Increasing venoarterial extracorporeal membrane oxygenation flow puts higher demands on left ventricular work in a porcine model of chronic heart failure

Venoarterial extracorporeal membrane oxygenation (VA ECMO) represents a well-established method that can treat refractory but potentially recoverable cardiogenic shock and thus revert a dire prognosis. Temporarily, it can fully substitute the functions of the lungs and heart to maintain sufficient gas exchange and systemic blood circulation [1, 2]. VA ECMO forms a circulatory bypass by draining blood from the right atrium, passing it through the gas exchange unit, and returning the oxygenated blood into the thoracic aorta. Benefits of VA ECMO on resuscitability, improved survival, and improved neurologic outcomes have been proved both in practice [3, 4] and in animal experiments [5‐7].

However, it has also been observed that VA ECMO significantly affects systemic circulation and the workload of the heart [8‐13]. Increasing VA ECMO blood flow seems to negatively impact left ventricular (LV) performance; hence, it has been suggested to use only the lowest possible rate of circulatory support to protect myocardial function [8, 10]. Several causes for this negative effect have been repeatedly suspected. During high bypass flow, the inflow of blood reaches the aortic root and increases the mean aortic pressure and the LV afterload. These factors, together with preload and coronary perfusion, play key roles in heart performance. In patients with reduced myocardial contractile reserve, the increase in LV wall tension and end-diastolic pressure (EDP) due to excess afterload opposes the coronary perfusion [14] and exacerbates heart insufficiency [15]. In structural heart diseases, additional mitral regurgitation may contribute to rise in left atrial pressure; consequently, pulmonary congestion may appear as a feared complication of VA ECMO [16‐18].

Especially in this context of LV overload, evaluation of VA ECMO in experiments with heart failure (HF) models have become important before translations to clinical practice. LV performance has been studied on models with intact or acutely decompensated hearts [6, 8, 19‐22], but to our knowledge, there have been no experimental studies on hemodynamic effects of VA ECMO in conditions of chronic HF and its decompensation. Even though a significant part of VA ECMO clinical applications is for circulatory decompensation developed on grounds of previously present chronic heart disease, we still lack evidence from corresponding experiments. Furthermore, retrospective clinical studies have also revealed that outcome of patients treated by ECMO differs according to the “acuteness or chronicity” of cardiac disease [23].

Although there are reasons why the use of chronic HF models is scarce—time-consuming preparations, instability of heart rhythm, high mortality rates, ethical questions—the advantages of using chronic HF models are evident as they offer prolonged neurohumoral activation, general systemic adaptation, functional changes of cardiomyocytes, and structural alterations of heart valves [24‐27]. In this experimental work, a chronic HF animal model was developed by long-term fast cardiac pacing which produced tachycardia-induced cardiomyopathy (TIC). TIC was first recognized in 1913 [28] and later widely used in experiments [29]. It reliably mimics decompensated dilated cardiomyopathy with low cardiac output which persists also after cessation of pacing and allows investigation of chronic HF conditions [27, 30‐32].

The aim of our study was to describe hemodynamic and LV performance changes in a swine model of decompensated chronic HF, supported by gradual increase in flow of VA ECMO. We focused on LV pressure and volume changes as well as myocardial contractility and work. It is expected that VA ECMO should be able to supply enough systemic blood circulation, but the details of the negative effects on LV performance have yet to be evaluated.

Detailed results are summarized in Table 1 and Fig. 4. In the all animals in our study, fast ventricular pacing for 4–8 weeks generated TIC with signs of decompensated HF which was denoted by baseline values of cardiac output 2.9 ± 0.4 L/min at rest, severe dilation of all heart chambers, valves insufficiency, and systemic hypotension. Left ventricular EF evaluated by echocardiography was below 30% in all animals; initial mean heart rate of sinus rhythm was 100 ± 19 beats/min; dyssynchrony of LV contraction was obvious. Baseline SvO₂ value of 62 ± 8% at rest corresponded with inadequate tissue oxygen delivery in our model, and elevated CVP underlined congestion.

After connecting VA ECMO, stepwise increments of EBF from minimal to maximal flow led to gradual and dramatic changes in LV hemodynamic parameters. LVPP increased by 49% from 49 ± 15 mmHg to 55 ± 13, 61 ± 13, 66 ± 12, 74 ± 10, and 73 ± 11 mmHg (for EBF 0 to 5, P = 0.001), and EDP increased by 114% from 7 ± 2 mmHg to 8 ± 2, 10 ± 2, 11 ± 3, 13 ± 2, and 15 ± 3 mmHg (for EBF 0 to 5, P < 0.001). Every escalation of EBF emphasized LV dilation. ESV increased severely by 20% from 139 ± 17 mL to 143 ± 16, 148 ± 18, 152 ± 19, 164 ± 15, and 167 ± 15 mL and EDV by 15% from 189 ± 26 mL to 194 ± 27, 203 ± 29, 209 ± 30, 217 ± 29, and 218 ± 30 mL (both for EBF 0 to 5, P < 0.001).

On the other hand, SV and EF changed only with less significant mean differences and both reached highest values at EBF 3 L/min. In detail, SV changed from 51 ± 20 mL to 51 ± 20, 56 ± 20, 59 ± 20, 55 ± 21, and 52 ± 21 mL (P = 0.03), and EF from 25 ± 7% to 24 ± 6, 26 ± 7, 27 ± 7, 23 ± 6, and 21 ± 6% (P = 0.18). Mean HR tended to decline with every increase in EBF—from 101 ± 22 beats/min to 96 ± 19, 93 ± 17, 90 ± 13, 90 ± 14, and 86 ± 14 beats/min (for EBF 0 to 5, P = 0.34).

Left ventricular SW was calculated from measured pressure–volume loops and exhibited significant flow-dependent increases from 1434 ± 941 mmHg\(*\)mL to 1595 ± 987, 1867 ± 1102, 2014 ± 1062, 2105 ± 1060, and 1892 ± 1036 mmHg\(*\)mL (EBF 0 to 5, P = 0.04). However, preload independent index of LV contractility represented by dP/dt_max/EDV ratio showed no consistent trend during the ECMO protocol—from 2.2 ± 0.8 mmHg/s/mL to 2.2 ± 0.6, 2.4 ± 0.4, 2.5 ± 0.4, 2.8 ± 0.6, and 3.0 ± 0.9 mmHg/s/mL (EBF 0 to 5, P = 0.94).

Arterial blood flow increased with every increase of EBF—in total by 127% in the carotid and by 187% in the subclavian artery, while pulsatility indices dropped significantly. SvO₂ increased to 77 ± 3% with EBF 1 and reached > 80% with all higher EBF steps (P < 0.001). Similarly, rSO₂ values were low at baseline but increased promptly with ECMO flow. With increasing EBF, the average value of CVP did gradually fall, but not under 7 mmHg, avoiding ECMO underfilling (P = 0.001).

Postmortem autopsies did not reveal any shunt or other cardiac anomaly, but myocardial hypertrophy (heart weight 471 ± 127 g).

VA ECMO is being used as an ultimate method in cases of severe circulatory decompensation, but multiple clinical and experimental studies have documented adverse changes in LV function with the increased EBF [2, 8, 12, 18] for both cardiac [13] and respiratory [45] compromised patients. The incidences of these complications vary widely between 12 and 68% [13, 46‐48] and are still believed to be underreported [13, 46].

The goal of our experiment was to assess the response of hemodynamic parameters and LV workload to different levels of EBF. We chose a porcine model of TIC, developed by long-term fast ventricular pacing, over weeks resulting in symptoms and signs of anatomical, functional, and neurohumoral profiles similar to human chronic HF. As far as we are informed, this study is unique in that it describes the hemodynamic effects of VA ECMO in decompensated chronic HF model. An important feature of our protocol is also adequate lung ventilation which ensures that coronary arteries receive well-oxygenated blood during all rates of EBF. Therefore, reported changes in LV hemodynamics should not be attributed to coronary ischemia as a result of poor oxygenation.

With initiation of the extracorporeal circulation and stepwise increase of EBF, progressive dilation of LV was observed. Affected were both the end-diastolic (in total by 15% from EBF 0 to 5) and even more, the end-systolic volume (by 20%). This increase in LV dimensions was strongly pronounced between EBF 0 and EBF 4; by increasing the flow beyond this point, left ventricle did not significantly further dilate.

LV pressures demonstrated upward trends as well. LVPP increased by 49% but still remained abnormally low despite high EBF. LV EDP was initially elevated to 7 mmHg due to abnormal LV filling dynamics, but when compared to a model of acute heart failure [8], the EDP elevation during low EBF in TIC was pointedly lower (Figs. 5 and 6). This may be explained by the chronicity of our HF model which led to massive ventricular dilation without appreciable thickening of its wall, higher ventricular compliance, and thus less pronounced EDP elevation. With ECMO flows from EBF 0 to 5, EDP increased significantly further by over 114% leading to high preload and high end-diastolic wall tension, possibly opposing coronary perfusion [14].

On the other hand, SV and EF had a different course when increasing EBF. Both these measures of LV ejection showed increase from EBF 1 to EBF 3, but with higher EBF, their mean values decreased. Systolic ejection trended to maximum at middle levels and to minimum at the highest levels of VA ECMO support. Such a trend was not observed in acute cardiogenic shock model induced by regional myocardial hypoxemia where ejection continued to decline [8] but was observed in another model of global myocardial hypoxia supported by stepwise VA ECMO [49]. Interestingly, in the latter work, this trend was observed regardless of pulsatile or non-pulsatile ECMO flows.

SW, defined as the area of PV loop and representing the instant LV workload, was expected to be proportional to ventricular systolic pressure and SV [40, 41]. In our measurements, demands on the LV work measured by SW showed significant increase by 40% from EBF 0 to EBF 4, and with further escalation of ECMO flow to EBF 5, declined. In fact, SW is well respecting the trend of LVPP and SV.

The maximal positive LV pressure change is considered to be one of isovolumetric phase contractility indices. Due to severe LV dilation in our experiment, we used this index normalized to EDV, dP/dt_max/EDV, as recommended for its high sensitivity for global inotropic state, irrespective of ventricular loading conditions [39, 41, 43]. It has been confirmed that its linear relation is affected less by afterload compared to the end-systolic PV ratio [43]. As expected, the initial value in our TIC model was significantly below normal. Further across the stepwise protocol of EBF, the ratio of dP/dt_max/EDV continued to increase but did not meet statistical significance. For isovolumetric phase of contraction, this could imply that LV did not lose its ability to contract. In individuals where the dP/dt_max/EDV ratio did not decrease with higher EBF, this could be explained by unrestricted or at least, sufficient coronary perfusion and secondarily by preserved potential of contractility. A similarly designed study reported hemodynamic effects of VA ECMO, but applied to healthy canine circulation [14]. Although they reported reduction of coronary flow with increasing EBF, myocardial oxygen consumption was not reduced. In a similar sense, our data did not indicate critically inadequate myocardial perfusion on any of EBF levels.

Blood propelled to the aortic root, especially potentiated by aortic and mitral regurgitations, often leads to increased left atrial pressure [15]. Yet, in none of our cases, progression into pulmonary edema was observed, which could be due to the short-term duration of extracorporeal support and sufficient right heart unloading.

The combination of increasing SW and not declining dP/dt_max/EDV ratio demonstrate increased demands on LV work, placed by the high afterload, and this increase in LV work occurs concurrently with LV dilation [41].

As the metabolic rate and hematocrit remain unchanged, systemic organ perfusion can be predicted by arterial flows and mixed venous blood saturation. We previously reported linear correlation of tissue saturation and regional arterial flow in VA ECMO support of chronic HF. Both tissue saturation and regional arterial flow demonstrated significant increases with respect to EBF [7]. Low initial value of SvO₂ improved already at level of EBF 1, suggesting sufficient systemic perfusion with only minor extracorporeal support. Not surprisingly, with every higher EBF step, average SvO₂ and carotid arterial flow rose, but the pulsatility index gradually decreased, demonstrating loss of aortic pulse pressure and the dominance of ECMO blood flow in systemic circulation [5, 7]. Practically, the value of regional tissue saturation together with SvO₂ have the potential to serve as additional methods to secure vital organ perfusion in ECMO therapy guidance.

A recent study by our group described hemodynamic effects of VA ECMO on an innovative animal model of acute HF induced by hypoxic coronary perfusion [8, 20], and these findings were later confirmed by computer modeling [9]. Similar to our current design, the effects of increasing EBF on LV were reported. Figure 5 is depicting PV loops of acute and chronic model reactions to VA ECMO flow which can offer insights into hemodynamic changes in both models.

In comparison to our current study, baseline ESV and EDV in acute HF were reported to be of far smaller dimensions (46% for ESV, 58% for EDV) and the effect of faster EBF was different in acute HF—by increasing EBF, ESV increased by 31% (64 ± 11 mL to 83 ± 14 mL, P < 0.001), but EDV only by 10% (112 ± 19 mL to 123 ± 20 mL, P = 0.43, both for EBF 1 to 5). Likewise, increase of LVPP was more pronounced (by 67%; from 60 ± 7 to 97 ± 8, P < 0.001), but changes of EDP were only mild (17.2 ± 1.4 to 19.0 ± 2.9, P = 0.87, both for EBF 1 to 5). In both experiments, SW followed the same trend, reaching the highest value at EBF 4, and HR declined with every increment of EBF.

In Fig. 6, PV characteristics of this acute model are plotted against our results. Values of EDV, ESV, EDP, and LVPP for corresponding EBF steps reveal linear relations with dissimilar slopes (r = 0.94, r = 0.97, r = 0.92, and r = 0.97, respectively, all P < 0.05).

In another model of acute HF generated by hypoxic myocardial perfusion, Shen et al. reported a decline of dP/dt_max and LVPP associated with VA ECMO flow, but in their settings all of the coronary vascular bed received hypoxemic blood [22].

When summarized, in our chronic HF model, EDP was lower at baseline but increased more with higher EBF, and end-diastolic dilation was more pronounced. EF and SV were proved to decline only in the case of acute hypoxic HF. One possible explanation is that in our protocol, high ECMO flow improves the coronary perfusion with oxygen rich blood, in contrast to the acute models, where a major portion of LV myocardium is perfused by constant flow of hypoxic blood, stunned, and therefore cannot keep pace and eject against increasing afterload. Long-term adaptation to chronic HF conditions with humoral activation during TIC model induction could be another possible explanation.

Although the impact of LV overload on clinical outcomes is mostly unknown and the literary evidence limited, it is expected that it negatively affects patients’ recovery [13, 50] and in minority of patients, the status can further progress into pulmonary congestion or edema [13, 17, 18, 50]. In addition, recent small clinical and experimental series have shown that the severity of LV overload and distention correlates well with the rate of EBF [8, 11, 51].

In clinical settings, echocardiography can serve as a method of choice to evaluate the ventricular response to volume loading during or while weaning from ECMO [11, 50, 51]. Assessments of SV and filling pressures, together with pulmonary wedge pressure, lung fluid accumulation, and regional saturations, are available measures which should be operated on daily basis [13, 50, 52]. Having these in hands, partial flows of extracorporeal support in conjunction with aggressive inotropic support have been suggested in effort to prevent LV overload [7, 8, 13].

In the presented study we showed how VA ECMO increases the demands on LV work and that despite significant increase of EDP, negative effect on load-independent contractility was not observed, suggesting sufficient myocardial oxygenated blood supply. Several methods have been proposed to unload the ventricles under VA ECMO therapy but are associated with increased rate of complications and the clinical evidence is limited to few single center small cohorts and case reports [13, 16‐18, 50]. In a recent report, 15 out of 36 patients with LV distention on VA ECMO required some venting strategy to reduce wall stress [13]. It has also been shown that even a small venting catheter of 7 or 8 Fr in left atrium [47] or LV [16] seems sufficient and that its early applications result in lower mortality [18, 48]. However, the clinical relevance of these effects still remains uncertain.

We assume that to decrease the risk of LV overload, VA ECMO flow should be maintained at the lowest level securing adequate tissue perfusion.

Even though the study protocol was prepared carefully and the total number of studied animals was adequately reasonable, several limitations must be considered. First, real isovolumetric phase was absent on recorded PV loops, as it could only be seen in cases with intact heart valves. In real measurements, LV volume may never stay constant due to valve insufficiency. The aortic regurgitation in our experiments was caused by PV loop catheter and was not considered severe on echocardiography. Second, contractility indices are highly sensitive but are also known to have a low reproducibility for contractile status. Therefore, myocardial perfusion cannot be directly deducted. In future studies, coronary arterial flow or oxygen consumption should be assessed. A third limitation could be seen in the relatively brief duration of extracorporeal support used in this study compared to common durations of ECMO therapy. To test long-term effects, experiments should be extended, and lung fluid content assessed. Lastly, right ventricular workload should also be considered.

We confirm that increase in EBF during VA ECMO therapy increases demands on LV performance, and our work extends this observation to conditions of decompensated chronic HF. We conclude that to protect the myocardium from harm, EBF of VA ECMO should be set with respect to not only systemic perfusion but also to LV parameters.