Multiple different models have been developed to predict operative risk. The value of these models was recently reviewed by the United States Veterans Administration Quality Enhancement Research Initiative’s (QUERI) Evidence-based Synthesis Program (ESP) [18]. None of these models was felt to be truly reliable. Nevertheless, there are some commonalities that seem to cross all of the models. As such, some of the predictive models are listed in abridged form below. Note that these models address global risk and are not specifically addressing the risk of right heart failure which is discussed separately later in this chapter. One of the earliest risk models (using HeartMate VE data) was presented in the mid-1990s by Dr. Mehmet Oz [19] and suggested that the outcome was worse if:

Over the years, the risk scoring algorithm was modified by Rao [20] with some points being added and some being removed. For a point score >5, the perioperative mortality was 46 %, whereas if the point score was ≤5, the perioperative mortality was only 12 % (p < 0.001).

Historically, the most widely used score had been the Leitz–Miller score derived from a cohort of 222 HeartMate XVE recipients, all of whom were included in the Thoratec DT registry [21]. Nine preoperative risk factors were identified which predicted 90-day mortality and 1-year survival:

Low-risk patients were those whose point total was ≤16 (69 % 1-year survival), whereas high-risk patients were those whose point total was >16 (13 % 1-year survival).

In 2009, the Hopkins group reported in the Annals of Thoracic Surgery a comparison of five risk scoring algorithms [22], three of which are described above. The last two are not LVAD-specific risk assessment algorithms whatsoever—the APACHE II score¹ and the Seattle Heart Failure Model.² Surprisingly, the best predictor of postoperative outcome was the SHFM.

In 2013, Cowgers proposed an updated HeartMate II Risk Score based upon an analysis of 1,122 patients enrolled in the HeartMate II BTT and DT trials between the years 2005 and 2010 [23]. This was the first large-scale analysis evaluating risk factors in patients with axial flow rather than pulsatile-flow devices. Only five variables (based upon multivariable analysis) were shown to be predictive of outcome—older age, hypoalbuminemia, renal dysfunction, coagulopathy, and implantation at a less experienced center.

One of the most important challenges to face the transplant physician and surgeon regardless of whether the LVAD is being implanted for a “bridge” or “destination” indication is right heart failure. As the number and variety of LVADs increases, this issue will become increasingly problematic. In theory, one would anticipate that this problem would occur more frequently in non-ischemic cardiomyopathy patients with biventricular dysfunction than in ischemic patients with primarily left-sided dysfunction. To date, no tested/validated algorithms capable of consistently predicting the likelihood of RV dysfunction exist. Nonetheless, some algorithms have been proposed including an RV failure score published in early 2008 by Aaronson based upon an analysis of 197 implants performed at the University of Michigan, 68 of which were complicated by RV failure [24]. In this analysis, RV failure was defined as the need for postoperative intravenous inotropic support for >14 days, inhaled nitric oxide for >48 h, the need for right-sided circulatory support, or hospital discharge on an inotrope. Predictors of right heart failure based upon a multivariable logistic regression model were defined as:

Interestingly, the RAP and transpulmonary gradient were not independent predictors of RV failure in this study. At present, this algorithm is not widely used.

A number of other analyses, also published between 2008 and 2010, focused on similar variables [25, 26]. Unfortunately, none of these risk prediction models/scores has achieved widespread acceptance probably because their predictive accuracy remains only fair, at best. The additional models are summarized below:

In 2010, Kormos reviewed the outcomes of 484 patients implanted with a HeartMate II during the BTT trial [27]. Right ventricular failure was defined as the need for a right ventricular assist device (RVAD), the need for inotropic support for 14 or more days after implantation, and/or the need for inotropic support starting more than 14 days after implantation. Six percent of patients needed an RVAD, 7 % of patients needed nitric oxide, and 7 % of patients required late inotropic support. Multivariate analysis revealed that each of the following three variables predicted a poor outcome, with the least predictive being renal dysfunction.

A recent study done by Korabathina evaluated a newer hemodynamic index—the pulmonary artery pulsatility index (PaPi). Although PaPi is intended to identify patients with inferior wall myocardial infarctions at risk of developing right ventricular dysfunction it may have a role in the world of LVADs. The index is calculated as the pulmonary artery pulse pressure/right atrial pressure [28, 29].

One of the most recently described approaches to predicting risk of RV failure after LVAD implantation is to measure RV free wall peak longitudinal strain [30] or RV global longitudinal myocardial strain using velocity vector imaging [31, 32]. This is a new technique that may eventually become more widely used as the software becomes more widely available.

Preoperative hemodynamic, hepatic, and renal parameters do not always tell the whole story. Even with fairly acceptable parameters, some patients are still just too debilitated or “frail” to tolerate the stresses associated with implantation of an LVAD. How to identify those patients has been problematic. A similar problem faced the designers of the Medtronic CoreValve™ study in which the degree of frailty needed to be calculated to help define surgical risk [33, 34]. Ultimately, the CoreValve™ investigators adopted a frailty index to help quantitate that which is sometimes intangible. The frailty index was not intended to actually define who would or would not do well. It simply proposed some additional criteria to evaluate when considering a patient’s candidacy for a major operation such as an aortic valve replacement and may be applicable in one way or another to LVAD placement as well, although this has yet to be studied or proven. Nevertheless, frailty may be an important predictor of outcome postimplantation [35].

In addition to the various hemodynamic, hepatic, hematologic, and renal parameters touched upon above, a multitude of concurrent cardiac problems other than right heart failure can also complicate the preoperative situation. These include mitral, tricuspid, and aortic valve disease, coronary artery disease, and intracardiac shunts. All of these problems need to be recognized and potentially addressed at the time of device placement.

If at all possible, patients should not be taken to the operating room for device implantation with an elevated INR secondary to chronic warfarin therapy. Ideally, warfarin should be stopped for a sufficient period of time to permit the INR to normalize [11]. If anticoagulation is needed due to an underlying medical problem, the patient should be converted as soon as practical to an intravenous agent such as unfractionated heparin or in the presence of heparin-induced thrombocytopenia to a direct thrombin inhibitor. In those instances where device placement must be performed urgently, the use of preoperative vitamin K and fresh frozen plasma (FFP) must be considered. Even so, the risk of bleeding is significant. Intraoperative and postoperative FFP, cryoprecipitate, and platelets are likely to be needed. The safety and efficacy of recombinant factor VIIa in this situation has not been established [36].

Many of the LVAD recipients are undergoing device placement as treatment for chronic end-stage ischemic heart disease. As such, the probability is high that the patient will be on an antiplatelet agent such as aspirin, clopidogrel, or perhaps one of the newer agents such as prasugrel, rivaroxaban, or apixaban. As with warfarin, stopping the agents would be ideal. Realistically, however, this is not always advisable. Measuring platelet function by thromboelastography might provide some information as to the degree of platelet dysfunction but from a practical point of view will not change the plan. Adequate platelets, preferably single donor, should be available.

One hematologic issue that needs to be recognized and addressed prior to device implantation yet is often overlooked is the presence of a hypercoagulable state [37]. The most common cause is probably heparin-induced thrombocytopenia, but other causes such as anticardiolipin antibody syndrome, factor V Leiden mutation, proteins S and C deficiency, methylenetetrahydrofolate reductase mutation, and prothrombin 20210 gene mutation are not infrequently seen. Each of these antibodies, deficiencies, and/or mutations can increase the risk of thrombosis, not an insignificant concern in the presence of a mechanical device. Referral to a hematologist for specific treatment recommendations is appropriate [38].

Reducing the risk of right heart dysfunction is critical to ensuring a positive clinical outcome [39]. INTERMACS data clearly demonstrates more serious events and worse outcomes in patients requiring biventricular support [40, 41]. Multiple preoperative and intraoperative tricks and techniques have been proposed, some based on randomized trials and others based solely upon expert opinion and anecdotes [27, 42–48]. Almost all experts agree that the explanation for RV dysfunction is multifactorial and includes intrinsic contractility issues, geometric issues related to septal interdependence, and afterload. Each of these causes needs to be addressed.

Inotropic support in one form or another is commonly used during the early postoperative period [24]. Some programs prefer pure inotropes or vasodilators such as dobutamine and milrinone while others due to concerns over excess vasodilation and hypotension opt for beta-agonists such as isoproterenol in combination with epinephrine and a pressor such as dopamine or vasopressin. Anecdotal experience suggests that intravenous thyroid hormone (liothyronine) may be of some benefit in persistent right heart dysfunction [49, 50]. This has never been proven and is an off-label indication.

As noted, afterload in the form of an increased pulmonary vascular resistance may play a significant role in the development of post-LVAD RV dysfunction [51]. For this reason, hemodynamic measurements are routinely assessed in all patients being considered for LVAD implantation [52]. The presence of pulmonary hypertension, in and of itself, is generally not of significant concern. In fact, the ability to generate a systolic pulmonary pressure in excess of 35–40 mmHg may actually portend a good prognosis, especially in the presence of a low right atrial pressure. An elevated transpulmonary gradient or pulmonary vascular resistance due to long-standing heart failure or long-standing mitral regurgitation, for example, is more concerning. How to address this issue varies depending upon the etiology. In some cases inotropes and/or diuretics are needed. In other cases, preoperative pulmonary vasodilator therapy might be indicated.

Regardless of a patient’s baseline RV afterload, the hemodynamic situation facing the right ventricle in the operating room during and after initiation of LVAD support is greatly altered. Blood transfusions, intravenous pressors, increased RV venous return, epicardial ventricular pacing, and changes in RV shape and septal function all create an environment favoring RV dysfunction. For this reason, many centers (although not all) introduce inhaled nitric oxide (NO) or inhaled prostaglandins prior to or shortly after initiating LVAD support [53–57]. Both interventions have been shown to lower pulmonary artery pressure and pulmonary vascular resistance (PVR). No large-scale randomized trials have been performed. In a fairly recent randomized trial, however, it was demonstrated that the use of NO at 40 ppm in the perioperative period did not achieve significance for the endpoint of reduction of RV dysfunction although it did reduce time on mechanical ventilation, length of hospital or intensive care unit stay, and the need for RVAD support after LVAD placement [53]. A third option available to address postoperative pulmonary hypertension is phosphodiesterase type 5A inhibition which has also been shown to result in a significant decrease in PVR when compared with control patients [58].

It has been suggested that the risk of early postoperative RV failure after placement of an axial flow pump can be reduced by maintaining a relatively low RPM [42, 59]. No such suggestion has yet been made for centrifugal pumps. (The flatter HQ curve seen with centrifugal devices means that LV unloading may not change as much with small changes in RPM.) In theory, at least with axial pumps, small decreases in RPM mean less unloading of the LV which should help to maintain normal RV shape by minimizing septal shift. While there may be some truth to this, the benefit of low RPMs with respect to RV function must be balanced against the need for forward flow and the risk of pump thrombosis which may be increased if one runs the pump at a slower speed.

If the patient does develop RV failure, treatment options depend upon the timing and reason for the dysfunction [46, 60, 61]. For intraoperative problems, one must rule out unexpected technical challenges and errors. These include malposition of the apical conduit, aortic insufficiency, and twisted outflow grafts, all of which may result in inadequate LV unloading and increased RV afterload. Alternatively, perhaps the problem is simply a cold or ischemic RV. It is for the former reason that most surgeons now implant LVADs during normothermia [62]. It is for the latter reason that patients undergoing LVAD placement also undergo an assessment of coronary anatomy. If indicated, the RCA may be bypassed. Other causes of increased RV afterload include an elevated baseline PVR or an acutely elevated PVR due to bleeding and the subsequent administration of blood products.

Intraoperative surgical techniques adopted by many centers to reduce the risk of RV failure include maintaining normothermia, avoiding cardioplegia, intraoperative hemoconcentration (2–4 L), ensuring a balanced intraventricular septum by TEE, and avoiding air emboli. It should be self-evident that bleeding must be minimized. Clearly, this last recommendation can be challenging. As noted previously, patients should generally not be on antiplatelet agents such as aspirin, clopidogrel, prasugrel, rivaroxaban, and apixaban or anticoagulants such as warfarin, ticagrelor, and dabigatran in the immediate preoperative period.

How one weans from cardiopulmonary bypass (CPB) is critical [63]. The RV cannot be allowed to distend. The mean systemic blood pressure should be maintained at least at 65–70 mmHg. The CVP should probably be in the range of 13–17 mmHg although each case needs to be individually evaluated. Lower CVPs are not necessarily better as they may reflect an inadequate RV preload. Higher CVPs suggest RV failure. LVAD flow should be initiated while still on CPB. RPMs (for the HeartMate II device) should start at approximately 6,500 and then be increased over the next 10–15 min as CPB is weaned off. Some centers leave the operating room with an RPM of 8,000 while others prefer to leave the operating room with a higher RPM. Almost certainly, there is no single correct recommendation. RPMs should be adjusted to the clinical situation and guided by intraoperative TEE or early postoperative ImaCor hTEE™ assessment of septal position, RV size and function, and LV size. Regardless of the initial setting, within 3–6 h of arrival in the CTICU, the RPMs should be increased to at least 8,600 to ensure that forward flow through the device is adequate.

One thing is clear—disagreement exists among highly respected CT surgeons as to whether a higher RPM is better or worse when it comes to RV function. In those instances in which RV failure does develop intraoperatively, it may be possible to avoid the need for an RVAD by providing temporary RV support by placing an arterial cannula from the CPB circuit into the PA and perfusing the pulmonary circuit for an hour or so. Weaning should then be reattempted prior to committing the patient to a hybrid-type RVAD (CentriMag™ or TandemHeart™) [64].

Heparin (or a derivative) is often administered preoperatively to minimize the risk of intraventricular thrombus formation. Regardless of whether heparin was infusing preoperatively or not, it is normally administered in fairly high dose during LVAD implantation. Once the LVAD has been successfully implanted and the flow generated by the device shown to be adequate, protamine is administered to reverse the effects of heparin. The role of postoperative heparin for the HeartMate II device is less clear and seems to be institution specific [65, 66]. Whereas in the past heparin was routinely introduced once early bleeding was brought under control, more recent guidelines recommend heparin postoperatively when low-flow conditions exist or when otherwise medically indicated [67]. As a result of the guideline modifications, many centers now proceed directly to anticoagulation with warfarin. Some centers, however, still bridge patients in the early postoperative period with heparin electing to introduce warfarin between days 3 and 5 postoperatively. Occasionally, bivalirudin or argatroban is used in lieu of heparin in patients with heparin-induced thrombocytopenia, but this is not recommended by the manufacturer [68–73]. There is no experience with the use of other novel oral anticoagulants.

Ideally, anticoagulation should not be discontinued in LVAD recipients. However, there are more than a few case reports and anecdotes describing situations in patients with HeartMate II devices in which anticoagulation was discontinued of necessity and in some cases permanently, especially after major gastrointestinal bleeding [74–76]. Most of the time, at least with respect to the HeartMate II device, few untoward effects were noted. Nevertheless, discontinuing warfarin is not recommended in the absence of a strong and compelling reason.

In those instances where oral anticoagulation must be discontinued, the risk needs to be clearly reviewed with the patient. Some data is available to help guide that discussion. Boyle, in 2013, reviewed the bleeding and thrombosis rates in 956 HM II patients, all of whom were on anticoagulation. Although the data has yet to be formally published, the analysis demonstrated that female gender increased the risk for both ischemic and hemorrhagic strokes, gastrointestinal bleeding, and pump thrombosis. Perhaps unexpectedly, age ≤65 years was a risk factor for hemorrhagic stroke. Diabetes was a risk factor for ischemic stroke. Boyle concluded that men >65 years might be more able to tolerate a lower INR than other patient cohorts. Naturally, this data applies only to the HM II device and further confirmation is needed [77].

It is critical to realize that not all pumps are alike and that anticoagulation protocols are not interchangeable. For example, the manufacturer of the HeartMate II recommends maintaining an INR of 2.0 ± 0.5 [78], and the manufacturer of the HeartWare HVAD pump recommends maintaining an INR in the range of 2.0–3.0 but probably closer to 2.3–2.7. Likewise, the dose of aspirin is not necessarily identical either. HeartMate II recipients are not infrequently maintained on 81 mg per day, but HeartWare HVAD recipients may well need a slightly higher dose of 325 mg per day as there may be a higher pump thrombosis rate when lower doses of aspirin are used [76].

Not infrequently, patients with end-stage left ventricular dysfunction present with valvular heart disease either secondary to the LV dysfunction or as a cause of the LV dysfunction. Even mild aortic insufficiency can become a serious problem over time, especially with non-pulsatile devices in which the valve often does not open and thus faces a constant pressure of 80 mmHg or greater resulting in pansystolic regurgitation. Blood ejected through the LVAD will leak backwards through the incompetent aortic valve into the left ventricle and then into the LVAD. Pump output will be excellent; however, net forward flow may be moderately to severely compromised. Placement of an aortic valve prosthesis may be required but mechanical prostheses tend to clot [79, 80]. The outcomes with bioprostheses have been marginally better. Over-sewing the aortic valve (recognizing that pump thrombosis will likely be a fatal event should it occur) has also been tried with varying success [81–83].

In contrast to aortic insufficiency, mild to no more than moderate mitral insufficiency, in the presence of an unloaded left ventricle, is generally not a problem [84]. Moderate to severe or severe mitral insufficiency may require a concurrent annuloplasty. Different institutions handle this problem differently and no consensus exists on what to do with 3+ or 4+ mitral regurgitation [85].

Tricuspid regurgitation of more than a mild degree may worsen right ventricular function and indirectly compromise left-sided filling [86]. If feasible, consideration should be given to placement of a tricuspid annuloplasty ring or suture repair as it has been shown that at least in patients with significant tricuspid regurgitation concomitant tricuspid procedures are associated with improved early clinical outcomes [87].