The burden of haemocompatibility with left ventricular assist systems: a complex weave

Contemporary durable mechanical circulatory support devices, based on axial or centrifugal continuous blood flow and most aptly referred to as left ventricular assist systems, have resulted in a paradigm shift from crisis support to long-term performance-based goals in patients with advanced heart failure.1 Progress in engineering has occurred by miniaturization of left ventricular assist systems, reduction in moving parts and in some cases, frictionless pumps that do not rely on mechanical bearings or hydrodynamic forces to operate optimally.2 , 3 Observationally, low pulsatile characteristics of blood flow of these newer circulatory pumps have introduced unique complications that are broadly considered to represent abnormalities in haemocompatibility.4–6 In this context, the word haemocompatibility refers to the consequences of either a harmonious relationship or an adverse interaction between the artificial pump interface and the activation or destruction of circulating blood elements. Clinically, this term extends well beyond this discrete definition into observed organ specific pathological consequences. Indeed, the holy grail for a more haemocompatible pump is to induce a physiological state as close to a biological mimic of native cardiac pulsatile physiology as possible, without incurring a penalty on extra-cardiac organ function. The constellation of haemolysis, pump thrombosis leading to malfunction, development of gastrointestinal bleeding, and strokes (haemorrhagic or thromboembolic) have all been suggested to represent evidence for an aberrant haemocompatible physiological state.7 More recent observations, however, have challenged the linear logic of these observations, arguing instead that these are complexly interwoven systems with suppression of one while activation of another ensues. We have observed a dissociation of these various adverse events between different left ventricular assist systems. Indeed, the axial continuous flow pump, the Heartmate II St. Jude Medical, Inc. (now Abbott), is associated with a trend to a higher rate of haemolysis and pump thrombosis than the centrifugal continuous flow pump, the HeartWare device (based on indirect comparisons), although various series show this adverse effect to be quite prominent with both devices.8–10 This centrifugal flow pump may exhibit a nominally lower rate of pump thrombosis leading to surgical reoperations but demonstrates higher rates of stroke-related morbidity than the axial flow pump.11 Both devices are associated with similarly high rates of bleeding complications, particularly gastrointestinal bleeding, a complication predicted by the development of an acquired Von Willebrand defect. A novel pump, the fully magnetically levitated centrifugal flow device, the Heartmate 3 St. Jude Medical (now Abbott), has been shown to eliminate haemolysis and pump thrombosis, but maintains the same rates of strokes and gastrointestinal bleeding as compared with the Heartmate II device in the short-term, even though observations suggest that it does not shear high molecular weight multimers with the same proportion as the other devices.5 , 12 , 13 This dissociation is depicted in Table 1, from trials that have randomized devices allowing for a parallel comparison of these adverse effects. There are other devices with innovative design features such as the EVAHEART (Nagano, Japan)14 and DURAHEART (Terumo Heart, Michigan, USA)15 systems, which have not been evaluated in comparative effectiveness trials. The EVAHEART uses a hydrodynamic contactless centrifugal pump design, retains pulsatility, and reduces degradation of Von Willebrand Factor but is limited by neurological adverse events.16 The DURAHEART was the first magnetically levitated centrifugal system and demonstrated absence of pump thrombosis but was limited by its larger profile and unacceptable drive-line characteristics with high infection rates.15 Thus, engineering innovation must account for clinical needs. How do we reconcile these observed dissociations and discrepancies in the context of the term ‘haemocompatibility’? Is one device not less or more haemocompatible? Does linearity have to be justified by demonstration of alterations in all components of the clinically observed adverse blood-device interface to define haemocompatibility? Differences between ‘haemocompatibility related events’ in left ventricular assist systems Six month outcomes from randomized clinical trials.11 , 13 + = modest risk (1–5%). ++ = significant risk (>5%–10%). +++ = marked risk (>10%). Pump thrombosis leading to reoperation and pump exchange. (note that the DURAHEART and EVAHEART are not included due to absence of randomized comparative effectiveness studies14–16). Differences between ‘haemocompatibility related events’ in left ventricular assist systems Six month outcomes from randomized clinical trials.11 , 13 + = modest risk (1–5%). ++ = significant risk (>5%–10%). +++ = marked risk (>10%). Pump thrombosis leading to reoperation and pump exchange. (note that the DURAHEART and EVAHEART are not included due to absence of randomized comparative effectiveness studies14–16). Observations of adverse sequelae of the pump and blood circulatory interface arise from a multiverse of abnormalities that include (a) direct consequences of increased shear stress on circulating blood elements,17 (b) physiological re-adjustment to a low-pulsatile or non-pulsatile state with alterations within the microvasculature,18 (c) heightened systemic inflammation,19 (d) differential mechanical interactions between patients and the device,20 and (e) adverse effects of treatment applied for maintaining optimal function, particularly antiplatelet and anticoagulation therapy.7 Any one of these aberrations or a multitude of these could be operative in determining the observed generic outcomes that are crudely bucketed into a notion of haemocompatibility (Table 2). As of now, we do not have clinically validated biomarkers or tests that can sufficiently predict net haemocompatibility. Although beyond the scope of this discussion, the presence or absence of pulsatility, its magnitude and effects on the peripheral circulation cannot be under-emphasized in the consideration of haemocompatibility and its eventual expression.20 Due to the highly variable nature of each of these potential attributes, it should not be assumed that a pump that exhibits improvement in one dimension will necessarily shift all variables in a positive direction. This very recognition is essential to prevent us from being trapped into assumptions that dissuade us from further exploration into the mechanisms that underlie the observed dissociation between these various observed aberrations. Let’s take neurological complications as a vivid representative of the complex weave of haemocompatibility interactions.21 , 22 Three sentinel factors operate in an interwoven manner including (1) pre-existing factors (prothrombotic milieu, pre-implant atrial fibrillation, and prior stroke), (2) pump-related factors intra-pump (gaps between housing and impeller, inner surface texture and peri-pump thrombosis, device type (bearing versus levitation), and (3) peri-pump factors (endothelial and microvascular dysfunction, vascular fragility, infection, blood pressure, alterations in antiplatelet or anticoagulation management). Thus, in every case of a neurological event, there is a fundamental change in haemocompatibility in concert with a dynamic change in management in multiple domains such as blood pressure control, pump flow alteration or anticoagulation therapy. Similarly, a gastrointestinal bleeding episode may be related to acquired von Willebrand disease, development of microvascular disorders such as angiodysplasia, or represent loss of mucosal integrity from oxidative stress. Thus, multiple potential pathological circumstances coalesce to result in the observed clinical outcome. The sheer complexity of the physiological haemocompatibility fabric is illustrated in Figure 1. Potential mechanisms causing physiological dysfunction in the patient-LVAS interface Haemolysis and platelet activation Oxidative stress Increased circulating microparticles Loss of high molecular weight multimers of vWF Elevated angiogenic factors Increased muscle sympathetic nerve activity via a baroreceptor-mediated pathway Increased vascular stiffness Elevated CRP and cytokines Prothrombosis Abnormal loading due to inflow and outflow cannula positions Aortic regurgitation Elevated blood pressure Altered coagulation pathways Altered rheology by therapeutic intent Haemolysis and platelet activation Oxidative stress Increased circulating microparticles Loss of high molecular weight multimers of vWF Elevated angiogenic factors Increased muscle sympathetic nerve activity via a baroreceptor-mediated pathway Increased vascular stiffness Elevated CRP and cytokines Prothrombosis Abnormal loading due to inflow and outflow cannula positions Aortic regurgitation Elevated blood pressure Altered coagulation pathways Altered rheology by therapeutic intent vWF, Von Willebrand Factor; CRP, C Reactive Protein; (Data adapted from references 16–20). Potential mechanisms causing physiological dysfunction in the patient-LVAS interface Haemolysis and platelet activation Oxidative stress Increased circulating microparticles Loss of high molecular weight multimers of vWF Elevated angiogenic factors Increased muscle sympathetic nerve activity via a baroreceptor-mediated pathway Increased vascular stiffness Elevated CRP and cytokines Prothrombosis Abnormal loading due to inflow and outflow cannula positions Aortic regurgitation Elevated blood pressure Altered coagulation pathways Altered rheology by therapeutic intent Haemolysis and platelet activation Oxidative stress Increased circulating microparticles Loss of high molecular weight multimers of vWF Elevated angiogenic factors Increased muscle sympathetic nerve activity via a baroreceptor-mediated pathway Increased vascular stiffness Elevated CRP and cytokines Prothrombosis Abnormal loading due to inflow and outflow cannula positions Aortic regurgitation Elevated blood pressure Altered coagulation pathways Altered rheology by therapeutic intent vWF, Von Willebrand Factor; CRP, C Reactive Protein; (Data adapted from references 16–20). The sheer complexity of the physiological haemocompatibility fabric is illustrated in this figure, wherein multiple potential pathological circumstances coalesce to result in the observed clinical outcome. While we should recognize that a dissociation may exist in the components that underlie haemocompatibility, and efforts to discern opportunities for therapeutic intervention must be constantly engaged, we nevertheless must have a clinically relevant approach to appreciating the burden of haemocompatibility related complications to inform clinical decision making and understanding the totality of the burden of haemocompatibility. A clinical score that includes pump and peri-pump factors should ideally be developed that allows us to determine the total burden of haemocompatibility related adverse effects, to determine the optimal patient and device haemocompatibility interface, particularly as we evaluate outcomes from clinical trial experiences. Such a tiered score could include the totality of haemocompatibility related clinical events, weighted by their clinical relevance. In such a hierarchal system, we could classify three categories of events: (a) Tier I (Mild) – ≤2 bleeding episodes >30 days post implant, gastrointestinal or other bleeding requiring hospitalization (e.g. epistaxis) unrelated to supra-therapeutic anticoagulation, or a suspected pump thrombosis that requires hospitalization (successfully medically treated), or non-stroke related neurological events, or arterial thromboembolism (b) Tier II (Moderate) - >2 bleeding episodes (>30 days post implant, gastrointestinal or other bleeding requiring hospitalization (e.g. epistaxis) unrelated to supra-therapeutic anticoagulation), or a non-disabling stroke and, (c) Tier III Moderately – Severe (A) to Severe (B) – (A) pump malfunction due to pump thrombosis that leads to a reoperation or (B) a disabling stroke or death due to haemocompatibility related causes. Events in the third tier are typically used in adjudicating primary end points of trials with left ventricular assist devices (but also include deaths due to non-haemocompatibility related causes such as operative deaths, right heart failure or sepsis). Since deaths are difficult to classify into discrete categories and represent the worst possible outcome, it may be prudent to include any death that is clearly related to a haemocompatibility event or uncertain in aetiology in scoring a Tier III haemocompatibility event. One could exclude an operative death, a death due to an infection or death from right heart failure in ascertaining this endpoint. The total proportion of these three discrete components could define the net burden of haemocompatibility for a group or for an individual (Table 3). Proposed classification of haemocompatibility related events (Mild, score 1 per event) ≤2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Suspected pump thrombosis episode that requires hospitalization (successfully medically treated) Non-stroke related neurological events arterial thromboembolism not resulting in organ loss Early bleeding should not be considered as a haemocompatibility related event since it is likely due to post-surgical causes and stress related episodes A suspected pump thrombosis episode is one where haemolysis or pump power spikes coupled with or without pump malfunction require in-hospital management but intensified therapy is successful in resolving the event Non-stroke related events include seizures, transient ischemic events or acute cognitive deficits (Moderate, score 2 per event) >2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Non-disabling stroke (haemorrhagic or ischemic) arterial thromboembolism resulting in organ loss > 2 gastrointestinal bleeds require disruption in net antiplatelet and anticoagulant exposure leading to destabilization of the haemostatic axis Non-disabling strokes can be adjudicated by using the modified Rankin score of ≤ 3 – patients can typically function independently with level of disability (A) (Moderately severe, score 3 per event) (B) (Severe, score 4 per event) Pump malfunction due to pump thrombosis leading to reoperation for removal or replacement Disabling stroke Death due to a haemocompatibility aetiology or inconclusive (unknown or multiple causes) Pump malfunctions due to electrical faults must be excluded but this end point should include those cases where rescue transplantation was performed The modified Rankin score is > 3 indicating that assistance is needed for performing activities of daily living It is sometimes difficult to ascertain if a death is due to haemocompatibility complications and ideally one should exclude deaths due to right heart failure and sepsis or those clearly adjudicated to no relate to haemocompatibility; however, if inconclusive or uncertain of what set up the cascade leading to death, we propose scoring against this end point (Mild, score 1 per event) ≤2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Suspected pump thrombosis episode that requires hospitalization (successfully medically treated) Non-stroke related neurological events arterial thromboembolism not resulting in organ loss Early bleeding should not be considered as a haemocompatibility related event since it is likely due to post-surgical causes and stress related episodes A suspected pump thrombosis episode is one where haemolysis or pump power spikes coupled with or without pump malfunction require in-hospital management but intensified therapy is successful in resolving the event Non-stroke related events include seizures, transient ischemic events or acute cognitive deficits (Moderate, score 2 per event) >2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Non-disabling stroke (haemorrhagic or ischemic) arterial thromboembolism resulting in organ loss > 2 gastrointestinal bleeds require disruption in net antiplatelet and anticoagulant exposure leading to destabilization of the haemostatic axis Non-disabling strokes can be adjudicated by using the modified Rankin score of ≤ 3 – patients can typically function independently with level of disability (A) (Moderately severe, score 3 per event) (B) (Severe, score 4 per event) Pump malfunction due to pump thrombosis leading to reoperation for removal or replacement Disabling stroke Death due to a haemocompatibility aetiology or inconclusive (unknown or multiple causes) Pump malfunctions due to electrical faults must be excluded but this end point should include those cases where rescue transplantation was performed The modified Rankin score is > 3 indicating that assistance is needed for performing activities of daily living It is sometimes difficult to ascertain if a death is due to haemocompatibility complications and ideally one should exclude deaths due to right heart failure and sepsis or those clearly adjudicated to no relate to haemocompatibility; however, if inconclusive or uncertain of what set up the cascade leading to death, we propose scoring against this end point Based on the aggregate of Intensity, one can develop the ‘Total Haemocompatibility Score’ by assigning a weight of 1, 2 or 3 (for Tier III A) and 4 (For Tier III B) depending on the type of event. The worst event score should be applied but in a cumulative aggregate manner. Therefore, a patient that has 3 gastrointestinal bleeds (score 2) followed by a non-disabling stroke (score 2) and then a reoperation to replace the device (score 3) would have a score of 7 (based on total events/patient). Proposed classification of haemocompatibility related events (Mild, score 1 per event) ≤2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Suspected pump thrombosis episode that requires hospitalization (successfully medically treated) Non-stroke related neurological events arterial thromboembolism not resulting in organ loss Early bleeding should not be considered as a haemocompatibility related event since it is likely due to post-surgical causes and stress related episodes A suspected pump thrombosis episode is one where haemolysis or pump power spikes coupled with or without pump malfunction require in-hospital management but intensified therapy is successful in resolving the event Non-stroke related events include seizures, transient ischemic events or acute cognitive deficits (Moderate, score 2 per event) >2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Non-disabling stroke (haemorrhagic or ischemic) arterial thromboembolism resulting in organ loss > 2 gastrointestinal bleeds require disruption in net antiplatelet and anticoagulant exposure leading to destabilization of the haemostatic axis Non-disabling strokes can be adjudicated by using the modified Rankin score of ≤ 3 – patients can typically function independently with level of disability (A) (Moderately severe, score 3 per event) (B) (Severe, score 4 per event) Pump malfunction due to pump thrombosis leading to reoperation for removal or replacement Disabling stroke Death due to a haemocompatibility aetiology or inconclusive (unknown or multiple causes) Pump malfunctions due to electrical faults must be excluded but this end point should include those cases where rescue transplantation was performed The modified Rankin score is > 3 indicating that assistance is needed for performing activities of daily living It is sometimes difficult to ascertain if a death is due to haemocompatibility complications and ideally one should exclude deaths due to right heart failure and sepsis or those clearly adjudicated to no relate to haemocompatibility; however, if inconclusive or uncertain of what set up the cascade leading to death, we propose scoring against this end point (Mild, score 1 per event) ≤2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Suspected pump thrombosis episode that requires hospitalization (successfully medically treated) Non-stroke related neurological events arterial thromboembolism not resulting in organ loss Early bleeding should not be considered as a haemocompatibility related event since it is likely due to post-surgical causes and stress related episodes A suspected pump thrombosis episode is one where haemolysis or pump power spikes coupled with or without pump malfunction require in-hospital management but intensified therapy is successful in resolving the event Non-stroke related events include seizures, transient ischemic events or acute cognitive deficits (Moderate, score 2 per event) >2 gastrointestinal or other bleeding episodes (>30 days post implant) requiring hospitalization (e.g. epistaxis unrelated to supra-therapeutic anticoagulation) Non-disabling stroke (haemorrhagic or ischemic) arterial thromboembolism resulting in organ loss > 2 gastrointestinal bleeds require disruption in net antiplatelet and anticoagulant exposure leading to destabilization of the haemostatic axis Non-disabling strokes can be adjudicated by using the modified Rankin score of ≤ 3 – patients can typically function independently with level of disability (A) (Moderately severe, score 3 per event) (B) (Severe, score 4 per event) Pump malfunction due to pump thrombosis leading to reoperation for removal or replacement Disabling stroke Death due to a haemocompatibility aetiology or inconclusive (unknown or multiple causes) Pump malfunctions due to electrical faults must be excluded but this end point should include those cases where rescue transplantation was performed The modified Rankin score is > 3 indicating that assistance is needed for performing activities of daily living It is sometimes difficult to ascertain if a death is due to haemocompatibility complications and ideally one should exclude deaths due to right heart failure and sepsis or those clearly adjudicated to no relate to haemocompatibility; however, if inconclusive or uncertain of what set up the cascade leading to death, we propose scoring against this end point Based on the aggregate of Intensity, one can develop the ‘Total Haemocompatibility Score’ by assigning a weight of 1, 2 or 3 (for Tier III A) and 4 (For Tier III B) depending on the type of event. The worst event score should be applied but in a cumulative aggregate manner. Therefore, a patient that has 3 gastrointestinal bleeds (score 2) followed by a non-disabling stroke (score 2) and then a reoperation to replace the device (score 3) would have a score of 7 (based on total events/patient). It is important to appreciate that a tiered classification system as proposed is imperfect but represents a point for standardized discussion between varied experiences across centres and trials, to appreciate the burden of haemocompatibility related complications. The proposed score should be examined and further refined in ongoing clinical trials. Such a validated score could allow for more meaningful between device comparisons and serve to form the basis for a discussion of informed consent with patients. It would also allow for a standardized language in the field to help us evaluate era effects as devices are introduced into the clinical realm. We must persist in efforts to reduce the net burden of complications by understanding mechanisms of disease that improve clinical outcomes by enhancing durability of the device–patient interface. No single score will adequately address the clinical need completely but serve to provide a more inter-related clinical view of haemocompatibility, which will only be appropriately addressed with what ultimately matters to patients and society – cost of care minimization and improved quality adjusted life years prolonged by left ventricular assist systems in advanced heart failure. Conflict of interest: Dr. Mehra reports consulting activity for St. Jude Medical, Inc., Medtronic, Stealth Bio therapeutics, Johnson and Johnson (Janssen) and Mesoblast. In each case, either clinical trial steering committees or DSMB or trial design consulting was performed.