Estimated actuarial survival (continuous-flow LVAD therapy vs. pulsatile-flow LVAD therapy ). Modified with permission from Hindawi Publishing Corporation 
Comparison of adverse event rates of continuous-flow LVADs and pulsatile-flow LVADs . *Significant difference (p < 0.05) between continuous-flow adverse event rate and pulsatile-flow adverse event rate. Modified with permission from Hindawi Publishing Corporation 
Mechanical cardiac assist devices have become significantly more reliable, but still have problems. The most common complications of long-term VAD implantation include pump thrombosis, mechanical failure necessitating device replacement, stroke, LVAD-related infection, sepsis, bleeding requiring blood transfusion, cardiac arrhythmia, renal failure, and aortic valve insufficiency. Each of these issues will be addressed in the following sections. Complications after the implantation of continuous-flow devices will be emphasized because those devices accounted for more than 90% of pumps implanted in 2015.
Pump Replacement and Thrombosis
The greater durability of continuous-flow LVADs led to an increase in the use of device implantation to treat heart failure. In the current era, more LVADs are implanted than hearts are transplanted. According to the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS ) database, between 2012 and 2014, the number of continuous-flow pumps implanted as a bridge to transplantation increased from 404 to 734, and the number implanted as destination therapy increased from 983 to 1108 .
Rates of mechanical failure markedly decreased in the continuous-flow era. In the pivotal study by Slaughter et al.  in 2009 that compared pulsatile-flow and continuous-flow devices , pump replacement for malfunction (including events such as a driveline fracture) was necessary for 9% of implanted continuous-flow pumps and 34% of implanted pulsatile-flow pumps. There were no primary pump or bearing failures in patients with a continuous-flow LVAD. However, pump thrombosis occurred in 4% of patients with continuous-flow pumps but in none of the patients with pulsatile-flow devices .
The findings of other studies have shown the reliability of continuous-flow devices and freedom from replacement. In one retrospective study , replacement of the device for failure or thrombosis was required for only 3.8% of patients. However, studies of earlier continuous-flow devices, such as the Jarvik 2000 and the DuraHeart, have documented device replacement rates as high as 14% . In the latter series, thrombosis was the most common indication for replacement (66%), followed by driveline infection (10%), and other problems (22%).
When VAD replacement is required, outcomes are not compromised. Of 469 patients who underwent 546 continuous-flow LVAD implantations from December 1999 to December 2013, 14% required the exchange of one continuous-flow LVAD for another . Survival was not significantly different between the exchange and non-exchange groups (Fig. 17.3).
Kaplan-Meier analysis of survival (VAD exchange vs. no VAD exchange). Survival from time of primary VAD implantation to time of most recent follow-up. Modified with permission from Elsevier 
In 2013, two separate reports documented an increase in the incidence of HeartMate II pump thrombosis that seemed to have begun in 2011 [11, 12]. These reports estimated the risk of pump thrombosis at 6 months to be 6–12%, a substantially greater risk than was previously reported. Another report indicated that the HeartWare device may also have a higher-than-expected pump thrombosis rate of 8% . Many of the events occurred within 6 months of implantation. Although these rising rates of thrombosis were not initially associated with a decrease in survival, they were associated with substantial morbidity, the need for pump exchange, and increased healthcare costs.
Suggested explanations for the increase in pump thrombosis rates included mechanical defects in the devices and suboptimal inflow cannula geometry , but these theories were not independently supported. Other possible contributing factors included clinical management issues, such as the use of lower levels of anticoagulation to minimize gastrointestinal (GI) bleeding and the reduced use of heparin bridging after implantation to minimize postoperative bleeding. The lowering of pump speeds to facilitate aortic valve opening and minimize aortic valve commissural fusion may have also been a possible contributing factor .
After these reports, management strategies were changed to emphasize heparin bridging after device implantation and to increase postoperative anticoagulation. Despite the increased awareness of the problem, actuarial freedom from pump thrombosis decreased progressively from 2009 to 2013. Freedom from pump thrombosis at 6 months fell from 98% in 2010 to 92% in 2013 . In 2014, freedom from pump thrombosis improved to a level approaching that of 2011. Subsequent analyses of data provided by the INTERMACS database suggested that pump thrombosis rates remain elevated when compared to those reported before 2011 . In this INTERMACS analysis , the most important predictors of pump thrombosis and pump exchange were age at the time of implantation and body mass index. Patients younger than age 72 and those with a body mass index greater than 25 kg/m2 were at increased risk. In this analysis, pump thrombosis was associated with an elevated mortality risk of 18% at 1 month after the first thrombosis and 37% at 1 year after the first thrombosis.
To prevent pump thrombosis, methods of early detection and intervention have been adopted, including the monitoring of lactate dehydrogenase (LDH) levels as a measure of hemolysis and possible thrombosis. In addition, nonoperative management with anticoagulant and thrombolytic therapy has been introduced, with success rates approaching 80% in some series with centrifugal continuous-flow pumps . Nonoperative therapeutic strategies have included the use of heparin, direct thrombin inhibitors , and platelet glycoprotein IIb/IIIa inhibitors , as well as local and systemic thrombolytic therapy with tissue plasminogen activator (tPa).
Compared with pulsatile-flow pumps , continuous-flow pumps offer greater durability and have been a clear step forward in improving LVAD outcomes. Although mechanical failure is much less common in patients with continuous-flow pumps, pump thrombosis occurs in 4–10% of these patients, posing substantial challenges. Appropriate attention to anticoagulation protocols is important. In many cases, monitoring LDH levels and early intervention (beginning with augmented anticoagulation and thrombolytic therapy) can successfully prevent pump thrombosis. If pump replacement is necessary, acceptable results and survival can be anticipated .
Stroke and Thromboembolic Events
Cerebrovascular events are a substantial cause of morbidity and mortality after implantation of either a continuous-flow or a pulsatile-flow VAD. In 2009, a randomized trial in which continuous-flow and pulsatile-flow LVADs were compared showed that the incidence of stroke was 17% in the continuous-flow group and 14% in the pulsatile-flow group. However, the number of strokes per patient year was only 0.13 in the continuous-flow group and 0.22 in the pulsatile-flow group. This incidence of stroke in the continuous-flow group was similar to that in patients with end-stage heart failure who do not have mechanical support .
Other studies have documented similar stroke rates after continuous-flow LVAD implantation . A review of 150 patients who underwent continuous-flow LVAD implantation revealed a stroke rate of 18% . The anticoagulation protocol in this study included aspirin (81 mg/day) and warfarin (target international normalized ratio [INR] of 2.0–2.5). In the 32 patients who died, stroke was the second most common cause of death (n = 8). Six patients had hemorrhagic strokes, and two had embolic strokes.
Risks factors for stroke include high blood pressure, infection, pump thrombosis, GI bleeding, aortic cross-clamping with cardioplegic arrest during implantation, and insufficient or excessive anticoagulation.
A review of 100 consecutive continuous-flow pump implantation cases revealed a stroke rate of 12% . Patients who had a stroke had a significantly higher prevalence of diabetes (66% vs. 41%), previous stroke (17% vs. 5%), and use of aortic cross-clamping with cardioplegic arrest during LVAD implantation (50% vs. 20%) than did patients without stroke (Figs. 17.4 and 17.5). Notably, the mean INR at the time of stroke was subtherapeutic in all four patients who had embolic strokes. Mortality within 30 days of stroke was 25%. A University of Minnesota review of 230 patients in whom a HeartMate II continuous-flow LVAD was implanted as a bridge to transplant revealed a stroke rate of 17%. Of those stroke cases, 49% were embolic and 52% were hemorrhagic. Diabetes and hypertension were not risk factors for stroke in this study, but prior cardiac surgery and infection were associated with a higher risk of stroke. Stroke compromises survival [18–20]: 12 months after implantation, survival in the stroke group was significantly lower (71%) than that in the non-stroke group (82%) (Fig. 17.6) .
Freedom from stroke (clamped or arrested vs. unclamped). Modified with permission from Wolters Kluwer 
Freedom from stroke (patients with diabetes vs. patients without diabetes). Modified with permission from Wolters Kluwer 
In a Columbia University Medical Center review of data from 301 patients who underwent implantation with continuous-flow LVADs from 2008 to 2015, strokes occurred in 40 (13%). The study emphasized appropriate characterization of the type of stroke. A clear distinction was made between ischemic stroke and primary intracerebral hemorrhagic stroke, which probably have different etiologies. Unlike patients in other studies, these patients were classified as having ischemic stroke, ischemic stroke with hemorrhagic conversion, or intracerebral hemorrhagic stroke after the careful review of the clinical presentation and radiologic findings. In this study, ischemic stroke—presumably caused by embolic disease originating from the LVAD—was the most frequent cause of death, occurring in 32 of the 40 patients who had a stroke; the remaining 8 patients had an intracerebral hemorrhage. On the basis of the Columbia protocol, continuous-flow LVAD patients were maintained on a regimen of aspirin (81 mg/day) and warfarin (target INR of 2.0–2.5). However, a substantial proportion of stroke patients had an INR at the time of stroke that was either subtherapeutic or above the target range. For several patients, warfarin had been discontinued for various reasons. In-hospital mortality was 50% after intracerebral hemorrhagic stroke and 28% after ischemic stroke. Survival correlated with the severity of the stroke, which was assessed clinically by using the National Institutes of Health Stroke Scale. Ischemic stroke patients often recovered sufficiently to proceed to transplant or discharge .
Identifying optimal anticoagulation strategies continues to be an important area of investigation. Studies by John et al.  and Katz et al.  showed that reduced amounts of anticoagulation may be acceptable. In the study by John et al. , 45 patients who received the HeartMate II were anticoagulated with aspirin and warfarin, but 41 of those patients had a mean INR less than 2.0. Among the 21 patients who had a mean INR less than 1.6, only one stroke occurred . In the intriguing study by Katz et al. , 94% of patients receiving reduced anticoagulation (i.e., only aspirin, only warfarin, or no agents at all) because of bleeding complications were free from ischemic stroke at 1 year .
Early thrombus formation after LVAD implantation is likely a risk factor for stroke. To minimize this risk, appropriate attention to anticoagulation with heparin is important in the early postoperative period. With prolonged circulatory support , the risk of stroke may be increased by comorbidities such as atrial fibrillation, appropriate aortic valve closure, and the anatomic position of the inflow cannula, device, and outflow graft. Stroke is a multifactorial problem that develops in approximately 10–20% of patients with continuous-flow LVADs. Hemorrhagic stroke is typically more lethal than ischemic stroke.