Despite its role as a cornerstone of the management of severe cardiogenic shock, VA-ECMO continues to be associated with a high mortality and complication rate. Complications associated with impaired LV ejection and elevated LV filling pressure – including thrombus formation and pulmonary oedema – are well recognized. Understanding the haemodynamic and circuit-related factors that contribute to elevation of LV filling pressure is critical to the rational, pre-emptive application of mitigation strategies, including mechanical unloading devices. In the absence of hemodynamic data beyond in-silico simulations, use of these devices is guided by expert opinion. This has resulted in a high degree of inter-institutional variability, and a sharp recent increase in the use of percutaneous left ventricular assist devices (pVADs) with their associated increased cost and complication rates7. More robust physical data – in-vitro and in-vivo – is needed.
In this context, the primary findings of this MCL study are: 1) under VA-ECMO support, aortic pressure and LV contractility independently correlate with LAP; 2) the effect of aortic pressure is mediated by LV contractility; 3) VA-ECMO flow rate does not independently affect LAP; and 4) direction of VA-ECMO flow does not affect LAP.
AoP and LV contractility independently predict LAP during VA-ECMO
Using an in-silico model, Dickstein demonstrated that, assuming a constant Frank-Starling relationship, the increase in pulmonary capillary wedge pressure (PCWP) following VA-ECMO initiation is dependent on both baseline LV function and the degree of AoP elevation6. In our study, LV contractility and AoP were independent predictors of LAP, confirming the findings of Dickstein in a physical model of the circulation. Of note, the consistency of findings between Dickstein’s model and ours occurred despite our model not incorporating Frank-Starling forces.
Our results extend those of Dickstein, by demonstrating an incremental steepening of the slope of the linear AoP-LAP relationship with reduction in LV contractility. These findings suggest that LV contractility acts as an effect modifier on the AoP-LAP relationship, pointing to a synergistic effect between LV contractility and AoP on the risk of elevated LV filling pressure. In the clinical setting, these results highlight the important role of inotropes and vasodilators – in combination where possible – to reduce LV pressures on VA-ECMO support. In our study, the slope of the AoP-LAP relationship in the setting of LV impairment was 0.40, suggesting that a reduction in AoP by 12.5mmHg is sufficient to achieve a clinically meaningful, 5mmHg reduction in LAP. Importantly however, clinical use of vasodilators and inotropes as first-line therapies can be limited by vasoplegia or ongoing myocardial ischaemia, respectively1. In-vivo pre-clinical and clinical data are needed to determine a dose-response relationship between afterload and LV filling pressure reduction, and to determine the optimal target AoP and inotrope dose.
Flow rate and direction do not predict LAP during VA-ECMO
When AoP was held constant, increased ECMO flow rate had no effect on LAP. Furthermore, in multivariable analysis, ECMO flow did not independently predict LAP. The putative effect of ECMO flow on LAP – demonstrated numerically in our study only when AoP was uncontrolled – can therefore be attributed solely to its effect on AoP. Changing the direction of ECMO return flow similarly had no effect on LAP. Taken together, these findings challenge the conventional orthodoxy that it is the retrograde nature of ECMO flow – and by extension, the amount of retrograde ECMO flow – that directly imparts afterload on the LV, therefore causing LV distension and increased filling pressure. Rather, consistent with the arguments of Dickstein6, our study suggests it is increased AoP, irrespective of the origin of this increase, that primarily drives changes in filling pressure. Clinically, our findings suggests that, rather than reducing ECMO flow to reduce LAP as is standard practice, the same effect could be achieved simply through pharmacologic reduction in SVR without sacrificing ECMO flow.
Limitations
This MCL was not designed to simulate complex regulatory adaptations, such as the baroreceptor reflex and Frank-Starling mechanism, nor the ability of VA-ECMO to produce adequate myocardial and peripheral tissue oxygenation and correct metabolic disturbance. Numerical models, hybrid MCLs and in-vivo studies may more accurately represent these biological effects14. Additionally, the ventricles in our MCL were connected ‘in-series’, without a common interventricular septum. Therefore, we were unable to examine the effects of ventricular interdependence, which may significantly attenuate the effects of VA-ECMO on LAP through changes in ventricular compliance15–18. Further studies simulating direct biventricular interactions and exploring the effects of VA-ECMO on biventricular pressure-volume relationships are needed. Finally, we used a HeartWare HVAD in our VA-ECMO circuit instead of a dedicated ECMO pump, although these are both centrifugal pumps and obey similar physiological and engineering principles.
Implications and Future Directions
These findings highlight the critical role of afterload pressure and LV contractility in determining LV preload under VA-ECMO support. In terms of clinical translation, they highlight the critical and synergistic role of vasodilators and inotropes as first-line therapies to prevent pulmonary oedema in this setting. Conversely, our results do not support the practice of altering ECMO flow in order to reduce LAP. Pre-clinical and clinical in-vivo studies are needed to establish a dose-response relationship of reduced afterload and positive inotropy on LAP, determine optimal therapeutic targets, and identify patients most likely to require escalation to more invasive measures such as mechanical unloading.