Characterization and timing of gastrointestinal bleeding in continuous flow left ventricular assist device recipients.

BACKGROUND AND AIMS: Heart failure is one of the leading causes of morbidity and mortality in the United States. The advent of left ventricular assist devices (LVAD) has improved the survival and quality of life in patients with end stage heart failure. Gastrointestinal bleeding (GIb) remains one of the limitations of LVADs.
METHODS: A single center, retrospective review of records was performed for patients who underwent LVAD implantation between 2010 and 2015. All patients who survived more than 30 days were followed till March 2016 and are described below.
RESULTS: A total of 79 patients were included in the study. The rate of GIb was 34.1% (27 patients) with a mean time to bleed of 267 days. Older patients were more likely to bleed. Upper GI bleeding was the source of bleeding in 54% patients. Arteriovenous malformations (AVM) were the source of bleeding in 74% bleeders and 80% of these patients had de novo AVM formation. 14/27 (51%) patients had a re-bleeding event. Thrombotic events were 4.5 times more likely to occur in patients who also had a GI bleed.
CONCLUSIONS: GI bleeding in LVAD patients is common with the source of bleeding more commonly being in the upper GI tract. GI bleeding may occur as early as 10 days post procedure, despite previous negative screening endoscopies. There is an increased risk of thrombotic events in patients who have experienced a GI bleed.

Introduction

Heart failure (HF) is a leading cause of hospitalization and death in the United States and one in five people with HF develop end stage heart disease [1]. Resultantly, left ventricular assist devices (LVADs) have demonstrated significant clinical utility through prolonging life in these patients and have evolved into a dual therapy option for patients as either a bridge to cardiac transplantation (BTT) [2, 3, 4, 5, 6, 7] or a long term therapy option known as “destination therapy (DT)”. Second generation continuous flow (CF) LVADs, have resulted in significant improvement in survival and quality of life in patients [8, 9, 10].

Despite the obvious benefits, gastrointestinal (GI) bleeding is one of the main complications of LVAD therapy [11, 12], with a pooled prevalence of 23 % in patients with CF-LVADs [13]. The pathophysiology behind the increased GI bleeding risk appears multifactorial, however, the primary contributing factor is believed to result from the development of acquired (type 2A) von Willebrand syndrome (due to shearing of the vWF polymers into monomers causing a functional deficiency), unmasking of subclinical arteriovenous malformations [14], and ongoing need for systemic anticoagulation and antiplatelet therapy in CF LVAD recipients. The most common site of GI bleeding is generally in the upper GI tract with arteriovenous malformations being the most common etiology [15, 16]. GI bleeding is often substantial in these patients, requiring multiple endoscopic procedures, reduction of anticoagulation and anti-platelet therapy. Given the expanding role of LVADs as BTT or DT, larger numbers of patients will be at risk of device associated adverse events, including GI bleeding for longer periods of time [17, 18, 19].

Thus far, identified patient related risk factors for GI bleeding with LVAD placement include older age, elevated INR, low platelet count and a history of prior GI bleeding [15]. While the associated risk of GI bleeding after LVAD implantation is clear, uncertainty remains around the timing of GI bleeding after implantation [13, 15, 16, 20]. In our study, we aimed to determine the rate of GI bleeding in patients with LVADs and potential periods of high risk, risk factors predisposing to GI bleeding and risk of thrombotic events in patients with GI bleeding.

Methods

The study was performed at Saint Luke's Mid America Heart Institute, a tertiary cardiac care center with regional expertise in cardiac transplantation and mechanical assist device implantation. A retrospective review of hospital records was performed for patients who underwent LVAD implantation from 2010-2015 for either bridge to transplant or destination therapy. Patients were considered only if they had followed up in our institution for at least 1 year post LVAD implantation. Patients were considered to have GI bleeding if they had one or more of the following: guaiac positive stools with a drop of hemoglobin (Hb) > 2 g/dL, hematemesis, hematochezia or melena, active bleeding or blood within the GI tract at time of endoscopy with transfusion of packed red blood cells. Recurrent bleeding was defined as more than one episode of GI bleeding after LVAD implantation. Upper and lower GI bleeding were defined as bleeding above and below the Ligament of Treitz respectively. Bleeding events were considered after seven days of LVAD placement to exclude peri-operative bleeding events. Patient records were followed until March 2016, or death, whichever event occurred first. This study was reviewed and approved by the Saint Luke's Hospital Institutional review board.

Data collected about LVAD details included type of LVAD, indication for implantation; patient demographics including age and sex, comorbidities including diabetes, CKD and pre LVAD history of GI bleed; laboratory values including most recent pre-implant hemoglobin, platelet count, INR, creatinine level, time to first GI bleed (in days) post LVAD implant, pump speed at the time of the bleeding event, time to recurrent GI bleed (in days), thromboembolic events, lowest hemoglobin, platelets and INR at time of bleed, number of PRBC and FFP units transfused, endoscopic and radiologic procedures performed and measures to stop bleeding.

Statistical analysis

All statistical analysis was performed using STATA v 14.0. Continuous variables were compared using Student's t-test and categorical variables were compared using Chi square test. A p-value less than 0.05 was considered significant. All data is reported using 95% confidence interval (CI). Logistic regression analysis was performed to identify independent risk factors for bleeding. Kaplan Meier survival curves for GI bleeding were plotted for the entire time of follow-up.

Results

Patient characteristics

A total of 110 patients received LVADs between 2010 and 2015, however 31 patients were excluded from the analysis due to short survival times or short length of follow up (less than 1 year) after implantation (<30 days). Of the remaining 79 subjects included in the analysis, the mean age was 64 (±11) years. 58 (73%) patients were male, 15 (19%) had baseline GERD, 33 (41%) had diabetes, 65 (82%) had hypertension, 18 (22%) had chronic kidney disease, 4 (5%) had a history of GI bleeding (Table 1). The mean duration of follow-up was 1.67 ± 1.4 years. Devices implanted were Heartmate II (Thoratec Corp, Pleasanton, CA, USA) (67 patients) and Heartware HVAS (HeartWare®, Inc., Framingham, MA, USA) (12 patients).

Table 1

Patient characteristics.

Variable	Patients without GI bleeding events N = 52 (66%)	Patients with GI bleeding events N = 27 (34%)	P value
LVAD type
Heartmate II	42 (80%)	25 (92%)	0.16
Heartware	10 (19%)	2 (7%)
Sex
Male	39 (75%)	19 (70%)	0.66
Comorbidities
GERD	8 (15%)	7 (25%)	0.260
Diabetes Type II	21 (40%)	12 (44%)	0.73
Hypertension	41 (78%)	24 (88%)	0.27
Chronic kidney disease	12 (23%)	6 (22%)	0.93
History of GI bleeding	3 (5%)	1 (3%)	0.69
Device strategy
Bridge to transplant	24 (44%)	3 (11%)	0.003
Destination therapy	28 (53%)	24 (88%)	0.002
Pre-operative anticoagulation and antiplatelet therapy
Apixaban	1 (1%)	0	0.47
Aspirin	35 (67%)	26 (96%)	0.003
Warfarin	2 (3%)	5 (18%)	0.03
Clopidogrel	5 (9%)	3 (11%)	0.83
Age	59 ± 14.5	69 ± 7.83	0.013
Hemoglobin	11.20 ± 2.14	10.54 ± 1.47	0.16
INR	2.27 ± 0.67	2.26 ± 0.54	0.99
Platelet count	257 ± 93	245 ± 78	0.57
Creatinine	1.04 ± 0.4	1.18 ± 0.3	0.19

Gastrointestinal bleeding in LVAD subjects

Of the 79 subjects included in the study, gastrointestinal bleeding occurred in 27 (34.1%) patients with a total of 46 bleeding events during 1,589 patient months (132.4 patient years) of follow up. The median time to first bleed was 60 days with 17 patients experiencing their first GI bleeding event within the first 90 days. DT LVAD recipients were around 6 times more likely to bleed as compared to BTT recipients (unadjusted OR 6.34, p = 0.0020) when time on therapy was not adjusted for, however after adjustment for time, there was no significant difference between bleeding in DT versus BTT patients.

Risk factors for bleeding

Warfarin preoperatively increased the odds of bleeding by 5.68 fold (OR 5.68, p = 0.031) and patients who had a bleeding event were on an average 10 years older than without bleeding events (69 vs. 59 years old). There was no significant association between sex, pump speed, underlying comorbidities, INR, platelets, hemoglobin or concomitant clopidogrel and risk of bleeding.

Re-bleeding events

Fourteen (52%) patients experienced re-bleeding events (Table 2). Diabetes was a significant risk factor for re-bleeding in our patient population, with diabetic patients 3 times more likely to re-bleed compared to non-diabetic patients (RR 3.125 p = 0.0031). Re-bleeders also required more pRBC transfusions (OR 1.77, p = 0.03). There was no significant relationship between the risk of re-bleeding and age, sex, other underlying co-morbidities (hypertension, chronic kidney disease, GERD). The severity of bleeding, as measured by number of packed red cells transfused, was not associated with time of onset of bleeding (defined as less than 30 days, 30–180 days and more than 180 days) or underlying patient characteristics.

Table 2

Single vs multiple bleeding events.

	Patients with single bleeding episode (n = 13)	Patients with multiple bleeding episode (n = 14)	P value
Age at time of 1st bleed	70.38 ± 7.3	69.42 ± 8.2	0.75
INR at the time of presentation	2.18 ± 0.5	2.3 ± 0.5	0.60
RBCs transfused (units)	1.69 ± 1.8	4.42 ± 3.17	0.01
FFPs transfused (units)	0.53 ± 0.9	1.35 ± 1.8	0.16
Baseline Hemoglobin	11.4 ± 0.5	10.07 ± 0.3	0.03
Diabetes mellitus	2 (15%)	10 (71%)	0.003
Hypertension	12 (92%)	12 (85%)	0.58
GERD	3 (23%)	4 (28%)	0.75
CKD	4 (30%)	2 (14%)	0.30

Source of bleeding and therapeutic interventions

In 25 (54%) events, bleeding was identified to be from upper GI source. 20 (74%) patients bled from arteriovenous malformations (AVM) and earliest time for bleed due to AVM was 10 days post implant (Figure 1). 16 (80%) patients with AVMs detected on endoscopy post bleeding had prior upper and lower gastrointestinal (GI) endoscopies within 3 months prior to LVAD implantation with no AVMs reported (Table 3).

Figure 1

Kaplan-Meier curve comparing rates of GI bleeding in patients with destination therapy and bridge to transplantation LVADs.

Table 3

Sources of GI bleeding.

Bleeding Source	% of bleed
AVMs	62%
Gastritis	11%
Hemorrhoids	9%
Colon polyps	6%
Diverticulosis	3%
Peptic ulcer disease	3%
Stomach polyps	6%

A total of 71 endoscopic procedures were performed to evaluate GI bleeding, the preliminary approach being upper followed by lower GI endoscopies or both (Table 4). Subsequent approaches included capsule studies (n = 13) and anterograde and retrograde double balloon enteroscopy (n = 13) including both anterograde and retrograde. Bleeding was resolved by decreasing pump speed and temporarily withholding anticoagulation in 10 (37%) patients. 7 (26%) patients needed additional pharmacotherapy beyond endoscopic interventions to control their bleeding (5 were treated with Octreotide and 2 with additional Thalidomide).

Table 4

Numbers of procedures performed.

Procedure	Number Performed
EGD	18 (22%)
Colonoscopy	3 (3%)
EGD with colonoscopy	24 (30%)
Capsule enteroscopy	13 (16%)
Technetium scan	8 (10%)
Double balloon endoscopy	13 (16%)
Total number	79

Thrombotic events

There were a total of 16 thrombotic events (14 pump thromboses and 2 cerebrovascular accidents (CVAs). Out of these, 10 occurred in patients who had either had a GI bleeding event previously or a bleed afterwards. The median time to a thrombotic event after the first GI bleed was 310 days. There was an increased incidence of thrombotic events in GI bleeders, with thrombotic events 4.5 times more likely to occur in patients who also had a GI bleeding episode (OR 4.50, 95% CI 1.4 to 14.31, p = 0.0106), though this was not temporally associated with the bleeding event. There was no difference between the means of lowest INRs at time of thrombotic event between subjects with or without GI bleeds.

Mortality

34 (43%) out of the 79 patients died, 20 (60%) of these were non-bleeders, Bleeding or re-bleeding did not adversely impact survival in our study.

Discussion

Gastrointestinal bleeding in LVAD patients is a known adverse event and, in our study, occurred at a rate of 34.1%. Older age and use of warfarin were identified as significant risk factors for bleeding in these subjects. Re-bleeding remains a common occurrence in patients with CF-LVADs, with almost half of the patients with a bleeding event experiencing re-bleeding, with 7 (26%) patients up experiencing 3 or more re-bleeding episodes. Uniquely, we report a dichotomous rate of bleeding with majority of the first bleeding events 17 (63%), occurring within the first 90 days after LVAD implantation with the earliest bleed as early as 10 days after device implantation. Finally, thrombotic events were 4 times more likely to occur in patients with a GI bleed.

In addition to the above findings, we also demonstrate a very high rate (80%) of possible de novo GI tract AVM formation/unmasking of subclinical AVMs in GI bleeding patients, with 60% of the first bleeding events attributed to subclinical AVMs. The majority (74%) of patients in this study bled from AVMs that were discovered using an aggressive approach towards diagnosing obscure GI bleeding, including capsule endoscopies and subsequent double balloon enteroscopy in this patient cohort. The formation of AVMs appears to mostly likely be due to increased intraluminal pressure within the blood vessels and lowered pulse pressure leading to intestine wall ischemia and development of AVMs [21]. The rate and temporal trends of development of new or transformation of subclinical AVMs to clinically significant ones with CF-LVAD implantation have not been previously described and hence there is lack of consensus on treatment strategies to reduce re-bleeding events [12].

In this study, we report two distinct time periods during which GI bleeding is more likely to occur, one occurring within 3 months post-implantation and the second period happening later, at almost 2 years post implantation. Other studies have described various times to bleeding ranging from 65 to 128 days after LVAD implantation (16,20). It is possible that these distinct periods are due to different etiologies of GI bleeding, however further study is needed to confirm this.

The approach to the management of GI bleeders in our patient population, started with temporarily withholding warfarin and decreasing pump speed (which increases left ventricular preload resulting in more left ventricular ejection thereby restoring some element of pulsatility to the flow). The preliminary diagnostic approach in patients who were stable, was to initially perform an upper GI endoscopy followed by lower GI endoscopy and a capsule endoscopy if the source of bleeding was not identified (obscure GI bleeding). This was then followed by double balloon enteroscopy (both anterograde and retrograde as needed) based on results from the capsule study. In cases where bleeding continued, long term octreotide therapy was used and in a small minority of patients thalidomide was used (Figure 2). Lowering the INR goal to 1.5 (baseline goal 2–3) in patients at high risk for bleeding was also carried out. Continued bleeding necessitated the discontinuation of warfarin in 2 of our patients. While there is no formalized comparison of this management strategy to establish efficacy, our approach remains in concordance with that described in literature elsewhere [13,15].

Figure 2

Algorithm for management of GI bleeding.

While our findings expand previous literature describing GI bleeding with CF-LVAD placement, there are several limitations to our study. First, given the single center retrospective design of the study, there may have been underreporting of events if the patients did not get periodically evaluated for GI bleeding or were admitted for a GI bleeding event elsewhere. Secondly, some patients were lost to follow up due to unclear reasons, at various times after implant, leading to loss of data. Finally, GI bleeding as defined by hematochezia or melena was not confirmed by medical staff and could have therefore been over reported in some cases.

In summary, patients receiving second generation LVADs have a significant risk of GI bleeding, important risk factors for the same include age, LVAD implantation as destination therapy and preoperative warfarin and aspirin use. Around 50% patients with a bleeding event experience re-bleeding and require multiple endoscopic procedures for the diagnosis and treatment of the bleeding. GI AVMs are the most common cause of the bleeding with de novo or newly unmasked subclinical AVMs identified in 80% patients with GI bleeding. The risk of thrombotic events, mainly stroke, is also elevated in patients who have experienced GI bleeding.

Declarations

Author Contribution statement

Devika Kapuria: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Taiyeb Khumri, Sanjeev Aggarwal, Christopher Koh: Conceived and designed the experiments; Wrote the paper.

Shariq Shamim, Rajiv Chhabra: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Pallavi Surana: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Salman Khan: Conceived and designed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Nabil Al-Khalisi: Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Competing interest statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.