Recirculation during veno-venous extra-corporeal membrane oxygenation – a simulation study

Cardiovascular simulation model

A closed-loop, real-time simulation model was developed consisting of 27 vascular segments, 6 in the pulmonary circulation, and 21 in the systemic circulation, the 4 cardiac chambers with corresponding valves, the pericardium and intrathoracic pressure (Fig. 1). The cardiovascular simulation model has been published elsewhere (12), although minor modifications have been implemented since publication. One pulmonary arterial compartment was added to better represent the elastic properties of the pulmonary arteries, and the functions determining vascular pressures during low volumes and vascular collapse were changed to better represent the pressure-volume relation during hypovolemia and/or large swings in intrathoracic pressure. In short, the cardiac chambers are represented as time-varying elastances and the closed-loop vascular system segments are characterized by non-linear resistances, compliances, inertias, and viscoelastances. Valves open and close depending on pressure gradients.

The simulations in this article were performed with a normal cardiac function, but an increase in pulmonary vascular resistance (3.4 Wood units) and a pulmonary shunt fraction of 50% mimicking a typical clinical VV-ECMO patient in our unit with moderate pulmonary hypertension (Tab. I) and arterial oxygen saturation below 80% without ECMO support, despite ventilator therapy with 100% oxygen. Intrathoracic pressure was set to zero to avoid variability due to circulatory changes during the respiratory cycle. Pressures, flows, volumes and saturations in every compartment are updated with 4000 Hz.

ECMO simulation

ECMO flow was fixed at the set continuous flow rate of 0 to 5 l/min, with a selection of clinically relevant cannulation sites: 1) right atrium to inferior vena cava (RA→VCI), 2) inferior vena cava to right atrium (VCI→RA), and 3) superior+inferior vena cava to right atrium (VCI+VCS→RA) (Figs. 2a-c). The first two cannulation modes are the most common according to the ELSO database (13). The third mode (VCI+VCS→RA) resembles the commercially available Wang-Zwische/Avalon cannula (AvalonElite; Avalon Laboratories Rancho Dominguez, CA USA) (9). When two draining cannulas were used (VCI+VCS→RA) flow was equally divided between the two. Since ECMO flow was constant, neither elastic nor inertial properties of the tubings were included in the simulation. Resistance of the tubings were set to values creating realistic pressure drops as seen in our institution. Although these pressures are important for the function of the pump, they do not affect the physiology of the patient, since flow is constant (mimicking a roller pump) so the values are therefore not presented.

Oxygen transport

The oxygen-carrying capacity of blood C (ml O₂/liter blood) in the circulation of the patient was calculated according to equation 1 (14), where Hb is the hemoglobin level (g/l) and Sat is the oxygen saturation (%) of the vascular compartment (mL)

C = 0.0134 ⋅ H b ⋅ S a t      Eq. [1]

Physically dissolved oxygen was only taken into account in post-oxygenator blood, where oxygen partial pressure, pO₂, is extremely high (typical 30-40 kPa), corresponding to 5% of total oxygen content (See supplementary material available online at www.artificial-organs.com). In the vascular compartments of the simulated patient, saturations were below 94%, corresponding to a pO₂ of <9 kPa (<1.5% dissolved oxygen), which in this study was considered negligible. The oxygen saturation was considered homogenous in each compartment and exchange of oxygen between compartments proportional to flow. A uniform hemoglobin level of 114 g/l was used in all simulations, since clinical mean hemoglobin level is 114 g/l (Tab. III). Total oxygen consumption excluding the heart was set to 250 ml/min. Cardiac oxygen consumption was calculated according to Suga et al (15) depending on the mechanical workload of the heart (in the study between 18 and 19 ml/min). Recirculation, R, in the model was calculated according to equation 2 modified from Walker et al (16) as explained in the supplementary material, where S_preoxO₂ is the oxygen saturation in the drainage cannula of the ECMO-system, S_postoxO₂ is the oxygen saturation in the returning cannula of the ECMO system, and S_in is the oxygen saturation in the veins drained by the cannulated vessel segment.

R = S p r e o x O 2 − S i n 1.04 ⋅ S p o s t o x O 2 − S i n      Eq. [2]

Simulation of circulatory failure in VV-ECMO due to increase in pulmonary vascular resistance

Pulmonary vascular resistance was increased from 3.4 Wood units to 12.2 Wood units in 20 steps, by decreasing the radius of the pulmonary resistance arteries by 2.0% per step, mimicking a clinical scenario where a patient with VV-ECMO deteriorates and becomes a candidate for veno-arterial support (17) or septostomy (18). VV-ECMO flow (VCI+VCS→RA) was kept constant at 4 l/min. This is often in real life accompanied by (or maybe more correctly caused by) worsening of pulmonary function (“white out”), but to simplify interpretation of data the pulmonary shunt was kept constant at 50% in the simulation. Heart rate was kept constant at 100 bpm. No autonomic reflexes or adaptive mechanisms were included in the simulation.

Clinical patients

Eleven consecutive adult VV-ECMO cases were selected retrospectively from our clinical database in the year of 2012 (Tab. I). Mean age was 45 ± 18 and 7/11 were males. Eight of the eleven patients survived. Mean SAPS-3 score at admission was 77 ± 15 (Tab. I), suggesting an expected hospital mortality rate of about 70% (19, 20). Venous drainage was accomplished by a 23 F 25 cm cannula placed with its tip in the right atrium via the jugular vein in all patients. Oxygenated blood was returned into one femoral vein through a 17 F to 21 F 18 cm cannula. Patients converted to VA-ECMO within the first 48 hours were excluded. Arterial oxygen saturations and oxygen saturations in the draining cannula of the ECMO system (S_preoxO₂) were collected together with ECMO flow rates and base-line patient characteristics between 23 and 27 hours from admission.

Calculations and statistics

The program version used was Aplysia CardioVascular Lab 5.5.0.11 (Aplysia Medical AB, Stockholm, Sweden). Mean values in the model were calculated as running arithmetic means. All data were collected at end-diastole after at least 5 min simulation to allow for steady-state conditions regarding hemodynamics and oxygen transport. No intrathoracic pressure changes were included in the simulation. The oxygenation of the patients’ non-shunting pulmonary capillary blood was set to 99.4%.

Simulation hemodynamics

Heart rate was kept at 100 bpm during all simulations. Mean systemic arterial blood pressure was 77 mmHg and mean pulmonary artery pressure was 22 mmHg in all simulations. The minimal differences seen in cardiac output can be explained by the slight changes in filling pressures created by the venous ECMO circulation. These differences are realistic, but not big enough to permit clinical detection and do not affect the conclusions in this study. See Table II.

1) VV-ECMO Simulation. Right atrial outflow and inferior vena cava return

Recirculation increased seemingly linearly up to 48% at ECMO blood flow 5 l/min. About 70% of the patients’ oxygen uptake was supplied by the ECMO circuit at this flow. Mixed venous oxygen saturation (S_vO₂) was about 21% lower than the saturation in the venous tubings of the ECMO circuit (S_preoxO₂). Arterial saturation was 89% (Figs. 3a-e).

Fig. 3 -

Simulation results with different ECMO flows and cannulation modes. a) shows recirculation, b) mixed venous oxygen ­saturation, c) ECMO preoxygenator oxygen saturation, d) systemic arterial oxygen saturation, and e) ECMO oxygen transfer.

4) VV-ECMO Simulation. Vena cava inferior+superior outflow and right atrial return. Circulatory failure due to progressive increase in pulmonary vascular resistance

Cardiac output decreases from 5.5 l/min to 3.4 l/min when pulmonary vascular resistance increases from 3.4 Wood units to 12.2 Wood units (Fig. 4b). Mean systemic arterial blood pressure declines from 77 mmHg to 54 mmHg and mean pulmonary arterial blood pressure increases from 22 mmHg to 42 mmHg (Figs. 4b and 5). Recirculation increases from 17% to 31%, while S_vO₂ decreases from 61% to 43% (Fig. 4a). Pre-oxygenator saturation (S_preoxO₂) decreases 5% from 68% to 63% (Fig. 4a). The difference between S_vO₂ and pre-oxygenator saturation (S_preoxO₂) increases from 7% to 20% (Fig. 4a).

Fig. 4 -

Simulation results in circulatory deterioration due to pulmonary hypertension. a) shows recirculation, mixed venous oxygen saturation and preoxygenator oxygen saturation; b) shows mean systemic arterial pressure, cardiac output, mean pulmonary arterial pressure and pulmonary vascular resistance. The horizontal axis is an arbitrary time scale, showing the development of pulmonary hypertension and right heart failure in 20 steps as described in the text.

Fig. 5 -

Left (black) and right (grey) ventricular pressure-volume loops with pulmonary vascular resistance 3.4 Woods unit (upper panel) and 12.2 Woods unit (lower panel). Dilatation of the right ventricle is seen in the lower panel with a reduction of the left ventricular volume due to lower right heart stroke volume, the constraining effect of the pericardium and ventricular septal interaction.

Clinical patient data

Oxygen data in the patients closely corresponded to simulated data derived from interpolations in Figures 3a-e (the simulated values are shown in parentheses in the following text). Mean arterial saturation was 86 ± 8% (88%) and venous saturation (S_preoxO₂) was 75 ± 3% (78%). Detailed hemodynamic data were not available, but by assuming oxygen extraction of 30%, we calculated an estimated S_vO₂ (also assuming ­similar saturations in superior and inferior caval veins, see supplementary material available online at www.artificial-organs.com), and were thereby able to calculate an estimated mean recirculation rate of 42 ± 6% (43%). A substantial difference between the estimated S_vO₂ of 56 ± 8% (56%) and venous saturation (S_preoxO₂) of 75 ± 3% (78%) was seen in all patients (Tab. III). Mean oxygen transfer in patients was 163 ± 35 ml/min (178 ml/min).

Pre-oxygenator saturation is an unpredictable measure of oxygenation

Simulation data shows that venous pre-oxygenator saturation (S_preoxO₂) is an unreliable measure of treatment efficiency in veno-venous ECMO. Oxygen saturation in venous drainage blood (S_preoxO₂) may both increase and decrease in patients with circulatory deterioration. The slight decrease in venous pre-oxygenator saturation seen in our simulation of low cardiac output due to right heart failure (Figs. 4a and b and 5) illustrates the low sensitivity of this parameter to life-threatening changes in hemodynamics. This is explained by a combination of increasing recirculation and lowered S_vO₂. In the most extreme form of circulatory deterioration, cardiac arrest, pre-oxygenator saturation in VV-ECMO will be 100% without any benefit to the patient. Together with available clinical experience and sparse scientific data, this suggests that arterial oxygenation parameters rather than venous saturations should be used for optimizing therapy (ECMO flow rate and cannula position).

Cannulation modes

In our clinical cohort, all patients were treated with drainage from the right atrium and reinfusion into the inferior caval vein. The reason for this clinical tradition is anecdotal experience from the 1990s suggesting better drainage and flow stability. However, both simulation data and a clinical study (7) indicate that lower flow may be needed to oxygenate the patient with flow from the femoral vein to the atrium or with a Wang-Zwische/Avalon cannula (9). Furthermore, maximal flow was higher in the clinical study when draining from the femoral/caval vein and reinfusing into the atrium (7).

Oxygen extraction

The figure used in our calculations of 30% oxygen extraction is based on normal physiology, where arterial saturation is 95% to 100% and S_vO₂ is 65% to 70% with normal metabolism and cardiac output. Critical ill patients may vary considerably in both cardiac output and metabolism, but an oxygen extraction of 30% is still a reasonable estimate (7), since the most severe cases of septic distributive shock and cardiac failure are excluded in a cohort of patients with VV-ECMO.

Limitations

Many clinical features of veno-venous ECMO are captured by the simulation. However, there are some limitations. Our simulation model does not include changes in pCO₂, bicarbonate, pH, and base excess, since focus is on recirculation and oxygen transport. The shift of the oxygen dissociation curve due to these factors is, however, usually less important for oxygenation than recirculation in these patients.

It has been shown clinically that dissolved oxygen is of importance for oxygen delivery when high levels of oxygen are provided to the patient either in the ventilator or in the ECMO circuit (8, 16). We have considered dissolved oxygen in post-oxygenator blood where pO₂ may be extremely high, but not included this in oxygen transport calculations in the simulated patient, since it is of minor importance below a pO₂ of 9 kPa to 10 kPa as is seen in VV-ECMO patients.

Recirculation in simulation of VV-ECMO with drainage from inferior+superior caval vein and return of oxygenated blood in the right atrium is higher than values reported from clinical measurements in patients with the Wang-Zwische/Avalon cannula (9). This may be explained by limitations of the model, where return of blood mixes instantly with atrial blood without a directed flow jet towards the tricuspid annulus. It must be understood, however, that despite an optimal cannula position some recirculation will occur because of a continued return flow during systole as well, when the tricuspid valve is closed.