We present a reproducible ovine model of severe cardiopulmonary failure supported by VA-ECMO. CS was induced by ethanol injection and confirmed by significant changes in pre-defined hemodynamic (SAP < 90 mmHg), metabolic (lactate > 4 mmol/L) and echocardiographic criteria (LVEF < 30 %), consistent with CS definitions from contemporary clinical trials and guidelines . These were further corroborated by pathognomonic biochemical, macroscopic, and histopathological features. Acute respiratory failure was induced by lowering mechanical ventilatory support and confirmed by pre-defined hypoxic criteria in ABG analysis.
Heart failure models should be induced to reflect various types of clinical situations, especially for mechanical circulatory support devices (such as ECMO) for profound cardiogenic shock. A systematic review by our group found that 7 out of the 18 studies investigating CS were unsuccessful at meeting a wide range of CS diagnostic criteria and would rather be described as acute heart failure models, rather than cardiogenic shock models . Methodological limitations of large animal trials can risk the clinical translation of their findings, hence particular attention was paid to confirm the extent of shock in all animals [25, 26].
Confirming site-specific injury: Electrophysiology
Ventricular fibrillation is the most frequent side-effect of previous experimental cardiogenic shock models, leading to death in up to 50% of animals [16, 17, 21, 27]. To reduce the loss of animals, antiarrhythmics were administered before ethanol injection, and this resulted in ventricular fibrillation in only one animal. Complete heart block developed in one animal, likely due to severe ethanol-induced injury to the atrio-ventricular conduction system. To further characterize this model electrophysiologically, both endo- and epicardial electroanatomic substrate mapping were performed in one animal. This showed that almost 20% of the endocardial LV surface had electrograms attenuated to an extent consistent with scar. This attenuation was not confluent but was seen in discrete zones of regional endocardial and epicardial LV myocardium corresponding to injection sites. The electrograms in these zones were attenuated but did not exhibit fractionation or isolated delayed components. This is consistent with the histological demonstration of dense confluent cellular necrosis in the injected regions without intervening surviving myofiber bundles. Epicardial voltages were relatively preserved, consistent with a zone of subepicardial myocyte preservation related to the depth of ethanol injection. Thus, formation of myocardial necrosis is directly related to injury site. Increased number of sites injected relates to expansion of scar area, allowing for titration of left-ventricular failure.
Advantages of ethanol-based model of CS
In the model presented, ethanol injection resulted in an isolated and loco-regional left ventricular wall motion abnormality, with changes being limited to injection site. This achieves multiple desirable goals:
Injury can be created regionally, and wall motion abnormality can be localized directly,
The extent of injury is potentially titratable – researchers can study decreasing levels of heart failure,
The pattern of injury can replicate a desired clinical scenario (right heart infarction, anterolateral infarction etc.) with high precision in multiple animals.
Direct ethanol injections for the development of right heart failure have been previously demonstrated by Thomaz et al, with isolated right heart failure induced by direct ethanol injection in 13 dogs . This study showed a 74% transmural infarction area, limited to the right ventricle, assessed on the 14th day after injury was induced. To date, only one study by Marques and collaborators has reported the use of intramural left ventricle ethanol injections, unfortunately with limited and inconclusive description of outcomes and no use of VA-ECMO to achieve considerable levels of cardiac failure . Several experimental heart failure models have been published with intracoronary ethanol injections [30, 31]. However, the pathogenesis of these models is essentially distinct from an intramural injection model as the former achieves LV injury by promoting a persistent and extensive constitution of thrombi distally to the injection site, while the latter induces myocardial necrosis without intraluminal thrombi [32, 33]. Effects of intraluminal injection disperse to the respective coronary vascular bed, while intramural injection limits the injury to its specific site. A site-specific injury pattern achieved with our model enables researchers to create specific clinical scenarios and allows for pre-clinical validation in a randomized setting, ultimately giving clinicians better guidance when it comes to choosing the correct treatment for the right patient.
Novel animal model of CS within the context of previously developed models of CS
Our model differs from previously published models, in its ability to reproduce a stable and controllable CS.
To date, pre-clinical heart failure models have employed a range of different animals (primarily sheep, pigs, and dogs) and methods (e.g., coronary ligature or occlusion, rapid pacing, myocardial hypoxia) . The majority of patients who develop acute CS requiring MCS have an overwhelming AMI; therefore, unlike chronic heart failure models (e.g., rapid pacing, mitral regurgitation models), experimental acute heart failure models have predominantly been created through coronary artery occlusions (surgical or intravascular). However, the latter can frequently lead to critical and unpredictable adverse events, such as refractory hemodynamic instability caused by ventricular arrhythmias or sudden cardiac arrest . Other models, such as global hypoxia, esmolol-induced LV failure or carbon monoxide poisoning have limited translatability, as they induce global LV failure which is rarely seen in AMI patients. Therefore, in comparison with previous models, our model overcomes limitations of achieving a myocardial injury which is either excessively severe or marginal, without possibility to titrate the degree of injury.
Several limitations of our model should be highlighted. Firstly, whilst on peripheral VA-ECMO, ejection fraction decreases as afterload increases. Thus, during ECMO, left ventricular outflow tract velocity time integral could be a reliable estimate of heart function, independently of the afterload. Due to dissimilarities between the human and sheep thorax anatomy, measurement of outflow tract velocity time integral was not obtained. Yet, to minimize the effect of increased afterload, we obtained our echocardiographic assessments at baseline and at CS, using only 1 L/min of VA-ECMO support to minimize afterload, as is often applied in clinical practice to assess the possibility of weaning patients from VA-ECMO . Secondly, only six animals were evaluated, yet consistent results were found amongst all the evaluated animal subjects. Thirdly, to achieve ethanol administration, the thorax was opened, which is known to modify the heart-lung interactions in mechanically ventilated patients . Indeed, opening the thorax will cancel the effect of MV on stressed vascular volume, in particular the increase in RV afterload, decrease in LV afterload and venous return. Nevertheless, an open chest model was essential to titrate and evaluate injury in these animals. Finally, we carried out cardiac mapping in only one animal, due to the extensive costs of these assessments. Although we obtained critical findings, further corroboration of electrophysiological impairments associated with ethanol injections should be obtained in future studies.
Prospects for future studies in CS: VA-ECMO and increased left ventricular overload
As previously outlined, although ECMO can provide essential restoration of end-organ perfusion, this technology is not devoid of potential hazards. Of these, it’s increasingly recognized that peripherally instituted VA-ECMO can hamper cardiac recovery as it exerts stress and strain on the left ventricular myocardium . This so-called cardiac overload is clinically relevant, since increased filling pressures can promote pulmonary oedema and formation of intraventricular thrombi with embolic events, which can drastically worsen patient outcomes irrespective of the use of VA-ECMO [38, 39]. The effect of increasing VA-ECMO flow on left-ventricular dimensions has been comprehensively assessed by Ostadal et al. in a porcine model of cardiogenic shock . The authors reported an increase of left ventricular volumes as ECMO flow was incrementally increased. Similar findings were present in our model: EDA and ESA increased significantly after full-flow ECMO support was initiated. Importantly, use of unloading devices, i.e. Impella is associated with decreased mortality at the cost of increased bleeding rates . Thus, in conjunction with continuous measurement of intracardiac pressures and volumes, our model could be crucial for future studies investigating innovative unloading technologies for VA-ECMO during CS.