Assessment of Right Heart Function during Extracorporeal Therapy by Modified Thermodilution in a Porcine Model

In compliance with the Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 1996) and Swiss National Guidelines, the Commission of Animal Experimentation of Canton Bern, Switzerland approved the study as independent substudy of an experiment about extracorporeal membrane oxygenation gas exchange (BE111/18); the manuscript adheres to the applicable Animal Research: Reporting of In Vivo Experiments (ARRIVE) Guidelines. The experiments were performed within roughly 12 h, with the anesthesia and surgery taking place in the morning and thereafter measurements in the afternoon. Further details on the method including anesthesia and experimental protocol are provided in the Supplemental Digital Content (https://links.lww.com/ALN/C425). The data that support the findings of this study are available from the corresponding author upon reasonable request.

Sixteen healthy pigs (Sus scrofa domesticus, Schweizer Edelschwein, age 12 to 15 weeks depending on individual growth for a median weight of 45.5 [42 to 47] kg, 10 female) were anaesthetized and ventilated (volume controlled ventilation, positive end-expiratory pressure 5 cm H₂O, fractional inspired oxygen tension 0.60, tidal volume 10 ml/kg body weight). After sternotomy and heparinization, an extracorporeal membrane oxygenation circuit (Stockert SCPC Centrifugal Pump, Germany & Capiox FX15 Oxygenator, Terumo, USA) was implanted via the right atrium and ascending aorta. Ultrasonic flow probes were placed around the pulmonary artery main trunk, the left pulmonary artery, and the arterial extracorporeal membrane oxygenation cannula. The thoracic cavity and pericardium were closed.

Sweep gas flow, extracorporeal membrane oxygenation blood flow, pulmonary blood flow with respective temperatures, ventilator parameters, carotid artery, and atrial pressures were recorded (100 Hz) in Labview (National Instruments, USA and Soleasy, Alea Solutions, Switzerland).

Two standard pulmonary artery catheters with rapid response thermistors (CCCombo 7778 F, Edwards Lifesciences, USA) provided thermodilution signals in the pulmonary artery and extracorporeal membrane oxygenation circuit (fig. 1). After calibration of catheter constants with five injections of 3, 5, and 10 ml each (iced saline 0.9%) into the extracorporeal membrane oxygenation circuit (4 l/min flow), extracorporeal membrane oxygenation flow was reduced in 1-l steps to 1 l/min. At each flow five 10-ml injections into the right atrium followed. Indicator backwash was assessed by direct injection into the right ventricle (fig. 2). Because the weaning steps had to follow a clinically applicable procedure, they could not be randomized, nor could injections be blinded. All animals underwent the weaning steps sequentially.

The Stewart-Hamilton equation states that cardiac output (CO) is inversely proportional to the area under the temperature change and directly proportional to the injectate volume^8,13  (where BT = body temperature, ΔBT = thermodilution signal, IV = injectate volume, IT = injectate temperature):

Our model is based on the hypothesis that the injectate volume is divided into the extracorporeal membrane oxygenation circuit and the lung circuit, depending on the different flows. After determining the catheter constant c (see equation 5 below), we define two different injectate volumes (IV_{extracorporeal membrane oxygenation} and IV_Lung), where IV_{extracorporeal membrane oxygenation} plus IV_Lung equals the total injected volume, such as (BF = blood flow):

The catheter constant c is a composite constant of the heat capacity factor K₁ and a correction factor C_T,^8,11  where:

σ₀ is the specific heat of the injectate (Joules · kg⁻¹ · °C⁻¹) whereas ρ₀ refers to the specific gravity of the injectate (unity).⁸  σ_B and ρ_B refer to the same parameters for blood. K₁ describes the expected temperature exchange between blood and indicator. C_T scales this factor to the intrinsic properties of the catheter used (priming volume, distance from injection port to thermistor and injection volume¹¹ ). Because K₁ is a composite of the constants for specific heat and gravity, C_T × K₁ is usually combined into one catheter constant c (equation 4’).

Catheter constants were calculated from direct injections into the extracorporeal membrane oxygenation circuit, where backwash is impossible and blood flow is known accurately:

There is a linear dependency of the catheter constant c and injectate volume¹¹  (Supplemental Digital Content, formula 2, https://links.lww.com/ALN/C425). Therefore, a linear regression was applied to find the dependency of the catheter constant on injection volume (formula 6), where formula 5 yields a numerical solution for formula 6.

For right atrial injections, formula 3 is solved for injection volume (IV_ECMO) using extracorporeal membrane oxygenation blood flow, derived c_iv and the extracorporeal membrane oxygenation thermodilution curve. IV_Total – IV_ECMO represents the injection volume passing the lung circuit:

This allows pulmonary blood flow calculations for all individual thermodilution curves:

Right ventricular ejection fraction was modeled by the exponential decay of the pulmonary thermodilution signal.^14–16  Right ventricular behavior was described using ejection fraction, end-diastolic and end-systolic volumes, central venous and pulmonary artery pressure and pulmonary vascular resistance.

All data were analyzed using Matlab R2019a (MathWorks, USA) with extension for Bland-Altman plots.¹⁷  Normality was assessed using Shapiro-Wilk’s test and QQ-plots. Data are presented as mean and SD for normally distributed data and median [minimum to maximum] for individual regressions. Correlation coefficients were calculated using Pearson’s square (r²). Linear regression (least square fit) was used to calculate the catheter constants, to assess agreement between calculated and true blood flow and to describe the relationships between ventricular volumes. One-Way ANOVA for repeated measurements (within-subject factor weaning step) was performed to assess the differences in right ventricular volumes, pulmonary vascular resistance, central venous and pulmonary artery pressures, and ejection fraction between extracorporeal membrane oxygenation flow reductions after assessment of sphericity with Mauchy’s test and Greenhouse-Geyser correction when appropriate. Post hoc pairwise comparison between groups were performed with Bonferroni correction. Bland-Altman analysis was used to evaluate agreement between methods with additional corrections for proportional bias and additional limits of agreement defined as ± 1.96 × root mean square error.¹⁸ P < 0.05 was considered significant with two-tailed testing. Injections were averaged after removing outliers (defined as 1.5 interquartile ranges above the upper or below the lower quartile). The sample size of a total of 16 animals was calculated for the main study protocol (extracorporeal membrane oxygenation gas exchange). No specific power calculation was conducted for this substudy.

We performed 785 thermal injections in 16 animals. Data from the first three pilot animals (135 injections) were excluded. We analyzed a total of 643 of 650 per protocol thermodilution injections. Seven injections were excluded because of missing pulmonary blood flow during 4 l/min extracorporeal membrane oxygenation (five injections) and calculation of negative pulmonary blood flow (two injections). Negative pulmonary blood flow resulted from calculated IV_ECMO greater than 10 ml.

Average constants from the extracorporeal membrane oxygenation injections were 4.5 ± 0.6 (3 ml IV), 5.1 ± 0.5 (5 ml IV), and 5.3 ± 0.4 (10 ml IV, P < 0.001). Linear regression yielded c = 0.09 × IV + 4.39 (P < 0.001, individual median and range r² = 0.41 [0.05 to 0.81], fig. 3). Injections into the right ventricle led to thermal changes in the extracorporeal membrane oxygenation thermistor in 7 of 13 animals (26 of 195 injections [13%]). Because of this heat loss into the extracorporeal membrane oxygenation circuit, we only used extracorporeal membrane oxygenation injections to derive the catheter constants.

Mean extracorporeal membrane oxygenation blood flow during the reduction steps (fig. 2) was 4,010 ± 119 ml/min, 3,019 ± 80.8 ml/min, 2,001 ± 58.4 ml/min, and 951 ± 86 ml/min, respectively, as specified per protocol. Calculation of indicator partitioning always yielded a value between 0 and 10 ml, as indirect proof of the robustness of the calculations. Injection volumes drawn into the extracorporeal membrane oxygenation circuit were reduced with reductions in extracorporeal membrane oxygenation blood flow to a mean of 7.1 ± 1.4 ml, 5.6 ± 1.3 ml, 3.4 ± 1.2 ml, and 1.2 ± 0.5 ml, respectively (fig. 4A, P < 0.001). The temperature integrals at the extracorporeal membrane oxygenation decreased with decreasing flow, counterintuitive to classic thermodilution (fig. 4B), owing to decreasing amounts of indicator, which were redirected toward the pulmonary circulation.

The calculated indicator partitions were used to calculate pulmonary blood flow according to formula 6 (fig. 5). Measured blood flow and calculated blood flow correlated highly (median r² = 0.85 [0.76 to 0.93] for individual animals, P < 0.001, fig. 5A). Averaging five thermodilution injections (outliers removed) gave a mean bias of −6 ± 47 ml and limits of agreement at 668 ± 166 (fig. 5C). Bias and increasing blood flow correlated weakly (r² = 0.15, P = 0.005), meaning that pulmonary blood flow is overestimated at high and underestimated at low lung flows. We therefore corrected the Bland Altman Plot with proportional bias¹⁸  (fig. 5C), yielding limits of agreement of 623 ml and an overall corresponding percentage error of 32.1%. The percentage error varied between flow groups with an error of 55.5% at 4 l/min extracorporeal membrane oxygenation flow, 35.4% at 3 l/min extracorporeal membrane oxygenation flow, 29.0% at 2 l/min extracorporeal membrane oxygenation flow, and 24.3% at 1 l/min extracorporeal membrane oxygenation flow.¹⁹ 

The calculated change in pulmonary blood flow between each extracorporeal membrane oxygenation blood flow reduction tracked the measured change in pulmonary blood flow with perfect concordance (100%, fig. 5B, individual median r² = 0.56 [0.02 to 0.98]). The least significant change²⁰  to be detected was 135 ml.

Right ventricular ejection fraction increased from 15.6% to 18.2% with extracorporeal membrane oxygenation flow reductions with a marked increase in measured stroke volume (table 1, fig. 6). The resulting increase in end-diastolic volume showed a highly linear correlation with end-systolic volumes (individual median r² = 0.99 [0.88 to 0.99], P < 0.001) and calculated stroke volume (fig. 6B, individual correlations for each animal: median r² = 0.96 [0.63 to 0.97], P < 0.001). Heart rate remained unchanged (r² = 0.00, P = 0.454). The high correlations were verified with the stroke volumes measured by the flow probes (individual median r² = 0.99 [0.98 to 0.99], P < 0.001 and 0.93 [0.70 to 0.98], P < 0.001, for end-systolic and stroke volumes, respectively, fig. 6, C and D). Central venous pressure remained constant during weaning (table 1) and over various end-diastolic volumes (fig. 6E). Pulmonary vascular resistance decreased significantly with increasing pulmonary and decreasing extracorporeal membrane oxygenation flow (table 1 and fig. 6F).

The Supplemental Digital Content (https://links.lww.com/ALN/C425) provides all individual plots for all injections and animals together with a step-by-step explanation of calculations.

The thermodilution signal at the extracorporeal membrane oxygenation circuit from an injection into the right atrium can be used to adequately determine the amount of injectate passing the pulmonary circulation, which allows accurate calculations of native cardiac output. Due to its accuracy and validation by multiple studies, thermodilution has become the gold standard in clinical cardiac output measurement.²¹  Because indicator loss into the extracorporeal circuit has been described for veno-venous extracorporeal membrane oxygenation¹⁰  with invalid results for cardiac output, thermodilution during extracorporeal treatment is not recommended.¹²  We verified the indicator loss for a veno-arterial setup even at low extracorporeal flows. The amount of indicator loss is flow dependent, thereby inverting the classic indirect proportionality of flow magnitude and area under the temperature curve.⁸  Three corollaries follow: (1) Calculation of catheter constants should be performed with injections solely into the extracorporeal membrane oxygenation circuit, where the indicator will fully pass the thermistor. The variation of calculated catheter constants is a result of the high precision flow probe (± 2% accuracy according to the manufacturer) and within reported margin of error for thermodilution.^22,23  Our method allows an in vivo calibration of catheter constants using repetitive measurements. (2) The deliberate injection into the right atrium as mixing chamber and an expected partition of the indicator with thermistors in both circuits ensure the recording of the total indicator volume. Blood flow estimation during extracorporeal membrane oxygenation therapy becomes possible, because indicator loss is controlled for and calibrated catheter constants for changing indicator volumes on the basis of a linear correlation between the indicator volume and the value of the constant are known. (3) Conventional thermodilution techniques are unable to determine cardiac output nor is the trending ability of these techniques applicable in daily practice because of the dynamic change of indicator volume with changes in blood flow.

In this animal model investigating clinically relevant extracorporeal membrane oxygenation blood flows, our method stays reliable with acceptable agreement to a high precision flow probe at the pulmonary main trunk. Owing to the similarities between the cardiovascular systems of pigs and humans, we expect that our model behaves similarly in humans. Further testing is warranted. There is a small overestimation with increasing pulmonary blood flow, a known phenomenon for thermodilution^23–25  and probably not clinically relevant. The trending ability (i.e., the directional flow change) is always aligned between the measured and calculated flow (fig. 5B), albeit with varying correlations. The least significant change of our method (i.e., the smallest detectable change of flow which could be considered statistically significant) is small. This allows detection of flow changes even at low blood flows. Our estimations are within clinically useful and acceptable limits of agreement.^20,26  We used a high-precision flow probe with a very high accuracy and measured at relatively low flows, which in turn may have influenced the percentage error of our method to its disfavor.²⁷  The method achieves a lower percentage error at higher blood flows, where it would be of particular interest for predicting weaning success. Whether repeated calibrations could increase accuracy should be further studied.

In veno-arterial extracorporeal membrane oxygenation, the extracorporeal membrane oxygenation circuit unloads the right ventricle. Reducing extracorporeal membrane oxygenation flow will inevitably lead to a right ventricular volume load. This leads to a simultaneous increase in end-diastolic and end-systolic volumes (fig. 6A). The slope of this figure represents 1 minus the ejection fraction, which remains constant within a narrow margin during extracorporeal membrane oxygenation flow reduction, in accordance with previously reported findings, where end-systolic and end-diastolic volumes shared the same relationship.¹⁵  Our data show that the right ventricular function is uniform in this population of healthy animals. The linear relationship between the end-systolic and -diastolic volume remains when reassessed with the flow probe measurements.

Although the calculated ejection fraction changes significantly, the overall change remains at 2.6%, which cannot be clinically relevant. Therefore, end-diastolic volume has to increase with the consecutive filling of the right ventricle to guarantee adequate cardiac output. This may have relevant consequences for weaning trials guided by echocardiography, where dilation is usually seen as a marker of failure, if accompanied by paradoxical septal motion or decreased functional parameters.^28,29  Based on what we show here for extracorporeal membrane oxygenation and what other studies have shown for afterload effects,³⁰  right ventricular dilation is a prerequisite to increase stroke volume. The decreasing pulmonary vascular resistance with increasing pulmonary flow refutes that this ventricular dilation is caused by increased afterload, but is explained by distention and pulmonary capillary recruitment with increasing pulmonary blood flow.³¹ 

With volume state constant, a rise in central venous pressure as indication of right ventricular failure¹⁵  did not occur. The ventricles did not reach their performance limit. The ventricular dilation may be attributed to increases in venous return.^15,32  One might suspect mathematical coupling between end-systolic and end-diastolic volumes: First, because end-diastolic volume is substantially larger than stroke volume and end-systolic volume is defined as end-diastolic volume minus stroke volume, end-diastolic volume and end-systolic volume will correlate inherently. Second, the exponential decay of the thermodilution signal is described by the function e^−kt. This function also determines the area under the temperature curve, which is inversely proportional to the blood flow and therefore stroke volume. We remained suspicious of the high correlations shown in our results and reported by other studies.³³  Nevertheless, the correlations remain very high and significant when independently measured instead of calculated stroke volumes are used (fig. 6, C and D). In these analyses, a direct mathematical relationship between ejection fraction and stroke volume calculations is impossible. With the use of two independent measurement methods, the suspected mathematical coupling has been previously addressed and contradicted.³⁴  Such linear behavior of right ventricular function has been previously described^15,31,35  and differs from left ventricular function curves.³⁶ 

The technical setup necessary for our method would only need an additional rapid response thermistor in the extracorporeal membrane oxygenation circuit and a standard pulmonary artery catheter. After calibrating the catheter constants using the extracorporeal membrane oxygenation circuit, injections into the right atrium could be performed to accurately monitor cardiac output. Guiding therapy based on exact cardiac output measurements and excellent trending could improve overall management of these patients.³⁷  In veno-arterial extracorporeal membrane oxygenation, an analysis of right ventricular function during extracorporeal membrane oxygenation weaning would be of great benefit. A quantification of right ventricular function may be important, because continuous unloading by the extracorporeal membrane oxygenation can mask right ventricular failure³⁸  and standard echocardiographic parameters may not be reliable during extracorporeal membrane oxygenation.³⁹  Right ventricular ejection fraction has predictive value for weaning success.⁴⁰  The possibility to continuously calculate right ventricular ejection fraction and filling volumes during extracorporeal membrane oxygenation weaning might allow recognition of a deteriorating clinical course and weaning failure. Right ventricular failure during extracorporeal membrane oxygenation weaning could simply be expressed in a change of the end-diastolic volume versus end-systolic volume slope toward one (where end-diastolic volume would equal end-systolic volume) and an increase in central venous pressures.^15,41 

There are several limitations to our study. First, we have evaluated a healthy animal population in a highly standardized experimental setup, where we demonstrated high precision and acceptable accuracy. The diagnostic and therapeutic value in human patients with severe diseases, impaired ventricular function or severe valvular dysfunction needs to be investigated. Computer simulations may aid in the planning of such experiments.⁴²  Second, our experiments were performed in the setting of veno-arterial extracorporeal membrane oxygenation. Whether a similar technique could be used in veno-venous extracorporeal membrane oxygenation would need to be studied. Third, our method can assess right ventricular filling volumes and functions. These findings could be verified by independent measurements of stroke volume and pulmonary vascular resistance. However, a direct visualization of ejection fraction using echocardiography and reporting transmural atrial or ventricular pressures may strengthen our conclusions. The ejection fractions reported here are lower than expected for pigs⁴³  and might be attributed to relative hypovolemia or catheter positioning, known confounders of the thermodilution based method.^44,45  Fourth, at maximum extracorporeal membrane oxygenation flow settings and therefore low pulmonary blood flow, the limits of agreement are within the range of cardiac output. However, cardiac output measurement at these flow settings might not be clinically relevant, because relevant unloading of the right ventricle and low cardiac output must be expected. We have calculated our percentage error using data from all flow settings and it remains close to 30%, as an accepted clinical limit.²⁶  Its value is influenced by the low flow states, where the constant limits of agreement are compared to a low flow. The stepwise calculation of the percentage error for every single setting shows that it decreases with increasing pulmonary flow.¹⁹  This is clinically important, because the closer pulmonary blood flows are to flows necessary for successful weaning, the more accurate the method becomes. On the other hand, increasing right ventricular load may increase tricuspid regurgitation, which we did not evaluate and may lead to underestimation of the thermodilution cardiac output,⁴⁶  while the estimates of ejection fraction may remain accurate.⁴⁷  Finally, we investigated pulmonary thermodilution and our findings cannot necessarily be extrapolated to transpulmonary thermodilution, where indicator loss might be less relevant because the indicator will be measured in a central artery behind the arterial extracorporeal membrane oxygenation canula. Still, this would allow estimations of combined extracorporeal membrane oxygenation flow and cardiac output because the injectate would pass both circuits and its sum would be measured in the femoral artery. Therefore, individual measurements of pulmonary flow are by definition not possible through transpulmonary thermodilution. A recent study found that cardiac index parameters when measured by transpulmonary thermodilution remained unchanged after onset of veno-venous extracorporeal membrane oxygenation.⁴⁸  However, indicator loss was not assessed directly. In contrast, we show a substantial partition of indicator throughout all jugular injections, and we reference our calculated blood flow to a high accuracy flow probe.

An adaption of the classical thermodilution technique with temperature measurements in both circuits results in a reliable method of estimating cardiac output and right ventricular function during veno-arterial- extracorporeal membrane oxygenation therapy. This might provide an easily applicable and readily available tool to monitor the clinical course and guide therapy for patients on veno-arterial extracorporeal membrane oxygenation.

The study was supported by grants from the “Stiftung für Forschung in Anästhesiologie und Intensivmedizin” (Bern, Switzerland; Nr 26/2018) awarded to Drs. Bachmann and Berger and by the “Fondation Johanna Dürmüller-Bol” (Muri bei Bern, Switzerland; Nr 481) awarded to Dr. Bachmann.

The Department of Intensive Care Medicine, Bern University Hospital, University of Bern (Bern, Switzerland) has, or has had in the past, research contracts with Abionic SA, AVA AG, CSEM SA, Cube Dx GmbH, Cyto Sorbents Europe GmbH, Edwards Lifesciences LLC, GE Healthcare, ImaCor Inc., MedImmune LLC, Orion Corporation, Phagenesis Ltd. and research and development/consulting contracts with Edwards Lifesciences LLC, Nestec SA, Wyss Zurich. The money was paid into a departmental fund. No author received any personal financial gain. The Department of Intensive Care Medicine has received unrestricted educational grants from the following organizations for organizing a quarterly postgraduate educational symposium, the Berner Forum for Intensive Care (until 2015): Abbott AG, Anandic Medical Systems, Astellas, AstraZeneca, Bard Medica SA, Baxter, B | Braun, CSL Behring, Covidien, Fresenius Kabi, GSK, Lilly, Maquet, MSD, Novartis, Nycomed, Orion Pharma, Pfizer, Pierre Fabre Pharma AG (formerly known as RobaPharm). The Department of Intensive Care Medicine has received unrestricted educational grants from the following organizations for organizing bi-annual postgraduate courses in the fields of critical care ultrasound, management of extracorporeal membrane oxygenation and mechanical ventilation: Abbott AG, Anandic Medical Systems, Bard Medica SA., Bracco, Dräger Schweiz AG, Edwards Lifesciences AG, Fresenius Kabi (Schweiz) AG, Getinge Group Maquet AG, Hamilton Medical AG, Pierre Fabre Pharma AG (formerly known as RobaPharm), PanGas AG Healthcare, Pfizer AG, Orion Pharma, Teleflex Medical GmbH.