Continuous-flow left ventricular assist devices (LVADs) are used as bridge-to-transplant, bridge-to-recovery, or destination therapy in patients with end-stage heart failure. Since 2006, more than 17,000 continuous-flow LVADs have been implanted, according to the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) database, and overall outcomes have improved.1
The causes of advanced heart failure vary greatly,2 and how continuous-flow LVAD support affects outcomes has not been thoroughly explored. Some studies have shown that survival times are shorter in patients with ischemic cardiomyopathy (ICM) than in those with nonischemic cardiomyopathy (NICM); patients with ICM tend to be older and have more comorbidities, widespread atherosclerotic disease, multivessel coronary artery disease, and arrhythmias associated with sudden death.3–8 Other studies have produced conflicting results regarding whether the cause of heart failure is an independent predictor of death after continuous-flow LVAD implantation.9–11 Few investigators have explored the differences between the types and the incidence of postimplant adverse events based on the distinctly different causes of heart failure.
Therefore, the effect of heart failure cause on outcomes in patients who receive continuous-flow LVADs as bridge-to-transplant or destination therapy warrants further investigation. In this study, we compared the overall survival and incidence of adverse events after continuous-flow LVAD implantation in patients with ICM versus NICM.
Patients and Methods
We performed a single-center retrospective review of all patients who underwent primary implantation of a continuous-flow LVAD—either the HeartMate II (Thoratec, an Abbott company) or the HeartWare HVAD (Medtronic)—from November 2003 through March 2016. Patients were divided into 2 groups on the basis of a standardized definition of ICM developed for clinical research.12 The criteria for inclusion in the ICM group were a history of myocardial infarction, coronary revascularization (coronary artery bypass grafting or percutaneous coronary intervention), ≥75% stenosis of the left main or proximal left anterior descending coronary artery, or ≥75% stenosis of ≥2 epicardial vessels.12 The NICM group included all other patients. The incidence of ICM in the study population was calculated for the periods from 2003 through 2009 and from 2010 through 2016. The Institutional Review Board of CHI St. Luke's Health–Baylor St. Luke's Medical Center approved the study (Protocol H38751). The requirement for informed consent was waived because the study was retrospective.
From each patient's medical record, we documented demographic information, preoperative characteristics including laboratory INTERMACS profile13 and comorbidities, hemodynamic and echocardiographic data, operative details, postoperative adverse events, and survival outcomes. Severity of valve regurgitation was assessed by color-flow imaging in all cases and quantitatively graded on a 4-point scale (severe = 3, moderate = 2, mild = 1, and none = 0),14 as documented in the medical record. Operative characteristics included device model (HeartMate II or HeartWare HVAD), cardiopulmonary bypass use and time, aortic cross-clamp use and time, and concomitant procedures performed during the index LVAD implantation. Concomitant procedures included coronary artery bypass grafting, patent foramen ovale closure, and valvular repair or replacement.
Outcome variables included the incidence of postoperative adverse events (including events per patient-year) and overall survival rates through 6 years of follow-up after LVAD implantation. Readmission was defined as a return to the hospital within 30 days of discharge from the index admission. Neurologic dysfunction was defined as a new neurologic deficit associated with abnormal neuroimaging findings and was categorized as either ischemic or hemorrhagic. Patients were considered to have gastrointestinal bleeding if they had one or more of the following: guaiac-positive stool, hematemesis, melena, active bleeding at the time of endoscopy or colonoscopy, or blood in the stomach at the time of endoscopy or colonoscopy. Patients were considered to have an infection if they had one or more of the following: a driveline infection that required surgical treatment, a pump infection that required surgical treatment, or bacteremia (confirmed by positive blood cultures from 2 separate sites). Acute kidney injury within 7 days of LVAD implantation was defined according to the RIFLE (risk, injury, failure, loss of kidney function, and end-stage kidney disease) classification as an abrupt doubling of the serum creatinine level or a 50% reduction in the estimated glomerular filtration rate.15
Statistical Analysis
Patient demographics, preoperative and operative characteristics, hemodynamic and echocardiographic data, and postoperative adverse events were compared between the 2 groups in a univariate analysis. Continuous variables were reported as mean ± SD. Tests of normality were performed on each variable. Normally distributed variables were compared by using the Student t test; variables that were not normally distributed were compared by using the Mann-Whitney U test. Categorical variables were reported as number and percentage and were compared by using the Pearson χ2 test (or the Fisher exact test if expected counts were not sufficiently large). We used χ2 analysis to evaluate the difference in the number of patients who had postoperative adverse events.
Overall survival rates in the ICM and NICM groups were compared by using Kaplan-Meier analysis and the log-rank test, with long-term survival defined as 6 years after LVAD implantation. Cox proportional hazards analysis was used to identify independent predictors of mortality. Proportional hazards assumptions were tested by using Schoenfeld residuals. Preoperative and operative variables with P values <0.20 were included in the univariate analysis. Variables with a P value <0.20 in the univariate Cox analysis were included in our final multivariate model. Variables with a variance inflation factor >10 were removed to avoid introducing collinearity. Heart failure cause was forced into our model. P values <0.05 were considered statistically significant. All analyses were performed with SPSS, version 20 (SPSS, an IBM company).
Results
A total of 526 patients underwent primary implantation of a continuous-flow LVAD during the study period, including the HeartMate II in 403 (76.6%) and the HeartWare HVAD in 123 (23.4%). The ICM group included 256 patients (48.7%); the NICM group, 270 (51.3%) (Table I). From 2003 through 2009, 52 patients (43.7%) had ICM; from 2010 through 2016, 204 (50.1%) had ICM (P=0.22).
The mean age of the ICM group was significantly greater than that of the NICM group, and the ICM group included a larger percentage of men, as well as smokers. Before LVAD implant, more of the ICM patients had hypertension, diabetes, peripheral vascular disease, and previous sternotomy. The ICM group also had significantly higher mean levels of blood urea nitrogen and aspartate aminotransferase. The groups were similar in terms of the percentage of patients who were bridged to transplant (P=0.45), preoperative complete blood count, hemodynamic measurements, and distribution of INTERMACS profiles. Operatively, the groups were similar in terms of cardiopulmonary bypass use, cardiopulmonary bypass time, aortic cross-clamp use, and aortic cross-clamp time.
Of the 270 patients in the NICM group, 190 (70.4%) had idiopathic NICM (Table II). Among the remaining patients with NICM, the most common causes were viral myocarditis in 20 (7.4%), familial dilated cardiomyopathy in 16 (5.9%), valvular dysfunction in 12 (4.4%), and doxorubicin toxicity in 9 (3.3%).
The postoperative outcomes in both groups were largely similar. However, 16 patients (5.9%) in the NICM group recovered LV function to the extent that the LVAD could be explanted, compared with 5 patients (2.0%) in the ICM group (P=0.02) (Table III). The ICM group had greater percentages of patients who were bridged to transplant (20.7% vs 17.0%), continued receiving LVAD support (32.8% vs 33.7%), or died (39.1% vs 35.2%).
In our analysis of postoperative adverse events, the 2 groups had similar rates of neurologic dysfunction, infection, acute kidney injury, and 30-day readmission (Table IV). Gastrointestinal bleeding was more frequent in the ICM group (31.3%) than in the NICM group (22.6%) (P=0.03). To better characterize these events, we evaluated the site and cause of bleeding in each event (Table V). Notably, the ICM group had higher rates of bleeding in both the upper (28.1% vs 18.1%; P=0.007) and lower (23.0% vs 13.3%; P=0.004) gastrointestinal tracts. In particular, the ICM group had higher rates of gastrointestinal bleeding in the small intestine (4.7% vs 1.5%; P=0.03) and colon (10.2% vs 5.6%; P=0.049). The ICM group also had higher rates of gastrointestinal bleeding from arteriovenous malformations (20.7% vs 11.9%; P=0.006) and ulcers (16.4% vs 9.3%; P=0.01).
The ICM group also had a higher rate of diverticulosis (5.9% vs 3.3%), but this difference was not statistically significant (P=0.17).
Finally, Kaplan-Meier analysis revealed that overall survival did not differ between the ICM and NICM groups (P=0.24, log-rank) (Fig. 1). Furthermore, multivariate Cox proportional hazards analysis revealed that the cause of heart failure was not an independent predictor of death (P=0.08) (Table VI). Significant independent predictors included older age (P <0.001), total bilirubin level (P=0.001), previous sternotomy (P=0.03), and concomitant procedures during LVAD implantation (P=0.048).
Discussion
In this study, we found that the cause of heart failure did not independently predict long-term survival after continuous-flow LVAD implantation. However, we also found that gastrointestinal bleeding, particularly from arteriovenous malformations and ulcers, occurred significantly more often in the ICM group (P=0.03). Multivariate Cox proportional hazards analysis showed that older age, previous sternotomy, total bilirubin level, and concomitant procedures during LVAD implantation were independent predictors of death. In addition, the mortality rate was higher for patients with ICM (hazard ratio [HR]=1.32) (P=0.08). Larger studies of this difference are warranted.
The overall incidence of ICM in our study population (50.1%) was comparable to that among the 3,511 patients in the Nationwide Inpatient Sample database who had an LVAD implanted in the period from 2011 through 2014 (53.5%).9 Our ICM group was older and included more men when compared with the NICM group. The patients with ICM also had fewer comorbidities (for example, tobacco use, peripheral vascular disease, diabetes mellitus, and hypertension). This finding was expected because these comorbidities are risk factors for the development of ICM.16 A greater percentage of patients with ICM had undergone a previous sternotomy, typically for coronary artery bypass grafting. (In the NICM group, previous sternotomy was typically for valvular repairs and replacements to treat valvular cardiomyopathy). Although viral and familial dilated cardiomyopathy and valvular disease were common causes of heart failure in our NICM cohort, by far the most frequent cause was idiopathic (70.4%). This was not unusual because the cause of NICM remains unknown in many cases17; regardless of the cause of chronic end-stage heart failure, invasive and costly diagnostic studies are usually not warranted in patients who are being considered for LVAD support because the results will not alter their care and management. The rarest causes in our NICM cohort were related to exposure to chemotherapeutic agents (for example, doxorubicin), peripartum cardiomyopathy, alcohol and cocaine use, viral infection or Chagas disease, and congenital disorders (for example, left ventricular noncompaction and congenital malformation).
A significantly greater percentage of NICM patients than ICM patients recovered LV function to the extent that they could have their LVADs explanted (P=0.02). This finding concurs with an analysis by Goldstein and colleagues,18 who concluded that LV recovery is more likely in young patients with NICM. Studies of myocardial response to LVAD therapy have shown that LV unloading may improve cardiomyocyte contractile function and normalize LV geometry and neurohormonal function.19–21 Other studies suggest that reverse remodeling is less likely in patients with ICM, perhaps because of chronic and irreversible scarring after myocardial infarction, which may hamper recovery.18,22,23 Of the NICM causes analyzed in this study, several are acute and may be more reversible and prone to reverse remodeling (although this may apply only to certain causes such as viral and postpartum cardiomyopathy). In contrast, some types of chronic NICM (those with hereditary or valvular causes, for example) may resist reverse remodeling and recovery. Additional studies are warranted to understand fully the optimal conditions for reverse remodeling and the best timing for bridge-to-recovery strategies.
An unexpected and interesting finding was that postoperative gastrointestinal bleeding, particularly from arteriovenous malformations and ulcers, was more common in patients with ICM than in those with NICM. In our study, age at implantation may have contributed to this difference. This is consistent with the suggestion by Kawabori and associates24 that age at implantation is the primary predictor of postoperative gastrointestinal bleeding. Of note, we also saw an increase over time in the proportion of LVAD recipients with ICM—from 43.7% (2003–2009) to 50.1% (2010–2016) (P=0.22).
The association of gastrointestinal bleeding from arteriovenous malformations with nonpulsatile flow and anticoagulation has been established.24–26 One theory is that the aortic stenosis effectively caused by axial flow prevents aortic valve opening and increases the shear stress caused by the pump, leading to acquired von Willebrand disease.27 Other studies suggest that reduced pulsatility reduces intestinal mucosal perfusion, causing ischemia and formation of friable new vessels.28 A recently proposed theory, related to the pathophysiology of arteriovenous malformation–related bleeding in continuous-flow LVAD recipients, is that angiodysplasia may be part of the natural pathophysiology of advanced heart failure. Patel and colleagues29 reported a significantly higher prevalence of angiodysplasia (P=0.009) in patients with advanced heart failure than in a control group. Clinically, our findings suggest that gastrointestinal bleeding may be encountered more frequently in LVAD recipients as the proportion of those who are older and have ICM grows. Consequently, our results warrant wider investigation to identify preventive strategies to reduce the incidence of such bleeding in this patient population.
Limitations
Our study had several limitations. First, it was a nonrandomized, retrospective, single-institution study and therefore inherently limited by its design. Second, the cause of the cardiomyopathy in most of our NICM patients was unknown, which may have confounded our results. Last, our classifications of heart failure causes in this study may not mirror the various classifications that have been used in other studies on this topic.
Conclusions
We investigated postimplant complications and long-term survival in patients with ICM or NICM who underwent continuous-flow LVAD support. The cause of heart failure was not a significant independent predictor of survival after continuous-flow LVAD implantation (HR=1.32; P=0.08), and patients with ICM were more likely to have gastrointestinal bleeding, particularly from arteriovenous malformations and ulcers.