of blood volume in men and women.122 Responses may also differ according to the different LBNP-protocols, which describe level and duration of the negative pressure applied. The term LBNP-tolerance refers to the individual differences in response to

LBNP, i.e. time to presyncope, and is a function of LBNP-level and time.119 42

Figure 7: Subject placed in the lower body negative pressure chamber during non-invasive positive pressure ventilation. Cardiac output is measured with suprasternal Doppler.

7.3 Right ventricular pacing (RVP)

In study II, cardiac output was reduced by right ventricular pacing (Medtronic 5388 Dual

Chamber Temporary Pacemaker, Medtronic, Minneapolis, USA). Temporary right ventricular pacing leads are routinely applied after cardiac surgery and endovascular procedures such as transcatheter aortic valve implantations (TAVI), as AV-blocks or bradyarrhythmias may occur. Reducing cardiac output is relevant mostly for research purposes. Previously published models for reduction of cardiac output in humans have mainly been aimed at reducing preload, and include head-up-tilt59,123, blood withdrawal for haemodilution124 and LBNP.79,81 When applied to sinus rhythm, right ventricular pacing induces reductions in SV of approximately

20% due to loss of atrial contribution125 and dyssynchrony.126 This equals the reduction in SV 43

during moderate hypovolaemia. However, the impact of RVP can be larger in patients with systolic dysfunction (e.g. reduced ejection fraction due to myocardial infarction) or diastolic dysfunction due to hypertension or aortic stenosis. We are not familiar with previous use of right ventricular pacing to induce rapid and transient reductions in cardiac output for the evaluation of haemodynamic monitoring equipment.

7.4 Passive leg raise (PLR)

Passive leg raise was used to increase preload in study II. During passive leg raise, the patient changes his position from semirecumbent with the upper body raised 45° degrees and flat legs, to a position with a flat back and legs elevated 45°. This simulates a fluid load of approximately 300 ml in an adult of 70 kg.127 Provided that both ventricles operate on the steep part of the Frank-Starling curve, increased preload will lead to an increase in SV.101 The increase occurs within the first minute of the manoeuvre. The test has been validated in numerous studies. A meta-analysis concluded with a pooled sensitivity of 85% and specificity of 91% with a threshold of 10% increase in cardiac output.128

The effect of PLR must be measured by monitoring cardiac output or surrogates such as

108,112 EtCO2 and VtCO2. ∆PP predicts fluid responsiveness in the PLR-model, but specificity and especially sensitivity is lower than when cardiac output is measured directly.128 The use of arterial pressure changes to evaluate the effect of PLR is not recommended, as sensitivity is low, even if specificity is good.128,129 The main advantage of the passive leg raise method is its applicability to spontaneously breathing patients with low and irregular tidal volumes, as well as to patients on mechanical ventilation.130 PLR may also be used in patients with arrhythmia and low pulmonary compliance.101 It has been suggested that the PLR manoeuvre improves ventilation/perfusion-ratio by reducing West zone I-conditions in ARDS-patients 44

ventilated with high PEEP.131 As PLR represents an endogenous and reversible fluid load, any potentially negative effects of an actual fluid challenge are avoided.

The method has few, but important limitations. Elevated intracranial pressure, instability in the columna and deep vein thrombosis in the lower extremities are considered contraindications to the method. Although debated127, it has also been suggested that the method is less accurate in patients with abdominal hypertension, as increased abdominal pressure may affect mobilisation of venous reserves.132,133

7.5 Modification of respiration: Non-invasive positive pressure ventilation (NPPV), continuous positive airway pressure (CPAP) and positive expiratory pressure (PEP)

In study I, NPPV was applied in the intermittent positive pressure ventilation (IPPV) mode

(Evita 4, Drӓger Medizintechnik, Lübeck, Germany). This is a volume controlled mode with continuous mandatory ventilation (VC – CMV). During IPPV, a volume target is set along with PEEP, FiO2, flow rate and respiratory rate. A pressure limit may also be set to avoid over-inflation. There is no synchronisation with spontaneous breathing efforts. The intrathoracic pressure changes and heart-lung-interactions during non-invasive IPPV should be similar to those during invasive mechanical ventilation, provided no leakage from the mask and the absence of spontaneous breathing activity.

In study III, different levels of continuous positive airway pressure (CPAP) and positive expiratory pressure (PEP) were applied. CPAP (Evita 4, Drӓger Medizintechnik, Lübeck,

Germany) provides positive pressure throughout the respiratory cycle. It alleviates respiratory work and prevents expiratory collapse of minor airways, thus increasing functional residual capacity. Clinically, it is used to improve oxygenation preoperatively or during respiratory distress, and to prevent atelectasis postoperatively and during intensive care. By increasing 45

intrathoracic pressure, CPAP also reduces left ventricular afterload analogue to mechanical positive pressure ventilation. CPAP levels of 0 - 10 cm H2O are common in clinical use.

The application of PEP increases intrathoracic pressure during expiration only. The positive pressure is a product of air flow and resistance, and depends on patient effort. Like CPAP,

PEP may prevent atelectasis and the development of pneumonia postoperatively by keeping distal airways open during expiration. In study III we used a PEP mouthpiece (Armstrong

Medical Ltd, Coleraine, Northern Ireland) with two different resistors to amplify intrathoracic pressure swings. Resistance levels of 5 and 10 cmH2O were chosen because they are frequently used in clinical practice, and readily accepted by most patients.

7.6 Cardiac output measurements

In papers I and III, SV was measured by an experienced operator with a handheld 2 MHz pulsed wave Doppler probe placed in the suprasternal notch. It calculates SV from blood flow velocity in the ascending aorta, 1-2 cm above the aortic valve. An angle of insonation of 20° and an aortic diameter of 20 mm is assumed.134 The oesophageal Doppler (OD) used in paper

II measures blood flow velocity in the descending aorta by a continuous 4 MHz probe (DP-12 probe; Cardio Q; Deltex Medical, Chichester, UK). SV is calculated by multiplying the distance travelled by the blood during systole (velocity time integral; VTI) with aortic cross- sectional area.135 The probe is easily inserted into the oesophagus of anaesthetised intubated patients. Although operator dependent, studies have shown low intra- and inter-observer variability after a limited amount of practice.136 OD measures flow continuously, and is suitable for monitoring rapid and transient cardiac output changes.11 The method agrees well with thermodilution.137 OD reliably tracks changes in cardiac output during surgery and critical illness137, and haemodynamic optimisation using OD has been shown to reduce postoperative complications and hospital stay after major surgery.138 Contraindications for the 46

use of oesophageal Doppler include oesophageal pathology or anatomical anomalies (e.g. after oesophageal surgery).

The measurement of cardiac output using OD is based on several assumptions.135 Firstly, the distribution of blood to the upper (brachiocephalic and coronary arteries, 30%) and lower

(70%) body is assumed to be constant. This assumption is reasonable in stable anaesthetised patients, but questionable during shock and critical illness.11 Secondly, the aortic diameter is not measured, but estimated based on normograms from height, weight and age, and assumed constant. Thirdly, the VTI is dependent on the insonation angle of the Doppler beam to the blood flow which is assumed constant at 45°. A violation of these assumptions may lead to significant measurement errors when values are expressed in absolute terms.135 When values are given as relative changes from baseline such as in paper II, only a change in the assumed variables during interventions will affect the results.

7.7 The volume-clamp method for blood pressure measurement

In study I and III, continuous blood pressure was measured non-invasively using the volume- clamp method (Finapres 2300, Ohmeda, Madison, WI, USA). The volume clamp method is based on the photoplethysmographic technology. A pneumatic finger cuff applies sufficient pressure to keep blood volume in the finger arteries constant throughout the cardiac cycle.

The required pressure equals intra-arterial pressure.139 Good agreement between pulse pressure variations calculated from intra-arterial and non-invasive blood pressure curves has been found postoperatively.140 Cardiac output measurements based on pulse contour analysis of the non-invasive curve have also been shown to agree well with transpulmonary thermodilution141 and oesophageal Doppler142 perioperatively. During intensive care the results are less convincing, which has been explained with increased vasomotor tone in these patients.139 This is a limitation of the photoplethysmographic technology which also manifests 47

itself when the photoplethysmographic waveform is used for the calculation of ∆POP.93

Difficulties obtaining a signal during hypoperfusion represents another limitation of the volume clamp method.139

7.8 Sampling of haemodynamic data

Study I and III was performed in a physiological research laboratory. Haemodynamic data including blood pressure waveforms were sampled from analogue outputs of the monitoring equipment at 400 Hz in study I and 300 Hz in study III to custom-made software (Regist 3;

Morten Eriksen, University of Oslo, Oslo, Norway). Photoplethysmographic waveforms were exported from the analogue output of the pulse oximeter at 400 Hz to Signal Express Version

14.0.0 (National Instruments, Austin, Texas, USA). Data was converted in an analogue-to- digital-converter (NIDAQPad-6015, National Instruments) before storage and further analysis.

High frequency sampling ensured that all troughs and peaks were included.

Study II was performed in a clinical setting in a cardiothoracic recovery unit. Haemodynamic data was sampled from analogue outputs of standard clinical monitoring equipment (GE Solar

8000i, GE Healthcare, Chicago, Illinois, USA) at 400 Hz to SignalExpress 14.0.0 after conversion in an analogue-to-digital-converter.

7.9 Calculation of respiratory variations in pulse pressure (∆PP) and the photoplethysmographic waveform amplitude (∆POP)

The photoplethysmographic waveforms were derived from a commercial pulse oximeter

(Masimo Radical 7, Version 7.3.1.1, Masimo Corp, Irvine, CA, USA) in both study I and study III.

For study I, ∆PP and ∆POP were calculated from the blood pressure and photoplethysmographic waveforms in a custom made program in LabView (National 48

Instruments, Austin, TX, USA). Each respiratory cycle based on thoracic impedance from the

ECG-leads was manually delimitated, and the maximal and minimal amplitude of the blood pressure and photoplethysmographic waveform within that respiratory cycle identified by the program. These values were then inspected and manually accepted.

For study III, the respiratory, arterial and photoplethysmographic curves were reconstructed in

R (R Foundation for Statistical Computing, Vienna, Austria). The maximum and minimum amplitudes within one respiratory cycle were identified and ∆PP and ∆POP calculated using different “packages” in R (Figure 8). Again, the curves and resulting values were individually inspected before they were accepted for further analyses.

Figure 8. Calculation of ∆PP (previous page) and ∆POP (this page) at LBNP 60 mmHg for paper III. The last four respiratory cycles (thin grey line) represent breathing after the application of PEP=10. Large green dots indicate the delimitation of respiratory cycles. Minimal and maximal PP- and POP-values within one cycle were used for the calculation of ∆PP and ∆POP.

7.10 Calculation of end-tidal (EtCO2) and volumetric (VtCO2) carbon dioxide

In study II, EtCO2 was sampled from the analogue output of a side-stream capnograph at 400

Hz and converted to a digital format for further analyses. The calculation of eliminated CO2 requires simultaneous measurements of flow and EtCO2. VtCO2 was calculated in a custom made program in LabView by aligning mainstream flow-curves and side-stream EtCO2- curves, and integrating their product over time (Figure 9). As VtCO2 is calculated from

EtCO2-values, the two variables are not unrelated. Side-stream measurements are more susceptible to measurement errors caused by dampening of the flow- or CO2-curves than

143 mainstream devices. Modern ventilators can measure CO2 as well as flow proximally, 50

which reduces the risk of inaccuracies due to long sampling tubes and alignment difficulties.

Figure 9: Calculation of VtCO2 from CO2- and flow-curves (green and pink, respectively). The alignment of

CO2- and flow curves results in a new curve (white), from under which the integral giving VtCO2 is calculated.

Yellow curve is ECG, and red curve arterial blood pressure.

7.11 Statistical methods

7.11.1 Linear mixed models

Most traditional statistical models, for instance linear regression and AN(C)OVA, are based on the assumption of independent observations. However, when several observations are obtained from the same person at different points in time (repeated measurements), this assumption is violated. The observations are no longer independent, as the observed value in one individual at one point will not be independent of the value observed at a later time. This 51

was of particular concern in studies I and III, where multiple measurements were made at successive LBNP-levels.

Linear mixed models (LMM) is a framework of statistical methods that handle dependent observations, such as repeated measurements.144 LMM consider the fact that the errors of the repeated observations in the same subjects are not independent, and the correlation between them is included in the model. In order to decide which model to use for the analysis, the hierarchy of levels that may affect outcome must be considered. Different types of predictor variables are reported as fixed or random effects. Fixed effects refer to the explanatory variables that typically affect population means, which are normally the factors of interest.

Random effects are levels that are independent of the explanatory variable, and report the variability between measurement units. Subjects may be one such unit, and differences between subjects may be regarded as random variation. Hence, subject effects are typically random.144,145

Another advantage of mixed models is that they maximise power of incomplete data sets.

Rather than performing complete case analysis, discarding the other observations if one data point is missing, LMM use all available observations. This was important in study I with only

14 subjects, of which not everyone completed the LBNP-protocol.

A generalised version of mixed models may be used for non-normally distributed data.145

7.11.2 Correlation

Correlation is a mathematical measure of the co-variation between two variables, most commonly used to assess the association between continuous variables.146 Pearson`s correlation is used to investigate linear associations of normally distributed data. Spearman`s 52

correlation assesses the association between the rank of the observations, and is used for non- normally distributed data. Spearman`s correlation also handles non-linear associations. Perfect positive association equals r = 1 (for Spearman`s rho: ρ) and perfect negative association r = -

1. No association equals r = 0. Correlation alone does not allow inference of causation, as other factors may contribute to the association. The correlation coefficient expresses to what extent the change in one variable is explained by a change in the other. This can be calculated as ρ2 x 100.146

7.11.3 Receiver operating characteristics plot (ROC-plot)

A variable`s ability to discriminate between two conditions may be expressed in a sensitivity- specificity-analysis, and graphically displayed in a ROC-plot.147 The predictive variable is continuous, and the outcome must be dichotomous or dichotomised, for instance into groups of preload responsive and preload unresponsive. The true positive rate (y-axis; sensitivity) is plotted against the false positive rate (x-axis; 1 - specificity). The area under the resulting curve shows the test`s ability to discriminate between two conditions. If several variables are tested, AUC-values are used to rank their respective abilities as predictive tools. An AUC- value of 0.5 indicates that the variable has no discriminative ability, whereas an AUC-value of

1 equals perfect discriminative ability. The optimal threshold value may be determined by

Youdens index (J = sensitivity + specificity -1 = sensitivity - false positive rate). Likelihood ratios, which compare the ratios between the probabilities of a given test result in patients who do and do not have a certain condition146, and positive and negative predictive values may be calculated. As not all conditions are easily divided into two groups by one cut-off value, a “grey zone” between two cut-off values has been suggested.42 The lower cut-off value may be used to exclude the outcome, and the higher to include the outcome, whereas 53

the values in the middle are considered unable to diagnose an outcome with sufficient certainty.

8 RESULTS

8.1 Study I

The aim of study I was to investigate to what extent ∆PP, ∆POP, PVI and PI reflect hypovolaemia during spontaneous breathing and NPPV. ∆PP increased significantly with increasing LBNP-level during spontaneous breathing (1.2%, 95% CI 0.5 to 1.8, p < 0.001), and this increase was nearly doubled by the interaction between LBNP-level and NPPV (1.0%,

95% CI 0.1 to 1.9, p = 0.033). There was no association between LBNP-level and the photoplethysmographic variables ∆POP and PVI, regardless of respiratory mode. PI decreased significantly with progressive hypovolaemia (-0.1%, 95% CI -0.4 to 0.0, p = 0.018), but was not affected by the application of NPPV.

8.2 Study II

The aim of study II was to investigate whether exhaled CO2 measured as EtCO2 or VtCO2 reflect sudden moderate changes in cardiac output. Both EtCO2 and VtCO2 consistently tracked cardiac output changes during both reduction and increase in cardiac output. The change in VtCO2 for a given change in cardiac output was substantially larger than the corresponding change in EtCO2. However, the changes between VtCO2, EtCO2 and cardiac output were only correlated during RVP when cardiac output was reduced, and the correlations were modest (ρ = 0.53; p = 0.002 for VtCO2 and ρ = 0.47; p = 0.006 for EtCO2).

A ROC curve analysis revealed discriminative ability for EtCO2 only, and only for the reduction in cardiac output, with an AUC of 0.80 (95% CI 0.62 to 0.92) versus 0.68 (95% CI

0.50 to 0.83) for VtCO2. During PLR, the changes in VtCO2 and EtCO2 were not significantly correlated with the change in cardiac output, and none of the variables had AUC-values significantly different from 0.5. 55

8.3 Study III

The aim of study III was to explore whether the application of different levels of PEP and

CPAP to spontaneously breathing subjects would affect the ability of ∆PP and ∆POP to reflect progressive hypovolaemia. ∆PP increased significantly with progressive LBNP-level, during all values of PEP and CPAP. The increase (slope) in ∆PP was additionally affected by the transition from PEP = 0 (baseline) to PEP = 10 (loge (∆PP) = 0.082, 95% CI 0.006 to 0.16, p = 0.028). ∆POP increased significantly with progressive LBNP-level during PEP = 5 and

PEP = 10, but not during PEP = 0. The increase (slope) in ∆POP was not affected by any changes in PEP or CPAP. ∆CVP reflected LBNP-level during all values of PEP and CPAP.

The increase (slope) in ∆CVP was significantly affected by the transition from PEP = 0 to

PEP = 5 (∆CVP = 0.50 (95% CI 0.063 to 0.94, p = 0.018) and from PEP = 0 to PEP = 10

(∆CVP = 0.63 (95% CI 0.20 to 1.1, p = 0.001). CPAP-level did not affect any of the variables.

9 DISCUSSION

9.1 Main results

9.1.1 The ability of ∆PP, ∆POP, PVI and PI to reflect hypovolaemia during spontaneous

breathing and non-invasive positive pressure ventilation (NPPV)

- Do ∆PP, ∆POP, PVI and PI reflect hypovolaemia during spontaneous

breathing and non-invasive positive pressure ventilation (NPPV)?

In spontaneously breathing subjects, ∆PP has mostly been investigated in models where some amplification of intrathoracic pressure changes has been attempted. ∆POP has been investigated for the prediction of fluid responsiveness in the PLR-model, but during hypovolaemia primarily other features of the photoplethysmographic waveform such as amplitude height, width and area have been investigated. All variables related to the photoplethysmographic waveform (∆POP, PVI and PI) are discussed below, as PVI and PI were investigated in study I only. The performance of ∆PP will be described more detailed in the discussion of study III.

In study I, we found that ∆PP increased with increasing LBNP-level during spontaneous breathing, and that this increase was amplified by the application of NPPV. There was no association between ∆POP or PVI and LBNP-level, and in contrast to ∆PP, neither ∆POP nor

PVI were affected by the application of NPPV. PI reflected LBNP-level, but was not affected by NPPV.

The fact that NPPV augmented the reduction in SV and the increase in ∆PP induced by LBNP indicates that NPPV did induce intrathoracic pressure changes, and altered SV by reducing venous return and increasing right ventricular afterload. This finding is interesting and may have potential clinical implications, as NPPV is frequently used in critically ill patients. 57

NPPV does not require the same amount of respiratory strength as other methods which are used to modify intrathoracic pressure in spontaneously breathing, such as forced breathing manoeuvres. Importantly, tidal volumes were high during NPPV, which probably contributed to the significant effect of positive pressure ventilation as such. NPPV affected ∆PP only in the presence of hypovolaemia, meaning that the subjects had to be hypovolaemic in order for

NPPV to affect ∆PP.

We also observed a significant association between ∆PP and LBNP-level during unmodified spontaneous breathing. That indicates an amplification of intrathoracic pressure oscillations also in the absence of NPPV, which may be related to the LBNP-model as such. In conscious humans, a reduction in preload may lead to compensatory deep inspiration, which both improves right ventricular filling and increases tidal volumes.119 Similar effects are observed when respiratory rate is reduced, which is a previously tested model of amplifying pleural pressure oscillations in spontaneously breathing.58,59 Slower respiratory rate induces a longer inspiratory phase, which in spontaneously breathing subjects increases tidal volumes and allows increased right ventricular filling. It has also been suggested that tidal volumes increase as a compensatory response to maintain minute ventilation.58 In a small study on healthy volunteers undergoing LBNP, a reduction in respiratory rate increased SV variations and ∆PP. ∆PP increased with increasing LBNP-level also before the reduction in respiratory rate, but variations were smaller.58 These compensatory mechanisms may explain the significant association between ∆PP and LBNP-level during otherwise unmodified spontaneous breathing, which was also found in study III (during PEP and CPAP = 0).

As ∆POP and PVI, like ∆PP, depend on changes in preload, we could have expected these variables to reflect LBNP-level like ∆PP did. There may be several reasons why ∆POP and

PVI were not associated with LBNP-level. Firstly, whereas ∆POP and PVI depend on changes 58

in left ventricular SV analogue to ∆PP, they are to a larger extent determined by other factors as well, such as vascular tone. Sympathetic activity increases during hypovolaemia and may affect the magnitude of ∆POP. The finger signal is especially sensitive to vasomotion because acral skin has arteriovenous anastomoses that are sympathetically innervated.148 Hence, ∆POP and PVI perform better in the absence of stimuli such as pain, stress or hypovolaemia. Good agreement between ∆POP and ∆PP has been found in anesthetised patients before or after surgery.149,150 During open abdominal surgery, the correlations between ∆PP and ∆POP, and

∆POPs ability to predict fluid responsiveness have been rather poor, with large variability for

∆POP.151-153 Large variability for ∆POP has also been found in ICU-patients93, and for PVI in a previous study on spontaneously breathing healthy volunteers during LBNP. In that study,

PVI reflected hypovolaemia on a group level, but could not discriminate between normo- and hypovolaemia in the individual, due to considerable overlap.89 PVI was also largely unaffected by increases in tidal volume and the application of PEEP, separately or in combination. PVI is however a more highly processed variable than ∆POP, and has demonstrated some ability to detect circulatory changes during spontaneous breathing.88,91

Secondly, the amplitude, and thus ∆POP, is affected by slower respiratory induced changes in the baseline of the photoplethysmographic signal.154 Studies investigating whether these baseline variations alone reflect hypovolaemia have conflicting results.80,84,94

Thirdly, our study had limited power. Whereas sample size was sufficient to show an association between LBNP-level and ∆PP, it was possibly not sufficient to exclude an association with ∆POP or PVI. ∆POP and PVI increased with increasing LBNP-levels, but with large confidence intervals. As the variability in ∆POP and PVI is higher than for ∆PP

89,93,124, a larger sample size is required to identify an association between LBNP and ∆POP. 59

Previous studies on the photoplethysmographic waveform during spontaneous breathing have primarily focused on other aspects of the signal. Amplitude height, width and area have been found to reflect small preload changes at an early stage, while blood pressure and heart rate remained normal.33,79-81 Regional differences in these variables have been found. In healthy subjects undergoing progressive LBNP, the finger photoplethysmogram showed early reductions in amplitude height, width and area, whereas the ear photoplethysmogram only changed after significant hypovolaemia.81 Whereas the amplitude reductions in ear and forehead signals primarily reflect the reduction in SV, it has been suggested that the reduction in finger signal amplitude is more a measure of vasomotion than of SV reduction, and reflects the vasoconstriction that occurs during mild hypovolaemia.72,75,79 When investigating ∆POP and PVI for haemodynamic evaluation, the sympathetic reduction in finger amplitude represents noise, and ∆POP and PVI are perhaps better obtained from the ear or forehead signal.85

PI was significantly associated with LBNP-level, but not ventilation mode. This is in line with previous studies, where PI has been found to reflect reduced perfusion and increased sympathetic activity induced by hypovolaemia.82,155 Recently, the change in the PI during

PLR was found to predict fluid responsiveness with high sensitivity and specificity in spontaneously breathing patients.83 As increased sympathetic activity may also be the result of other stimuli, PI has been explored as an indicator of pain, and found to reflect pain in both healthy volunteers and critically ill patients.82,156

The results from study I showed that ∆PP and PI, but not ∆POP or PVI, reflected hypovolaemia induced by LBNP both during spontaneous breathing and NPPV. The results must be seen in relation to measurement site, photoplethysmographic feature, and sample size.

9.1.2 The ability of EtCO2 and VtCO2 to reflect hypovolaemia 60

- Do VtCO2 and EtCO2 track changes in cardiac output in mechanically

ventilated patients, and do the changes in VtCO2 and EtCO2 correspond with

the changes in cardiac output?

In study II we investigated the ability of VtCO2 and EtCO2 to reflect changes in cardiac output, and the association between the changes. We found that VtCO2 and EtCO2 tracked both reductions and increases in cardiac output. However, the magnitudes of the changes were only correlated during reductions in cardiac output, and the correlations were modest. ROC analyses found a discriminating ability for EtCO2 during the reduction in cardiac output only.

The fact that the changes in VtCO2 and EtCO2 were only moderately correlated with the changes in cardiac output, and only during RVP, is possibly explained by the nature of the relationship between cardiac output and exhaled CO2. This is believed to be closer in unsteady states.105,157 During low flow states, oxygen consumption depends on oxygen delivery

10,111 (DO2). If pulmonary oxygen uptake is intact, DO2 is determined by cardiac output. Hence, the product of oxygen consumption, VtCO2, becomes an indirect measure of the adequacy of oxygen delivery, i.e. cardiac output. If O2-delivery falls below the critical level, metabolism and, consequently, CO2-production drops. This means that a large reduction in cardiac output, which brings cardiac output below the critical level, will lead to larger relative reductions in

10,106,111 VtCO2 and EtCO2. In addition, when cardiac output is reduced, less CO2 is brought to the lungs for elimination, and pulmonary dead space increases as a consequence of under-

105,111 perfused lung areas. Air from un-perfused parts of the lung dilutes the expired CO2, reducing the amount of VtCO2 and EtCO2 in the expired air. The observed reduction in

VtCO2 and EtCO2 is probably explained by the two latter factors in our study, as most of the patients only experienced moderate reductions in cardiac output. By contrast, during normal 61

or high flow, the relationship between DO2, oxygen consumption and VtCO2 is weaker.

Ventilation rather than perfusion limits the amount of CO2 eliminated by the lungs, so that

EtCO2 and VtCO2 reflect the adequacy of ventilation rather than perfusion. This was probably one of the reasons why we found no significant correlations between the changes in cardiac output and the changes in VtCO2 and EtCO2 during PLR in our postoperative, circulatory stable patient group. By contrast, previous studies on circulatory unstable patients have found

101,108 112 good predictive abilities for EtCO2 and VCO2. Another factor that may have contributed to discrepant results was that in these studies, calculations were based on maximal increases in EtCO2 or VtCO2 during PLR, whereas calculations were based on mean values from 1 min measurement periods of PLR in our study, which are lower.

The moderate correlation coefficients of 0.53 for VtCO2 and 0.47 for EtCO2 during RVP mean that only a part of the reduction in VtCO2 and EtCO2 was explained by the change in cardiac

2 146 output (ρ x 100; 28 % for VtCO2 and 22% for EtCO2). Consequently, it is reasonable that also the diagnostic value of VtCO2 and EtCO2 in this context was limited. However, the correlation between the changes in cardiac output and changes in VtCO2 was similar to the correlation with EtCO2, but a predictive ability was demonstrated only for EtCO2. The low

45,112 AUC-values found for VtCO2 also contradict previous findings. One possible explanation for this finding could be the so-called spectrum effect, which describes how the dispersion of indicator values (here: EtCO2 and VtCO2) in different study populations may affect AUC-

158,159 values. In our postoperative population, the distribution of VtCO2 and EtCO2-values was relatively narrow, which may have led to lower AUC-values.

Previous studies have shown that VCO2/VtCO2 and EtCO2 consistently reflect changes in cardiac output across different populations, designs and measurement methods.45,101,105,106,108,112 They have been found superior to other dynamic variables in 62

predicting fluid responsiveness in patients ventilated with tidal volumes < 8 ml/kg45,101,112,160 and arrhythmia.160 This indicates a robust relationship between pulmonary flow and exhaled

10,105,161 104,106,108,111 CO2. Both a logarithmic and a linear relationship between the changes in cardiac output and the changes in exhaled CO2 have been suggested. Stronger correlations with cardiac index and greater diagnostic accuracy has been described for VCO2 than EtCO2,

45,112 indicating that VCO2 may be a better predictor of fluid responsiveness. We found similar correlations between the changes in VtCO2 and EtCO2 and the changes in cardiac output, but the changes in VtCO2 were significantly larger than the changes in EtCO2 for a given change in cardiac output. The magnitude of the changes in cardiac output, VtCO2 and EtCO2 was relatively constant during both reductions and increases in cardiac output. Roughly, a 20% change in cardiac output led to a 10% change in VtCO2 and a 5% change in EtCO2. These changes are similar to the increases found by Young et al.112 As one objection to the clinical use of EtCO2 has been that the changes are small, consistently larger changes in VtCO2/VCO2 could be of clinical interest. However, the fact that a discriminative ability was demonstrated only for EtCO2 in study II prevents us from concluding that VtCO2 is superior to EtCO2 in reflecting cardiac output changes. Of note, even the small change of 5% in EtCO2 which was found by us and others is several times higher than the least significant change.101,108

Both variables tracked cardiac output, which is relevant for clinical use. VtCO2 or EtCO2 may reveal whether a sudden drop in blood pressure is caused by a reduction in vascular resistance with maintained cardiac output, in which case VtCO2 and EtCO2 will remain unchanged, or an actual reduction in cardiac output, in which case VtCO2 and EtCO2 will drop as well.

Based on the results from study II, VtCO2 and EtCO2 track changes in cardiac output. The association between the changes in cardiac output, EtCO2 and VtCO2 is stronger when cardiac output is reduced. 63

9.1.3 The ability of ∆PP and ∆POP to detect hypovolaemia during positive expiratory

pressure (PEP) and continuous positive airway pressure (CPAP)

- Do ∆PP and ∆POP reflect hypovolaemia in spontaneously breathing subjects

during positive expiratory pressure (PEP) and continuous positive airway

pressure (CPAP)?

In Paper III we investigated ∆PP and ∆POP during progressive hypovolaemia and different levels of PEP and CPAP. We found that both ∆PP and ∆POP increased with progressive hypovolaemia. The increase in ∆PP was amplified by the application of the highest PEP-level.

The increase in ∆POP was not affected by any changes in respiratory resistance. Preload changes induced by different LBNP- and PEP-levels were reflected in ∆CVP. The application of CPAP did not affect any of the variables.

Both experimental and clinical studies indicate that dynamic variables perform better during spontaneous breathing if pleural pressure swings are amplified.58,59,162,163 The best performance during spontaneous breathing has been demonstrated during a deep inspiratory breathing manoeuver, where ∆PP predicted fluid responsiveness with a sensitivity of 90 % and a specificity of 100%.56 Other studies have also found good predictive ability for ∆PP during a deep inspiratory manoeuvre.163 Deep inspiratory manoeuvres require cooperative patients with a certain respiratory strength. While these studies are valuable proofs of concept, their clinical significance may be debated. Different forms of expiratory resistance on the other hand are applied in clinical practice, and have been shown to enable the prediction of preload dependency in animal162 and human studies.57,59,123 Dahl et al found that the application of a resistor enabled ∆PP to predict fluid responsiveness in spontaneously breathing anaesthetised pigs.162 In a study on healthy volunteers during head-up-tilt, the best predictive ability for ∆PP was obtained by a combination of expiratory resistance and reduced 64

respiratory rate.59 Monge Garcia et al demonstrated how the Valsalva manoeuvre enabled the prediction of fluid responsiveness with high sensitivity and specificity in spontaneously breathing patients.57 Our results showing that the increase in ∆PP was augmented with the application of high PEP are in line with these studies.

∆PP reflected LBNP-level during spontaneous breathing in study I, and during PEP 0 and

CPAP 0 (baseline) in study III, when no external modification of airway pressure was applied.

As described in the discussion of study I, this is possibly explained by deep inspiration which may occur in conscious humans as a compensatory response to the reduction in preload. A deep or prolonged inspiratory phase allows larger impact of the negative pleural pressure on venous return and SV.89 The fact that the increase in ∆PP was significantly augmented by

NPPV in study I and PEP 10 in Study III indicates that pleural pressure amplification was achieved in both models, and that these amplifications were captured by ∆PP.

Differences in CPAP-level were not reflected by any of the variables, although both ∆PP and

∆POP increased with progressive LBNP also during CPAP. This indicates that respiration with larger tidal volumes induced by hypovolaemia rather than the application of CPAP was responsible for the association between LBNP-level and ∆PP and ∆POP. This is supported by the fact that changes induced by LBNP and PEP, but not CPAP, were reflected in ∆CVP. In other words, CPAP did not induce changes in venous return that were sufficient to affect

∆CVP, ∆PP or ∆POP. This supports the notion that the magnitude of the intrathoracic pressure variations rather than respiratory mode as such affects the dynamic variables` ability to reflect volume status. Whereas PEP only increases expiratory resistance while maintaining the inspiratory decrease in RAP, CPAP reduces the gradient for venous return by applying positive airway pressure also during inspiration. 65

In mechanically ventilated patients with spontaneous breathing activity, the predictive ability of ∆PP has been poor.54,55 This is likely due to a mixture of mechanical and spontaneous breathing patterns, which have opposite effects on SV.54 High respiratory rate, leading to a dissociation of haemodynamic effects on left and right SV, may contribute to the poor results in patients with unregulated spontaneous breathing activity.164

∆POP reflected LBNP-level in study III, but not in study I. This supports the assumption that limited sample size could be the major reason for the lack of association between LBNP-level and ∆POP in study I. It is conceivable that breathing through a mask or resistor in study III, even without resistance (PEP or CPAP = 0), could induce breathing patterns that produce larger variations in venous return than spontaneous breathing alone in study I. However, as

∆POP failed to reflect LBNP-level also during NPPV in study I, the first explanation is more likely.

∆CVP was not a primary endpoint in our study, but included in order to elucidate the effects of LBNP and respiratory mode on venous return (Figure 9). ∆CVP was more sensitive to changes in PEP-level than both ∆PP and ∆POP. Measured on the right side of the heart,

∆CVP reflects changes in venous return, but unlike ∆PP and ∆POP, it is not directly affected by ventricular afterload. Like other static variables, CVP does not reliably reveal preload dependency.36 The respiratory variations in CVP (∆CVP) on the other hand has been shown to predict fluid responsiveness during mechanical48 ventilation and during deep inspiratory manoeuvers in spontaneously breathing patients.165 During deep inspiration, the increase in pressure gradient is limited by the pressure in the large extrathoracic veins. As these are surrounded by atmospheric pressure, they collapse if CVP goes below 0 mmHg.28 This will decrease right ventricular preload and, consequently, ∆CVP and SVV. Although absolute

CVP-values may not be used to reveal preload dependency, they may still be helpful for 66

haemodynamic evaluation. When cardiac output decreases, CVP may give an indication as to whether this is caused by insufficient venous return (CVP low) or impaired cardiac function

(CVP high).27

Figure 9. Respiratory variations in CVP (∆CVP) during expiratory resistance of 10, 0 and 5 cm H2O. Green dots represent the maximal and blue dots the minimal values of CVP within one respiratory cycle. The cyan line represents the respiratory signal derived from the ECG-leads.

The results of study III indicate that both ∆PP and ∆POP reflect hypovolaemia induced by

LBNP during different levels of CPAP and PEP, but that only high levels of PEP induce additional changes in SV. These changes are recognised only by ∆PP. Thus, this study indicates that ∆PP may be used for the detection of hypovolaemia during high levels of PEP.

9.2 Methodological considerations

9.2.1 Study populations

9.2.1.1 Choice of study populations

Study I and study III were performed on healthy volunteers because we wanted to compare the ability of dynamic variables to reflect changes in cardiac output during different respiratory modes, without potentially disturbing factors. The model enabled us to investigate 67

the effects of hypovolaemia and different respiratory modes in the absence of pain, medication and comorbidity, which could have affected results. That was of particular importance for the evaluation of ∆POP, which is strongly affected by factors that alter vasomotor tone. However, as healthy volunteers may differ from a hospital population, external validity may be limited. Cardiac dysfunction and reduced lung compliance may affect the performance of ∆PP.66,166 Hence, results in a hospital population treated with NPPV,

PEP or CPAP may be different than demonstrated in our studies on healthy volunteers.

In study II we investigated patients after cardiopulmonary bypass. As the relationship between cardiac output and eliminated CO2 is disturbed by irregular ventilation, the study had to be performed on patients during controlled ventilation. Patients after cardiac surgery were chosen because they had ventricular pacemaker leads after surgery, which enabled rapid and transient reductions in cardiac output. This patient group is equipped with invasive monitoring and is frequently investigated in haemodynamic research. Low temperature may occur following cardiopulmonary bypass, but all patients were above 36° before cardiopulmonary bypass was terminated. Interventions lasted maximum 1 minute, during which metabolism is unlikely to change substantially. Thus, it is unlikely that CO2-production or -elimination was affected by an increase in body temperature during experiments postoperatively.

9.2.1.2 Sample size calculation

When conducting a study, sample size should be estimated in advance for two reasons: Firstly, adequate sample size is necessary to detect an effect, for instance a difference between two groups, and reject the null-hypothesis with adequate power. Secondly, the inclusion of more participants than necessary may be unethical, especially when study participation entails a risk. 68

Sample size estimation requires the knowledge of several factors: a priori estimated effect size, standard deviation, significance level (α) and power (1-β). α refers to the probability of making a Type I-error, which is rejecting the null hypothesis when it is actually true (there is no difference). β refers to the probability of making a Type II-error, which is to accept a non- significant result when the null-hypothesis is false, and there actually is a difference. α is usually set at 5% (p-value of 0.05). β depends on the size of the effect, or difference, as well as the sample size. 1-β describes the power of the study, which is the probability that a study of a certain size will detect a true difference or effect as statistically significant, so that the null-hypothesis may be rejected.146 The smaller the effect size, the larger the sample size must be.

Study I was an experimental study performed on a limited number of subjects (n = 14). When the study was planned, few if any similar studies had been published. Hence, effect size was largely unknown, and no conventional sample size calculation was performed in advance.

Although comparable in size to other studies on healthy volunteers in the LBNP-model59,79,81, the number of subjects may represent a limitation. Whereas it was sufficiently powered to show a statistical association between ∆PP and LBNP-level, it was probably not sufficiently powered to exclude an association between ∆POP and LBNP-level.

Study II was originally planned as a method comparison study, comparing the ability of expiratory CO2 and non-invasive pulse contour analysis to estimate changes in cardiac output.

Sample size was estimated a priori based on the intended comparison of two methods in their ability to reflect cardiac output changes, indicating a required number of 30-40 subjects.

However, we encountered practical difficulties with the pulse contour measurements (specific to the design, and not the device as such), and decided to abort this part of the study. When presenting the intervention part of the study, we decided to display confidence intervals rather 69

than performing a post-hoc power analysis, in accordance to the recommendations in the

CONSORT guidelines.167 The magnitude of the confidence intervals reflects the power of the study, as confidence intervals decrease with increasing number of study participants.

Study III was conducted on healthy volunteers who were also participating in another study in the LBNP-model the same day. In this study, the association between cerebral oxygen tension and flow in the internal and external carotid arteries during hypovolaemia induced by LBNP was investigated. Both studies were planned simultaneously, but sample size estimation was based on calculations from the study investigating determinants of cerebral oxygen tension, and no separate power analysis was performed for study III.

9.2.1.3 Randomisation

In study II we had two interventions, passive leg raise and right ventricular pacing. As ventricular pacing was performed by clinically engaged department consultants, timing of the intervention had to be adapted to their availability, and the order of interventions varied in a non-randomised fashion. Thus, a potential carry-over effect, where the effect of the first intervention impacts the effect of the second, cannot be excluded. Ideally, patients should have been randomly assigned to a specific order of interventions; PLR - RVP or RVP - PLR.

Carry-over effects were minimised by allowing minimum 5 min pause between interventions.145

In study III, we also had two interventions; PEP and CPAP. In this study, the subjects were randomised in blocks of six to a specific order of interventions, which also decided the order of interventions at the following LBNP-level.

9.2.2 Models

9.2.2.1 Lower body negative pressure (LBNP) as a model of central hypovolaemia 70

The advantages of the LBNP-model are that hypovolaemia may be induced non-invasively, controlled and reversibly. The LBNP-model has previously been used to investigate different features of the dynamic variables in spontaneously breathing subjects.79,81,89 However, there may be some differences in the effects of hypovolaemia induced by LBNP and hypovolaemia due to for instance haemorrhage following trauma, as the sympathetically activated responses may be more pronounced after injury and pain. It is an advantage of the model that the effect of hypovolaemia can be investigated without “disturbance” from altered sympathetic activity due to pain or anaesthetic drugs, but the association between hypovolaemia and the dynamic variables, especially ∆POP, may be different in conscious bleeding humans.

In paper I and III, negative pressures in incremental steps of 20 mmHg to 80 mmHg were applied. These LBNP-levels equal blood losses of approximately 400 ml to 1400 ml.121 The fact that the method in our studies led to reduced preload was demonstrated by significant decreases in SV and cardiac output in both paper I and paper III.

We did not measure respiratory rate or tidal volume during unmodified spontaneous breathing in study I. As we observed significant associations between ∆PP and LBNP-level which we believe may have been related to respiratory rate and / or tidal volume, this would have been of interest. The LBNP-model alters ventilation by shifting the diaphragma downwards; thus increasing functional residual capacity and forced vital capacity.119 The increase in tidal volumes that occurs as a compensatory response to preload reduction may thus be augmented when the preload reduction is induced by LBNP.

9.2.2.2 Non-invasive positive pressure ventilation (NPPV) to induce respiratory

variations in stroke volume

Controlled mechanical ventilation of conscious healthy volunteers is challenging, and different positive pressure ventilation modes were tried in pilots. A volume controlled mode such as intermittent positive pressure ventilation (IPPV-mode) facilitated the standardisation of tidal volumes, and compliance with the ventilator was better than with synchronised intermittent mandatory ventilation or pressure-control modes. Airway pressures and leakage were monitored closely to ensure maximal compliance with the ventilator. Particular care was taken to minimise leakage, as leakage would have relieved airway pressure and reduced the respiratory impact on preload. Active respiratory movements in the subjects would have offset the effects of positive pressure ventilation. Thus, sequences with spontaneous breathing activity were removed before data analysis.

9.2.2.3 Right ventricular pacing for the reduction of cardiac output

Pacemaker leads are routinely applied in patients undergoing cardiac surgery. According to institutional practice, the pacemaker leads are being tested before and after the thoracic cavity is closed, and later postoperatively in patients who turn out to be pacemaker dependent. In study II, the postoperative pacemaker test was performed on all included patients as part of the intervention. The method was suitable for creating sudden, moderate reductions in cardiac output which could be captured by oesophageal Doppler and were sufficient to induce changes in exhaled CO2. Patients were included prior to surgery. Only patients with ejection fraction ≥ 40% were included. Patients who turned out to be pacemaker dependent, or were deemed haemodynamically unstable by the attending cardiothoracic anaesthesiologist after surgery, were excluded from study participation.

9.2.3 Measurements and calculations: Sources of potential errors

In paper II, the results were mainly given as relative (%) changes from baseline for several reasons. Firstly, it enabled direct comparisons between the changes in cardiac output, VtCO2 and EtCO2, which have different units. In addition, a given EtCO2- or VtCO2-value does not equal a specific value of cardiac output, as eliminated CO2 depends on several factors that may vary over time and between patients. Secondly, the use of oesophageal Doppler is based on several assumptions, as described in the methods section. Some oesophageal Doppler monitors contain an M-mode echo transducer which enables measurement of aortic diameter, but the oesophageal Doppler we used assumes the size of the aortic diameter. It has been shown that this diameter changes with MAP168, and we cannot exclude a similar effect during the endogenous fluid load induced by the passive leg raise manoeuver. The PLR-manoeuver may also dislocate the OD-probe. A small difference in insonation angle between Doppler beam and aortic flow may lead to substantial measurement errors.135 Thus, great care was taken to prevent dislocation of the probe during PLR, and no measurements were obtained during the actual change in position; only immediately after. A violation of these assumptions would have had larger impact if the changes had been expressed in absolute values. When the changes are calculated in relative terms, only a change in the Doppler position or the aortic diameter during the actual intervention would have affected the estimates. However, it is possible that the differences observed in cardiac output at baseline were due to measurement errors.

There were potential sources of error related to the calculation of ∆PP and ∆POP in studies I and III, and of VtCO2 in study II. Different custom made computer programs were used for the calculation of ∆PP, ∆POP and VtCO2. However, both the calculation of ∆PP and ∆POP in studies I and III, and the calculation of VtCO2 based on flow- and capnography curves in study II, required visual inspection and selection of all values going into the final dataset. The 73

delimitation of the respiratory cycles and alignment of respiratory curves to flow -, blood pressure and photoplethysmographic curves were also performed manually. Manual data handling introduces the possibility of error and bias, which could possibly be avoided with an all-automatic routine. However, automated algorithms must be validated.

As study II was performed postoperatively shortly before the patients were intended to wake up, no neuromuscular blocker was administered before measurements. Hence, irregular breaths had to be eliminated from the data records manually. Similarly, spontaneous breathing efforts during NPPV had to be eliminated to avoid the inclusion of cycles containing errors which would reduce study validity. Manual data selection contains the possibility of bias and may lead to false positive results if only the “best” cycles are selected for further analyses.

Differences in methodology are frequently cited to explain discrepant study results, not least in studies investigating the photoplethysmographic signal. Differences in measurement time, selection of respiratory cycles and methods for estimating maximal and minimal curve peaks vary between studies, and may affect both reliability and validity. The differences between a

“raw”, unfiltered data set and a manually filtered one have been demonstrated for ∆POP.

Addison et al99 compared a subset of manually filtered data with the original dataset which included all measurements, and applied stepwise improvements in the algorithm analysing

∆POP. Whereas the manually filtered dataset required only simple algorithm improvements, more advanced algorithms were needed to improve the correlation between ∆PP and ∆POP in the unfiltered dataset. Correction for low perfusion in particular led to significant improvement in correlation. Similarly, when including all measurements over a longer period instead of manually selecting respiratory cycles, Landsverk et al found larger inter- and intra- individual variability in ∆POP compared to ∆PP in mechanically ventilated intensive care patients.93 74

A related source of error is the lack of blinding. The primary investigator was not blinded to the interventions when performing the manual calculations of ∆PP, ∆POP and VtCO2.

Generally, a hypothesized expectation may impact calculations. The use of software programs which automated the major part of data analysis and calculation reduced the potential impact of this methodological limitation in our studies, but ideally, calculations should be performed by someone blinded to intervention status. Alternatively, inter-observer variability may be assessed by comparing the calculations of two independent investigators. Our group has used this approach in several previous studies on ∆PP and ∆POP. Inter-observer variability has been low, and we have no reason to assume that this would be any different for the present studies.

9.2.4 Statistical methods

9.2.4.1 Linear mixed models

The two major strengths of LMM are their ability to handle repeated measurements and datasets with missing observations. Both were the case in studies I and III. We used mixed models for the data analysis in papers I and III because we had multiple observations from each subject, which may be assumed to be correlated and thus violate the assumptions for other linear models.145 LMM is used for repeated measurements as it considers the correlation between observations from the same subjects at different measurement points.144 Due to differences in LBNP-tolerance, the number of observations differed between subjects in study

I and III. Thus, the maximisation of power despite missing observations was another major advantage of LMM in these studies.

In study I, an interaction term between LBNP and respiratory mode (NPPV) was included.

The addition of an interaction variable is necessary when the effect or value of one variable is affected by the effect or value of another, which was the case for LBNP and NPPV. 75

9.2.4.2 Correlation

Spearman`s correlation was used in study II to investigate the relationship between cardiac output, EtCO2 and VtCO2. Of note, correlation tests association, but not agreement. Thus, in study II, we investigated the association between the magnitudes of the changes in three variables, but cannot infer from that to what degree the variables agree over time. For the comparison of different methods of cardiac output measurements, other analyses such as

Bland Altman should be provided. Also, whether a change in one variable may be predicted by the change in another is better assessed using other methods, such as linear regression or receiver operating characteristics (ROC) plots. The moderate correlations found in study II were somewhat surprising given the very clear pattern of how the variables followed each other during and after interventions, and may be an illustration of the limitation of correlation as a method to test the association between these variables.

9.2.4.3 Receiver operating characteristics (ROC) plots

Whereas the correlation between changes following an intervention is interesting, we are often interested in a variable`s ability to predict or detect a given outcome. ROC-plots describe the ability of a continuous variable to predict a dichotomised outcome. This analysis is frequently used in the prediction of fluid responsiveness. It may also be used to evaluate a variable`s ability to detect a reduction of a given magnitude, such as the ability of VtCO2 and

EtCO2 to detect a change in CO of ≥15%. In study II, we found that the correlation between

VtCO2 and cardiac output was similar or even stronger than the correlation between EtCO2 and cardiac output, but that only EtCO2 was able to detect a reduction of ≥ 15% in cardiac output, according to the ROC curves. This may reflect limitations in the receiver operating 76

characteristics plots as described in the methods section, but it also illustrates how different methods of analyses may affect results.

9.3 Ethical considerations

All three studies were approved by the regional committee of ethics in medicine; REK Sør-

Øst, before data collection began, with the permission numbers 2009/2180 (Study I),

2013/1605 (Study II) and 2015/344 (Study III). Oral and written informed consent was obtained prior to inclusion.

The ethical considerations differ for the different studies. The LBNP-model used in study I and III induces actual central hypovolaemia with reduction in cardiac output and blood pressure, and the risk of presyncope or circulatory collapse. The application of positive pressure ventilation or respiratory resistance adds to the physical strain related to LBNP. In healthy volunteers this model has been shown to be safe, as hypovolaemia is reversed immediately at signs of discomfort or impending circulatory collapse. The release of the

LBNP leads to immediate normalisation of cardiac output and blood pressure. During experiments, a consultant anaesthesiologist as well as a physiologist well acquainted with the

LBNP-model and symptoms of circulatory collapse were present at all times. Medical history was obtained from the subjects before experiments. However, disease unknown to the participants would probably not have been revealed. The same experiment could not have been performed on patients or subjects with cardiovascular comorbidity, primarily due to the danger of ischemic insults following hypovolaemia. As the LBNP-levels applied in our studies induces significant reductions in blood flow, pregnant women were excluded from participation. 77

The volunteers consisted mainly of students and colleagues of the authors of papers I and III.

Hence, there was a relation, however loose, between the participants and researchers. Students were not individually approached for inclusion, but informed about the possibility of parttaking in the studies. As study participation was quite time consuming, they were rewarded with a moderate amount of money in study I and a gift card in study III.

The ethical considerations for study II were twofold: Firstly, we included patients who were scheduled for open heart surgery, and were thus already seriously ill. A request to consider participation in a study shortly before surgery may have led to additional mental strain for some of the patients. Out of 42 approached patients, 40 agreed to participate. Secondly, RVP led to a reduction in SV. Hence, only patients with ejection fraction over 40% were included, and the pacing sequence was limited to 30 sec. Pacing was performed by an experienced cardiothoracic anaesthesiologist after the procedure had been approved by the anaesthesiologist responsible during surgery. No adverse effects related to the pacing were observed. The study was planned in agreement with the hospital`s cardiologists and cardiothoracic surgeons.

9.4 Future perspectives

Dynamic variables for haemodynamic evaluation and the prediction of fluid responsiveness have been extensively studied in mechanically ventilated patients. There are also several smaller studies on spontaneously breathing healthy volunteers where breathing rate or depth is manipulated. Based on these studies, dynamic variables appear to have some merit in spontaneously breathing subjects, but clinical studies during for instance NPPV or PEP are lacking. The potential for the dynamic variables in these settings should be investigated. 78

Exhaled carbon dioxide, especially EtCO2, has been studied for the prediction of fluid responsiveness, but it is possible that the small changes associated with moderate changes in cardiac output limit their use perioperatively. Clinical studies investigating outcome in patients where perioperative fluid management has been guided by EtCO2 or VCO2 continuously displayed by the ventilator would be of interest. Also, the potential for exhaled

CO2 for haemodynamic evaluation in spontaneously breathing subjects could be further explored.

10 CONCLUSIONS

- Do ∆PP, ∆POP, PVI and PI reflect hypovolaemia during spontaneous

breathing and non-invasive positive pressure ventilation (NPPV)?

Based on the findings in paper I, we may conclude that ∆PP and PI reflected hypovolaemia induced by LBNP during spontaneous breathing and NPPV. The increase in ∆PP, but not PI, was amplified with the application of NPPV. Neither ∆POP nor PVI reflected hypovolaemia in this model. However, less emphasis should be placed on the negative findings of this study because of the limited sample size.

- Do VtCO2 and EtCO2 track changes in cardiac output in mechanically

ventilated patients, and do the changes in VtCO2 and EtCO2 correspond with

the changes in cardiac output?

Based on the findings in paper II, we conclude that VtCO2 and EtCO2 tracked cardiac output changes induced by right ventricular pacing and passive leg raise. The changes in VtCO2 and

EtCO2 were correlated with the changes in cardiac output only when cardiac output was reduced, and correlations were modest.

-Do ∆PP and ∆POP reflect hypovolaemia in spontaneously breathing subjects

during positive expiratory pressure (PEP) and continuous positive airway

pressure (CPAP)?

Based on the findings in paper III, we conclude that ∆PP and ∆POP reflected hypovolaemia induced by LBNP during different levels of CPAP and PEP. An increase in PEP augmented the increase in ∆PP, but not ∆POP. The application of CPAP did not affect any of the variables significantly. 80

11 REFERENCES

1. Guly HR, Bouamra O, Spiers M, Dark P, Coats T, Lecky FE, Trauma A, Research N. Vital signs and estimated blood loss in patients with major trauma: testing the validity of the ATLS classification of hypovolaemic shock. Resuscitation 2011; 82: 556-9. 2. Cecconi M, De Backer D, Antonelli M, Beale R, Bakker J, Hofer C, Jaeschke R, Mebazaa A, Pinsky MR, Teboul JL, Vincent JL, Rhodes A. Consensus on circulatory shock and hemodynamic monitoring. Task force of the European Society of Intensive Care Medicine. Intensive Care Med 2014; 40: 1795-815. 3. Cannon JW. Hemorrhagic Shock. N Engl J Med 2018; 378: 370-79. 4. Beekley AC, Sebesta JA, Blackbourne LH, Herbert GS, Kauvar DS, Baer DG, Walters TJ, Mullenix PS, Holcomb JB, st Combat Support Hospital Research G. Prehospital tourniquet use in Operation Iraqi Freedom: effect on hemorrhage control and outcomes. J Trauma 2008; 64: S28-37; discussion S37. 5. Vincent JL, De Backer D. My paper 20 years later: effects of dobutamine on the VO(2)/DO(2) relationship. Intensive Care Med 2014; 40: 1643-8. 6. Gutierrez G, Reines HD, Wulf-Gutierrez ME. Clinical review: hemorrhagic shock. Crit Care 2004; 8: 373-81. 7. Reljin N, Zimmer G, Malyuta Y, Shelley K, Mendelson Y, Blehar DJ, Darling CE, Chon KH. Using support vector machines on photoplethysmographic signals to discriminate between hypovolemia and euvolemia. PLoS One 2018; 13: e0195087. 8. Levick RJ. An Introduction to Cardiovascular Physiology 5E. 2009. 9. Vincent JL. Understanding cardiac output. Crit Care 2008; 12: 174. 10. Dubin A, Murias G, Estenssoro E, Canales H, Sottile P, Badie J, Baran M, Rossi S, Laporte M, Palizas F, Giampieri J, Mediavilla D, Vacca E, Botta D. End-tidal CO2 pressure determinants during hemorrhagic shock. Intensive Care Med 2000; 26: 1619-23. 11. Teboul JL, Saugel B, Cecconi M, De Backer D, Hofer CK, Monnet X, Perel A, Pinsky MR, Reuter DA, Rhodes A, Squara P, Vincent JL, Scheeren TW. Less invasive hemodynamic monitoring in critically ill patients. Intensive Care Med 2016; 42: 1350-9. 12. Monnet X, Anguel N, Naudin B, Jabot J, Richard C, Teboul JL. Arterial pressure-based cardiac output in septic patients: different accuracy of pulse contour and uncalibrated pressure waveform devices. Crit Care 2010; 14: R109. 13. Saugel B, Wagner JY, Scheeren TW. Cardiac output monitoring: less invasiveness, less accuracy? J Clin Monit Comput 2016; 30: 753-55. 14. Hamzaoui O, Monnet X, Richard C, Osman D, Chemla D, Teboul JL. Effects of changes in vascular tone on the agreement between pulse contour and transpulmonary thermodilution cardiac output measurements within an up to 6-hour calibration-free period. Crit Care Med 2008; 36: 434-40. 15. Slagt C, Malagon I, Groeneveld AB. Systematic review of uncalibrated arterial pressure waveform analysis to determine cardiac output and stroke volume variation. Br J Anaesth 2014; 112: 626-37. 16. Huang SJ. Measuring cardiac output at the bedside. Curr Opin Crit Care 2019; 25: 266-72. 17. Thiele RH, Bartels K, Gan TJ. Cardiac output monitoring: a contemporary assessment and review. Crit Care Med 2015; 43: 177-85. 18. Pearse RM, Harrison DA, MacDonald N, Gillies MA, Blunt M, Ackland G, Grocott MP, Ahern A, Griggs K, Scott R, Hinds C, Rowan K, Group OS. Effect of a perioperative, cardiac output-guided hemodynamic therapy algorithm on outcomes following major gastrointestinal surgery: a randomized clinical trial and systematic review. JAMA 2014; 311: 2181-90. 81

19. Osawa EA, Rhodes A, Landoni G, Galas FR, Fukushima JT, Park CH, Almeida JP, Nakamura RE, Strabelli TM, Pileggi B, Leme AC, Fominskiy E, Sakr Y, Lima M, Franco RA, Chan RP, Piccioni MA, Mendes P, Menezes SR, Bruno T, Gaiotto FA, Lisboa LA, Dallan LA, Hueb AC, Pomerantzeff PM, Kalil Filho R, Jatene FB, Auler Junior JO, Hajjar LA. Effect of Perioperative Goal-Directed Hemodynamic Resuscitation Therapy on Outcomes Following Cardiac Surgery: A Randomized Clinical Trial and Systematic Review. Crit Care Med 2016; 44: 724-33. 20. Corcoran T, Rhodes JE, Clarke S, Myles PS, Ho KM. Perioperative fluid management strategies in major surgery: a stratified meta-analysis. Anesth Analg 2012; 114: 640-51. 21. Henderson WR, Griesdale DE, Walley KR, Sheel AW. Clinical review: Guyton--the role of mean circulatory filling pressure and right atrial pressure in controlling cardiac output. Crit Care 2010; 14: 243. 22. Michard F, Teboul JL. Using heart-lung interactions to assess fluid responsiveness during mechanical ventilation. Crit Care 2000; 4: 282-9. 23. Gelman S. Venous function and central venous pressure: a physiologic story. Anesthesiology 2008; 108: 735-48. 24. Wijnberge M, Sindhunata DP, Pinsky MR, Vlaar AP, Ouweneel E, Jansen JR, Veelo DP, Geerts BF. Estimating mean circulatory filling pressure in clinical practice: a systematic review comparing three bedside methods in the critically ill. Ann Intensive Care 2018; 8: 73. 25. Cecconi M, Aya HD, Geisen M, Ebm C, Fletcher N, Grounds RM, Rhodes A. Changes in the mean systemic filling pressure during a fluid challenge in postsurgical intensive care patients. Intensive Care Med 2013; 39: 1299-305. 26. Yastrebov K, Aneman A, Slama M, Kokhno V, Luchansky V, Orde S, Hilton A, Lukiyanov D, Volobueva I, Sidelnikova S, Polovnikov E. The stop-flow arm equilibrium pressure in preoperative patients: Stressed volume and correlations with echocardiography. Acta Anaesthesiol Scand 2019. 27. Magder S. Heart-Lung interaction in spontaneous breathing subjects: the basics. Ann Transl Med 2018; 6: 348. 28. Funk DJ, Jacobsohn E, Kumar A. The role of venous return in critical illness and shock-part I: physiology. Crit Care Med 2013; 41: 255-62. 29. Guyton AC. Determination of cardiac output by equating venous return curves with cardiac response curves. Physiol Rev 1955; 35: 123-9. 30. Maas JJ, Geerts BF, van den Berg PC, Pinsky MR, Jansen JR. Assessment of venous return curve and mean systemic filling pressure in postoperative cardiac surgery patients. Crit Care Med 2009; 37: 912-8. 31. Pinsky MR. Heart lung interactions during mechanical ventilation. Curr Opin Crit Care 2012; 18: 256-60. 32. Michard F. Changes in arterial pressure during mechanical ventilation. Anesthesiology 2005; 103: 419-28; quiz 49-5. 33. Alian AA, Shelley KH. Respiratory physiology and the impact of different modes of ventilation on the photoplethysmographic waveform. Sensors (Basel) 2012; 12: 2236-54. 34. De Backer D, Heenen S, Piagnerelli M, Koch M, Vincent JL. Pulse pressure variations to predict fluid responsiveness: influence of tidal volume. Intensive Care Med 2005; 31: 517-23. 35. Yang X, Du B. Does pulse pressure variation predict fluid responsiveness in critically ill patients? A systematic review and meta-analysis. Crit Care 2014; 18: 650. 36. Michard F, Teboul JL. Predicting fluid responsiveness in ICU patients: a critical analysis of the evidence. Chest 2002; 121: 2000-8. 37. Marik P, Monnet X, Teboul J-L. Hemodynamic parameters to guide fluid therapy. Annals of Intensive Care 2011; 1: 1-9. 38. Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, Gerlach H, Moreno R, Carlet J, Le Gall JR, Payen D, Sepsis Occurrence in Acutely Ill Patients I. Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 2006; 34: 344-53. 82

39. Pinsky MR. Functional hemodynamic monitoring. Crit Care Clin 2015; 31: 89-111. 40. Marik PE, Cavallazzi R, Vasu T, Hirani A. Dynamic changes in arterial waveform derived variables and fluid responsiveness in mechanically ventilated patients: a systematic review of the literature. Crit Care Med 2009; 37: 2642-7. 41. Michard F, Boussat S, Chemla D, Anguel N, Mercat A, Lecarpentier Y, Richard C, Pinsky MR, Teboul JL. Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure. Am J Respir Crit Care Med 2000; 162: 134-8. 42. Cannesson M, Le Manach Y, Hofer CK, Goarin JP, Lehot JJ, Vallet B, Tavernier B. Assessing the diagnostic accuracy of pulse pressure variations for the prediction of fluid responsiveness: a "gray zone" approach. Anesthesiology 2011; 115: 231-41. 43. Coudray A, Romand JA, Treggiari M, Bendjelid K. Fluid responsiveness in spontaneously breathing patients: a review of indexes used in intensive care. Crit Care Med 2005; 33: 2757-62. 44. Michard F, Chemla D, Richard C, Wysocki M, Pinsky MR, Lecarpentier Y, Teboul JL. Clinical use of respiratory changes in arterial pulse pressure to monitor the hemodynamic effects of PEEP. Am J Respir Crit Care Med 1999; 159: 935-9. 45. Tusman G, Groisman I, Maidana GA, Scandurra A, Arca JM, Bohm SH, Suarez-Sipmann F. The Sensitivity and Specificity of Pulmonary Carbon Dioxide Elimination for Noninvasive Assessment of Fluid Responsiveness. Anesth Analg 2016; 122: 1404-11. 46. Geerts BF, Aarts LP, Groeneveld AB, Jansen JR. Predicting cardiac output responses to passive leg raising by a PEEP-induced increase in central venous pressure, in cardiac surgery patients. Br J Anaesth 2011; 107: 150-6. 47. Huang CC, Fu JY, Hu HC, Kao KC, Chen NH, Hsieh MJ, Tsai YH. Prediction of fluid responsiveness in acute respiratory distress syndrome patients ventilated with low tidal volume and high positive end-expiratory pressure. Crit Care Med 2008; 36: 2810-6. 48. Cherpanath TG, Geerts BF, Maas JJ, de Wilde RB, Groeneveld AB, Jansen JR. Ventilator- induced central venous pressure variation can predict fluid responsiveness in post-operative cardiac surgery patients. Acta Anaesthesiol Scand 2016; 60: 1395-403. 49. Michard F. Volume management using dynamic parameters: the good, the bad, and the ugly. Chest 2005; 128: 1902-3. 50. Lee JH, Jeon Y, Bahk JH, Gil NS, Hong DM, Kim JH, Kim HJ. Pulse pressure variation as a predictor of fluid responsiveness during one-lung ventilation for lung surgery using thoracotomy: randomised controlled study. Eur J Anaesthesiol 2011; 28: 39-44. 51. Biais M, Bernard O, Ha JC, Degryse C, Sztark F. Abilities of pulse pressure variations and stroke volume variations to predict fluid responsiveness in prone position during scoliosis surgery. Br J Anaesth 2010; 104: 407-13. 52. Pinsky MR. Probing the limits of arterial pulse contour analysis to predict preload responsiveness. Anesth Analg 2003; 96: 1245-7. 53. Vos JJ, Kalmar AF, Struys MM, Wietasch JK, Hendriks HG, Scheeren TW. Comparison of arterial pressure and plethysmographic waveform-based dynamic preload variables in assessing fluid responsiveness and dynamic arterial tone in patients undergoing major hepatic resection. Br J Anaesth 2013; 110: 940-6. 54. Heenen S, De Backer D, Vincent JL. How can the response to volume expansion in patients with spontaneous respiratory movements be predicted? Crit Care 2006; 10: R102. 55. Monnet X, Rienzo M, Osman D, Anguel N, Richard C, Pinsky MR, Teboul JL. Passive leg raising predicts fluid responsiveness in the critically ill. Crit Care Med 2006; 34. 56. Preau S, Dewavrin F, Soland V, Bortolotti P, Colling D, Chagnon JL, Durocher A, Saulnier F. Hemodynamic changes during a deep inspiration maneuver predict fluid responsiveness in spontaneously breathing patients. Cardiol Res Pract 2012; 2012: 191807. 83

57. Monge Garcia MI, Gil Cano A, Diaz Monrove JC. Arterial pressure changes during the Valsalva maneuver to predict fluid responsiveness in spontaneously breathing patients. Intensive Care Med 2009; 35: 77-84. 58. Zollei E, Bertalan V, Nemeth A, Csabi P, Laszlo I, Kaszaki J, Rudas L. Non-invasive detection of hypovolemia or fluid responsiveness in spontaneously breathing subjects. BMC Anesthesiol 2013; 13: 40. 59. Bronzwaer AS, Ouweneel DM, Stok WJ, Westerhof BE, van Lieshout JJ. Arterial Pressure Variation as a Biomarker of Preload Dependency in Spontaneously Breathing Subjects - A Proof of Principle. PLoS One 2015; 10: e0137364. 60. Reuter DA, Bayerlein J, Goepfert MS, Weis FC, Kilger E, Lamm P, Goetz AE. Influence of tidal volume on left ventricular stroke volume variation measured by pulse contour analysis in mechanically ventilated patients. Intensive Care Med 2003; 29: 476-80. 61. Mesquida J, Kim HK, Pinsky MR. Effect of tidal volume, intrathoracic pressure, and cardiac contractility on variations in pulse pressure, stroke volume, and intrathoracic blood volume. Intensive Care Med 2011; 37: 1672-9. 62. Myatra SN, Prabu NR, Divatia JV, Monnet X, Kulkarni AP, Teboul JL. The Changes in Pulse Pressure Variation or Stroke Volume Variation After a "Tidal Volume Challenge" Reliably Predict Fluid Responsiveness During Low Tidal Volume Ventilation. Crit Care Med 2017; 45: 415-21. 63. Muller L, Louart G, Bousquet PJ, Candela D, Zoric L, de La Coussaye JE, Jaber S, Lefrant JY. The influence of the airway driving pressure on pulsed pressure variation as a predictor of fluid responsiveness. Intensive Care Med 2010; 36: 496-503. 64. Liu Y, Wei LQ, Li GQ, Yu X, Li GF, Li YM. Pulse Pressure Variation Adjusted by Respiratory Changes in Pleural Pressure, Rather Than by Tidal Volume, Reliably Predicts Fluid Responsiveness in Patients With Acute Respiratory Distress Syndrome. Crit Care Med 2016; 44: 342-51. 65. Vallee F, Richard JC, Mari A, Gallas T, Arsac E, Verlaan PS, Chousterman B, Samii K, Genestal M, Fourcade O. Pulse pressure variations adjusted by alveolar driving pressure to assess fluid responsiveness. Intensive Care Med 2009; 35: 1004-10. 66. Lakhal K, Ehrmann S, Benzekri-Lefevre D, Runge I, Legras A, Dequin PF, Mercier E, Wolff M, Regnier B, Boulain T. Respiratory pulse pressure variation fails to predict fluid responsiveness in acute respiratory distress syndrome. Crit Care 2011; 15: R85. 67. Rex S, Schalte G, Schroth S, de Waal EE, Metzelder S, Overbeck Y, Rossaint R, Buhre W. Limitations of arterial pulse pressure variation and left ventricular stroke volume variation in estimating cardiac pre-load during open heart surgery. Acta Anaesthesiol Scand 2007; 51: 1258-67. 68. Piccioni F, Bernasconi F, Tramontano GTA, Langer M. A systematic review of pulse pressure variation and stroke volume variation to predict fluid responsiveness during cardiac and thoracic surgery. J Clin Monit Comput 2017; 31: 677-84. 69. De Backer D, Taccone FS, Holsten R, Ibrahimi F, Vincent JL. Influence of respiratory rate on stroke volume variation in mechanically ventilated patients. Anesthesiology 2009; 110: 1092-7. 70. Mahjoub Y, Pila C, Friggeri A, Zogheib E, Lobjoie E, Tinturier F, Galy C, Slama M, Dupont H. Assessing fluid responsiveness in critically ill patients: False-positive pulse pressure variation is detected by Doppler echocardiographic evaluation of the right ventricle. Crit Care Med 2009; 37: 2570-5. 71. Hoiseth LO, Hoff IE, Hagen OA, Landsverk SA, Kirkeboen KA. Dynamic variables and fluid responsiveness in patients for aortic stenosis surgery. Acta Anaesthesiol Scand 2014; 58: 826-34. 72. Tusman G, Bohm SH, Suarez-Sipmann F. Advanced Uses of Pulse Oximetry for Monitoring Mechanically Ventilated Patients. Anesth Analg 2017; 124: 62-71. 73. Elgendi M. On the analysis of fingertip photoplethysmogram signals. Curr Cardiol Rev 2012; 8: 14-25. 74. Partridge BL. Use of pulse oximetry as a noninvasive indicator of intravascular volume status. J Clin Monit 1987; 3: 263-8. 84

75. Awad AA, Ghobashy MA, Ouda W, Stout RG, Silverman DG, Shelley KH. Different responses of ear and finger pulse oximeter wave form to cold pressor test. Anesth Analg 2001; 92: 1483-6. 76. Nilsson LM. Respiration signals from photoplethysmography. Anesth Analg 2013; 117: 859-65. 77. Cannesson M, Attof Y, Rosamel P, Desebbe O, Joseph P, Metton O, Bastien O, Lehot JJ. Respiratory variations in pulse oximetry plethysmographic waveform amplitude to predict fluid responsiveness in the operating room. Anesthesiology 2007; 106: 1105-11. 78. Natalini G, Rosano A, Taranto M, Faggian B, Vittorielli E, Bernardini A. Arterial versus plethysmographic dynamic indices to test responsiveness for testing fluid administration in hypotensive patients: a clinical trial. Anesth Analg 2006; 103: 1478-84. 79. McGrath SP, Ryan KL, Wendelken SM, Rickards CA, Convertino VA. Pulse oximeter plethysmographic waveform changes in awake, spontaneously breathing, hypovolemic volunteers. Anesth Analg 2011; 112: 368-74. 80. Scully CG, Selvaraj N, Romberg FW, Wardhan R, Ryan J, Florian JP, Silverman DG, Shelley KH, Chon KH. Using time-frequency analysis of the photoplethysmographic waveform to detect the withdrawal of 900 mL of blood. Anesth Analg 2012; 115: 74-81. 81. Alian AA, Galante NJ, Stachenfeld NS, Silverman DG, Shelley KH. Impact of central hypovolemia on photoplethysmographic waveform parameters in healthy volunteers. Part 1: time domain analysis. J Clin Monit Comput 2011; 25: 377-85. 82. Hoiseth LO, Hisdal J, Hoff IE, Hagen OA, Landsverk SA, Kirkeboen KA. Tissue oxygen saturation and finger perfusion index in central hypovolemia: influence of pain. Crit Care Med 2015; 43: 747-56. 83. Beurton A, Teboul JL, Gavelli F, Gonzalez FA, Girotto V, Galarza L, Anguel N, Richard C, Monnet X. The effects of passive leg raising may be detected by the plethysmographic oxygen saturation signal in critically ill patients. Crit Care 2019; 23: 19. 84. Nilsson L, Goscinski T, Lindenberger M, Lanne T, Johansson A. Respiratory variations in the photoplethysmographic waveform: acute hypovolaemia during spontaneous breathing is not detected. Physiol Meas 2010; 31: 953-62. 85. Chu H, Wang Y, Sun Y, Wang G. Accuracy of pleth variability index to predict fluid responsiveness in mechanically ventilated patients: a systematic review and meta-analysis. J Clin Monit Comput 2016; 30: 265-74. 86. Delerme S, Renault R, Le Manach Y, Lvovschi V, Bendahou M, Riou B, Ray P. Variations in pulse oximetry plethysmographic waveform amplitude induced by passive leg raising in spontaneously breathing volunteers. Am J Emerg Med 2007; 25: 637-42. 87. Delerme S, Castro S, Freund Y, Nazeyrollas P, Josse MO, Madonna-Py B, Rouff E, Riou B, Ray P. Relation between pulse oximetry plethysmographic waveform amplitude induced by passive leg raising and cardiac index in spontaneously breathing subjects. Am J Emerg Med 2010; 28: 505-10. 88. Keller G, Cassar E, Desebbe O, Lehot JJ, Cannesson M. Ability of pleth variability index to detect hemodynamic changes induced by passive leg raising in spontaneously breathing volunteers. Crit Care 2008; 12: R37. 89. Nilsson LM, Lindenberger DM, Hahn RG. The effect of positive end-expiratory pressure and tripled tidal volume on pleth variability index during hypovolaemia in conscious subjects: a volunteer study. Eur J Anaesthesiol 2013; 30: 671-7. 90. Sato M, Kunisawa T, Kurosawa A, Sasakawa T. Pulse oximetry-derived pleth variability index can predict dexmedetomidine-induced changes in blood pressure in spontaneously breathing patients. J Clin Anesth 2016; 34: 318-24. 91. Kuwata S, Suehiro K, Juri T, Tsujimoto S, Mukai A, Tanaka K, Yamada T, Mori T, Nishikawa K. Pleth variability index can predict spinal anaesthesia-induced hypotension in patients undergoing caesarean delivery. Acta Anaesthesiol Scand 2018; 62: 75-84. 92. Dorlas JC, Nijboer JA. Photo-electric plethysmography as a monitoring device in anaesthesia. Application and interpretation. Br J Anaesth 1985; 57: 524-30. 85

93. Landsverk SA, Hoiseth LO, Kvandal P, Hisdal J, Skare O, Kirkeboen KA. Poor agreement between respiratory variations in pulse oximetry photoplethysmographic waveform amplitude and pulse pressure in intensive care unit patients. Anesthesiology 2008; 109: 849-55. 94. Gesquiere MJ, Awad AA, Silverman DG, Stout RG, Jablonka DH, Silverman TJ, Shelley KH. Impact of withdrawal of 450 ml of blood on respiration-induced oscillations of the ear plethysmographic waveform. J Clin Monit Comput 2007; 21: 277-82. 95. Hoiseth LO, Hoff IE, Hagen OA, Kirkeboen KA, Landsverk SA. Respiratory variations in the photoplethysmographic waveform amplitude depend on type of pulse oximetry device. J Clin Monit Comput 2016; 30: 317-25. 96. Feldman JM. Can clinical monitors be used as scientific instruments? Anesth Analg 2006; 103: 1071-2. 97. Shelley KH. Photoplethysmography: beyond the calculation of arterial oxygen saturation and heart rate. Anesth Analg 2007; 105: S31-6, tables of contents. 98. Addison PS. A review of signal processing used in the implementation of the pulse oximetry photoplethysmographic fluid responsiveness parameter. Anesth Analg 2014; 119: 1293-306. 99. Addison PS, Wang R, Uribe AA, Bergese SD. Increasing signal processing sophistication in the calculation of the respiratory modulation of the photoplethysmogram (DPOP). J Clin Monit Comput 2015; 29: 363-72. 100. Anderson CT, Breen PH. Carbon dioxide kinetics and capnography during critical care. Crit Care 2000; 4: 207-15. 101. Monnet X, Bataille A, Magalhaes E, Barrois J, Le Corre M, Gosset C, Guerin L, Richard C, Teboul J-L. End-tidal carbon dioxide is better than arterial pressure for predicting volume responsiveness by the passive leg raising test. Intensive Care Med 2013; 39: 93-100. 102. Garnett AR, Ornato JP, Gonzalez ER, Johnson EB. End-tidal carbon dioxide monitoring during cardiopulmonary resuscitation. JAMA 1987; 257: 512-5. 103. Pernat A, Weil MH, Sun S, Tang W. Stroke volumes and end-tidal carbon dioxide generated by precordial compression during ventricular fibrillation. Crit Care Med 2003; 31: 1819-23. 104. Weil MH, Bisera J, Trevino RP, Rackow EC. Cardiac output and end-tidal carbon dioxide. Crit Care Med 1985; 13: 907-9. 105. Ornato JP, Garnett AR, Glauser FL. Relationship between cardiac output and the end-tidal carbon dioxide tension. Ann Emerg Med 1990; 19: 1104-6. 106. Shibutani K, Muraoka M, Shirasaki S, Kubal K, Sanchala VT, Gupte P. Do changes in end-tidal PCO2 quantitatively reflect changes in cardiac output? Anesth Analg 1994; 79: 829-33. 107. Wahba RW, Tessler MJ, Beique F, Kleiman SJ. Changes in PCO2 with acute changes in cardiac index. Can J Anaesth 1996; 43: 243-5. 108. Monge Garcia MI, Gil Cano A, Gracia Romero M, Monterroso Pintado R, Perez Madueno V, Diaz Monrove JC. Non-invasive assessment of fluid responsiveness by changes in partial end-tidal CO2 pressure during a passive leg-raising maneuver. Ann Intensive Care 2012; 2: 9. 109. Jacquet-Lagreze M, Baudin F, David JS, Fellahi JL, Hu PB, Lilot M, Piriou V. End-tidal carbon dioxide variation after a 100- and a 500-ml fluid challenge to assess fluid responsiveness. Ann Intensive Care 2016; 6: 37. 110. Toupin F, Clairoux A, Deschamps A, Lebon JS, Lamarche Y, Lambert J, Fortier A, Denault AY. Assessment of fluid responsiveness with end-tidal carbon dioxide using a simplified passive leg raising maneuver: a prospective observational study. Can J Anaesth 2016; 63: 1033-41. 111. Isserles SA, Breen PH. Can changes in end-tidal PCO2 measure changes in cardiac output? Anesth Analg 1991; 73: 808-14. 112. Young A, Marik PE, Sibole S, Grooms D, Levitov A. Changes in end-tidal carbon dioxide and volumetric carbon dioxide as predictors of volume responsiveness in hemodynamically unstable patients. J Cardiothorac Vasc Anesth 2013; 27: 681-4. 86

113. Suarez-Sipmann F, Bohm SH, Tusman G. Volumetric capnography: the time has come. Curr Opin Crit Care 2014; 20: 333-9. 114. Siobal MS. Monitoring Exhaled Carbon Dioxide. Respir Care 2016; 61: 1397-416. 115. Arango-Granados MC, Zarama Cordoba V, Castro Llanos AM, Bustamante Cristancho LA. Evaluation of end-tidal carbon dioxide gradient as a predictor of volume responsiveness in spontaneously breathing healthy adults. Intensive Care Med Exp 2018; 6: 21. 116. Elfil M, Negida A. Sampling methods in Clinical Research; an Educational Review. Emerg (Tehran) 2017; 5: e52. 117. Hisdal J, Toska K, Walloe L. Design of a chamber for lower body negative pressure with controlled onset rate. Aviat Space Environ Med 2003; 74: 874-8. 118. Cooke WH, Ryan KL, Convertino VA. Lower body negative pressure as a model to study progression to acute hemorrhagic shock in humans. J Appl Physiol (1985) 2004; 96: 1249-61. 119. Goswami N, Blaber AP, Hinghofer-Szalkay H, Convertino VA. Lower Body Negative Pressure: Physiological Effects, Applications, and Implementation. Physiol Rev 2019; 99: 807-51. 120. Convertino VA, Cooke WH, Holcomb JB. Arterial pulse pressure and its association with reduced stroke volume during progressive central hypovolemia. J Trauma 2006; 61: 629-34. 121. Hinojosa-Laborde C, Howard JT, Mulligan J, Grudic GZ, Convertino VA. Comparison of compensatory reserve during lower-body negative pressure and hemorrhage in nonhuman primates. Am J Physiol Regul Integr Comp Physiol 2016; 310: R1154-9. 122. Lindenberger M, Olsen H, Lanne T. Lower capacitance response and capillary fluid absorption in women to defend central blood volume in response to acute hypovolemic circulatory stress. Am J Physiol Heart Circ Physiol 2008; 295: H867-73. 123. Dahl M, Hayes C, Steen Rasmussen B, Larsson A, Secher NH. Can a central blood volume deficit be detected by systolic pressure variation during spontaneous breathing? BMC Anesthesiol 2016; 16: 58. 124. Pizov R, Eden A, Bystritski D, Kalina E, Tamir A, Gelman S. Arterial and plethysmographic waveform analysis in anesthetized patients with hypovolemia. Anesthesiology 2010; 113: 83-91. 125. Hamby RI, Aintablian A. Preload reduction with right ventricular pacing: effects on left ventricular hemodynamics and contractile pattern. Clin Cardiol 1980; 3: 169-77. 126. Saito M, Kaye G, Negishi K, Linker N, Gammage M, Kosmala W, Marwick TH, Protect-Pace i. Dyssynchrony, contraction efficiency and regional function with apical and non-apical RV pacing. Heart 2015; 101: 600-8. 127. Monnet X, Marik PE, Teboul JL. Prediction of fluid responsiveness: an update. Ann Intensive Care 2016; 6: 111. 128. Monnet X, Marik P, Teboul J-L. Passive leg raising for predicting fluid responsiveness: a systematic review and meta-analysis. Intensive Care Med 2016; 42: 1935-47. 129. Cavallaro F, Sandroni C, Marano C, La Torre G, Mannocci A, De Waure C, Bello G, Maviglia R, Antonelli M. Diagnostic accuracy of passive leg raising for prediction of fluid responsiveness in adults: systematic review and meta-analysis of clinical studies. Intensive Care Med 2010; 36: 1475-83. 130. Teboul J-L, Monnet X. Prediction of volume responsiveness in critically ill patients with spontaneous breathing activity. Curr Opin Crit Care 2008; 14: 334-39. 131. Fougeres E, Teboul JL, Richard C, Osman D, Chemla D, Monnet X. Hemodynamic impact of a positive end-expiratory pressure setting in acute respiratory distress syndrome: importance of the volume status. Crit Care Med 2010; 38: 802-7. 132. Mahjoub Y, Touzeau J, Airapetian N, Lorne E, Hijazi M, Zogheib E, Tinturier F, Slama M, Dupont H. The passive leg-raising maneuver cannot accurately predict fluid responsiveness in patients with intra-abdominal hypertension. Crit Care Med 2010; 38. 133. Malbrain ML, Reuter DA. Assessing fluid responsiveness with the passive leg raising maneuver in patients with increased intra-abdominal pressure: be aware that not all blood returns! Crit Care Med 2010; 38: 1912-5. 87

134. Eriksen M, Walloe L. Improved method for cardiac output determination in man using ultrasound Doppler technique. Med Biol Eng Comput 1990; 28: 555-60. 135. Berton C, Cholley B. Equipment review: new techniques for cardiac output measurement-- oesophageal Doppler, Fick principle using carbon dioxide, and pulse contour analysis. Crit Care 2002; 6: 216-21. 136. Valtier B, Cholley BP, Belot JP, de la Coussaye JE, Mateo J, Payen DM. Noninvasive monitoring of cardiac output in critically ill patients using transesophageal Doppler. Am J Respir Crit Care Med 1998; 158: 77-83. 137. Dark PM, Singer M. The validity of trans-esophageal Doppler ultrasonography as a measure of cardiac output in critically ill adults. Intensive Care Med 2004; 30: 2060-6. 138. Wakeling HG, McFall MR, Jenkins CS, Woods WG, Miles WF, Barclay GR, Fleming SC. Intraoperative oesophageal Doppler guided fluid management shortens postoperative hospital stay after major bowel surgery. Br J Anaesth 2005; 95: 634-42. 139. Saugel B, Cecconi M, Wagner JY, Reuter DA. Noninvasive continuous cardiac output monitoring in perioperative and intensive care medicine. Br J Anaesth 2015; 114: 562-75. 140. Lansdorp B, Ouweneel D, de Keijzer A, van der Hoeven JG, Lemson J, Pickkers P. Non-invasive measurement of pulse pressure variation and systolic pressure variation using a finger cuff corresponds with intra-arterial measurement. Br J Anaesth 2011; 107: 540-5. 141. Broch O, Renner J, Gruenewald M, Meybohm P, Schottler J, Caliebe A, Steinfath M, Malbrain M, Bein B. A comparison of the Nexfin(R) and transcardiopulmonary thermodilution to estimate cardiac output during coronary artery surgery. Anaesthesia 2012; 67: 377-83. 142. Chen G, Meng L, Alexander B, Tran NP, Kain ZN, Cannesson M. Comparison of noninvasive cardiac output measurements using the Nexfin monitoring device and the esophageal Doppler. J Clin Anesth 2012; 24: 275-83. 143. Jaffe MB. Respiratory Gas Analysis-Technical Aspects. Anesth Analg 2018; 126: 839-45. 144. Harrison XA, Donaldson L, Correa-Cano ME, Evans J, Fisher DN, Goodwin CED, Robinson BS, Hodgson DJ, Inger R. A brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ 2018; 6: e4794. 145. Seltman HJ. Experimental design and analysis. Online at: http://www stat cmu edu/, hseltman/309/Book/Book pdf 2012. 146. Altman DG. Practical statistics for medical research: CRC press, 1990. 147. Kallner A. Bayes' theorem, the ROC diagram and reference values: Definition and use in clinical diagnosis. Biochem Med (Zagreb) 2018; 28: 010101. 148. Wyller VB, Godang K, Morkrid L, Saul JP, Thaulow E, Walloe L. Abnormal thermoregulatory responses in adolescents with chronic fatigue syndrome: relation to clinical symptoms. Pediatrics 2007; 120: e129-37. 149. Cannesson M, Besnard C, Durand PG, Bohe J, Jacques D. Relation between respiratory variations in pulse oximetry plethysmographic waveform amplitude and arterial pulse pressure in ventilated patients. Crit Care 2005; 9: R562-8. 150. Natalini G, Rosano A, Franceschetti ME, Facchetti P, Bernardini A. Variations in arterial blood pressure and photoplethysmography during mechanical ventilation. Anesth Analg 2006; 103: 1182-8. 151. Hengy B, Gazon M, Schmitt Z, Benyoub K, Bonnet A, Viale JP, Aubrun F. Comparison between respiratory variations in pulse oximetry plethysmographic waveform amplitude and arterial pulse pressure during major abdominal surgery. Anesthesiology 2012; 117: 973-80. 152. Hoiseth LO, Hoff IE, Skare O, Kirkeboen KA, Landsverk SA. Photoplethysmographic and pulse pressure variations during abdominal surgery. Acta Anaesthesiol Scand 2011; 55: 1221-30. 153. Solus-Biguenet H, Fleyfel M, Tavernier B, Kipnis E, Onimus J, Robin E, Lebuffe G, Decoene C, Pruvot FR, Vallet B. Non-invasive prediction of fluid responsiveness during major hepatic surgery. Br J Anaesth 2006; 97: 808-16. 88

154. Alian AA, Shelley KH. Photoplethysmography. Best Pract Res Clin Anaesthesiol 2014; 28: 395- 406. 155. van Genderen ME, Bartels SA, Lima A, Bezemer R, Ince C, Bakker J, van Bommel J. Peripheral perfusion index as an early predictor for central hypovolemia in awake healthy volunteers. Anesth Analg 2013; 116: 351-6. 156. Hasanin A, Mohamed SAR, El-Adawy A. Evaluation of perfusion index as a tool for pain assessment in critically ill patients. J Clin Monit Comput 2017; 31: 961-65. 157. Nassar BS, Schmidt GA. Capnography During Critical Illness. Chest 2016; 149: 576-85. 158. Hoiseth LO, Hagemo JS. Predicting fluid responsiveness in whom? A simulated example of patient spectrum influencing the receiver operating characteristics curve. J Clin Monit Comput 2018; 32: 215-19. 159. Saugel B, Flick M, Bendjelid K, Critchley LAH, Vistisen ST, Scheeren TWL. Journal of clinical monitoring and computing end of year summary 2018: hemodynamic monitoring and management. J Clin Monit Comput 2019; 33: 211-22. 160. Lakhal K, Nay MA, Kamel T, Lortat-Jacob B, Ehrmann S, Rozec B, Boulain T. Change in end- tidal carbon dioxide outperforms other surrogates for change in cardiac output during fluid challenge. Br J Anaesth 2017; 118: 355-62. 161. Jin X, Weil MH, Tang W, Povoas H, Pernat A, Xie J, Bisera J. End-tidal carbon dioxide as a noninvasive indicator of cardiac index during circulatory shock. Crit Care Med 2000; 28: 2415-19. 162. Dahl MK, Vistisen ST, Koefoed-Nielsen J, Larsson A. Using an expiratory resistor, arterial pulse pressure variations predict fluid responsiveness during spontaneous breathing: an experimental porcine study. Crit Care 2009; 13: R39. 163. Hong DM, Lee JM, Seo JH, Min JJ, Jeon Y, Bahk JH. Pulse pressure variation to predict fluid responsiveness in spontaneously breathing patients: tidal vs. forced inspiratory breathing. Anaesthesia 2014; 69: 717-22. 164. De Backer D, Pinsky MR. Can one predict fluid responsiveness in spontaneously breathing patients? Intensive Care Med 2007; 33: 1111-13. 165. Magder S, Georgiadis G, Cheong T. Respiratory variations in right atrial pressure predict the response to fluid challenge. J Crit Care 1992; 7: 76-85. 166. Monnet X, Bleibtreu A, Ferre A, Dres M, Gharbi R, Richard C, Teboul JL. Passive leg-raising and end-expiratory occlusion tests perform better than pulse pressure variation in patients with low respiratory system compliance. Crit Care Med 2012; 40: 152-7. 167. Moher D, Hopewell S, Schulz KF, Montori V, Gotzsche PC, Devereaux PJ, Elbourne D, Egger M, Altman DG. CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials. BMJ 2010; 340: c869. 168. Monnet X, Chemla D, Osman D, Anguel N, Richard C, Pinsky MR, Teboul JL. Measuring aortic diameter improves accuracy of esophageal Doppler in assessing fluid responsiveness. Crit Care Med 2007; 35: 477-82.

12 PAPERS

Hindawi Publishing Corporation Critical Care Research and Practice Volume 2014, Article ID 712728, 9 pages http://dx.doi.org/10.1155/2014/712728

Research Article Respiratory Variations in Pulse Pressure Reflect Central Hypovolemia during Noninvasive Positive Pressure Ventilation

Ingrid Elise Hoff,1,2 Lars Øivind Høiseth,2,3 Jonny Hisdal,4 Jo Røislien,1,5 Svein Aslak Landsverk,2 and Knut Arvid Kirkebøen2,3

1 Norwegian Air Ambulance Foundation, Holterveien 24, 1441 Drøbak, Norway 2 DepartmentofAnaesthesiology,OsloUniversityHospital,P.O.Box4956,Nydalen,0424Oslo,Norway 3 Faculty of Medicine, University of Oslo, P.O. Box 1072 Blindern, 0316 Oslo, Norway 4 Department of Vascular Medicine, Oslo University Hospital, P.O. Box 4956, Nydalen, 0424 Oslo, Norway 5 Department of Biostatistics, Institute of Basic Medical Sciences, University of Oslo, P.O. Box 1072 Blindern, 0316 Oslo, Norway

Correspondence should be addressed to Ingrid Elise Hoff; [email protected]

Received 14 September 2013; Revised 9 December 2013; Accepted 22 December 2013; Published 19 February 2014

Academic Editor: Djillali Annane

Copyright © 2014 Ingrid Elise Hoff et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background. Correct volume management is essential in patients with respiratory failure. We investigated the ability of respiratory variations in noninvasive pulse pressure (ΔPP), photoplethysmographic waveform amplitude (ΔPOP), and pleth variability index (PVI) to reflect hypovolemia during noninvasive positive pressure ventilation by inducing hypovolemia with progressive lower body negative pressure (LBNP). Methods. Fourteen volunteers underwent LBNP of 0, −20, −40, −60, and −80 mmHg for 4.5 min at each level or until presyncope. The procedure was repeated with noninvasive positive pressure ventilation. We measured stroke volume (suprasternal Doppler), ΔPP (Finapres), ΔPOP,andPVIandassessedtheirassociationwithLBNP-levelusinglinearmixed model regression analyses. Results. Stroke volume decreased with each pressure level (−11.2 mL, 95% CI −11.8, −9.6, 𝑃 < 0.001), with an additional effect of noninvasive positive pressure ventilation− ( 3.0 mL, 95% CI −8.5, −1.3, 𝑃 = 0.009). ΔPP increased for each LBNP-level (1.2%, 95% CI 0.5, 1.8, 𝑃 < 0.001) and almost doubled during noninvasive positive pressure ventilation (additional increase 1.0%, 95% CI 0.1, 1.9, 𝑃 = 0.003). Neither ΔPOP nor PVI was significantly associated with LBNP-level. Conclusions. During noninvasive positive pressure ventilation, preload changes were reflected by ΔPP but not by ΔPOP or PVI. This implies that ΔPP may be used to assess volume status during noninvasive positive pressure ventilation.

1. Introduction to this uncertainty, the applicability of dynamic variables is currently limited to patients on controlled mechanical Δ Pulse pressure variations ( PP) reflect volume status during ventilation. However, with increasing data on negative conse- mechanical ventilation [1]. Respiratory variations in the quences of intubation and mechanical ventilation, noninva- photoplethysmographic waveform amplitude (ΔPOP) and sive positive pressure ventilation (NPPV) is frequently used the pleth variability index (PVI) are proposed as noninvasive in emergency departments and intensive care units. Patients alternatives [2, 3]. In spontaneously breathing subjects, only treated with NPPV are often on the verge of respiratory passive leg raise and the end-expiratory occlusion test have failure, and correct fluid management is essential. been shown to consistently reflect preload dependency4 [ , 5]. The aim of the present study was therefore to explore However, the literature is divided on two major issues con- the ability of dynamic variables to reflect graded hypov- cerning the usefulness of dynamic variables: whether they are olemia during NPPV. Lower body negative pressure (LBNP) applicable during spontaneous breathing [6, 7]andwhether is a well-established model for central hypovolemia and the photoplethysmographic waveform derived variables are preload reduction [10, 11]. This noninvasive model in healthy useful alternatives to pulse pressure variation [8, 9]. Due volunteers enables investigation of arterial pressure and 2 Critical Care Research and Practice

Spontaneous breathing Noninvasive positive pressure ventilation

Baseline Baseline −20 −20 −40 −40 −60 LBNP −60 LBNP −80 (mmHg) −80 (mmHg)

Time Time 4.5 4.5 4.5 4.5 4.5 (min) 4.5 4.5 4.5 4.5 4.5 (min)

Figure 1: Schematic illustration of the experimental protocol. Each level was kept for 4.5 min. LBNP: lower body negative pressure. photoplethysmographic waveform derived variables without 2.2. Data Acquisition and Analysis. Data were recorded over pain or medication, factors which could influence the results. the total interval of each LBNP-level, that is, 4.5 min. Data We hypothesized that the reduction in stroke volume (SV) from all completed LBNP-levels are included in the analysis. induced by increasing levels of LBNP would be aggravated Calculations were made from data sampled and averaged with the application of NPPV and that the reduction in SV over 10 consecutive respiratory cycles without arrhythmia. wouldbereflectedinthedynamicvariablesΔPP, ΔPOP, and Respiratory movements were recorded with a custom-made PVI. air flowmeter. During NPPV tidal volume, respiratory rate, airway pressures, leakage, and spontaneous breathing activity were continuously measured and recorded every 10 s using 2. Materials and Methods commercial software (VentView 2.0, Drager¨ Medical Ag & After approval by the regional ethics committee (REK Sør- Co, Lubeck,¨ Germany). In order to investigate the effects of Øst, ref.no 2009/2180, December 2009), written informed NPPV the data was manually filtered after recording and 10 consecutive respiratory cycles (1 min) without excessive consent was obtained from 14 healthy volunteers (7 male, 7 > female, aged 28 ± 7years,height177± 10 cm, and weight 71 ± spontaneous breathing ( 5% of the corresponding minute volume per respiratory cycle) were identified. Hemodynamic 13 kg (mean ± SD)). The subjects were instructed to refrain measurements from the corresponding minute were then from alcohol or caffeinated drinks 24 hours prior to partic- analyzed. ΔPP and ΔPOPwerecalculatedinacustom ipation. Pregnant women and subjects using cardiovascular made program (Labview 8.2; National Instruments, TX, medication were not included. USA), according to Michard [13]. One respiratory cycle was manually delimitated and the program displayed 2.1. Experimental Protocol. Subjects were in the supine posi- corresponding blood pressure and photoplethysmographic tion during experiments, which were performed in room waveform curves. The photoplethysmographic waveform, temperature. LBNP was applied by a custom made LBNP- PVI, and PI were obtained from a finger clip (Masimo chamber previously described [12]andinducedbystepwise Radical 7, version 7.3.1.1, Masimo Corp., Irvine, CA, USA) suction of air out of the chamber. After baseline measure- onthethirdfingeroftherighthand,whichwascoveredto ments, subjects underwent consecutive LBNP-pressures of prevent temperature loss and disturbance of the signal from −20, −40, −60, and −80 mmHg. Each level was kept for ambientlight.PVIandPIwerecalculatedaccordingtothe 4.5 min. After minimum 15 min rest, the procedure was manufacturer’s algorithms (http://www.masimo.com/pdf/ repeated with NPPV. NPPV was applied via a face mask whitepaper/LAB4583A.pdf; 13.09.2013). Averaging period with intermittent positive pressure (IPPV-mode), tidal vol- forPIwassetto2sinthepulseoximeter.PVIandPI ume 10 mL/kg ideal weight, positive end-expiratory pressure were downloaded using the TrendCom software (Masimo) (PEEP) = 0 cm H2O, fraction of inspired oxygen 0.21, and and averaged over 1 min. Continuous arterial pressure was respiratory frequency of 10–12/min (Evita 4, Drager¨ Mediz- obtained noninvasively at heart level from the left third finger intechnik GmbH, Lubeck,¨ Germany). Spontaneous breathing (Finometer, FMS Finapres Medical Systems BV, Amsterdam, and mask leakage were minimized by thorough mask adjust- The Netherlands). SV was obtained continuously with mentandbyensuringcompliancewiththeventilationmode suprasternal Doppler (SD-100, GE Vingmed Ultrasound, before data recordings. The protocol was discontinued if Horten, Norway) by an experienced operator. An angle ∘ one of the following events occurred: systolic blood pressure of 20 and a diameter of the aortic valve of 20 mm were <70 mmHg, a sudden decrease in systolic blood pressure assumed in the calculation of SV from the velocity-time ≥15 mmHg, a decrease in heart rate (HR) ≥15 beats/min, integrals. Heart rate (HR) was obtained from a standard dizziness, sweating, or nausea. The experimental protocol is 3-lead electrocardiogram (ECG). ECG, arterial pressure, and illustrated in Figure 1. photoplethysmographic waveforms were sampled at 400 Hz. Critical Care Research and Practice 3

Table 1: Dynamic variables and hemodynamic data during spontaneous breathing and noninvasive positive pressure ventilation.

Subjects completing ΔPP (%) ΔPOP (%) PVI (%) PI (%) LBNP-level the LBNP-level (𝑛) SB NPPV SB NPPV SB NPPV SB NPPV SB NPPV Baseline 14 12 7.4 ± 3.1 7.1 ± 2.4 12.4 ± 5.4 13.5 ± 10.0 18.5 ± 6.9 18.4 ± 10.0 2.8 ± 2.1 3.2 ± 2.0 20 14 12 7.6 ± 2.5 7.1 ± 1.9 12.7 ± 6.1 14.0 ± 7.5 18.8 ± 8.0 17.8 ± 11.6 2.4 ± 1.9 2.5 ± 1.4 40 13 10 8.3 ± 8.0 9.8 ± 4.4 13.4 ± 8.0 14.4 ± 6.7 18.6 ± 7.2 20.1 ± 10.1 2.2 ± 1.6 2.7 ± 1.4 60 12 10 9.3 ± 3.8 12.1 ± 5.2 12.2 ± 4.1 14.4 ± 6.2 21.1 ± 8.3 22.1 ± 9.9 2.3 ± 1.4 2.5 ± 1.1 80 11 9 12.6 ± 7.1 15.7 ± 6.0 16.6 ± 5.5 18.4 ± 6.6 22.5 ± 8.5 26.9 ± 10.4 2.1 ± 1.2 2.1 ± 1.3 Subjects completing SV (mL) HR (beats/min) MAP (mm Hg) PP (mm Hg) LBNP-level the LBNP-level (𝑛) SB NPPV SB NPPV SB NPPV SB NPPV SB NPPV Baseline 14 12 80 ± 13 79 ± 12 59 ± 8 57 ± 6 78 ± 11 78 ± 17 58 ± 12 62 ± 9 20 14 12 73 ± 14 67 ± 14 58 ± 8 59 ± 8 76 ± 11 75 ± 18 60 ± 14 60 ± 17 40 13 10 60 ± 13 53 ± 15 64 ± 9 67 ± 11 77 ± 13 76 ± 20 57 ± 14 55 ± 17 60 12 10 50 ± 13 43 ± 11 73 ± 11 78 ± 14 77 ± 12 77 ± 19 51 ± 11 52 ± 16 80 11 9 35 ± 12 32 ± 8 89 ± 17 93 ± 20 77 ± 14 72 ± 23 45 ± 9 45 ± 19 Data are mean ± SD. LBNP: lower body negative pressure; SB: spontaneous breathing; NPPV: noninvasive positive pressure ventilation; ΔPP: pulse pressure variation; ΔPOP: photoplethysmographic waveform variation; PVI: pleth variability index; PI: perfusion index; HR: heart rate; SV: stroke volume; MAP: mean arterial pressure; PP: pulse pressure.

2.3. Statistics. The primary endpoint was the change in ΔPP, successful testing but completed the LBNP-series during ΔPOP,PVI, and PI following the transition from spontaneous spontaneous breathing. Data from these two subjects are breathing to noninvasive positive pressure ventilation. Data included in the analysis of the effect of LBNP on hemo- are given as mean ± SD unless otherwise stated. The associ- dynamic variables. Hemodynamic data are summarized in ations between the independent variables LBNP and NPPV Table 1 and mean values shown in Figure 2.Resultsfromthe and the dependent variables HR, mean arterial pressure linear mixed model analyses are given in Table 2. Different (MAP), SV, ΔPP, ΔPOP, PVI, and PI were analyzed by linear effects of LBNP and NPPV on hemodynamic variables are mixed model regression analyses. Linear mixed model is a illustrated in Figure 3. generalization of traditional linear regression, which adjusts SV decreased significantly with progressive levels of for the correlation between repeated measurements within LBNP, with a mean reduction of 11.2 mL (95% CI −11.8, −9.6, each subject and finds the best linear fit to the data across 𝑃 < 0.001) between each LBNP-level. After application of all individuals. The model maximizes power by utilizing all NPPV,SV was significantly lower at all LBNP-levels− ( 3.0 mL, data despite missing observations following the premature 95% CI −8.5, −1.3, 𝑃 = 0.009)comparedtospontaneous termination of the LBNP-protocol in some subjects. The breathing (Figure 3(b)). ΔPP was significantly affected both effect estimates describe the mean effect of LBNP onthe by LBNP-level alone and by the interaction between LBNP hemodynamic variables when going from one LBNP-level to and NPPV (Figure 3(c)). Whereas ΔPP increased by 1.2% the next. The difference in hemodynamic variables with and (95% CI 0.5, 1.8, 𝑃 < 0.001) between each LBNP-level during without NPPV was assessed by adding the interaction term spontaneous breathing, the application of NPPV led to an LBNP∗NPPV in the regression model. Introduction of an additional increase of 1.0% (95% CI 0.1, 1.9, 𝑃 = 0.033) interaction term is necessary where the effect of one variable during LBNP, almost a doubling. NPPV alone did not affect (LBNP) is affected by the presence or value of another variable ΔPP significantly, meaning that NPPV only led to an increase (NPPV). Results are given as coefficients (beta-values) with in ΔPP during hypovolemia induced by LBNP. Neither 95% confidence intervals (CI). 𝑃 values below 0.05 were LBNP nor NPPV altered ΔPOP or PVI significantly. PI considered statistically significant. As no similar studies had decreased significantly with progressive LBNP-levels, but was previously been published, a conventional power analysis was not affected by NPPV. Heart rate was significantly affected not performed. Due to the experimental nature of the study, by LBNP alone. MAP did not change significantly during the the number of study subjects was limited to 14. Statistical LBNP-protocol or following the transition from spontaneous calculationswereperformedinSPSS19.0(SPSSInc.,Chicago, breathing to NPPV (Figure 3(a)). Pulse pressure decreased IL, USA) and R 2.12 (R Foundation for Statistical Computing, significantly between each LBNP-level− ( 3.5 mmHg, 95% CI 2011). −4.5, −2.0, 𝑃 < 0.001) but was unaffected by NPPV.

3. Results 4. Discussion The number of subjects completing each level is shown in The main finding in this experimental study is that ΔPP Table 1. Two subjects failed to comply with NPPV despite consistently reflected progressive central hypovolemia during 4 Critical Care Research and Practice # # value value 0.001 𝑃 𝑃 < PP significantly Δ 2.0) − NPPV). 2.5, 1.3) 0.519 0.4, 0.2) 0.504 0.4, 0.0) 0.018 0.2, 1.0) 0.196 4.5, 3.2, 5.7) 0.586 − − ∗ − − − − 0.6 ( 0.1 ( 0.1 ( 3.5 ( − − − − POP: photoplethysmographic waveform Δ ing and NPPV, and NPPV led to a decrease in SV value Estimate (95% CI) value Estimate (95% CI) : statistically significant. 𝑃 𝑃 # NPPV), which means that NPPV only affected essure; ∗ PP: pulse pressure variation; Δ 1.5, 0.6) 0.415 2.2, 0.8) 0.347 4.0, 2.4) 0.609 0.4 ( 3.1, 4.0) 0.822 1.7 ( 0.5, 1.4) 0.337 0.4, 2.3) 0.186 − − − − − − 80 mmHg). Estimates of NPPV-effects are constant and independent of LBNP-level. 0.4 ( 1.0 ( 0.7 ( − − − − 60, and − 40, # − 20, value Estimate (95% CI) value Estimate (95% CI) 0.001 − 𝑃 𝑃 < ect estimates of LBNP and NPPV separately and in combination (LBNP NPPV: interaction between LBNP and NPPV; ∗ POP (%) PVI (%) PI (%) Δ 1.0, 1.4) 0.777 1.1 ( 0.3, 1.3) 0.233 1.0 ( 1.7, 3.9) 0.435 4.1, 4.7) 0.898 0.4 ( 0.2, 3.5) 0.091 − − − − − V. This meansthat LBNP led todecrease a SV in during both spontaneous breath 7.3 (6.0, 8.6) 0.2 ( 0.7 ( 0.3 ( PP is due to a significantinteraction between LBNP and NPPV (LBNP Δ # # # # value Estimate (95% CI) value Estimate (95% CI) 0.001 0.001 0.009 𝑃 𝑃 < < < PP (%) 9.6) 1.3) Δ SV (mL) HR (beats/min) MAP (mm Hg) PP (mm Hg) − − NPPV-effects are given as changes per 20 mmHg change in LBNP ( ∗ 2.7, 1.5) 0.571 1.0 ( 2.3, 0.7) 0.305 1.6 ( 8.5, 11.8, − − − − 0.6 ( 0.7 ( 3.0 ( − 11.2 ( − − − Estimate (95% CI) Estimate (95% CI) Table 2: Data from the generalized mixed model analyses showing eff NPPV 1.0 (0.1, 1.9) 0.033 NPPV ∗ ∗ LBNP LBNPNPPV LBNP NPPV 1.2 (0.5, 1.8) LBNP Estimates of LBNP-effects and LBNP There are separate statistically significant effects of both LBNPduring hypovolemia.and PI NPPV on S is significantly affectedLBNP: lower by LBNP body alone. negative pressure; NPPV: noninvasive positive pressure ventilation; LBNP both during normo- and hypovolemia. In contrast, the effect of NPPV on amplitude variation; PVI: pleth variability index; PI: perfusion index; SV: stroke volume; HR: heart rate; MAP: mean arterial pressure; PP: pulse pr Critical Care Research and Practice 5

30 30 50

25 25 40

20 20 30 15 15 PP (%) PVI (%) POP (%) POP

∆ 20 10 ∆ 10

10 5 5

0 0 0 0 −20 −40 −60 −80 0 −20 −40 −60 −80 0−20−40 −60 −80 LBNP (mmHg) LBNP (mmHg) LBNP (mmHg) (a) (b) (c) 6 120 120

5 100 100 4 80

3 60 80 SV (mL) 2 40 HR (beats/min) 60

Perfusion index (PI) (%) Perfusion 1 20

0 0 40 0 −20 −40 −60 −80 0−20−40 −60 −80 0 −20 −40 −60 −80 LBNP (mmHg) LBNP (mmHg) LBNP (mmHg) (d) (e) (f) 100 80

70 80 60

60 50

40 MAP (mmHg) 40 Pulse pressure (mmHg) Pulse pressure 30

20 20 0 −20 −40 −60 −80 0 −20 −40 −60 −80 LBNP (mmHg) LBNP (mmHg) (g) (h) Figure 2: Line charts of mean values at each LBNP-level for ΔPP, ΔPOP, PVI, PI, SV, HR, MAP, and PP. Open circle: measurements during spontaneous breathing. Full circle: measurements during NPPV. 1 SD illustrated with one-sided error bars for clarity. ΔPP: respiratory variations in pulse pressure, ΔPOP: respiratory variations in the photoplethysmographic waveform amplitude, PVI: pleth variability index, PI:perfusionindex,SV:strokevolume,HR:heartrate,MAP:meanarterialpressure,PP:pulsepressure,NPPV:noninvasivepositivepressure ventilation, LBNP: lower body negative pressure.

NPPV.The application of NPPV accentuated the reduction in 4.1. SV and Pulse Pressure. The effects of LBNP on neuro- SV induced by LBNP.This accentuation was reflected by ΔPP humoral and sympathetic neural activity have been thor- but not by ΔPOP or PVI. There were no associations between oughly described [10]. Increased vasomotor tone during early LBNP-levels and ΔPOP or PVI. Neither ΔPOP nor PVI hypovolemia preserves systemic blood pressure and may in changed significantly with the transition from spontaneous combination with increased heart rate mask reduced SV. ventilation to NPPV. We observed this compensatory response in the present 6 Critical Care Research and Practice Hemodynamic variable Hemodynamic Hemodynamic variable Hemodynamic Hemodynamic variable Hemodynamic

0 −80 0 −80 0 −80 LBNP-level LBNP-level LBNP-level SB SB SB NPPV NPPV NPPV (a) (b) (c)

Figure 3: Illustration of different effects of LBNP and NPPV on hemodynamic variables. (a) No effect of LBNP or NPPV alone orin combination (for instance MAP), (b) independent effects of both LBNP and NPPV (stroke volume), and (c) independent effect of LBNP which increases in combination with NPPV (interaction; ΔPP). No effect of NPPV alone. LBNP: lower body negative pressure, NPPV: noninvasive positive pressure ventilation, SB: spontaneous breathing, MAP: mean arterial pressure, ΔPP: pulse pressure variation. study. Pulse pressure decreased with progressive LBNP-levels, sensitivity was low. In a study on spontaneously breathing but the decrease in pulse pressure occurred later than the pigs ΔPP was significantly higher during hypovolemia, but decrease in SV.Similar results were reported in another study ΔPP only predicted fluid responsiveness when an expiratory on spontaneously breathing volunteers undergoing progres- resistor was added [20]. Whereas none of these studies sive LBNP [14]. NPPV leads to an accentuated reduction in show that ΔPP predicts fluid responsiveness with sufficient SV at all levels of LBNP except baseline. This is in accordance sensitivity, they indicate that pulse pressure variations may with the known impact of increased intrathoracic pressure on reflect hypovolemia during spontaneous breathing. This is venous return during central hypovolemia [15]. The opposite in line with our findings, but in addition we show how effect was demonstrated by Ryan et al. [16]whofoundthat ΔPP increases following the transition from spontaneous breathing through an inspiratory threshold device preserved breathing to NPPV. The substantial increase in ΔPP dur- SV by reducing intrathoracic pressure, which increased ing NPPV indicates that NPPV alters intrathoracic pres- LBNP-tolerance in healthy volunteers. sure similarly to invasive mechanical ventilation, provided compliance with the ventilator and minimal spontaneous 4.2. ΔPP, ΔPOP, PVI, and PI. Hemodynamic effects of breathing. Heenen et al. investigated dynamic variables in controlled mechanical ventilation [10]andtheabilityof patients with spontaneous breathing movements and found ΔPPtoreflecthypovolemiaduringmechanicalventilation that ΔPP varied substantially at baseline with no significant have previously been demonstrated [17]. There are also differences between fluid responders and nonresponders, and studies indicating that dynamic variables like stroke volume areas under the receiver operating characteristics curves were variation and vena collapsibility index may be useful during low [7]. No standardization of respiratory rate, tidal volumes, spontaneous respiration [6]. Heart-lung interactions differ or airway pressure was attempted in this study, which may in substantially between spontaneous breathing and mechanical part explain their results. By manually filtering the data after ventilation, and insufficient changes in intrathoracic pressure recordings we were able to investigate the effects of NPPV due to low, irregular tidal volumes and irregular respiratory alone, undisturbed by excessive spontaneous breathing. rates impede the use of respiratory induced variables to There are conflicting reports on ΔPOP as an alternative to evaluate preload or preload dependency. Whereas mechan- ΔPP to reflect hypovolemia and predict fluid responsiveness ical inspiration reduces right ventricular filling and increases [8, 21]. Differences in vasomotor tone, measurement sites, and right ventricular afterload, spontaneous inspiration increases measurement methodology contribute to discrepant results. both right ventricular filling and right ventricular afterload. Promising studies on variables derived from the photo- Deep inspiration could possibly induce sufficient changes plethysmographic waveform have mainly been performed in intrathoracic pressure to be reflected in SV and pulse in stable patients during anesthesia, over short periods of pressure. Preau´ et al. [18]foundthatΔPP predicted fluid time, and in the absence of sympathetic triggers such as responsiveness in spontaneously breathing patients undergo- advanced hypovolemia, surgery, pain, and stress [8, 22]. In ing a deep inspiratory maneuver but with lower sensitivity mechanically ventilated patients undergoing stepwise blood than reported in studies on mechanically ventilated patients. withdrawal, Pizov et al. [17] found that arterial waveform Similarly, Soubrier et al. [19] tested volume responsiveness variables detected mild hypovolemia earlier and more consis- in spontaneously breathing patients and found that ΔPP tently than photoplethysmographic variables. Interestingly, and variations in systolic blood pressure at baseline were correlations between ΔPOP and ΔPP improved with increas- significantly higher in responders than in nonresponders, but ing hypovolemia of up to 20% of estimated blood volume, Critical Care Research and Practice 7 with ΔPOP showing the largest variability. These were all thereductioninPIinthepresentstudy,andPIhasbeen stable patients with no signs of circulatory failure. As shown suggested as an early marker of the physiologic responses that in Table 1 and Figure 2,wefoundthatΔPOP increased with occur during hypovolemia [33]. We experienced temporary progressive LBNP, but due to large confidence intervals this loss of PVI signals in several subjects both during normo- and increase was not statistically significant. Other studies have hypovolemia, although the photoplethysmographic wave- demonstrated wide limits of agreement between ΔPP and form remained well defined. PI was <1inallcases.Theloss ΔPOP during surgery and intensive care [9, 21, 23, 24]. The of PVI-signal has been described earlier [3]. physiology behind the photoplethysmographic signal is very complex, influenced by cardiac and autonomic as well as 4.3. Methodological Considerations. First, even with mea- respiratory factors [25]. Due to rich innervation and large surements on several levels in each subject our sample size vascular plexuses the finger is very sensitive to vasomotion is limited. It is however comparable to other studies using [26]. Using spectral analysis on both spontaneously breathing the LBNP-model [34, 35] and we were able to demonstrate and mechanically ventilated patients, Shelley et al. [27]found significant effects of both hypovolemia and NPPV on SV and a correlation between estimated blood loss and ventilatory ΔPP. Second, we wanted to investigate the physiology in a effectsintheearsignalbutnotinthefingersignal.Webelieve completely noninvasive experimental model in order to avoid that increased sympathetic activation leading to vasocon- disturbing factors such as pain, agitation, and medication. striction is the main reason why ΔPOP to a lesser extent than Whereas pulse pressure variations are normally investigated ΔPP reflected hypovolemia in our study and largely explains using an invasive line, the Finapres technology has been the discrepancy between this study and others performed on validated for arterial pressure waveform analysis [21, 36]and stable patients during normovolemia. blood pressure measurements [37].Third,whiletheLBNP- In addition to different measurement sites, the use of model allows investigation of reversible central hypovolemia, different features of the photoplethysmographic signal may it is not suitable for conventional fluid responsiveness testing explain conflicting reports. In a LBNP-model, McGrath et using intravenous fluid boluses. However, in this model fluid al. [28] studied the correlation between changes in SV and responsiveness is “tested” by termination of LBNP, which changes in pulse width, amplitude, and area of an unfiltered inevitably leads to increased preload and the restoration of signal and found that it differed for all measurement sites. hemodynamic variables [10]. Finally, application of NPPV to The lowest correlation was between changes in SV and pulse healthy, nonsedated subjects is challenging. The major issues amplitudeinthefinger,thefeaturemostcommonlyused were to minimize spontaneous breathing and leakage. High when calculating respiratory changes in the photoplethysmo- tidal volumes, PEEP = 0 and IPPV-mode proved necessary graphic waveform. Commercial finger pulse oximeters are to ensure compliance with NPPV and keep tidal volumes frequently used in clinical practice and should be investi- stable. Despite a compensatory increase in frequency, a gated. However, probes from different manufacturers could reduction in tidal volumes to 8 mL/kg increased spontaneous also contribute to different results as sensitivity and signal breathing efforts and leakage. Previous studies indicate that processing differ [29]. We obtained both pulse pressure and airway pressures, ventilation mode, tidal volumes, and lung photoplethysmographic waveforms from noninvasive finger compliance affect the hemodynamic effects of mechanical probes. The two measurement methods are based on different ventilation and thus the performance of dynamic variables physiological principles (volume measurements in relatively [38–40]. Spontaneous breathing efforts varied in and between large finger arteries versus absorption of infrared light from subjects and might have affected central hemodynamics. the tissues, primarily reflecting microcirculatory changes). However, as we filtered the respiratory data obtained during It has previously been described how the complexity of the NPPV after recording we identified periods with high, stable photoplethysmographic signal renders it more susceptible to tidal volumes that were sufficient to result in measurable “noise” which may be physiological or technological artifacts preload changes. influencing the signal [29]. Our findings are in line with this explanation. 4.4. Conclusion. The main new finding in this experimental There was no significant relationship between PVI and Δ Δ LBNP-level in our study. Like other features of the pho- study on central hypovolemia is that PP, but not POP or toplethysmographic waveform PVI depends on stable per- PVI, is significantly associated with LBNP-level in healthy Δ fusion and vasomotion, as demonstrated in a recent study volunteers during NPPV. Clinically, this implies that PP where changes in vasomotor tone were induced by nore- maybeusedtoevaluatevolumestatusinpatientstreated pinephrine [30]. Another study showed that PVI failed with NPPV. Further, clinical studies are needed to clarify the to predict fluid responsiveness in mechanically ventilated potential for ΔPP in this setting. patients when PI was low (<4), whereas pulse pressure variations reliably did [31]. The PI value depends on the Conflict of Interests fraction of infrared light returning from the measurement site and represents the ratio of pulsatile (pulsating blood) The authors declare no conflict of interests. to nonpulsatile (nonpulsating blood, bone, and soft tissue) signal. PI is therefore affected by changes in vasomotor tone Acknowledgment as this affects the ratio between pulsatile and nonpulsatile blood [32]. Increased vasomotor tone and low SV may explain This work received departmental funding only. 8 Critical Care Research and Practice

References [16]K.L.Ryan,W.H.Cooke,C.A.Rickards,K.G.Lurie,and V. A. Convertino, “Breathing through an inspiratory threshold [1] P. E. Marik, X. Monnet, and J.-L. Teboul, “Hemodynamic device improves stroke volume during central hypovolemia in parameters to guide fluid therapy,” Annals of Intensive Care,vol. humans,” Journal of Applied Physiology, vol. 104, no. 5, pp. 1402– 1, article 1, 2011. 1409, 2008. [2] M. Shamir, L. A. Eidelman, Y. Floman, L. Kaplan, and R. Pizov, [17]R.Pizov,A.Eden,D.Bystritski,E.Kalina,A.Tamir,andS. “Pulse oximetry plethysmographic waveform during changes in Gelman, “Arterial and plethysmographic waveform analysis in blood volume,” British Journal of Anaesthesia,vol.82,no.2,pp. anesthetized patients with hypovolemia,” Anesthesiology,vol. 178–181, 1999. 113, no. 1, pp. 83–91, 2010. [3] M. Cannesson, B. Delannoy, A. Morand et al., “Does the pleth [18] S. Preau,´ F. Dewavrin, V. Soland et al., “Hemodynamic changes variability index indicate the respiratory-induced variation in during a deep inspiration maneuver predict fluid responsive- the plethysmogram and arterial pressure waveforms?” Anesthe- ness in spontaneously breathing patients,” Cardiology Research sia and Analgesia,vol.106,no.4,pp.1189–1194,2008. and Practice, vol. 2012, Article ID 191807, 8 pages, 2012. [4] J.-L. Teboul and X. Monnet, “Detecting volume responsiveness [19] S. Soubrier, F. Saulnier, H. Hubert et al., “Can dynamic indica- and unresponsiveness in intensive care unit patients: two tors help the prediction of fluid responsiveness in spontaneously different problems, only one solution,” Critical Care,vol.13,no. breathing critically ill patients?” IntensiveCareMedicine,vol.33, 4, p. 175, 2009. no. 7, pp. 1117–1124, 2007. [5] X. Monnet, D. Osman, C. Ridel, B. Lamia, C. Richard, and J.-L. [20] M. K. Dahl, S. T. Vistisen, J. Koefoed-Nielsen, and A. Larsson, Teboul, “Predicting volume responsiveness by using the end- “Using an expiratory resistor, arterial pulse pressure variations expiratory occlusion in mechanically ventilated intensive care predict fluid responsiveness during spontaneous breathing: an unit patients,” Critical Care Medicine,vol.37,no.3,pp.951–956, experimental porcine study,” Critical Care,vol.13,no.2,article 2009. R39, 2009. [6] M. J. Lanspa, C. K. Grissom, E. L. Hirshberg, J. P. Jones, [21] H. Solus-Biguenet, M. Fleyfel, B. Tavernier et al., “Non- and S. M. Brown, “Applying dynamic parameters to predict invasive prediction of fluid responsiveness during major hepatic hemodynamic response to volume expansion in spontaneously surgery,” British Journal of Anaesthesia,vol.97,no.6,pp.808– breathing patients with septic shock,” Shock,vol.39,no.2,pp. 816, 2006. 155–160, 2013. [22] M. Cannesson, O. Desebbe, P. Rosamel et al., “Pleth variabil- [7] S. Heenen, D. de Backer, and J.-L. Vincent, “How can the ity index to monitor the respiratory variations in the pulse response to volume expansion in patients with spontaneous oximeter plethysmographic waveform amplitude and predict respiratory movements be predicted?” Critical Care,vol.10,no. fluid responsiveness in the operating theatre,” British Journal of 4, article R102, 2006. Anaesthesia,vol.101,no.2,pp.200–206,2008. [8]M.Cannesson,Y.Attof,P.Rosameletal.,“Respiratoryvaria- [23] L. Ø. Høiseth, I. E. Hoff, Ø. Skare, K. A. Kirkebøen, and tions in pulse oximetry plethysmographic waveform amplitude S. A. Landsverk, “Photoplethysmographic and pulse pressure to predict fluid responsiveness in the operating room,” Anesthe- variations during abdominal surgery,” Acta Anaesthesiologica siology, vol. 106, no. 6, pp. 1105–1111, 2007. Scandinavica,vol.55,no.10,pp.1221–1230,2011. [24] S. A. Landsverk, L. O. Hoiseth, P. Kvandal, J. Hisdal, O. Skare, [9] R. H. Thiele, D. A. Colquhoun, K. T. Forkin, and M. E. Durieux, and K. A. Kirkeboen, “Poor agreement between respiratory “Assessment of the agreement between photoplethysmographic variations in pulse oximetry photoplethysmographic waveform and arterial waveform respiratory variation in patients under- amplitude and pulse pressure in intensive care unit patients,” going spine surgery,” Journal of Medical Engineering & Technol- Anesthesiology,vol.109,no.5,pp.849–855,2008. ogy,vol.37,no.7,pp.409–415,2013. [25] L. M. Nilsson, “Special article: respiration signals from photo- [10] W. H. Cooke, K. L. Ryan, and V. A. Convertino, “Lower body plethysmography,” Anesthesia & Analgesia,vol.117,no.4,pp. negative pressure as a model to study progression to acute 859–865, 2013. hemorrhagic shock in humans,” Journal of Applied Physiology, [26]A.A.Awad,M.A.M.Ghobashy,W.Ouda,R.G.Stout,D.G. vol.96,no.4,pp.1249–1261,2004. Silverman, and K. H. Shelley, “Different responses of ear and [11] V. A. Convertino, W. H. Cooke, and J. B. Holcomb, “Arterial finger pulse oximeter wave form to cold pressor test,” Anesthesia pulse pressure and its association with reduced stroke volume and Analgesia,vol.92,no.6,pp.1483–1486,2001. during progressive central hypovolemia,” Journal of Trauma, [27] K. H. Shelley, D. H. Jablonka, A. A. Awad, R. G. Stout, H. vol. 61, no. 3, pp. 629–634, 2006. Rezkanna, and D. G. Silverman, “What is the best site for [12] J. Hisdal, K. Toska, and L. Walløe, “Design of a chamber measuring the effect of ventilation on the pulse oximeter for lower body negative pressure with controlled onset rate,” waveform?” Anesthesia and Analgesia,vol.103,no.2,pp.372– Aviation Space and Environmental Medicine,vol.74,no.8,pp. 377, 2006. 874–878, 2003. [28] S. P.McGrath, K. L. Ryan, S. M. Wendelken, C. A. Rickards, and [13] F. Michard, “Changes in arterial pressure during mechanical V. A. Convertino, “Pulse oximeter plethysmographic waveform ventilation,” Anesthesiology,vol.103,no.2,pp.419–428,2005. changes in awake, spontaneously breathing, hypovolemic vol- [14]A.T.Reisner,D.Xu,K.L.Ryan,V.A.Convertino,C.A. unteers,” Anesthesia and Analgesia,vol.112,no.2,pp.368–374, Rickards, and R. Mukkamala, “Monitoring non-invasive cardiac 2011. output and stroke volume during experimental human hypo- [29]X.Monnet,B.Lamia,andJ.-L.Teboul,“Pulseoximeterasa volaemia and resuscitation,” British Journal of Anaesthesia,vol. sensor of fluid responsiveness: do we have our finger on the best 106,no.1,pp.23–30,2011. solution?” Critical Care,vol.9,no.5,pp.429–430,2005. [15] M. R. Pinsky, “Cardiovascular issues in respiratory care,” Chest, [30] X. Monnet, L. Guerin, M. Jozwiak et al., “Pleth variability index vol. 128, no. 5, supplement 2, pp. 592S–597S, 2005. is a weak predictor of fluid responsiveness in patients receiving Critical Care Research and Practice 9

norepinephrine,” British Journal of Anaesthesia,vol.110,no.2, pp.207–213,2013. [31] O. Broch, B. Bein, M. Gruenewald et al., “Accuracy of the pleth variability index to predict fluid responsiveness depends on the perfusion index,” Acta Anaesthesiologica Scandinavica,vol.55, no. 6, pp. 686–693, 2011. [32] A. P. Lima, P. Beelen, and J. Bakker, “Use of a peripheral perfusion index derived from the pulse oximetry signal as a noninvasive indicator of perfusion,” Critical Care Medicine,vol. 30, no. 6, pp. 1210–1213, 2002. [33] M. E. van Genderen, S. A. Bartels, A. Lima et al., “Peripheral perfusion index as an early predictor for central hypovolemia in awake healthy volunteers,” Anesthesia & Analgesia,vol.116, no.2,pp.351–356,2013. [34] J.C.Puyana,M.Pinsky,U.Leuenberger,andJ.Berkow,“Model- based prediction of autoregulatory exhaustion in response to lower-body negative pressure-induced shock,” Journal of Trauma and Acute Care Surgery,vol.75,no.2,supplement2, pp. S190–S196, 2013. [35] L. M. Nilsson, D. M. Lindenberger, and R. G. Hahn, “The effect of positive end-expiratory pressure and tripled tidal volume on pleth variability index during hypovolaemia in conscious subjects: a volunteer study,” European Journal of Anaesthesiology,vol.30,no.11,pp.671–677,2013. [36] B. Lansdorp, D. Ouweneel, A. De Keijzer, J. G. Van Der Hoeven, J. Lemson, and P. Pickkers, “Non-invasive measurement of pulse pressure variation and systolic pressure variation using a finger cuff corresponds with intra-arterial measurement,” British Journal of Anaesthesia,vol.107,no.4,pp.540–545,2011. [37] J. R. Martina, B. E. Westerhof, J. van Goudoever et al., “Noninva- sive continuous arterial blood pressure monitoring with nexfin,” Anesthesiology,vol.116,no.5,pp.1092–1103,2012. [38] F. Vallee,J.C.M.Richard,A.Marietal.,“Pulsepressure´ variations adjusted by alveolar driving pressure to assess fluid responsiveness,” Intensive Care Medicine,vol.35,no.6,pp. 1004–1010, 2009. [39] D. de Backer, S. Heenen, M. Piagnerelli, M. Koch, and J.-L. Vin- cent, “Pulse pressure variations to predict fluid responsiveness: influence of tidal volume,” Intensive Care Medicine,vol.31,no. 4, pp. 517–523, 2005. [40] X. Monnet, A. Bleibtreu, A. Ferreetal.,“Passiveleg-raising´ and end-expiratory occlusion tests perform better than pulse pressure variation in Patients with low respiratory system compliance,” Critical Care Medicine,vol.40,no.1,pp.152–157, 2012.

Hindawi Critical Care Research and Practice Volume 2019, Article ID 6393649, 9 pages https://doi.org/10.1155/2019/6393649

Clinical Study Volumetric and End-Tidal Capnography for the Detection of Cardiac Output Changes in Mechanically Ventilated Patients Early after Open Heart Surgery

Ingrid Elise Hoﬀ ,1,2 Lars Øivind Høiseth ,2,3 Knut Arvid Kirkebøen ,2,4 and Svein Aslak Landsverk 2

1Norwegian Air Ambulance Foundation, P.O. Box 414 Sentrum, 0103 Oslo, Norway 2Department of Anaesthesiology, Oslo University Hospital, P.O. Box 4956 Nydalen, 0424 Oslo, Norway 3Section of Vascular Investigations, Oslo University Hospital, P.O. Box 4956 Nydalen, 0424 Oslo, Norway 4Faculty of Medicine, University of Oslo, P.O. Box 1072 Blindern, 0316 Oslo, Norway

Correspondence should be addressed to Ingrid Elise Hoﬀ; iehoﬀ@yahoo.no

Received 24 December 2018; Revised 13 April 2019; Accepted 3 May 2019; Published 30 May 2019

Academic Editor: Samuel A. Tisherman

Copyright © 2019 Ingrid Elise Hoﬀ et al. +is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background. Exhaled carbon dioxide (CO2) reflects cardiac output (CO) provided stable ventilation and metabolism. Detecting CO changes may help distinguish hypovolemia or cardiac dysfunction from other causes of haemodynamic instability. We investigated whether CO2 measured as end-tidal concentration (EtCO2) and eliminated volume per breath (VtCO2) reflect sudden changes in cardiac output (CO). Methods. We measured changes in CO, VtCO2, and EtCO2 during right ventricular pacing and passive leg raise in 33 ventilated patients after open heart surgery. CO was measured with oesophageal Doppler. Results. During right ventricular pacing, CO was reduced by 21% (CI 18–24; p < 0.001), VtCO2 by 11% (CI 7.9–13; p < 0.001), and EtCO2 by 4.9% (CI 3.6–6.1; p < 0.001). During passive leg raise, CO increased by 21% (CI 17–24; p < 0.001), VtCO2 by 10% (CI 7.8–12; p < 0.001), and EtCO2 by 4.2% (CI 3.2–5.1; p < 0.001). Changes in VtCO2 were significantly larger than changes in EtCO2 (ventricular pacing: 11% vs. 4.9% (p < 0.001); passive leg raise: 10% vs. 4.2% (p < 0.001)). Relative changes in CO correlated with changes in VtCO2 (ρ � 0.53; p � 0.002) and EtCO2 (ρ � 0.47; p � 0.006) only during reductions in CO. When dichotomising CO changes at 15%, only EtCO2 detected a CO change as judged by area under the receiver operating characteristic curve. Conclusion. VtCO2 and EtCO2 reflected reductions in cardiac output, although correlations were modest. +e changes in VtCO2 were larger than the changes in EtCO2, but only EtCO2 detected CO reduction as judged by receiver operating characteristic curves. +e predictive ability of EtCO2 in this setting was fair. +is trial is registered with NCT02070861.

1. Introduction infrequently used. Factors that limit their use may be the need for extra equipment, operator dependency, or costs. Haemodynamic deteriorations are frequent in many clinical Hence, simple, inexpensive, and preferably minimally in- situations but may initially be subtle and thus difficult to vasive methods to monitor CO or changes in CO are needed. detect. Estimation of cardiac output (CO) or CO changes Capnography is widely used in mechanically ventilated may help distinguish vasodilatation due to anaesthetic or patients. During constant ventilation and metabolism and in sedative drugs from impairment of cardiac function or the absence of lung disease, changes in exhaled carbon di- hypovolemia and help evaluate response to therapy. +us, oxide (CO2) reflect changes in pulmonary blood flow [3]. monitoring CO is recommended during major surgery [1] Exhaled CO2 can be expressed as end-expiratory partial and circulatory failure [2]. Non- or minimally invasive pressure (EtCO2), or as volume eliminated CO2 per min- CO monitoring methods are increasingly available but ute (VCO2) or per tidal volume (VtCO2). Volumetric 2 Critical Care Research and Practice capnography also provides information about pulmonary anaesthesiologist. Patients were ventilated in pressure- dead space and metabolism [3, 4]. Both volumetric and regulated volume control mode, tidal volumes 6–8 mL/kg waveform capnography are recommended in the guidelines predicted body weight, positive end-expiratory pressure 5–8 for mechanical ventilation [5] and several modern ventila- cmH2O, risetime 0.20 s, FiO2 as required for SpO2 >94%, tors provide VtCO2 and VCO2 as well as EtCO2 [3, 6, 7]. and frequency 9–13 breaths/min, adjusted to obtain an Measurement of EtCO2 is included in the Advanced EtCO2 between 32 and 38 mmHg before interventions (Evita Cardiac Life Support guidelines [8], as EtCO2 reflects ef- Infinity C 500, Dragerwerk¨ AG&Co, Lubeck,¨ Germany). fective heart compressions and return of spontaneous cir- Ventilation and medication were kept constant during in- culation after cardiac arrest [9]. EtCO2 has also been shown terventions (Table 1). to predict fluid responsiveness during passive leg raise (PLR) or after a fluid load [10–12] and is included in the 2014 guidelines on haemodynamic monitoring in circulatory 2.2. Data Acquisition and Analysis. EtCO2 was measured by using an analogue side-stream capnograph (Medlab CAP 10; shock [2]. However, the changes in EtCO2 following PLR or a fluid load are quite small (≈5%). +is could limit their Medlab GmbH, Stutensee, Germany) with infrared ab- clinical use, as small changes are difficult to distinguish from sorption technology and sampled at 400 Hz in SignalExpress random fluctuations. Some studies suggest that the changes 14.0.0 (National Instruments, Austin, Texas) after conversion in an analogue-to-digital converter (NIDAQPad-6015, in VCO2 following a preload challenge or increased positive end-expiratory pressure (PEEP) are larger [13, 14]. Good National Instruments). Flow was measured continuously by agreement has been shown between CO measurements by Drager¨ Infinity ID with hot-wire anemometer technology. thermodilution and volumetric capnography in both animal Haemodynamic data, including blood pressure obtained via [15, 16] and human studies [17, 18]. Recent clinical studies a 20G catheter in the left radial artery, were downloaded from GE Solar 8000i (GE Healthcare, Chicago, Illinois, US) on the relationship between exhaled CO2 and CO have mainly focused on the prediction of fluid responsiveness, and analysed in a custom-made program (LabView 2010, National Instruments). CO was measured with oesophageal and EtCO2 has been investigated more often than VCO2 Doppler (DP-12 probe; Cardio Q; Deltex Medical, Chi- [11–14]. Few studies have investigated both VtCO2 and chester, UK), which continuously measures flow velocity in EtCO2 during moderate reductions in CO, although the detection and evaluation of decreases in CO is of major the descending aorta and thus rapid changes in stroke interest both perioperatively and during intensive care. volume (SV) [21]. +e Doppler probe was thoroughly fixed In the present study, we used right ventricular pacing in the position that gave the best signal and maximum peak (RVP) to induce moderate reductions in CO. RVP reduces velocity of the aortic flow, and the signal was closely ob- CO by approximately 20% due to loss of atrial contribution served throughout experiments. SV measurements were [19] and dyssynchrony [20]. To the best of our knowledge, downloaded beat-by-beat by the serial output. RVP has not previously been used as a model to investigate non- or minimally invasive CO monitoring methods. 2.3. Calculation of VtCO2. +e volumetric capnograms were +e aim of this study was to investigate to what extent reconstructed from flow and EtCO2 curves for the calcu- VtCO2 and EtCO2 reflect sudden moderate reductions in lation of VtCO2, as the VtCO2 and VCO2 values from the CO induced by RVP as well as sudden moderate increases in ventilator could not be extracted for offline analyses. Digital CO induced by PLR. We hypothesised that VtCO2 and mainstream flow curves from the ventilator were continu- EtCO2 would reflect changes in CO, that the changes in ously sampled on a laptop computer using Medibus software VtCO2 would be larger than the changes in EtCO2, and that (Drager,¨ Dragerwerk¨ AG&Co, Lubeck,¨ Germany) and the changes in CO, EtCO2, and VtCO2 would be correlated. aligned with converted side-stream EtCO2 curves in a custom-made program in LabView, thereby accounting for 2. Methods the relative delay of 1–4 s of the side-stream capnogram [22]. +e products of the flow and EtCO2 curves over time were 2.1. Patients. +e study was approved by the regional ethics integrated, giving VtCO2 for each respiratory cycle. Re- committee 02/07/2014 (REC South-East, 2013/1605) and spiratory cycles containing nonpaced heartbeats during the registered in http://www.clinicaltrials.gov 02/23/2014 RVP sequence were omitted. (NCT02070861), prior to patient enrolment. Forty adult patients scheduled for open coronary artery bypass surgery or aortic valve replacements were included from April 2014 2.4. Study Design. +e experimental design is illustrated in to June 2015. Written informed consent was obtained prior Figure 1. Reduction in CO was obtained by right ventricular to surgery. Patients with atrial fibrillation or ejection fraction pacing. Epicardial pacemaker leads were established towards <40% and patients in whom oesophageal Doppler placement the end of surgery according to standard departmental was contraindicated were not included. +e study was practice. Pacing was induced by using an external pacemaker conducted in the cardiothoracic recovery unit of a university (Medtronic 5388 Dual Chamber Temporary Pacemaker, hospital 1-2 h after surgery. Patients were sedated with Medtronic, Minneapolis, USA). Pacing was performed by propofol 2-3 mg/kg/h according to departmental practice. one of the department’s cardiothoracic anaesthesiologists +ey were haemodynamically stable prior to interventions, similarly to the pacemaker test routinely performed in pa- as evaluated by the attending cardiothoracic tients who require postoperative pacing. Pace rate was set Critical Care Research and Practice 3

Table 1: Patient characteristics. 2.5. Data Analysis. CO, VtCO2, and EtCO2 were normally Variable Mean (SD) distributed assessed by the Shapiro–Wilk test. +e effect of RVP and PLR on each variable and the difference between Age (years) 65 ± 9 changes in VtCO and EtCO were tested using paired t- Gender, male/female, n (%) 29 (88)/4 (12) 2 2 Height (cm) 177 ± 8 tests. +e correlations between the relative changes from Weight (kg) 87 ± 12 baseline to interventions in CO, VtCO2, and EtCO2 were Tidal volume (mL·kg−1 predicted body weight) 6.8 ± 1.1 analysed using the Spearman test of correlation, as these PIP (cmH2O) 23 ± 2 changes were mainly not normally distributed. Precision was Respiratory rate (min−1) 12 ± 1 calculated from the baseline sequence as 1.96 × √(within- Procedure CABG/AVR, n (%) 26 (79)/7 (21) subject mean square) in a one-way ANOVA with subjects as COPD, n (%) 3 (9) factors [24] and presented relative to the grand mean value. Patients receiving nitroglycerin −1 −1 12 (36) We considered the average of 30 s a clinically reasonable 0–2.5 μg·kg ·min , n (%) measurement unit and divided the breath-to-breath pre- Patients receiving nitroprusside · −1· −1 2 (6) cision by √6 (corresponding to 12 breaths/min). Least 0.8–1.8 μg kg min , n (%) × Patients receiving norepinephrine significant change (LSC) was calculated as √2 precision 1 (3) 0.02 μg·kg−1·min−1, n (%) [25]. Analyses were performed in SPSS Statistics 24 (IBM, Patients receiving amiodarone 900 mg·24·h−1, Armonk, New York, USA). We originally planned the 1 (3) n (%) presented analyses as part of a study comparing two different Data are mean ± SD unless otherwise stated. PIP � peak inspiratory pres- CO measurement devices, and sample size was calculated for sure; PEEP � positive end-expiratory pressure; CABG � coronary artery the intended comparison. However, due to technical diffi- bypass grafting; AVR � aortic valve replacement; COPD � chronic ob- culties, that part of the study had to be aborted as we could structive pulmonary disease. not guarantee the validity of the data. No post hoc power analysis was undertaken for the present analyses, but confidence intervals are presented, according to the recommendations in the CONSORT guidelines [26]. A change in CO of 15% was considered clinically significant. Based on the results of a previous study, this corresponds to changes of approximately 7.5% in VtCO2 and 3.8% in EtCO2 [13]. Areas under the receiver operating characteristic (ROC) curves for Baseline 60 s Right ventricular Post-right ventricular EtCO2 and VtCO2 were calculated and compared in Med- pacing 30 s pacing 60 s Calc Software 18.11 (MedCalc Software bvba, Ostend, Belgium). +eir discriminative value was evaluated by their 5 min ability to detect a change in CO of 15%. p values <0.05 were considered statistically significant and all tests were two- tailed. Calculations and analyses were performed without blinding. Baseline 60 s Passive leg raise 60 s Post-passive leg raise 60 s 3. Results Figure 1: Study protocol. Sixty-second baseline measurements before 30 s of RVP and 60 s of PLR. +e sequence of the in- Two patients were included, but not studied, due to changes terventions varied, minimum 5 min apart. in the operative schedule. One patient was excluded due to postoperative bleeding and two because they were marginally higher than the patient’s own heart rate in order pacemaker-dependent after surgery. Two patients were to prevent spontaneous beats, but as low as possible to excluded because of disturbances in the acquired data sig- prevent increased heart rate from offsetting the intended nals. +us, 33 patients (29 men, 4 women) completed the reduction in SV. Calculations were made from measure- study (Figure 2). ments obtained during 30 s of uninterrupted RVP, ap- Figure 3 shows individual and mean values at all 6 proximately 6 breaths. Increases in CO were induced by measurement points. For all variables, there were statistically PLR, where the patient’s position was altered from semi- significant reductions in mean scores from BL to RVP and recumbent to horizontal with legs elevated 45°. +is ma- statistically significant increases from RVP to BL and from noeuvre represents an endogenous and reversible fluid BL to PLR (Table 2, Figure 3). +e confidence intervals of the challenge of approximately 300 ml, with maximal volume line plots in Figure 3 indicate that the study was not un- effect during the first minute [23]. +us, calculations were derpowered for the presented analyses. From BL to RVP, CO based on measurements from the initial 60 s after leg raise, was reduced by 21.0% (CI 18–24; p < 0.001), VtCO2 by 11% approximately 12 breaths. Interventions were minimum (CI 7.9–13; p < 0.001), and EtCO2 by 4.9% (CI 3.6–6.1; 5 min apart to ensure return to baseline (BL) before new p < 0.001). Relative changes in CO correlated significantly measurements. Sixty seconds of BL were recorded before with changes in both VtCO2 (ρ � 0.53; p � 0.002) and and after each intervention with calculations based on BL EtCO2 (ρ � 0.47; p � 0.006) (Figure 4). From BL to PLR, CO measurements before interventions. increased by 21% (CI 17–24; p < 0.001), VtCO2 by 10% (CI 4 Critical Care Research and Practice

Included (n = 40)

Excluded (n=7) (i) Changes in operative schedule (n =2) (ii) Postoperative bleeding (n =1) (iii) Postoperative pacemaker dependency (n =2) (iv) Disturbance in acquired signals (n =2)

Analysed (n = 33)

Figure 2: Flow chart inclusion.

EtCO Cardiac output VtCO2 2 9 26 5.2 5.0 8 24 4.8 7 22 4.6 20 6 4.4 18 , (ml) , (kPa) 2 4.2

5 2 16 4.0 4 VtCO 14 EtCO 3.8 3 12 Cardiac output, (l/min) 3.6 2 10 3.4 1 8 3.2 PM PM PM PLR PLR PLR BL PM BL PM BL PM BL PLR BL PLR BL PLR Post-PM Post-PM Post-PM Post-PLR Post-PLR Post-PLR (a) (b) (c) Figure 3: Lineplot. Individual (grey) and mean (black) values with 95% conﬁdence intervals for CO, VtCO2, and EtCO2 before, during, and after interventions. CO � cardiac output; EtCO2 � end-tidal carbon dioxide; VtCO2 � exhaled carbon dioxide per tidal volume; BL � baseline; RVP � right ventricular pacing; PLR � passive leg raise.

7.8–12; p < 0.001), and EtCO2 by 4.2% (CI 3.2–5.1; 4. Discussion p < 0.001). None of these changes were significantly correlated (Figure 4). Overall, the changes in VtCO2 were +e main findings of this study were that VtCO2 and EtCO2 significantly larger than the changes in EtCO2 (from BL to tracked sudden moderate reductions in CO. Both reductions RVP, 11% vs. 4.9% (p < 0.001); from BL to PLR, 10% vs. 4.2% and increases in CO with RVP and PLR coincided with re- (p < 0.001)). ductions and increases, respectively, in EtCO2 and VtCO2 Precision and LSC for 30 s baseline measurements were (Figure 3). +e magnitudes of the changes, however, were only 4.8% and 6.9%, respectively, for CO, 2.4% and 3.4% for correlated when CO was reduced, and correlations were VtCO2, and 1.5% and 2.1% for EtCO2. +us, all mean modest (Figures 4 and 5). According to the ROC analyses, only changes seen after the interventions were larger than the EtCO2 was able to discriminate changes in CO using a LSC. +e LSC for CO, VtCO2, and EtCO2 are indicated in threshold of 15% change and only the reduction during RVP Figures 4 and 5, respectively. According to the scatterplots (Figure 6). during RVP, a reduction in VtCO2 and EtCO2 larger than Young et al. [13] found VCO2 superior to EtCO2 for the LSC implicated a reduction in CO of more than 11% for predicting fluid responsiveness in the PLR model, and the all subjects. changes in VCO2 were substantially larger than the changes in ROC-plot analyses are shown in Figure 6. +e best dis- EtCO2. Tusman et al. [14] showed that a reduction in VCO2 criminative ability was found for EtCO2 (AUC 0.80; 95% CI following an increase in PEEP predicted fluid responsiveness 0.62–0.92, p � 0.003) during RVP, whereas the ROC curve for with better sensitivity and specificity than EtCO2. In our VtCO2 was not significantly different from 0.5. Neither EtCO2 study, the changes in CO during RVP appear to be slightly nor VtCO2 was able to discriminate changes in CO during PLR. stronger correlated with the changes in VtCO2 than with the Critical Care Research and Practice 5

Table 2: Haemodynamic data at baseline and during right ventricular pacing and passive leg raise. BL VP p value BL PLR p value CO (l/min) 4.86 ± 1.20 3.84 ± 1.08 <0.001 4.72 ± 1.17 5.65 ± 1.26 <0.001 SV (mL/min) 71 ± 20 54 ± 17 <0.001 69 ± 17 81 ± 19 <0.001 VtCO2 (ml) 17 ± 3 15 ± 3 <0.001 16 ± 3 18 ± 4 <0.001 EtCO2 (kPa) 4.3 ± 0.3 4.1 ± 0.4 <0.001 4.3 ± 0.3 4.5 ± 0.3 <0.001 MAP (mmHg)∗ 72 ± 6 62 ± 9 <0.001 72 ± 9 78 ± 8 <0.001 HR (beats/min) 69 ± 10 73 ± 10 <0.001 70 ± 10 71 ± 10 0.015

Data are presented in mean ± SD. BL � baseline; RVP � ventricular pacing; PLR � passive leg raise; CO � cardiac output; SV � stroke volume; VtCO2 � exhaled ∗ carbon dioxide per tidal volume; EtCO2 � end-tidal carbon dioxide; MAP � mean arterial pressure ( measurements from 21 patients); HR � heart rate.

5% 20%

0% 15% 2 2 –5% 10%

–10% 5% Change in EtCO Change in EtCO –15% 0%

–20% –5% –50% –40% –30% –20% –10% 0% 10% –10% 0% 10% 20% 30% 40% 50% Change in cardiac output Change in cardiac output RVP, ρ = 0.47 (0.15 to 0.70), p = 0.006 PLR, ρ = 0.22, (–0.14 to 0.52), p = 0.23

(a) (b) Figure 4: Scatterplot EtCO2. Correlation between mean relative changes in CO and EtCO2 from BL to RVP and PLR, respectively. Least signiﬁcant changes for CO and EtCO2 are indicated with shadows. CO � cardiac output; EtCO2 � end-tidal carbon dioxide; VtCO2 � exhaled carbon dioxide per tidal volume; BL � baseline; RVP � right ventricular pacing; PLR � passive leg raise; ρ � Spearman’s rho with conﬁdence intervals.

10% 30%

5% 25%

0% 20% 2 2 –5% 15%

–10% 10%

–15% 5% Change in VtCO Change in VtCO –20% 0%

–25% –5%

–30% –10% –50% –40% –30% –20% –10% 0% 10% –10% 0% 10% 20% 30% 40% 50% Change in cardiac output Change in cardiac output RVP, ρ = 0.53 (0.23 to 0.74), p = 0.002 PLR, ρ = –0.075 (–0.41 to 0.28), p = 0.68

(a) (b) Figure 5: Correlation between mean relative changes in CO and VtCO2 from BL to RVP and PLR, respectively. Least signiﬁcant changes for CO and VtCO2 are indicated with shadows. Dots are for RVP; circles are for PLR. CO � cardiac output; EtCO2 � end-tidal carbon dioxide; VtCO2 � exhaled carbon dioxide per tidal volume; BL � baseline; RVP � right ventricular pacing; PLR � passive leg raise; ρ � Spearman’s rho with conﬁdence intervals. 6 Critical Care Research and Practice

Right ventricular pacing Passive leg raise 1.0 1.0

0.8 0.8

0.6 0.6

0.4 Sensitivity Sensitivity 0.4

0.2 0.2

0.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1 – specificity 1 – specificity

p EtCO , AUC = 0.64 (95% CI 0.46 to 0.80), p =0.23 EtCO2, AUC = 0.80 (95% CI 0.62 to 0.92), = 0.003 2 p VtCO , AUC = 0.50 (95% CI 0.32 to 0.68), p =1.0 VtCO2, AUC = 0.68 (95% CI 0.50 to 0.83), = 0.13 2

(a) (b) Figure 6: ROC-plot. Receiver operating characteristic plots of VtCO2 and EtCO2 during right ventricular pacing and passive leg raise, respectively. EtCO2 � end-tidal carbon dioxide; VtCO2 � exhaled carbon dioxide per tidal volume; AUC � area under the curve.

Table 3: Sensitivity, speciﬁcity, likelihood ratios, and predictive values set at criterion value giving maximal Youden index. +e prevalence of responders and nonresponders was set to the ratio in the sample when calculating predictive values. AUC (95% CI) Criterion (%) Speciﬁcity (%) Sensitivity (%) +LR −LR PPV (%) NPV (%)

VtCO2 at RVP 0.68 (0.50 to 0.83) 5.2 83 56 1.9 0.3 83 56 EtCO2 at RVP 0.80 (0.62 to 0.92) 1.7 100 56 2.3 0.0 86 100 VtCO2 at PLR 0.50 (0.32 to 0.68) 4.2 92 44 1.7 0.19 82 67 EtCO2 at PLR 0.64 (0.46 to 0.80) 2.2 88 56 2.0 0.23 84 63 AUC � area under the curve; +LR � positive likelihood ratio; −LR � negative likelihood ratio; PPV � positive predictive value; NPV � negative predictive value; VtCO2 � exhaled carbon dioxide per tidal volume; EtCO2 � end-tidal carbon dioxide; RVP � ventricular pacing; PLR � passive leg raise. changes in EtCO2. Precision was better for EtCO2 than for in EtCO2 following a given change in CO, and correlations VtCO2, but this did not outweigh the larger effect of changes were similar. However, given that a diagnostic ability was in CO on VtCO2. +e ROC analyses, using a threshold of 15%, demonstrated only for EtCO2, the results do not support the indicate a stronger discriminative ability for EtCO2 than superiority of VtCO2 over EtCO2. In some of the studies, VtCO2, which appears contradictory to the previously VCO2 and EtCO2 were also found superior to pulse pressure mentioned findings. However, the criterion value giving the variations (PPVs) or stroke volume variations (SVVs) in the maximal Youden index for EtCO2 was low (Table 3), limiting presence of arrhythmia [29] or tidal volumes <8 mL/kg its use as a clinical cutoff value. +ere are also some limi- [14, 29]. +is is explained by the fact that PPV and SVV tations to the ROC analysis associated with the dispersion of are validated for the prediction of fluid responsiveness predictor values in the population which is investigated. +ese mainly in patients with tidal volumes ≥8 mL/kg and without limitations are previously described [27] and highlighted in a arrhythmia [30, 31]. However, as protective ventilation recent review [28] and should be considered when comparing becomes the norm, it is noteworthy that the same re- AUC values from different studies. strictions do not seem to apply for EtCO2 or VtCO2. In the studies by Monge Garc´ıa et al. [10] and Monnet +e physiologic relationship between exhaled CO2 and et al. [11], EtCO2 predicted fluid responsiveness with higher CO in dynamic states is previously described [15, 32]. Re- sensitivity and specificity than arterial pulse pressure, and duced pulmonary perfusion leads to reduced CO2 transport Jacquet-Lagreze et al. [12] found the same when comparing to the lungs and increased alveolar dead space; both resulting EtCO2 to MAP. +ese findings were confirmed in a recent in reduced CO2 elimination. With increased pulmonary study by Lakhal et al. [29], who in addition found that EtCO2 perfusion, more CO2 is brought to the lungs, underperfused assessed fluid responsiveness better than changes in systolic lung tissue is recruited, and CO2 elimination is increased. blood pressure and femoral blood flow did. In summary, Although reports of the nature of the relationship between while EtCO2 has been found superior to other widely used exhaled CO2 and CO differ [18, 32, 33], several studies have noninvasive indices, newer studies suggest that VCO2 and found significant correlations between changes in CO and VtCO2 could be superior to EtCO2. In the present study, the changes in EtCO2 after PLR [10, 11]. We believe there are changes in VtCO2 were substantially larger than the changes mainly two reasons why there were no correlations between Critical Care Research and Practice 7

EtCO2, VtCO2, and CO during PLR in our study. Firstly, limitations. Measurements are based on assumptions re- previous studies investigated patients with circulatory fail- garding the diameter of the aorta, angle of insonation, and ure, whereas our cohort was haemodynamically and met- fraction of CO that enters the descending aorta [35]. As we abolically stable. +e relationship between changes in CO measured relative changes, the results would only have and exhaled CO2 is stronger during unstable circulatory been affected if the assumed variables changed during states, e.g., in patients with reduced CO [6]. In steady states, experiments. Aortic diameter has been shown to change exhaled CO2 mainly depends on CO2 production. Lung after a fluid load [36], and we cannot exclude a similar effect perfusion and ventilation/perfusion ratio will be affected after PLR. +ese limitations suggest that oesophageal only marginally, if at all, by an increase in CO of 20% in Doppler may perform better as a monitor of CO trends euvolemic patients who are adequately ventilated. +is is in than of absolute values. +is may also explain why some line with the findings of Ornato et al [32], who in an animal patients in the present study demonstrated rather low CO study demonstrated that the correlation between changes in values despite being assessed as haemodynamically stable at CO and changes in EtCO2 decreased as CO reached normal baseline. or supranormal values, when pulmonary flow no longer For the description of metabolism, exhaled CO2 is represents a limitation to the CO2 elimination via the lungs. mostly expressed as VCO2, whereas both VCO2 and By contrast, we observed significant correlations between the VtCO2 have been used to describe the relationship be- relative reductions in CO, VtCO2, and EtCO2 when CO was tween exhaled CO2 and circulation [13, 15, 37]. We decreased during the RVP sequence, even though the change measured VtCO2 to enable a direct comparison with in CO was of similar magnitude. Secondly, the mean relative EtCO2, which is also measured breath-to-breath. As increase in EtCO2 during PLR in our study was 4.2%, which ventilation was kept constant throughout experiments, the is smaller than in previous studies which have reported an choice of VtCO2 over VCO2 should not affect the results, increase of >5%. As these studies were designed to study which may therefore be seen in relation to previous studies fluid responsiveness, EtCO2 was recorded during the investigating VCO2. +e absolute changes in VtCO2 are maximal haemodynamic changes following PLR. We sam- small. However, they are significantly larger than the pled CO, EtCO2, and VtCO2 over 1 min of PLR, and al- corresponding changes in EtCO2, which use is already though the main preload increase is likely to take place implemented in guidelines for haemodynamic evaluation. within that minute, the time span includes lower values that Modern ventilators display updated VCO2 values after dilute this effect. Also, it is possible that the position change each breath. For clinical use, changes in VCO2 may be during the PLR manoeuvre could affect CO2 elimination by easier to detect than changes in VtCO2, as they appear other mechanisms than the preload increase. +is could have larger. influenced the results. In a postoperative setting with hae- Any form of ventilation/perfusion mismatch may affect modynamically stable patients, the detection of a sudden the relationship between CO and exhaled CO2 [38]. Other decrease in CO, e.g., due to bleeding, is arguably more investigators have therefore excluded patients with pul- relevant than the prediction of preload responsiveness. monary dysfunction [14, 18]. Only three of our patients In the absence of CO monitoring, MAP is often used for (9.1%) had been diagnosed with chronic obstructive pul- haemodynamic assessment. As MAP is highly influenced by monary disease. However, it is possible that some had vascular resistance [34], it may be affected by anaesthetics, undiagnosed lung disease or postoperative pulmonary pain, hypovolemia, and hypothermia. Hypotension occurs dysfunction which may have affected our results. frequently in the operating room or intensive care unit and As mechanical ventilation alters pulmonary physiology can be due to a number of causes. By also considering changes and haemodynamics [39], further studies are necessary to in EtCO2 or VtCO2 in cases of decreasing blood pressure, the elucidate the performance of VtCO2 and EtCO2 in spon- clinician may be aided in their therapeutic decisions. taneously breathing patients.

5. Conclusion 4.1. Methodological Considerations. As departmental logis- tics had to be considered during data acquisition, the order VtCO2 and EtCO2 tracked reductions in cardiac output, but of interventions varied in a nonrandomised fashion. +e correlations between the changes were modest. Judged by possibility of carryover effects was minimised by ensuring receiver operating characteristic curves, a CO reduction was sufficient time between all interventions but cannot be only detected by EtCO2. Further studies are warranted to excluded. establish the role of exhaled CO2 as a clinical tool for +ere was a departmental change in monitoring detecting cardiac output changes in this setting. equipment during the study, and the available software did not allow export of invasive blood pressure data from the Data Availability new monitors to the computer. +us, MAP measurements were retrospectively obtainable from 21 patients only. +is +e data used to support the findings of this study are re- represents a limitation to the study. stricted by Oslo University Hospital in order to protect CO had to be monitored continuously as changes in CO patient privacy. Pseudonymised data are available from the induced by RVP and PLR are rapid and transient. However, corresponding author for researchers who meet the criteria CO measurement with oesophageal Doppler has some for access to confidential data. 8 Critical Care Research and Practice

Conflicts of Interest [14] G. Tusman, I. Groisman, G. A. Maidana et al., “+e sensitivity and specificity of pulmonary carbon dioxide elimination for +e authors declare that there are no conflicts of interest noninvasive assessment of fluid responsiveness,” Anesthesia & regarding the publication of this article. Analgesia, vol. 122, no. 5, pp. 1404–1411, 2016. [15] S. A. Isserles and P. H. Breen, “Can changes in end-tidal PCO2 Acknowledgments measure changes in cardiac output?,” Anesthesia & Analgesia, vol. 73, no. 6, pp. 808–814, 1991. +is work was funded by the Norwegian Air Ambulance [16] M. Klein, L. Minkovich, M. Machina et al., “Non-invasive Foundation, Oslo, Norway. measurement of cardiac output using an iterative, respiration- based method,” British Journal of Anaesthesia, vol. 114, no. 3, pp. 406–413, 2015. References [17] K. Shibutani, M. Muraoka, S. Shirasaki, K. Kubal, V. T. Sanchala, and P. Gupte, “Do changes in end-tidal PCO [1] R. M. Pearse, D. A. Harrison, N. MacDonald et al., “Effect of a 2 quantitatively reflect changes in cardiac output?,” Anesthesia perioperative, cardiac output-guided hemodynamic therapy & Analgesia, vol. 79, no. 5, pp. 829–833, 1994. algorithm on outcomes following major gastrointestinal [18] P. J. Peyton, “Continuous minimally invasive peri-operative surgery,” JAMA, vol. 311, no. 21, pp. 2181–2190, 2014. [2] M. Cecconi, D. De Backer, M. Antonelli et al., “Consensus on monitoring of cardiac output by pulmonary capnotracking: circulatory shock and hemodynamic monitoring: task force of comparison with thermodilution and transesophageal echo- the european society of intensive care medicine,” Intensive cardiography,” Journal of Clinical Monitoring and Computing, Care Medicine, vol. 40, no. 12, pp. 1795–1815, 2014. vol. 26, no. 2, pp. 121–132, 2012. [3] M. S. Siobal, “Monitoring exhaled carbon dioxide,” Re- [19] R. I. Hamby and A. Aintablian, “Preload reduction with right spiratory Care, vol. 61, no. 10, pp. 1397–1416, 2016. ventricular pacing: effects on left ventricular hemodynamics [4] F. Suarez-Sipmann, S. H. Bohm, and G. Tusman, “Volumetric and contractile pattern,” Clinical Cardiology, vol. 3, no. 3, capnography,” Current Opinion in Critical Care, vol. 20, no. 3, pp. 169–177, 1980. pp. 333–339, 2014. [20] M. Saito, G. Kaye, K. Negishi et al., “Dyssynchrony, con- [5] B. K. Walsh, D. N. Crotwell, and R. D. Restrepo, traction efficiency and regional function with apical and non- “Capnography/Capnometry during mechanical ventilation: apical RV pacing,” Heart, vol. 101, no. 8, pp. 600–608, 2015. 2011,” Respiratory Care, vol. 56, no. 4, pp. 503–509, 2011. [21] H. Møller-Sørensen, J. Cordtz, M. Østergaard, J. C. Nilsson, [6] B. S. Nassar and G. A. Schmidt, “Capnography during critical and K. L. Hansen, “Transesophageal Doppler reliably tracks illness,” Chest, vol. 149, no. 2, pp. 576–585, 2016. changes in cardiac output in comparison with intermittent [7] S. Verscheure, P. B. Massion, F. Verschuren, P. Damas, and pulmonary artery thermodilution in cardiac surgery patients,” S. Magder, “Volumetric capnography: lessons from the past Journal of Clinical Monitoring and Computing, vol. 31, no. 1, and current clinical applications,” Critical Care, vol. 20, no. 1, pp. 135–142, 2017. p. 184, 2016. [22] B. S. Kodali, “Capnography outside the operating rooms,” [8] M. F. Hazinski, J. P. Nolan, J. E. Billi et al., “Part 1: executive Anesthesiology, vol. 118, no. 1, pp. 192–201, 2013. summary: 2010 international consensus on cardiopulmonary [23] X. Monnet and J.-L. Teboul, “Passive leg raising: five rules, not resuscitation and emergency cardiovascular care science with a drop of fluid!,” Critical Care, vol. 19, no. 1, p. 18, 2015. treatment recommendations,” Circulation, vol. 122, no. 16, [24] J. M. Bland and D. G. Altman, “Measuring agreement in pp. S250–S275, 2010. method comparison studies,” Statistical Methods in Medical [9] S. M. Hartmann, R. W. D. Farris, J. L. Di Gennaro, and Research, vol. 8, no. 2, pp. 135–160, 1999. J. S. Roberts, “Systematic review and meta-analysis of end- [25] M. Cecconi, A. Rhodes, J. Poloniecki, G. Della Rocca, and tidal carbon dioxide values associated with return of spon- R. M. Grounds, “Bench-to-bedside review: the importance of taneous circulation during cardiopulmonary resuscitation,” the precision of the reference technique in method com- Journal of Intensive Care Medicine, vol. 30, no. 7, pp. 426–435, parison studies—with specific reference to the measurement 2015. of cardiac output,” Critical Care, vol. 13, no. 1, p. 201, 2009. [10] M. I. Monge Garc´ıa, A. Gil Cano, M. Gracia Romero, [26] D. Moher, S. Hopewell, K. F. Schulz et al., “CONSORT 2010 R. Monterroso Pintado, V. Perez´ Madueño, and J. C. D´ıaz explanation and elaboration: updated guidelines for reporting Monrove,´ “Non-invasive assessment of fluid responsiveness parallel group randomised trials,” BMJ, vol. 340, no. 1, p. c869, by changes in partial end-tidal CO2 pressure during a passive 2010. leg-raising maneuver,” Annals of Intensive Care, vol. 2, no. 1, [27] L. O. Hoiseth and J. S. Hagemo, “Predicting fluid re- p. 9, 2012. sponsiveness in whom? A simulated example of patient [11] X. Monnet, A. Bataille, E. Magalhaes et al., “End-tidal carbon spectrum influencing the receiver operating characteristics dioxide is better than arterial pressure for predicting volume curve,” Journal of Clinical Monitoring and Computing, vol. 32, responsiveness by the passive leg raising test,” Intensive Care no. 2, pp. 215–219, 2018. Medicine, vol. 39, no. 1, pp. 93–100, 2013. [28] B. Saugel, M. Flick, K. Bendjelid, L. A. H. Critchley, [12] M. Jacquet-Lagreze, F. Baudin, J. S. David et al., “End-tidal S. T. Vistisen, and T. W. L. Scheeren, “Journal of clinical carbon dioxide variation after a 100- and a 500-ml fluid monitoring and computing end of year summary 2018: he- challenge to assess fluid responsiveness,” Annals of Intensive modynamic monitoring and management,” Journal of Clin- Care, vol. 6, no. 1, p. 37, 2016. ical Monitoring and Computing, vol. 33, no. 2, pp. 211–222, [13] A. Young, P. E. Marik, S. Sibole, D. Grooms, and A. Levitov, 2019. “Changes in end-tidal carbon dioxide and volumetric carbon [29] K. Lakhal, M. A. Nay, T. Kamel et al., “Change in end-tidal dioxide as predictors of volume responsiveness in hemody- carbon dioxide outperforms other surrogates for change in namically unstable patients,” Journal of Cardiothoracic and cardiac output during fluid challenge,” British Journal of Vascular Anesthesia, vol. 27, no. 4, pp. 681–684, 2013. Anaesthesia, vol. 118, no. 3, pp. 355–362, 2017. Critical Care Research and Practice 9

[30] P. E. Marik, R. Cavallazzi, T. Vasu, and A. Hirani, “Dynamic changes in arterial waveform derived variables and fluid responsiveness in mechanically ventilated patients: a systematic review of the literature,” Critical Care Medicine, vol. 37, no. 9, pp. 2642–2647, 2009. [31] M. R. Pinsky, “Heart lung interactions during mechanical ventilation,” Current Opinion in Critical Care, vol. 18, no. 3, pp. 256–260, 2012. [32] J. P. Ornato, A. R. Garnett, and F. L. Glauser, “Relationship between cardiac output and the end-trial carbon dioxide tension,” Annals of Emergency Medicine, vol. 19, no. 10, pp. 1104–1106, 1990. [33] M. H. Weil, J. Bisera, R. P. Trevino, and E. C. Rackow, “Cardiac output and end-tidal carbon dioxide,” Critical Care Medicine, vol. 13, no. 11, pp. 907–909, 1985. [34] L. Ø Høiseth, J Hisdal, I. E Hoff, O. A Hagen, S. A Landsverk, and K. A Kirkebøen, “Tissue oxygen saturation and finger perfusion index in central hypovolemia: influence of pain,” Critical Care Medicine, vol. 43, no. 4, pp. 747–756, 2015. [35] C. Berton and B. Cholley, “Equipment review: new techniques for cardiac output measurement—oesophageal Doppler, Fick principle using carbon dioxide, and pulse contour analysis,” Critical Care, vol. 6, no. 3, pp. 216–221, 2002. [36] X. Monnet, D. Chemla, D. Osman et al., “Measuring aortic diameter improves accuracy of esophageal Doppler in assessing fluid responsiveness,” Critical Care Medicine, vol. 35, no. 2, pp. 477–482, 2007. [37] G. Tusman, F. Suarez-Sipmann, G. Paez, J. Alvarez, and S. H. Bohm, “States of low pulmonary blood flow can be detected non-invasively at the bedside measuring alveolar dead space,” Journal of Clinical Monitoring and Computing, vol. 26, no. 3, pp. 183–190, 2012. [38] P. Peyton, “Pulmonary carbon dioxide elimination for cardiac output monitoring in peri-operative and critical care patients: history and current status,” Journal of Healthcare Engineering, vol. 4, no. 2, pp. 203–222, 2013. [39] A. A. Alian and K. H. Shelley, “Respiratory physiology and the impact of different modes of ventilation on the photoplethysmographic waveform,” Sensors, vol. 12, no. 2, pp. 2236–2254, 2012.

III

Respiratory variations in pulse pressure and photoplethysmographic waveform amplitude during positive expiratory pressure and continuous positive airway pressure in a model of progressive hypovolemia

Ingrid Elise Hoff, MD1,2, Jonny Hisdal, PhD3,4, Svein Aslak Landsverk, MD, PhD2, Jo Røislien, PhD1, Knut Arvid Kirkebøen, MD, PhD4 and Lars Øivind Høiseth, MD, PhD2,3

1 Norwegian Air Ambulance Foundation, PO Box 414 Sentrum, 0103 Oslo, Norway

2 Department of Anesthesiology, Oslo University Hospital, PO Box 4956 Nydalen, 0424 Oslo, Norway

3 Section of Vascular Investigations, Department of Vascular Surgery, Oslo University Hospital, PO Box 4956 Nydalen, 0424 Oslo, Norway

4 Faculty of Medicine, University of Oslo, PO Box 1072 Blindern, 0316 Oslo, Norway

Corresponding author: Ingrid Elise Hoff, MD

Dept. of Anesthesiology

Oslo University Hospital

Postboks 4956 Nydalen

0424 Oslo, Norway

Tel + 47 22119690, Fax + 47 22119857

[email protected]

ORCID: 0000-0001-7001-9530

Word count: Abstract 250, Manuscript 3319

Acknowledgements: This work was funded by the Norwegian Air Ambulance Foundation.

ABSTRACT Purpose: Respiratory variations in pulse pressure (dPP) and photoplethysmographic waveform amplitude (dPOP) are used for evaluation of volume status in mechanically ventilated patients. Amplification of intrathoracic pressure changes may enable their use also during spontaneous breathing. We investigated the association between the degree of hypovolemia and dPP and dPOP at different levels of two commonly applied clinical interventions; positive expiratory pressure (PEP) and continuous positive airway pressure

(CPAP).

Methods: 20 healthy volunteers were exposed to progressive hypovolemia by lower body negative pressure (LBNP). PEP of 0 (baseline), 5 and 10 cmH2O was applied by an expiratory resistor and CPAP of 0 (baseline), 5 and 10 cmH2O by a facemask. dPP was obtained non- invasively with the volume clamp method and dPOP from a pulse oximeter. Central venous pressure was measured in 10 subjects. Associations between changes were examined using linear mixed-effects regression models.

Results: dPP increased with progressive LBNP at all levels of PEP and CPAP. The LBNP- induced increase in dPP was amplified by PEP 10 cmH20. dPOP increased with progressive

LBNP during PEP 5 and PEP 10, and during all levels of CPAP. There was no additional effect of the level of PEP or CPAP on dPOP. Progressive hypovolemia and increasing levels of PEP were reflected by increasing respiratory variations in CVP.

Conclusion: dPP and dPOP reflected progressive hypovolemia in spontaneously breathing healthy volunteers during PEP and CPAP. An increase in PEP from baseline to 10 cmH2O augmented the increase in dPP, but not in dPOP.

Key words: dynamic variables, spontaneous breathing, hypovolemia, continuous positive airway pressure, positive expiratory pressure

INTRODUCTION Dynamic variables, such as respiratory variations in pulse pressure (dPP) and the photoplethysmographic waveform amplitude (dPOP) are accurate indicators of preload changes, and predict fluid responsiveness during mechanical ventilation [1-3]. During mechanical ventilation, dPP is mainly caused by increased intrathoracic pressure in the inspiratory phase, leading to increased right atrial pressure (RAP). As RAP, which normally equals central venous pressure (CVP) [4], is the pressure opposing venous return, this leads to cyclic variations in venous return and thus stroke volume and pulse pressure, which are larger when the heart is preload responsive. It is generally assumed that these variations are inadequate to enable dynamic variables to reflect preload dependency in spontaneously breathing subjects. However, some studies indicate that the ability of dPP and dPOP to reflect volume status or predict fluid responsiveness improves when intrathoracic pressure variations are amplified with respiratory resistance [5, 6].

Positive expiratory pressure (PEP) and continuous positive airway pressure (CPAP) are two interventions which increase airway pressure and are frequently used to prevent atelectasis and respiratory failure in spontaneously breathing patients postoperatively and during critical illness. In these patients, evaluation of volume status is important, and whether PEP and CPAP may affect the ability of dPP and dPOP to detect hypovolemia is of clinical relevance, as the use of these respiratory interventions could also provide an opportunity to evaluate volume status.

This study aimed to investigate the ability of dPP and dPOP to track hypovolemia induced by lower body negative pressure (LBNP) in spontaneously breathing volunteers during different levels of PEP and CPAP. We further investigated whether associations between the dynamic variables and volume status were affected by the level of respiratory resistance. We also

4 aimed to explore whether level of respiratory resistance was reflected in the respiratory variations in CVP (dCVP). We hypothesized that the dynamic variables would increase with progressive hypovolemia and that the increases would be amplified with increasing levels of respiratory resistance.

METHODS

Subjects The study was approved by the Regional Committee for Medical and Health Research Ethics

(REC South East D, reference 2015/344) prior to inclusion. All procedures were in accordance with the ethical standards of the institutional and regional ethical research committees and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Twenty healthy adult volunteers (11 males, 9 females, aged 25(3) (mean[SD]) years, height 176(9) cm and weight 69(8) kg were included from November 2015 to

December 2016 after oral and written informed consent. Exclusion criteria were disease or disability requiring regular medication (except allergies), arrhythmia, pregnancy, history of syncope, and infection in the elbow crease. Participants refrained from caffeine consumption and excessive physical exercise on the day of the experiments.

Study protocol The experiments were performed between 8:00 AM and 4:00 PM in a vascular investigation lab air-conditioned to 20-22°C. Fig 1 illustrates the study protocol. Subjects were placed in the supine position in the LBNP chamber, which was sealed with a neoprene skirt at the level of the iliac crest as described previously [7]. LBNP leads to sequestering of blood in the lower abdomen and extremities, with a negative pressure of 80 mmHg corresponding to a blood loss of more than 1 liter [8]. Stepwise progressions in negative pressure from 0 (baseline) to

20, 40, 60 and 80 mmHg were induced. Each LBNP-level lasted 6.0 (5.3, 6.7) min (median

[25th, 75th percentiles], and the entire LBNP-exposure (from LBNP 20 mmHg) lasted 18 (14,

22) min. At each LBNP-level, after 1 min stabilization, PEP of 0 (baseline), 5 and 10 cmH2O was applied, followed by CPAP of 0 (baseline), 5 and 10 cmH2O in alternating order at different LBNP-levels. Subjects were randomized in blocks of six generated by pseudorandom numbers in Excel 2010 (Microsoft Office 365; Microsoft Corporation,

Redmond, Washington, USA) to start with different PEP or CPAP levels, thus also changing the order at the subsequent LBNP-levels. PEP was induced with an expiratory resistor

(Armstrong Medical Ltd, Coleraine, Northern Ireland), through which the subjects were instructed to exhale calmly. CPAP was provided by Drӓger Evita 4 (Drӓgerwerk, Lübeck,

Germany), and the subjects were instructed to breathe normally in the facemask. No instructions were given for respiratory rate, allowing the subjects to adjust the ventilation to avoid hyper – and hypoventilation. PEP 0 cmH20 was breathing through an empty mouthpiece, and CPAP 0 cmH2O was breathing through a facemask with CPAP set to 0 cmH2O on the ventilator. Each pressure level was applied for 5-6 breaths. LBNP was released and the protocol terminated at the subject`s request, or if the subject displayed signs of impending circulatory collapse such as sweating, nausea or dizziness or a sudden marked reduction in heart rate or mean arterial pressure. Only measurements from completed breathing sequences were used for analyses.

Data acquisition Non-invasive arterial pressure waveform was acquired by the volume-clamp method

(Finometer; FMS Finapres Mesurement Systems, Arnhem, The Netherlands) and exported with ECG at 300 Hz to custom-made software (Regist3; Morten Eriksen, University of Oslo,

Oslo, Norway). Photoplethysmographic waveform from a commercially available pulse oximeter attached to the right 2. finger (Masimo Radical 7, software 7.3.1.1, Masimo Corp.,

Irvine, CA, USA with probe LNOP DC-I; Masimo Corp.) was exported from the analog output at 400 Hz to SignalExpress 14.0.0 (National Instruments, Austin, Texas, USA). Aortic blood flow velocity was measured by suprasternal Doppler with a 2 MHz probe (SD-50; GE

Vingmed Ultrasound, Horten, Norway) and sampled at 300 Hz in Regist3. In 10 subjects, CVP was obtained from a central venous catheter (Secalon Seldy 16G, Argon Critical Care Systems,

Singapore), which was inserted via the left basilic vein to the left subclavian vein and connected to a pressure transducer (CODAN Critical Care GmbH, Forstinning, Germany), which was leveled and zeroed in the mid-axillary line. Correct position was verified by a typical central venous pressure waveform. CVP-measurements from peripherally inserted catheters and conventional central venous catheters have been shown to be highly correlated [9]. CVP waveforms were exported with ECG from the TramRac4A (General

Electric Healthcare) at 400 Hz to SignalExpress. Analog signals were exported as text files, time synchronized and handled in R 3.4.0 (R Foundation for Statistical Computing, Vienna,

Austria) using RStudio 1.0.143 (RStudio, Boston, MA, USA).

Signal analysis and calculations Stroke volume was obtained by calculating aortic flow velocity-time integrals gated by the R- peaks of the ECG, assuming an angle of 20° to the aortic blood flow and an aortic diameter of 20 mm [10]. Cardiac output was calculated by multiplying stroke volume with heart rate. dPP and dPOP were calculated by the formulas

% and

% ,

where PPmax and PPmin are the maximal and minimal pulse pressures and POPmax and POPmin are the maximal and minimal photoplethysmographic amplitudes within one respiratory cycle. Calculations were performed in R using the WaveletComp [11] and peakPick-packages

[12]. After downsampling to 40 Hz, respiratory variations in CVP (dCVP) were calculated as the absolute difference between the peak and trough of the CVP-pressure waveform within one respiratory cycle. All calculated values were plotted and visually inspected before being

8 accepted to the final dataset. Respiratory cycles with obvious disturbances (e.g. motion artifacts) were omitted. Details and examples of the calculations are presented in

Supplementary Material 1. For stroke volume, heart rate, mean arterial pressure and CVP, mean values, trimming the highest and lowest 5% to remove disturbances, were calculated for each LBNP-level.

Statistical analyses Sample size was estimated for another protocol performed on the same subjects the same day, and a separate power analysis was not performed for the currently presented results.

The number of subjects in the present study is comparable to other studies published in the same field. Confidence intervals of the present analyses are displayed according to the

CONSORT guidelines [13].

The potential effects of PEP and CPAP on the dynamic variables were analyzed separately.

The associations between the level of respiratory resistance and LBNP (explanatory variables), and the dynamic variables (outcome variables) were explored using linear mixed models (LMM) due to the clustering of data within subjects. Level of respiratory resistance

(baseline, 5 or 10 cmH20) was treated as a factor and LBNP-level as a continuous variable, including an interaction term between the two. When plotting LBNP-level on the x-axis and a dynamic variable on the y-axis, the slope represents the change in dynamic variable with a change in volume status, and thus the ability to reflect changes in volume status. The dynamic variables were right-skewed on the original scale, and were therefore loge- transformed before analysis. The results are presented back-transformed in the figures for clarity. Data were analyzed in R using RStudio. LMM was fitted using the glmmPQL function of “MASS” package [14], and estimates with confidence intervals for each LBNP and respiratory resistance level were calculated using the “glht”- function of the “multcomp ”-

9 package [15]. P-values were corrected for multiple post-hoc comparisons by the “single-step” method in the “multcomp” package. P-values < 0.05 were considered statistically significant.

RESULTS All subjects completed LBNP 20 mmHg, 19 subjects completed LBNP 40 mmHg, 13 subjects completed LBNP 60 mmHg and five subjects completed LBNP 80 mmHg.

Hemodynamic data are shown in Fig 2. Stroke volume, cardiac output and CVP were reduced from LBNP 0 mmHg at all LBNP-levels, whereas heart rate was increased from LBNP -40 mmHg. Mean arterial pressure did not change with progressive LBNP. dPP and dPOP with progressive LBNP and different levels of PEP or CPAP are presented in

Fig 3 and 4. dPP increased significantly with progressive LBNP both at baseline and during higher levels of PEP and CPAP. dPOP did not significantly increase with progressive LBNP during baseline PEP , but during PEP 5 and 10 cmH2O, and during all levels of CPAP. There was a significant difference between the regression slopes of dPP at baseline PEP and PEP 10 cmH20, indicating that the increase in dPP with progressive LBNP was amplified by the application of PEP 10 cmH20. The application of PEP 5 cmH20, CPAP 5 or 10 cmH20 did not induce any additional changes in dPP or dPOP (Table 1). dCVP increased significantly with increasing LBNP during all levels of PEP and CPAP (Table 2 and Fig 5). The increase in dCVP for any LBNP-level was significantly larger during PEP 5 cmH20 and PEP 10 cmH20 than during baseline PEP. By contrast, change in CPAP-level did not amplify the increase in dCVP.

DISCUSSION

The main findings of this experimental study were that the dynamic variables dPP and dPOP reflected progressive hypovolemia during all levels of CPAP and PEP, except for dPOP at baseline PEP. The application of PEP 10 cmH2O amplified the increase in dPP, but not in dPOP. Both progressive LBNP and the transition from baseline PEP to PEP 5 cmH2O and 10 cmH2O were associated with increases in dCVP.

During mechanical ventilation, several mechanisms contribute to respiratory induced changes in stroke volume [16]. The most important for the use of dynamic variables to diagnose volume status is the reduction in the gradient for venous return as RAP increases in the inspiratory phase. During mechanical ventilation, studies have shown that higher tidal volumes lead to an increase in dPP and SVV [17, 18]. Mesquida et al. altered tidal volume, chest wall compliance and cardiac function in dogs, and found that all changes in SVV and dPP were due to changes in right ventricular stroke volume following cyclic changes in venous return [19]. They concluded that changes in intrathoracic pressure, rather than changes in tidal volume as such, determine dPP and SVV.

Likewise, the use of dynamic variables to estimate volume status during spontaneous breathing relies on swings in intrathoracic pressure sufficiently large to produce notable changes in stroke volume over one respiratory cycle if the heart is working on the steep part of the Frank-Starling curve. Several maneuvers can amplify intrathoracic pressure changes during spontaneous breathing, such as deep, forced inspiration, or in- or expiration against a resistance. During deep spontaneous inspiration, RAP decreases and the gradient for venous return increases, increasing right ventricular filling. In a preload responsive heart, this leads to an increase in stroke volume which is visible as maximal pulse pressure after the pulmonary transit time of 2-3 s, normally during expiration. During mechanical ventilation,

12 intrathoracic and right atrial pressures increase in the inspiratory phase, which reduces the gradient for venous return and thus right ventricular filling. This leads to reduced left ventricular stroke volume and minimum pulse pressure a few heart beats later; normally during expiration. As these are opposite effects, a combination of the two forms of ventilation may neutralize the pressure changes specific to each respiratory mode, and reduce pulse pressure variations. We believe this explains why an increase in PEP-, but not

CPAP-level, affected dPP in the present study. PEP increases intrathoracic pressure during spontaneous expiration, and the gradient for venous return decreases [20]. By contrast,

CPAP increases airway pressures both during in- and expiration. This augments the reduction in preload and stroke volume during expiration, but the increase in venous return which normally occurs during spontaneous inspiration is offset by increasing intrathoracic pressure following insufflation of air under pressure. As a result, the maximal stroke volume and thus the variation in pulse pressure are smaller than when only an expiratory resistance is applied.

This is illustrated by the observed changes in CVP (Figure 4). The difference between the peaks and troughs in CVP (dCVP) at any given LBNP-level tended to increase with increasing

PEP, as only the maximal (expiratory) values increased. By contrast, during CPAP, the differences are smaller as both peaks and troughs tended to increase with increasing CPAP- levels, reflecting the continuously increased intrathoracic pressure.

Dynamic variables have shown some ability to diagnose hypovolemia or fluid responsiveness in spontaneously breathing subjects by manipulating respiratory pattern using Valsalva maneuvers [6], slow patterned breathing [21], a deep inspiratory maneuver [22] or forced inspiratory breathing [23]. However, in patients with hemodynamic instability, a forced respiratory maneuver decreased the diagnostic ability of dPP [24]. Airway pressures have also been amplified during spontaneous breathing by applying various respiratory resistors.

In a study on healthy volunteers using the LBNP-model, a good diagnostic ability for stroke volume variation was found during spontaneous breathing, but stroke volume variation was actually reduced with hypovolemia [25]. The diagnostic ability, and also the reduction with hypovolemia, was abolished by applying supported ventilation. On the other hand, in a porcine model, dynamic variables predicted fluid responsiveness well using an expiratory resistor of 7.5 cmH2O [5]. In a setup with the similar respiratory resistors on healthy volunteers inducing central hypovolemia using head-up tilt [26], the best ability to diagnose hypovolemia was found for systolic pressure variations using both inspiratory and expiratory resistors of 7.5 cmH2O. However, inspiratory resistors are rarely, if ever, used in routine clinical practice, and when using only an expiratory resistor, a statistically significant diagnostic ability was found for SVV, but not for dPP [26]. Also using head-up tilt, a diagnostic ability of dPP to detect hypovolemia was found using expiratory resistors of 7.5 cmH2O when combined with a respiratory rate of 6 breaths/min [27]. These studies indicate that using only an expiratory resistor (PEP) of less than 10 cmH2O may not impose sufficient intrathoracic pressure change to diagnose hypovolemia or fluid responsiveness unless further respiratory interventions such as increased tidal volume or reduced respiratory rate

(often implying increased tidal volumes) are also applied. However, our results suggest that by applying PEP as high as 10 cmH2O, the need for further respiratory adjustments may be reduced.

The photoplethysmographic waveform from which dPOP is calculated is complex, reflecting absorption of light from arterial, capillary and venous blood. Thus it is affected by both the cardiovascular, respiratory and autonomic nervous systems [28], and has been used to extract information about both fluid responsiveness [29], respiration [30] and pain [31]. dPOP appears to reflect volume status less consistently than dPP during mechanical

14 ventilation, which is generally explained by the complexity of the signal as well as proprietary processing and filtering algorithms [29, 32]. This may also explain why the transition to PEP 10 was reflected in dPP, but not in dPOP in the present study, and why dPOP remained unchanged with increasing LBNP-level during baseline PEP, whereas dPP increased. In a previous study, our group found a significant association between LBNP-level and dPP, but not dPOP [33], but this study was limited by small sample size. dPOP and the related automated variable Pleth Variability Index (PVI) have been found to decrease in response to passive leg raise. However, correlations with changes in cardiac index and the ability to predict fluid responsiveness were weak [34-36]. PVI was also found to increase with LBNP -40 mmHg when both PEEP 5 cmH2O and tripled tidal volume was applied, but not with either in isolation [37]. The fact that dPOP reflected volume loss in the present study may indicate that, as for dPP, amplification of pleural pressure swings improves the ability of dPOP to reflect hypovolemia.

Methodological considerations As this study was performed on healthy volunteers, its findings may not be valid for all patients. Breathing through a resistor requires patient cooperation, and the effect on intrathoracic pressure at a given resistance may differ in and between patients according to respiratory effort. We did not standardize respiratory rate, as we wanted to investigate the effects of CPAP and PEP during conditions that approximate ordinary clinical use of PEP and

CPAP. Baseline PEP and CPAP-levels of 0 cmH2O may not be regarded as regular spontaneous breathing, as breathing trough a mouthpiece or facemask as such may affect respiratory pattern. PEP and CPAP 0 cmH2O should rather be regarded as physiological baselines not intended for use in clinical practice, reflecting the experimental nature of this study.

The different respiratory interventions were of limited duration, as we feared that the advantages of longer intervention periods would be offset by earlier termination of the protocol in some subjects. Differences in respiratory and heart rate may also affect the impact of a forced respiratory maneuver on dPP and dPOP, as fewer heart beats per respiratory cycle give fewer stroke volumes in which to detect a difference [38].

For calculation of cardiac output, diameter of the aortic orifice and angle of insonation were not measured, but assumed. This probably led to a loss of accuracy, but as our analyses are based on changes and relative values of cardiac output, we do not believe this has had significant impact on our results.

To minimize the risk of complications in healthy volunteers, CVP was only measured in 10 subjects who had a prominent cubital vein entering the basilic vein, and insertion appeared technically uncomplicated. However, based on the estimates with confidence intervals, measurements from 10 subjects appear to suffice to illustrate the impact of different respiratory modes on CVP.

Calculations of the dynamic variables and dCVP were performed semi-automatically in R to increase reproducibility, as described in the Supplementary Material. As the algorithm has not been previously validated, all values were plotted and manually inspected to approximate the validity of manual calculation.

Different pulse oximeters use proprietary algorithms and produce different photoplethysmographic waveforms. Thus, dPOP from different pulse oximeters may give different results [39]. dPP was derived by the volume-clamp method, and dPP calculated from an invasively measured arterial pressure waveform may also differ that of the present study.

Conclusion

In this study, we found that dPP and dPOP reflected progressive hypovolemia during all levels of CPAP and PEP, except for dPOP at baseline PEP. The application of PEP 10 cmH2O amplified the increase in dPP, but not in dPOP. Both progressive LBNP and the transition from baseline PEP to PEP 5 cmH2O and 10 cmH2O were reflected in dCVP. Clinical studies may elucidate whether respiratory changes in pulse pressure or the photoplethysmographic waveform amplitude reflect volume status in patients during treatment with PEP or CPAP.

Conflict of interest: The authors declare that they have no conflicts of interest.

REFERENCES:

1. Shamir, M., et al., Pulse oximetry plethysmographic waveform during changes in blood volume. Br J Anaesth, 1999. 82(2): p. 178-81. 2. Pizov, R., et al., Arterial and plethysmographic waveform analysis in anesthetized patients with hypovolemia. Anesthesiology, 2010. 113(1): p. 83-91. 3. Cannesson, M., et al., Respiratory variations in pulse oximetry plethysmographic waveform amplitude to predict fluid responsiveness in the operating room. Anesthesiology, 2007. 106(6): p. 1105-11. 4. Berger, D. and J. Takala, Determinants of systemic venous return and the impact of positive pressure ventilation. Ann Transl Med, 2018. 6(18): p. 350. 5. Dahl, M.K., et al., Using an expiratory resistor, arterial pulse pressure variations predict fluid responsiveness during spontaneous breathing: an experimental porcine study. Crit Care, 2009. 13(2): p. R39. 6. Monge Garcia, M.I., A. Gil Cano, and J.C. Diaz Monrove, Arterial pressure changes during the Valsalva maneuver to predict fluid responsiveness in spontaneously breathing patients. Intensive Care Med, 2009. 35(1): p. 77-84. 7. Hisdal, J., K. Toska, and L. Walloe, Design of a chamber for lower body negative pressure with controlled onset rate. Aviat Space Environ Med, 2003. 74(8): p. 874-8. 8. Cooke, W.H., K.L. Ryan, and V.A. Convertino, Lower body negative pressure as a model to study progression to acute hemorrhagic shock in humans. J Appl Physiol (1985), 2004. 96(4): p. 1249-61. 9. Ricksten, S.E., et al., Estimation of central venous pressure by measurement of proximal axillary venous pressure using a "half-way" catheter. Acta Anaesthesiol Scand, 1986. 30(1): p. 13-7. 10. Eriksen, M. and L. Walloe, Improved method for cardiac output determination in man using ultrasound Doppler technique. Med Biol Eng Comput, 1990. 28(6): p. 555-60. 11. Roesch, A., H. Schmidbauer, and M.A. Roesch, Package ‘WaveletComp’, in The Comprehensive R Archive Network. 2014. 12. Westermark, P.O., peakPick: Peak Picking Methods Inspired by Biological Data. R package version 0.11. 2015. 13. Moher, D., et al., CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials. BMJ, 2010. 340: p. c869. 14. Ripley, B., et al., Package ‘mass’. CRAN Repos. Httpcran R-Proj. OrgwebpackagesMASSMASS Pdf, 2013. 15. Hothorn, T., F. Bretz, and P. Westfall, Simultaneous inference in general parametric models. Biometrical journal, 2008. 50(3): p. 346-363. 16. Michard, F., Changes in arterial pressure during mechanical ventilation. Anesthesiology, 2005. 103(2): p. 419-28; quiz 449-5. 17. Reuter, D.A., et al., Influence of tidal volume on left ventricular stroke volume variation measured by pulse contour analysis in mechanically ventilated patients. Intensive Care Med, 2003. 29(3): p. 476-80. 18. De Backer, D., et al., Pulse pressure variations to predict fluid responsiveness: influence of tidal volume. Intensive Care Med, 2005. 31(4): p. 517-23. 19. Mesquida, J., H.K. Kim, and M.R. Pinsky, Effect of tidal volume, intrathoracic pressure, and cardiac contractility on variations in pulse pressure, stroke volume, and intrathoracic blood volume. Intensive Care Med, 2011. 37(10): p. 1672-9. 20. Magder, S., Heart-Lung interaction in spontaneous breathing subjects: the basics. Ann Transl Med, 2018. 6(18): p. 348. 21. Zollei, E., et al., Non-invasive detection of hypovolemia or fluid responsiveness in spontaneously breathing subjects. Bmc Anesthesiology, 2013. 13.

22. Preau, S., et al., Hemodynamic changes during a deep inspiration maneuver predict fluid responsiveness in spontaneously breathing patients. Cardiol Res Pract, 2012. 2012: p. 191807. 23. Hong, D.M., et al., Pulse pressure variation to predict fluid responsiveness in spontaneously breathing patients: tidal vs. forced inspiratory breathing. Anaesthesia, 2014. 69(7): p. 717-22. 24. Soubrier, S., et al., Can dynamic indicators help the prediction of fluid responsiveness in spontaneously breathing critically ill patients? Intensive Care Med, 2007. 33(7): p. 1117-1124. 25. Elstad, M. and L. Walloe, Heart rate variability and stroke volume variability to detect central hypovolemia during spontaneous breathing and supported ventilation in young, healthy volunteers. Physiol Meas, 2015. 36(4): p. 671-81. 26. Dahl, M., et al., Can a central blood volume deficit be detected by systolic pressure variation during spontaneous breathing? BMC anesthesiology, 2015. 16(1): p. 58. 27. Bronzwaer, A.S., et al., Arterial Pressure Variation as a Biomarker of Preload Dependency in Spontaneously Breathing Subjects - A Proof of Principle. PLoS One, 2015. 10(9): p. e0137364. 28. Alian, A.A. and K.H. Shelley, Photoplethysmography. Best Pract Res Clin Anaesthesiol, 2014. 28(4): p. 395-406. 29. Addison, P.S., A review of signal processing used in the implementation of the pulse oximetry photoplethysmographic fluid responsiveness parameter. Anesth Analg, 2014. 119(6): p. 1293- 306. 30. Nilsson, L.M., Respiration signals from photoplethysmography. Anesth Analg, 2013. 117(4): p. 859-65. 31. Yang, Y., et al., Postoperative Pain Assessment Indices Based on Photoplethysmography Waveform Analysis. Front Physiol, 2018. 9: p. 1199. 32. Shelley, K.H., Photoplethysmography: beyond the calculation of arterial oxygen saturation and heart rate. Anesth Analg, 2007. 105(6 Suppl): p. S31-6, tables of contents. 33. Hoff, I.E., et al., Respiratory Variations in Pulse Pressure Reflect Central Hypovolemia during Noninvasive Positive Pressure Ventilation. Crit Care Res Pract, 2014. 2014: p. 712728. 34. Delerme, S., et al., Relation between pulse oximetry plethysmographic waveform amplitude induced by passive leg raising and cardiac index in spontaneously breathing subjects. Am J Emerg Med, 2010. 28(4): p. 505-10. 35. Delerme, S., et al., Variations in pulse oximetry plethysmographic waveform amplitude induced by passive leg raising in spontaneously breathing volunteers. Am J Emerg Med, 2007. 25(6): p. 637-42. 36. Keller, G., et al., Ability of pleth variability index to detect hemodynamic changes induced by passive leg raising in spontaneously breathing volunteers. Crit Care, 2008. 12(2): p. R37. 37. Nilsson, L.M., D.M. Lindenberger, and R.G. Hahn, The effect of positive end-expiratory pressure and tripled tidal volume on pleth variability index during hypovolaemia in conscious subjects: a volunteer study. Eur J Anaesthesiol, 2013. 30(11): p. 671-7. 38. Heenen, S., D. De Backer, and J.L. Vincent, How can the response to volume expansion in patients with spontaneous respiratory movements be predicted? Crit Care, 2006. 10(4): p. R102. 39. Hoiseth, L.O., et al., Respiratory variations in the photoplethysmographic waveform amplitude depend on type of pulse oximetry device. J Clin Monit Comput, 2016. 30(3): p. 317- 25.

Figure legends

Fig. 1 Protocol. Each LBNP-level lasted for approximately 6 min and consisted of 1 min stabilization and minimum 5 breaths at each level of PEP and CPAP. PEP- and CPAP-levels were assigned in randomized order. PEP: positive expiratory pressure, CPAP: continuous positive airway pressure

Fig. 2 Hemodynamic data. Individual hemodynamic data (grey lines) and estimates from the regression models (black lines, with 95 % confidence intervals) for each level of LBNP. P- values are compared to LBNP 0 mmHg. LBNP: lower body negative pressure

Fig. 3 The effects of LBNP and level of respiratory resistance on dPP. Boxplots of respiratory variations in PP from all observations in all subjects at each level of LBNP and respiratory resistance (left panel). Lines with ribbons are estimates and confidence intervals from the regression models (right panel). The slopes of the regression lines represent the change of dPP with a change in LBNP. Note that as the linear regressions were performed on the loge- values of dPP, the lines are curved on the original scale, as presented in the figure. dPP: respiratory variations in pulse pressure, PEP: positive expiratory pressure, CPAP: continuous positive airway pressure, LBNP: lower body negative pressure

Fig. 4 The effects of LBNP and level of respiratory resistance on dPOP. Boxplots of respiratory variations in POP from all observations in all subjects at each level of LBNP and respiratory resistance (left panel). Lines with ribbons are estimates and confidence intervals from the regression models (right panel). The slopes of the regression lines represent the change of dPOP with a change in LBNP. Note that as the linear regressions were performed on the loge- values of dPOP, the lines are curved on the original scale, as presented in the figure. dPOP:

20 respiratory variations in the photoplethysmographic waveform, PEP: positive expiratory pressure, CPAP: continuous positive airway pressure, LBNP: lower body negative pressure

Fig. 5 The effects of LBNP and level of respiratory resistance on CVP. Boxplots are from all observations in all subjects. Left panels show respiratory variations (difference between maximal and minimal CVP within each respiratory cycle) with different LBNP and PEP-/CPAP- levels. Right panels show the peaks and troughs within each respiratory cycle. The difference between the peaks and troughs gives the respiratory variations seen in the left panels. The lines with ribbons behind the boxplots are estimates and confidence intervals from the regression models. CVP: central venous pressure, dCVP: respiratory variations in central venous pressure, PEP: positive expiratory pressure, CPAP: continuous positive airway pressure, LBNP: lower body negative pressure

Table 1 The slope coefficients of dPP and dPOP for each LBNP-level (0, 20, 40, 60 and 80) during different levels of PEP and CPAP, and comparisons between the PEP and CPAP levels. dPP and dPOP-data are loge-transformed due to right skewness of the residuals. dPP: respiratory variations in pulse pressure, dPOP: respiratory variations in the photoplethysmographic waveform amplitude. LBNP: lower body negative pressure, PEP: positive expiratory pressure, CPAP: continuous positive airway pressure

Table 2 The slope coefficient of dCVP for each LBNP-level (0, 20, 40, 60 and 80) during different levels of PEP and CPAP, and comparisons between the PEP and CPAP levels. dCVP: respiratory variations in central venous pressure, LBNP: lower body negative pressure, PEP: positive expiratory pressure, CPAP: continuous positive airway pressure

Fig.1 Protocol

Fig.2 Hemodynamic data

Fig. 3 Respiratory variations in pulse pressure (dPP) during PEP and CPAP

Fig. 4 Respiratory variations in the photoplethysmographic waveform amplitude (dPOP) during PEP and CPAP

Fig. 5 Respiratory variations in central venous pressure (dCVP) during PEP and CPAP

Table 1. Effects of LBNP-, PEP- and CPAP-level on dPP and dPOP. Interventio PEP-/CPAP- Slope P- Difference to 0 P- Difference to P- n inlevel cm (95coefficient % CI) value (95%cmH2 0CI) value 5cmH20 (95% CI) value H2O

PEP 0 Loge 0.088 (0.032 to <0.001 5 (dPP) 0.14) 0.14 (0.087 to <0.001 0.054 (-0.023 to 0.27 10 0.20) 0.17 (0.12 to <0.001 0.13) 0.082 (0.006 to 0.028 0.028 (-0.047 to 0.76 0.22) 0.16) 0.10)

0 Loge(dPOP 0.051 (-0.01 to 0.13 5 ) 0.11) 0.10 (0.043 to <0.001 0.052 (-0.033 to 0.38 10 0.17) 0.08 (0.019 to 0.004 0.14) 0.029 (-0.057 0.81 -0.023 (CI -0.11 0.89 0.14) to 0.12) to0.062)

CPAP 0 Loge 0.12 (0.062 to <0.001 5 (dPP) 0.17) 0.16 (0.10 to <0.001 0.039 (-0.038 to 0.55 10 0.21) 0.19 (0.13 to <0.001 0.12) 0.069 (-0.008 0.097 0.03 (-0.047 to 0.74 0.24) to 0.15) 0.11)

0 Loge(dPOP 0.086 (0.025 to <0.002 5 ) 0.15) 0.08 (0.02 to <0.004 -0.005 (-0.088 to 1.0 10 0.14) 0.15 (0.09 to <0.001 0.078) 0.065 (0.019 to 0.19 0.069 (-0.014 to 0.14 0.21) 0.15) 0.15)

Table 2. Effects of LBNP-, PEP- and CPAP-level on dCVP. Intervention PEP-/CPAP- Slope P- Difference to 0 P-value Difference to P-value value levelin cm coefficient(95 % CI) cmH20 (95% CI) 5cmH20 (95% CI) H2O PEP 0 dCVP 0.49 (0.18 to <0.001 5 1.00.80) (0.68 to <0.001 0.50 (0.063 to 0.018 10 1.11.3) (0.81 to <0.001 0.94) 0.63 (0.20 to 0.001 0.13 (-0.31 to 0.88 1.43) 1.1) 0.56) CPAP 0 dCVP 0.99 (0.75 to <0.001 5 0.921.2) (0.69 to <0.001 -0.063 (0.39 to 0.96 10 0.781.2) (0.55 to <0.001 0.26) -0.21 (-0.53 to 0.35 -0.14 (-0.47 to 0.66 1.0) 0.12) 0.18)

26 Supplementary 1

Calculation of dPP and dPOP.

Arterial pressure and photoplethysmographic waveforms were exported as .txt-files and imported to R using RStudio.

The signals from the two sampling softwares (Regist3 and SignalExpress) were time- synchronized and given a common time-stamp.

The R-peaks of the ECG were detected and manually inspected with code using the

“wavelets”-package.

To locate the peaks within each heartbeat, a smoothed curve was created using the

“analyze.wavelet” and “reconstruct”-functions of the “WaveletComp”-package with a bandpass filter restricted to between 0.7 and 1.3 times the heart rate. The peak of this smoothed curve was located using the “peakpick”-function of the “peakPick”-package.

The peak of the original signal was found as the maximal value of the original waveform within ±0.3 s of the peak of the smoothed signal. The trough of the original signal was found as the minimal value within 0.4 s before the peak. Pulse pressure and photoplethysmographic amplitude was calculated as the difference between this maximal and minimal value.

Respiration was registered from the thorax impedance of the ECG-leads from the GE-Solar monitor. After smoothing the respiration signal using the “analyze.wavelet”-function with lower and upper periods of 2 and 20 s, respectively, the smoothed curve was plotted with the original, and lower and/or upper periods were adjusted if deemed necessary after visual inspection. The peaks of the smoothed signal were detected with the “peakpick”-function.

Maximal and minimal amplitudes for calculation of dPP and dPOP were identified between two peaks of the smoothed respiration signal.

All results were plotted and manually inspected before being entered to the final dataset with obviously erroneous values being removed or manually corrected.

Supplementary Figure 1: Example of the algorithm for calculating respiratory variations in pulse pressure from the Finometer, in this case for PEP 5 cmH2O at LBNP 40 mmHg. The black line is the original arterial pressure waveform. The peaks were found by locating the peaks of a smoothed signal (not shown), and thereafter locating and quantifying the peaks of the original waveform, presented as blue dots. The troughs of the original signal were found thereafter, presented as small green dots. The pulse pressure of each heartbeat was calculated as the difference between these two, presented as black numbers. The respiratory cycles were delimitated by finding the peaks of the respiratory signal of the ECG-

28 leads, presented as large green dots. The smallest and largest pulse pressures between two respiratory peaks were found, and pulse pressure variations calculated, presented as blue numbers. The values not representing respiratory cycles with resistance were removed after manual inspection. Time in seconds on the x-axis, arterial blood pressure in mmHg on the y- axis (labels from several plots overprojected in the example). dPOP was calculated using a corresponding algorithm.

Supplementary 2

Calculation of dCVP

Central venous pressure waveforms were exported as .txt-files and imported to R using

RStudio. The signal was downsampled from 400 to 40 Hz.

To calculate the variations with respiration, the signal was smoothed using the

“analyze.wavelet” and “reconstruct”-functions of the “WaveletComp”-package with a lower period set to 2 s. Peaks and troughs were located using the “peakpick” of the “peakPick” package.

All results were plotted and manually inspected before being entered to the final dataset with obviously erroneous values being removed or manually corrected. An example is presented below.

Supplementary Figure 2: Example of the algorithm for calculating respiratory variations in

CVP. The black line is the CVP-waveform downsampled to 40 Hz. The red line is the smoothed CVP-waveform after wavelet analysis and reconstruction. The local minima and maxima for each respiratory cycle are blue and green dots with values as presented, respectively. The difference between these values, representing the respiratory variation in

CVP, is presented in red. The cyan line is respiration from the ECG-leads. The numbers below the cyan line are comments entered during the experiments representing the respiratory resistance, in this case, the level of expiratory resistance in cmH20. The values not representing respiratory cycles with resistance were removed after manual inspection. Time in seconds on the x-axis, CVP in mmHg on the y-axis.