Severe atraumatic bleeding, especially of gastrointestinal origin, is an important and frequent complication of CF-LVAD therapy; the management requires a multidisciplinary approach. The history and physical exam should suggest the initial site of bleeding: patients with hematemesis or melena should be considered to have upper gastrointestinal bleeding and should undergo upper endoscopy (Gurvits et al. 2017), while patients with hematochezia can be investigated for suspected lower gastrointestinal bleeding, with the initial examination being colonoscopy (Carrozzini et al. 2019b). When endoscopy is inconclusive, other exams are necessary. CT can be used to identify LVAD-related bleeding complications and it is often diagnostic also for non-gastrointestinal bleeding, such as pericardial, pleural, abdominal wall, retroperitoneal and intracranial hemorrhage (Carrozzini et al. 2019a).
The management of gastrointestinal bleeding in patients with CF-LVAD is predominantly medical, with reduction of anticoagulation, and administration of concentrated red blood cells, octreotide and proton-pump inhibitors, although the data regarding their overall efficacy remain sparse (Sieg et al. 2017). If these treatments are not effective, surgery is considered, even if the surgical and anesthetic risk of these patients is considerably increased by their coagulation defects and by the presence of important comorbidities (Guha et al. 2015). Endovascular therapy is obviously less invasive with less anesthesia risk, and can be performed on patients unfit for surgery, as is often the case in those with cardiac disorders; moreover it is a repeatable and poorly invasive procedure.
As previously mentioned, these patients have a high risk of bleeding due to antithrombotic therapy and other pathophysiological mechanisms due to the continuous flow of the cardiac device; antithrombotic therapy can be corrected if overdosed, but usually not interrupted. This obviously makes the endovascular approach advantageous compared to the surgical one, because it requires just a small arteriotomy (4 or 5 French in our experience) without increasing the risk of further anaemia via access site wound.
The inability to suspend anticoagulant therapy affects the choice of embolics agents as some (eg coils) may not be as effective as others (eg gelfoam, NBCA); in most cases, in fact, we have chosen to combine the injection of permanent particulate embolization material with microcoils.
Few studies are available about the effectiveness of interventional radiology procedures in CF-LVAD patients: a case of Morito and others (Morito et al. 2014) reports the successful endovascular treatment of a patient with a cerebral hemorrhage caused by ruptured cerebral aneurysms, detected by angiography; the clinical study by Metha et al. (Mehta et al. 2015) reports that patients with LVADs presenting with gastrointestinal bleeding have fewer successful embolizations and a higher rate of clinical failure than the control group of patients.
In our experience the diagnostic angiography, performed after a positive contrast enhancement-CT, identified the bleeding site in 100% of cases. In 6/7 (85,7%) patients the embolization procedure performed immediately after the diagnostic angiography was therapeutic, with normalization of hemoglobin values and rapid recovery; only 1/7 (14,3%) patient was again subjected to embolization 21 days later, with a successful procedure. No patient presented complications.
Of course there are potential problems with endovascular approach. At the end of the procedure it is possible that, due to the anticoagulation, difficulties occur in the closure of the arteriotomy with the consequent need for closing devices: however in our experience the small caliper of the arteriothomy, a prolonged manual compression on the access site and the immobilization of the limb for the next 24 h were sufficient for the closure of the arteriotomy. In the case of intermittent bleeding it can happen that during the arteriography the spread of contrast medium is not evident: in our series the superselective arteriography of the branches previously identified at the CT-angiography always identified the sites of the bleeding, but in the case where this eventuality occurs we believe that a multidisciplinary approach with clinicians and surgeons is needed before proceeding with prophylactic embolization. Finally, as in all cases of gastrointestinal embolization, one of the most feared complications is the bowel infarction with subsequent need for surgical resection, further complicated in these cases by the high risk of bleeding and anesthesiology: fortunately this did not occur in our patients; however, we believe that the superselective embolization of the only branches responsible for bleeding reduces this risk.
Certainly not all sources of bleeding in CF-LVAD patients are susceptible to endovascular treatment. Mediastinal bleeding often occurs in the first week after implantation of the device, in most cases due to surgical complications (Nascimbene et al. 2016) and for this reason they are better controlled with the surgical approach; non-surgical bleeding, instead, usually arises later, in distant site bleeds and at the site of arteriovenous malformations: these are the cases in which the endovascular approach should be considered as the first line treatment.