Author + information
- Received July 30, 2014
- Revision received October 31, 2014
- Accepted November 14, 2014
- Published online April 1, 2015.
- Daniel J. Goldstein, MD∗∗ (, )
- Keith D. Aaronson, MD†,
- Antone J. Tatooles, MD‡,
- Scott C. Silvestry, MD§,
- Valluvan Jeevanandam, MD‖,
- Robert Gordon, MD¶,
- David R. Hathaway, MD#,
- Kevin B. Najarian, MS#,
- Mark S. Slaughter, MD∗∗,
- ADVANCE Investigators
- ∗Cardiovascular and Thoracic Surgery, Montefiore Medical Center, Bronx, New York
- †Division of Cardiovascular Medicine, University of Michigan, Ann Arbor, Michigan
- ‡Division of Cardiovascular Surgery, Christ Advocate Medical Center, Oak Lawn, Illinois
- §Division of Cardiothoracic Surgery, Washington University Hospital, St. Louis, Missouri
- ‖Department of Cardiac and Thoracic Surgery, University of Chicago, Chicago, Illinois
- ¶Division of Cardiology, Department of Medicine, Northwestern Memorial Hospital, Chicago, Illinois
- #Clinical Affairs and Biostatistics, HeartWare Inc., Boston, Massachusetts
- ∗∗Thoracic and Cardiovascular Surgery Division, Department of Surgery, University of Louisville, Louisville, Kentucky
- ↵∗Reprint requests and correspondence:
Dr. Daniel J. Goldstein, Montefiore Medical Center, 3400 Bainbridge Avenue, MAP 5, Bronx, New York 10367.
Objectives This study evaluated gastrointestinal bleeding (GIB) in patients receiving the HeartWare HVAD System (HeartWare Inc., Framingham, Massachusetts) in the pivotal BTT (Bridge to Transplant) trial and under the continued access protocol (CAP).
Background GIB has become a significant problem for recipients of continuous flow device left ventricular assist devices (CF-LVAD). The need for anticoagulation and antiplatelet therapies complicates the management of GIB.
Methods Bleeding events from 382 patients with advanced heart failure (140 patients enrolled in the BTT trial, and an additional 242 CAP patients) were analyzed. Post-implant anticoagulation consisted of heparin followed by warfarin at a target international normalized ratio of 2 to 3. Acetylsalicylic acid was recommended at 81 to 325 mg.
Results Overall, 59 of 382 (15.4%) patients experienced 108 GIB events (0.27 events per patient year). Mean time to first bleed was 273.1 days and 86.1% of events occurred beyond 30 days. Freedom from GIB was 84.1% at 1 year. Median international normalized ratio at the time of first bleed was 2.4 ± 1.4. The most common etiology of bleeding identified was arteriovenous malformation and the most common site was the small intestine. Repeat bleeding was infrequent, though GIB patients required more readmissions and developed nondevice infections more frequently. No patients required surgical intervention and no deaths directly related to GIB occurred.
Conclusions Recipients of the HeartWare Ventricular Assist Device System had an incidence of 0.27 GIB/patient year with a freedom from GIB of 84.1% at 1 year. All patients with GIB events were managed with medical and endoscopic therapies, although 31% of patients experienced a recurrence of GIB. No surgical intervention was required. GIB did not impact survival. (Evaluation of the HeartWare Left Ventricular Assist Device for the Treatment of Advanced Heart Failure [ADVANCE]; NCT00751972)
Remarkable improvement in survival and quality of life outcomes has been realized by the introduction of continuous flow left ventricular assist devices (CF-LVAD) for the management of advanced heart failure (1–3). While the benefits associated with enhanced reliability, driveline size, elimination of noise, and better battery duration are undisputed, the use of this technology has led to vexing clinical challenges occurring more frequently than with previous pulsatile technologies. Among these are hemolysis and pump thrombosis (4), de novo aortic insufficiency (5) and gastrointestinal bleeding (GIB) (6), with GIB singled out as the most common cause of readmission (7). The reported incidence in institutional series varies widely, from 5% to 30% (6,8–16). Because only bleeding meeting pre-specified criteria such as those requiring transfusions of ≥4 U within 7 days post-implant or those requiring reoperation in the landmark clinical trials of the HeartMate II (Thoratec Corporation, Pleasanton, California) (1,2) or HeartWare (3) devices (HeartWare Inc., Framingham, Massachusetts) were reported in earlier publications, we now report here the incidence and rate of GIB in the multicenter BTT (Bridge to Transplant) clinical trial of the HeartWare Ventricular Assist (HVAD) System.
The pathogenesis of GIB associated with LVAD implantation is not fully understood, although putative mechanisms include: 1) need for combined antiplatelet and antithrombotic therapy; 2) development of acquired von Willebrand syndrome with loss of high molecular weight multimers (17); and 3) reduced pulse pressure in the setting of high shear stress—akin to aortic stenosis in Heyde’s syndrome (18).
The HVAD System (HeartWare Inc.) is a miniaturized, implantable, continuous-flow blood pump. Two large pivotal trials have documented excellent clinical outcomes for recipients of this technology and the pump received Food and Drug Administration approval for the BTT trial indication (3,19). Comparative destination therapy clinical trials are ongoing.
The purpose of this investigation was to fully characterize GIB events in the entire cohort of 382 patients who were implanted with the HeartWare HVAD system as BTT who were part of the pivotal and continued access protocol (CAP) trials.
The study design of the HVAD BTT trial and the associated CAP has been described previously (3,20). In brief, the BTT trial was a prospective, 30-center clinical trial conducted in the United States that evaluated the HVAD as BTT trial therapy. The trial enrolled 140 patients with advanced heart failure who were eligible for heart transplant. Patients were compared with a contemporaneous group of patients enrolled in the INTERMACS registry (The Interagency Registry for Mechanically Assisted Circulatory Support) who received a commercially available LVAD as BTT therapy. Patients were followed until cardiac transplantation, device explant for recovery, death, or for at least 180 days after implantation, and follow-up would continue through 5 years after implantation. The BTT trial achieved its primary success endpoint, defined as non-inferiority to the INTERMACS control in terms of survival to 180 days on the original device or transplant or explant for recovery, with rates of 91% for the HVAD system and 90% for the INTERMACS registry (noninferiority p < 0.0001). Survival at 180 days was 94% for those with an HVAD and 90% for those in the control group. The Food and Drug Administration–approved CAP enrolled an additional 256 patients after completion of enrollment in the BTT trial. We present here an assessment of GIB events in the 382 patients (140 from the BTT trial and 242 from the CAP) implanted through November 2012 and followed through database lock in July 2013.
GIB was defined as any clinically suspected or documented suspicion of bleeding from the GI tract as indicated by a new drop in hemoglobin and/or the appearance of melena, hematochezia, hematemesis, or guaiac positive stools. All GIB events were recorded and characterized and are included in this analysis.
Anticoagulation was individualized and differed among centers. After device implantation, patients received bridging intravenous heparin. As patients became able to tolerate oral medications they transitioned from heparin to warfarin and acetylsalicylic acid (ASA), with a recommended target international normalized ratio (INR) of 2 to 3 and a recommended ASA dose of 81 to 325 mg.
The studies were conducted in compliance with Food and Drug Administration regulations for Good Clinical Practice, and were approved by each site’s Institutional Review Board. All patients or their authorized representatives provided informed consent.
GIB events were reviewed individually for additional details through manual review of the narrative reports. Descriptive statistics were used to describe source, location, and treatment. Survival is reported descriptively through Kaplan-Meier analysis, with follow-up censored at the time of heart transplantation or device explant for recovery, or withdrawal of consent or loss to follow-up. Overall survival was defined as freedom from death from any cause, with censoring at the time of heart transplant or explant for recovery. Competing outcomes were calculated by Kaplan-Meier nonparametric product limit actuarial method.
All adverse events, including those meeting the INTERMACS definitions were evaluated with respect to severity, expectedness, and device relatedness. Adverse events were reported both as the percentage of subjects affected and the rate per patient-year of follow-up.
Statistical comparisons were made between outcomes using log-rank t test, with significance determined at a p value of ≤0.05, with no adjustment for multiple comparisons. Adjustments based on baseline differences were made using the Cox proportional hazards model.
A total of 382 patients enrolled in the BTT or CAP populations between August 2008 and November 2012 are included in this report. Overall, patients had a mean age of 53.2 ± 11.7 years, were predominantly male and Caucasian, and had a mean body surface area of 2.0 ± 0.3 m2. New York Heart Association functional class IV heart failure classification was recorded for 95.8% of patients and 75.4% were in either INTERMACS Class 2 or 3. Table 1 compares demographic characteristics of patients with (n = 59) and without GIB events (n = 323). Patients with GIB had higher body mass index and creatinine, and more frequently had diabetes and an ischemic etiology of heart failure. Overall, patients who experience a GIB were less sick (mean INTERMACS profile 2.9 ± 1.3 vs. 3.3 ± 1.2; p = 0.014).
GIB events occurred in 59 of 382 (15.4%) HVAD recipients over a follow-up ranging from 36 to 1,339 days (median 365 days). A total of 108 GIB events were reported over 406.6 patient-years of HVAD support, representing 0.27 GIB events per patient year. Most of the events (93 of 108, 86.1%) occurred more than 30 days after LVAD implantation. The mean time to first event was 273.1 days. The event distribution is depicted in Figure 1. Among the 59 patients with at least 1 bleeding event, 39 of 59 (66%) had 1 GIB event and 20 of 59 (34%) had more than 1 GIB event (range 1 to 8). Freedom from a GIB event was 84.1% at 1 year (Figure 2). There was no site effect noted among the centers participating in this clinical trial (Figure 3).
INR values at the time of the GIB episode were available in 102 of 108 (94.4%) episodes. The mean INR at the time of first bleeding event was 2.2 ± 1.1. Mean INRs following initial post-implant discharge were significantly higher in patients who developed GIB than in those who did not (2.4 vs. 1.6, p ≤ 0.0001). Table 2 shows a comparison of the aspirin doses being taken by patients with and without GIB events. At the time of their event, patients with GIB were taking significantly higher doses of aspirin than those without GIB. Also, in patients with GIB taking aspirin, 4 patients were on dual antiplatelet therapy at the onset of their bleeding event (1 patient also on clopidogrel, and 3 others also taking dipyridamole).
In 78% of bleeding events, a specific lesion was suspected or identified. By far the most common lesion was an arteriovenous malformation (AVM) followed by an ulcer (includes gastric and duodenal). In 22% of instances, no active site could be identified (Table 3). The most common lesion site was the small intestine (Figure 4). The duration of LVAD support was longer for patients with GIB than those without GIB: 477.2 ± 363 days for the 59 patients with GIB vs. 372.6 ± 331.24 days for the 323 with no GIB, p = 0.0159. Among patients with GIB, at 365 days post-implantation, 58.2% of patients were alive on the original device, 27.4% had received a transplant, 5.3% required device exchange, and 9.1% had died. Among patients with no GIB, at 365 days post-implantation, 43.8% of patients were alive on the original device, 38.2% had received a transplant, 6.3% required device exchange, and 11.7% had died. Figures 5A and 5B show a competing risk analysis for these outcomes.
The management of the 108 bleeding events (the 59 patients who experienced a GIB event) is depicted in Figure 6. Transfusion of blood products was required in the overwhelming majority (94%) of bleeding events and in an 18.5% of bleeding events, no further therapy was required. An endoscopic procedure was used to manage bleeding in 52 of 108 (48%) cases. Endoscopic procedures used included cautery, clipping, thermal ablation, or argon plasma coagulation. No surgical procedures (i.e., laparotomy) were required to control GIB.
Survival and adverse events
There were no deaths directly related to GIB. One patient died 6 days following a GIB event. The GIB was attributed to intense anticoagulation to treat a pump thrombus event. The patient then had a pump exchange, developed right heart failure, and later died post-exchange due to vasodilatory shock. There was no statistically significant difference in overall survival between patients who experienced GIB and those who did not for either unadjusted survival or for survival adjusted for ischemic etiology and body surface area (Figure 7).
The rate and incidence of adverse events for the GIB and non-GIB populations are shown in Tables 4 and 5⇓. Patients who had GIB events had higher rates of bleeding requiring hospitalization (1.17 events/patient year vs. 0.07 events/patient year, p = 0.0001) and higher rates of non–device-related infections (0.74 events/patient year vs. 0.407 events/patient year, p = 0.0021). Although rates of sepsis were similar for GIB versus non-GIB, the incidence of sepsis was higher among patients with GIB (28.8% vs. 17.0%, p = 0.0452). The rates and incidence of other adverse events were not significantly different between patients with and without GIB.
Of the 59 patients who had a bleeding event, five had a subsequent thrombotic event (pump thrombosis, transient ischemic attack, ischemic cerebrovascular accident, or peripheral thrombosis). The mean time to first thrombus event following the GIB was 155.4 days. The history of each thrombotic event is described briefly in Table 6. In 50% of GIB events, therapy with 1 or more anticoagulant drugs was interrupted within 5 days after the event. Table 7 details the interruptions in anticoagulant therapy after GIB events.
The clinical introduction of miniaturized continuous flow left ventricular assist devices (CF-LVADs) drastically improved survival and quality of life outcomes for recipients of these technologies. Moreover, it led to a wider acceptance of mechanical support by patients, referring physicians and the heart failure community in general. This success has been tempered by the recognition of new adverse events not encountered with previous pulsatile first-generation technology, namely pump thrombosis, de novo aortic insufficiency, and GIB. The latter was the focus of this investigation.
Frazier et al. (21) first described gastrointestinal bleeding (from AVMs) in patients receiving a CF-LVAD—the Jarvik 2000 (Jarvik Heart, Inc., New York, New York) device—an intraventricular axial flow pump. Several single-institution reports began documenting a high incidence of GIB in recipients of these technologies and in fact, several have singled out GIB as the most frequent etiology for readmission following successful implantation (7,22).
The present study is, to our knowledge, the first multicenter investigation aimed at fully characterizing the vexing problem of GIB in recipients of this technology. Among 382 patients who received the HVAD System as a BTT, 15.4% developed 108 GIB events over nearly 407 years of support, for a rate of 0.27 GIB/PPY. Freedom from GIB events was 84.1% at 1 year and most events occurred beyond the perioperative period. This incidence is lower than that suggested by a recent review of all case reports and case series in the literature whereby 265 of 1,316 (20.1%) CF-LVAD recipients developed a GIB (6). In the few case series that have reported annualized GIB event rates, the reported rates were higher and involved mostly HeartMate II recipients. Stulak et al. (11) documented a rate of 0.45 GIB/patient year among 389 recipients of 4 different CF-LVADs while Crow et al. (8) described a rate of 0.63 GIB/patient year among their 55 recipients of nonpulsatile devices.
Time to first GIB event was 273 days, longer than the 5 months reported by Stulak et al. (11), or the 63 ± 62 days, described by Demirozu et al. (9), among the 19% of 172 patients who had GIB after receiving a HeartMate II device. French et al. (12) described a maximal hazard rate of GIB of 2.23 events/patient year at 21 days post-implantation.
We noted several interesting differences in baseline parameters of patients with GIB compared with those without GIB. On univariate analysis, the former were larger (higher body surface area, body mass index), were more likely to be diabetic, had worse renal function, and were as a group less sick. It is well known that renal dysfunction can alter platelet function by interfering with adhesion and aggregation. Our finding that GIB patients had worse baseline renal function supports this pathophysiology. However, because data regarding renal function at the time of GIB events is not available, it is difficult to attribute renal dysfunction as a definitive contributor as it is possible that in many patients with baseline dysfunction, renal function normalized after implantation of the LVAD, as it frequently occurs with restoration of optimal hemodynamics and end organ perfusion. Our findings in this regard, however, mimic those of Demirozu et al. (9), who noted a trend for patients with GIB to have higher baseline creatinine level.
The observation of ischemic etiology as a preoperative risk factor for GIB coincides with the finding by Boyle et al. (23), who found that ischemic etiology was a strong perioperative factor (hazard ratio: 1.35) in an analysis of 900 patients who received the HeartMate II LVAD.
While our data suggests a statistically significant increased risk of GIB for patients with larger body surface area/body mass index, the actual differences in these values (2.0 vs. 2.1 m2 and 29.9 vs. 27.9 kg/m2, respectively) is very small and likely clinically unimportant. The finding of diabetes as a preoperative risk factor for GIB is intriguing and is interesting to note a recent publication documenting diabetes as an independent risk factor for upper GIB in a large population study (24). As described previously (15,16,25), GIB events often recur, with 34% of our patients having at least 1 recurrence. The identification of AVMs as the most common etiology of GIB is also consistent with previous reports (9,10,21); this finding has been attributed at least partly to low pulsatility and shear stress–induced acquired von Willebrand syndrome resulting from continuous flow assist physiology. In fact, recipients of CF-LVADs with higher pulsatility have been shown to have a reduced incidence of nonsurgical bleeding events (14). In our series, nearly one-quarter of GIB events could not be localized. Many or all of these are likely due to occult AVMs of the small bowel, which are notoriously difficult to identify and treat.
Management of GIB events generally followed common clinical practice, with transfusions, cessation of antithrombotic therapy, antacid therapy, and diagnostic work-ups at the discretion of the caring physicians. The issue of timing an intensity of reintroduction of antithrombotic therapy following resolution of a GIB event is of paramount importance. Unfortunately, the present study did not capture this information and recommendations cannot be made. Moreover, although higher ASA doses and higher INR values were observed in persons with GIB, the clinical trial was not designed to determine optimal dosing of antithrombotic therapies.
Though no data is available to describe frequency of its use or success, octreotide, either in short- or long-acting form, has emerged as an additional tool in the armamentarium to treat occult GIB that is likely secondary to small bowel AVMs (26,27). While GIB events often required readmissions and incurred costs, no surgical interventions were required and no deaths were directly related to the GIB event. At 12 months, 85.6% of patients were alive with device or transplanted.
Comparison of the incidence of adverse event rates between patients who had GIB and those who did not suggests that the former are more likely to be readmitted and transfused and are more prone to develop localized nondevice infections and sepsis. The causal relationship between blood transfusions and subsequent infections is well established in cardiac surgical patients (28) thus our findings are not surprising. When these events are annualized to adjust for different follow up times however, only the differences in rehospitalization, localized nondevice infection and thromboembolic rates remain statistically significant. The lower rate of thromboembolic complications observed among persons with GIB, perhaps underscores the possibility that certain patients have an inherent propensity to bleed and are thus less likely to develop thrombotic complications.
Concern has arisen regarding the potential prothrombotic milieu created by the need to discontinue antithrombotic therapy in recipients of CF-LVADs. Notwithstanding the differences in devices and management strategies it is interesting to note that a recent report by Stulak et al. (11) documented a 0.31 thromboembolic event per patient year rate among persons with GIB with a median interval of 5 months after the GIB event. And while thromboembolic events were 7.4 times more likely to occur in patients with prior GIB, neither the GIB nor the thromboembolic event portended a lower survival (11). In our series, nearly two-thirds of patients had interruption of either antiplatelet or warfarin therapy. Five patients developed a thrombotic complication (range 11 days to 10 months) following a GIB event. In 2 of these instances, antithrombotic therapy had not been interrupted.
Several limitations to our analysis should be considered. First, the study was not randomized, and all patients received an HVAD. Any comparisons are only available through historical literature. Also, this was a post-hoc analysis, and the clinical trial was not powered to analyze GIB events.
In summary, GIB has emerged as a significant complication limiting the success of current CF-LVAD technologies. GIB events occurred in 16% of HVAD recipients, for an event rate of 0.27 GIB/patient year, which is lower than rates reported for other devices. While the exact etiology of GIBs has not been elucidated, low pulsatility and VAD-induced shear stress degradation of high molecular weight von Willebrand factor appear to contribute to the genesis. AVMs of the small bowel are the most common culprit lesions. The morbidity associated with the development of GIB in LVAD recipients suggests that preoperative endoscopic screening of patients with risk factors for GIB including advanced age, history of GIB, occult blood in the stool and/or microcytic anemia should be seriously considered. Conventional diagnostic evaluations and therapies are effective in managing these bleeding events, but recurrence is not uncommon. Fortunately, while the GIBs are associated with readmissions and a higher incidence of nondevice infections, survival is not affected. It is hoped that the introduction of pulsatility algorithms and wider gaps that reduce shear stress in next generation pumps will reduce rates of this vexing complication.
COMPETENCY IN MEDICAL KNOWLEDGE 1: GIB is a common complication in patients who have received a continuous flow left ventricular assist device.
COMPETENCY IN MEDICAL KNOWLEDGE 2: Possible etiologies of gastrointestinal bleeding in CF-LVAD recipients include low pulsatility and ventricular assist device–induced shear stress, which degraded high molecular weight von Willebrand factor.
COMPETENCY IN PATIENT CARE: Conventional diagnostic evaluations and therapies are effective in managing gastrointestinal bleeding in CF-LVAD recipients. Although GIB increases readmissions and nondevice infections, it does not affect survival.
TRANSLATIONAL OUTLOOK: Improvements in pump design and control algorithms to reduce shear stress and increase pulsatility are likely to reduce rates of CF-LVAD-related gastrointestinal bleeding.
The authors acknowledge Mary V. Jacoski, MS, and Edward K. Baldwin, PhD, of HeartWare Inc., for their assistance in the preparation of the manuscript.
HeartWare Inc. (Framingham, Massachusetts) was the sponsor of the study. Dr. Goldstein is a surgical proctor for and is on the advisory board of HeartWare Inc.; and is on the medical advisory board of Thoratec Inc. Dr. Aaronson has received grant and research support from HeartWare Inc. and Thoratec; and serves on the advisory board (without remuneration) for HeartWare Inc. and Thoratec. Dr. Tatooles has served as an investigator for HeartWare Inc. and Thoratec. Dr. Silvestry has served as a consultant for HeartWare Inc. and Thoratec. Dr. Jeevanandam has served as a consultant for HeartWare Inc. Dr. Hathaway was formerly an employee of and has served as a consultant for HeartWare Inc. Mr. Najarian is an employee of and owns stock in HeartWare Inc. Dr. Slaughter has received research grant support from HeartWare Inc. Dr. Gordon has reported that he has no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- arteriovenous malformation
- acetylsalicylic acid
- continued access protocol
- continuous flow left ventricular assist device
- gastrointestinal bleeding
- HeartWare ventricular assist device
- international normalized ratio
- Received July 30, 2014.
- Revision received October 31, 2014.
- Accepted November 14, 2014.
- American College of Cardiology Foundation
- Aaronson K.D.,
- Slaughter M.S.,
- Miller L.W.,
- et al.
- Cowger J.,
- Pagani F.D.,
- Haft J.W.,
- Romano M.A.,
- Aaronson K.D.,
- Kolias T.J.
- Wever-Pinzon O.,
- Selzman C.H.,
- Drakos S.G.,
- et al.
- Strueber M.,
- O'Driscoll G.,
- Jansz P.,
- Khaghani A.,
- Levy W.C.,
- Wieselthaler G.M.
- Hasin T.,
- Marmor Y.,
- Kremers W.,
- et al.
- Boyle A.,
- Jorde U.P.,
- Sun B.,
- et al.
- Coutance G.,
- Saplacan V.,
- Belin A.,
- Repesse Y.,
- Buklas D.,
- Massetti M.