AIM: To assess whether video capsule endoscopy (VCE) affects the outcomes of left ventricular assist devices (LVADs) recipients with gastrointestinal bleeding.

METHODS: This is a retrospective study of LVAD recipients with obscure gastrointestinal bleeding (OGIB) who underwent VCE at a tertiary medical center between 2005 and 2013. All patients were admitted and monitored with telemetry and all VCE and subsequent endoscopic procedures were performed as inpatients. A VCE study was considered positive only when P2 lesions were found and was regarded as negative if P1 or P0 were identified. All patients were followed until heart transplant, death, or the end of the study.

RESULTS: Between 2005 and 2013, 30 patients with LVAD underwent VCE. Completion rate of VCE was 93.3% and there was no capsule retention. No interference of VCE recording or the function of LVAD was found. VCE was positive in 40% of patients (n = 12). The most common finding was active small intestinal bleeding (50%) and small intestinal angiodysplasia (33.3%). There was no difference in the rate of recurrent bleeding between patients with positive and negative VCE study (50.0% vs 55.6%, P = 1.00) during an average of 11.6 ± 9.6 mo follow up. Among patients with positive VCE, the recurrent bleeding rate did not differ whether subsequent endoscopy was performed (50% vs 50%, P = 1.00).

CONCLUSION: VCE can be safely performed in LVAD recipients with a diagnostic yield of 40%. VCE does not affect recurrent bleeding in LVAD patients regardless of findings.

Approximately 50000 patients die from advanced heart failure in the United States each year, with a high mortality rate and life expectancy < 2 years when only medical therapies are utilized[1]. Although heart transplantation is a definitive therapeutic option for advanced heart failure, only 2200 heart transplants are performed annually due to donor shortage, leaving a large proportion of heart failure patients in need of an alternative therapy[2]. In recent years, Left Ventricular Assist Devices (LVADs) are used increasingly in this setting as destination therapy, bridge to transplantation, bridge to recovery, or bridge to decision in patients with advanced heart failure[3]. This approach has increased survival and improved quality of life in advanced heart failure patients[4].

Since the initial introduction of LVAD therapy, it is well documented that LVADs increase the risk of gastrointestinal bleeding (GIB), with as many as 20%-40% of LVAD recipients manifesting GIB[5-7]. The mechanism of GIB in LVAD recipients remains incompletely understood, but it is thought to be contributed to by development of angiodysplasia, acquired von Willebrand disease, persistent right ventricular dysfunction, and mucosal ischemia secondary to low pulse pressure[8,9]. GIB is further exacerbated by the use of anticoagulation in LVAD patients. Although previous reports found that upper GI tract is the most common site of GIB in LVAD recipients[10], obscure gastrointestinal bleeding (OGIB) remains a frustrating condition frequently encountered in this population.

Video capsule endoscopy (VCE) has made a significant impact on the evaluation of patients with OGIB, with a diagnostic yield of approximately 60%-70%[11]. However, it is a relative contraindication to use VCE in the setting of implanted electrical medical devices[12], and there is limited data on both the usage of VCE in LVAD associated OGIB, as well as the safety of VCE in LVAD patients[13-19].

We retrospectively evaluated our experience with VCE in LVAD patients with OGIB. The aims of our study were to determine the safety and diagnostic yield of VCE, and to assess the outcomes based on management driven by VCE in LVAD recipients.

Thirty LVAD patients underwent VCE over the 8-year study period. No patient was lost to follow up or excluded in this study. All patients were treated and all procedures were performed as inpatients. The mean age was 60.1 ± 10.2 years, and 20% of patients were female (Table 1). Thoratec HeartMate II LVADs were implanted in all of the patients except one patient who had a HeartWare HVAD. The Charlson comorbidity index, the history of GIB prior to LVAD implantation, the interval between LVAD implantation and GIB, and the history of overt GIB did not differ between patients with positive vs negative VCE studies. Twenty-three out of the thirty patients (76.7%) presented with overt OGIB: 21 with melena (70%), 2 with hematochezia (6.7%); whereas 7 patients (23.3%) presented with occult OGIB. Most of our patients received antiplatelets (86.7%) or anticoagulants (93.3%) on presentation. On average 3.2 ± 1.7 endoscopic procedures were performed within 4.1 ± 5.0 d prior to VCE, including 37 EGDs, 17 push enteroscopies, 40 colonoscopies, and 2 sigmoidoscopies. VCE was performed 6.2 ± 2.6 d after the presentation of GIB. VCE was placed endoscopically in 2 patients (6.7%) because one patient had a history of pyloric stenosis, and the other patient failed the swallow study. The mean small bowel transit time of VCE was 3.2 ± 1.1 h. VCE did not reach the cecum in 2 patients (6.7%) over the 8 h recording period, but there was no capsule retention. There was no electromagnetic interference of either VCE or LVAD identified in any patients.

Patients with positive VCE study stayed in the hospital longer than patients with negative VCE study (20.3 d vs 8.3 d, P = 0.04). Over the average 11.6 mo follow-up period, there was no statistically significant difference in the recurrent bleeding rate (50% vs 55.6%, P = 1.00), the number of endoscopies performed after VCE (1.8 ± 1.8 vs 1.7 ± 2.5, P = 0.97), or mortality rate (33.3% vs 33.3%, P = 0.90) between patients with positive and negative VCE. The total recurrent bleeding rate in this population was 53.3% (n = 16) and the presentation included melena (n = 12), hematochezia (n = 3) and anemia with positive fecal occult blood (n = 1). All 16 patients with recurrent bleeding were hospitalized and underwent transfusion and endoscopic procedures for managing recurrent GIB. The overall mortality rate in this study was 33.3% (n = 10): 7 patients died from underlying heart failure, 2 patients died from septic shock, one patient died from subdural hematoma, and none of the patients died from GIB. Four LVAD recipients underwent heart transplantation on average 4.3 mo after VCE and did not develop recurrent GIB afterwards. Before heart transplantation, VCE studies were positive in 2 patients (1 duodenal angiodysplasia and 1 jejunal angiodysplasia) and negative in 2 others.

The diagnostic yield of VCE to detect P2 lesions in this study was 40%. Table 2 demonstrates the locations and the findings of positive VCE studies. Small intestine was the most common site of positive VCE findings (75%). The predominant positive VCE findings in our study were small intestinal bleeding with no source or lesion identified (50%) and small intestinal angiodysplasias (33.3%). Eighteen VCE studies (60%) were negative, including 13 P0 lesions (12 normal, 1 small nodule) and 5 P1 lesions (red spots). The only patient who had a HeartWare HVAD implanted had a normal VCE study. Despite prior negative upper endoscopies performed 2 and 24 d prior to VCE, two lesions were found within reach of EGD by VCE: one gastric ulcer and one duodenal angiodysplasia (neither VCE was placed endoscopically). One VCE found active bleeding in the colon without the cause of bleeding identified. Angiodysplasia were found in 4 patients: 1 in the duodenum and 3 in the small intestine. In 2 patients where VCE failed to reach the cecum at the end of recording, VCE still detected the cause of GIB: one with gastric ulcer and one with small intestine angiodysplasia. Using logistic regression, we found that higher INR on presentation was associated with a higher probability of positive findings in VCE (OR = 3.62, 95%CI: 1.03-12.7, P = 0.04), adjusted for age, gender, and hemoglobin level.

Positive VCE studies led to further endoscopic evaluations in 6 patients out of 12 (50%): 6 push endoscopies and 3 single balloon enteroscopies. The other 6 patients with positive VCE did not have further endoscopies because they had no further bleeding and their hemoglobin had stabilized. During follow-up the overall recurrent bleeding rate in patients with positive VCE was 50% (6 out of 12). In addition, there was no difference in the recurrent bleeding rate whether subsequent endoscopic procedures were performed following positive VCE (3 out of 6 or 50% in each group, P = 1.00). Furthermore, after VCE, medications were adjusted in 7 out of 12 patients with positive VCE, and 8 out of 18 patients with negative VCE. These changes included discontinuation or decrease in the dose of aspirin and initiation of proton pump inhibitors. The change of medical management did not affect the rate of recurrent bleeding regardless of whether patients had a positive VCE (40% vs 57.1%, P = 1.00), or negative VCE (60% vs 50%, P = 1.00). The presentations of recurrent GIB were melena (n = 4), hematochezia (n = 1) and anemia with positive fecal occult blood (n = 1). The clinical course and management of LVAD recipients with positive VCE studies are detailed in Table 3.

In this retrospective study spanning 8 years, we demonstrated the safety of VCE and a 40% diagnostic yield of P2 lesions in LVAD recipients with OGIB. We found that the results of VCE were not associated with the rate of recurrent GIB, the number of endoscopic procedures performed, or mortality rate. In addition, the findings of VCE and the subsequent management did not affect the rate of recurrent GIB in LVAD patients.

This study identified 30 LVAD recipients undergoing VCE for OGIB. To our knowledge, this is the largest available series of VCE in LVAD patients in the literature[6,13-19,22] (Table 4). Our results show that VCE is safe to perform in LVAD recipients, without interference between VCE and LVAD and without capsule retention. One prior study found 2 cases of LVAD possibly interfering with capsule images, and suggested that the leads of VCE be placed away from LVAD[23]. Based on the results of this and other previous studies, the interference between VCE and LVAD is uncommon[14,16,23]. Consistent with our experience, a recent review article reported that VCE is unlikely to impair the function of cardiac pacemakers, implantable cardioverter defibrillators, and LVAD, although the authors cautioned that wireless telemetry may interfere with VCE recordings[24].

The diagnostic yield of VCE in LVAD patients in two previous reports was 31% (n = 13) and 80% (n = 5) (Table 4)[6,22]. With a larger sample size of 30, the diagnostic yield in our study is 40%. Since only P2 lesions were considered positive, our diagnostic yield reflects true clinically relevant findings. It is known that the diagnostic yield of VCE is higher when it is performed closer to the presenting GIB event[25,26], or in patients with overt GIB than with occult GIB[27,28]. It is also known that the diagnostic yield of VCE is comparable to double balloon enteroscopy (DBE) according to a meta-analysis[29]. A recent study of DBE in LVAD recipients who presented with overt OGIB found that the diagnostic yield of DBE was 69% when DBE was performed within 24 h of initial presentation[30]. We suspect that VCE would have a similar diagnostic yield if it is performed within 24 h of overt OGIB. In our study, VCE was performed on average 6.2 d after admission- after coagulopathy was corrected and after other endoscopic procedures failed to identify the cause of OGIB. Using logistic regression, we found that higher INR on presentation was associated with a higher probability of positive VCE. If OGIB is highly suspected in an LVAD patient with a supra-therapeutic INR on presentation, expediting VCE, possibly before coagulopathy is corrected and other endoscopies performed, may improve the diagnostic yield of VCE in this population.

In a general population with OGIB, > 80% of the bleeding sites are in the small bowel[11]. The most common location of positive VCE finding in our LVAD patients with OGIB is also small bowel (75%). The most common finding of positive VCE in our study was small bowel bleeding without the cause of bleeding identified, whereas small bowel angiodysplasia was the second most common finding. The source of bleeding can be difficult to identify by VCE in the setting of active bleeding since blood can obscure visualization and VCE cannot clear the visual field with water irrigation. Nevertheless, we surmise that angiodysplasia is the most likely cause of GIB in those with small bowel bleeding without lesions identified given that it has been shown to be the most common cause of GIB in LVAD patients[6,7,22]. Indeed, 3 patients with small bowel bleeding without lesion identified on VCE were later found to have bleeding small bowel angiodysplasias on subsequent enteroscopies. Of note, in our study 25% of positive VCE findings were within reach of EGD or colonoscopy, underscoring the elusive nature of these lesions and the importance of repeat examinations if necessary.

Our study is the first to report recurrent GIB rate in LVAD recipients after VCE. Although LVAD is thought to predispose patients to GIB through various mechanisms, the recurrent GIB rate (53%) in our LVAD population is similar to OGIB in non-LVAD patients[31]. Furthermore, there was no difference in the rate of recurrent GIB, the number of endoscopies performed after VCE, and the mortality rate, whether VCE studies were positive or negative. Even among patients with a positive VCE, further endoscopic intervention did not affect recurrent bleeding- the recurrent bleeding rate was 50% regardless of endoscopic intervention (Table 3). On the other hand, the negative VCE study in our cohort was not associated with lower recurrent bleeding rate. It is possible that OGIB may merely be a reflection of the underlying condition or hemodynamics of LVAD patients, and endoscopic procedures do not necessarily impact their natural courses. Endoscopic evaluation or therapy in this population therefore should be used judiciously, especially given that medical therapy with thalidomide or Octreotide has recently been reported to be effective in LVAD recipients with OGIB[32,33]. Further study is required to identify the characteristics of patients who may benefit from endoscopic intervention vs supportive care.

Of the 12 patients with positive VCE studies, only 6 patients had subsequent endoscopic procedures because the other 6 patients had no further GI bleeding and their hemoglobin had stabilized, again indicating a dissociation of VCE findings with patients’ clinical courses. Given that only 50% of patients with positive VCE had persistent GIB requiring further endoscopic intervention, one possible approach to treat LVAD recipients with OGIB based on our finding is to defer VCE and endoscopies until persistent GIB is observed. The current health care environment creates tremendous pressure on hospitals to shorten patients’ stay and expedite diagnostic procedures and treatment. For certain patients in this population, however, it may be more cost-effective to observe and provide supportive medical care without endoscopies. Future studies are needed to test this hypothesis and to determine which subset of LVAD recipients with OGIB would benefit from VCE and subsequent endoscopies.

None of LVAD recipients in this study died from GIB, consistent with a prior report[34]. All 4 patients who received heart transplantation in our study did not have recurrent bleeding episodes during the study period, which is similar to the result in a recent systematic review that reported 12 patients without recurrent GIB after heart transplantation[7]. All together, the data indicates that the changes of physiology and hemodynamics associated with LVAD are the causes of GIB, and this process and GIB can be reversed when LVAD is removed. Consistent with this concept, a previous study found that all LVAD recipients had reduced high molecular weight von Willebrand factor multimers which were normalized after patients received heart transplants[35].

Our study has limitations. Given that this is a retrospective study from a single tertiary medical center, it has weaknesses similar to other retrospective studies and should be carefully interpreted or generalized. In addition, although this study provides the largest series of VCE in LVAD recipients with OGIB, the sample size of 30 is still limited. We might underestimate the true rate of recurrent bleeding because the follow-up duration may not be long enough. Lastly, in our study VCE were interpreted by several gastroenterologists and the management of patients was not standardized. These variations may affect the results of our study.

In conclusion, our study shows that VCE can be safely performed in LVAD recipients with OGIB and with a 40% diagnostic yield. However, the results of VCE and the subsequent management driven by VCE did not affect the rate of recurrent GIB. Endoscopic intervention thus should be used judiciously in this patient population. An observation-and-supportive care approach could be an alternative way to treat LVAD patients with OGIB. Future studies should answer the questions of what subset of LVAD recipients with OGIB would benefits from endoscopic therapy, and what the most cost-effective way is to take care of this challenging group of patients.