Is There a Role for Empiric Gastroduodenal Artery Embolization in the Management of Patients with Active Upper GI Hemorrhage?

Abstract

Purpose

To assess the relative efficacy of empiric gastroduodenal artery (GDA) embolization in reducing recurrent hemorrhage compared to image-guided targeted embolization.

Methods

Data were retrospectively collected for consecutive patients who had catheter angiography for major upper gastrointestinal hemorrhage from May 2008 to November 2010 (n = 40). The total number of cases were divided into two main groups according to angiographic findings: those that demonstrated a site of hemorrhage on catheter angiography (group 1, n = 13), and those where the site of hemorrhage was not identified on catheter angiography (group 2, n = 27). Group 2 was then further divided into patients who received empiric embolization (group 2a, n = 20) and those who had no embolization performed after angiography (group 2b, n = 7).

Results

The technical and clinical success rates for embolization in groups 1 and 2a were, respectively, 100 vs. 95 %, and 85 vs. 80 %. There was no statistical significance in the recurrent hemorrhage rate, reintervention rate, or 30 day mortality between targeted and empiric embolization groups. There were no complications attributed to embolization within this study cohort.

Conclusion

Cases of duodenal-related major upper gastrointestinal hemorrhage where no embolization is performed have poor outcome. Empiric embolization of the GDA in patients with major upper gastrointestinal hemorrhage refractory to endoscopic treatment appears to be a safe and effective treatment, with low reintervention rates and good clinical outcome comparable to patients where the site of hemorrhage is localized and embolized with computed tomographic angiography or catheter angiography and embolized.

Introduction

Endoscopy is the primary intervention for the diagnosis and treatment of upper gastrointestinal hemorrhage (UGIH). The majority of patients with nonvariceal upper gastrointestinal hemorrhage (NVUGIH) can be effectively treated and will stop bleeding with endoscopic treatment [1]. There is, however, a subset of patients in whom endoscopy fails to arrest the UGIH or who are considered to be at high risk for further major hemorrhage. Although traditionally surgery has been considered as the next treatment option for these patients, many of these patients have considerable comorbidity that puts them at high risk for surgery [2]. In our institution, embolization is considered as the first-line management of UGIH after failed endoscopy/endoscopic management to control the hemorrhage. Surgery is mostly reserved as the last treatment modality for cases of bleeding despite embolization. Localization of the point of hemorrhage on either computed tomography angiography (CTA) or catheter angiography allows image-guided embolization. However, when CTA or catheter angiography fails to identify the artery that is the culprit, in our institution, this has been extended to embolization guided by the site of hemorrhage identified on endoscopy or known from previous surgical findings. In most cases, the site of hemorrhage is from the gastroduodenal artery (GDA) territory and results from a duodenal ulcer. This form of embolization has been termed empiric embolization. The purpose of this study was to assess the relative efficacy and safety of empiric GDA embolization compared to targeted embolization.

Material and Methods

Institutional board approval and ethical approval were not required for this retrospective study.

Terminology

The terminology used in the study was as follows. Targeted embolization refers to embolization of the culprit artery that can be identified on angiography by active contrast extravasation. Empiric embolization refers to patients where the bleeding territory is known from previous endoscopic or surgical findings but was not identified by catheter angiography, and who undergo embolization of the assumed arterial territory [which was always the GDA and/or the inferior pancreaticoduodenal artery (IPDA) in our series]. Technical success was defined as absence of contrast extravasation on angiography with successful occlusion of the intended vessel and its branches with no demonstrable collateral flow, or, in the case of empiric GDA, complete embolization of this vessel. Clinical success was defined as a combination of technically successful embolization with an improvement in the patients hemoglobin, no evidence of further hemorrhage (or decrease in hemoglobin) after embolization, and no requirement for further intervention within the first 4 weeks of embolization.

We conducted a retrospective review of consecutive patients who underwent catheter angiography for major NVUGIH at our institution between May 2008 and November 2010. Data were collected from a secure Picture Archiving System (PACS) database, electronic medical records, and the interventional radiology database. A total of 41 procedures were performed on 40 patients with NVUGIH of duodenal origin.

The total number of cases were initially divided into two main groups according to angiographic findings: those that demonstrated a site of hemorrhage on catheter angiography (group 1, n = 13), and those where no bleeding point/active hemorrhage was demonstrated on angiography (group 2, n = 27). Group 2 was then further divided into patients who underwent empiric embolization (group 2a, n = 20) and those who had no embolization performed after angiography (group 2b, n = 7).

The following variables were obtained: age, sex, endoscopic diagnosis, hemodynamic stability/instability (defined as a systolic blood pressure <100 mmHg, heart rate >100 beats/min or clinical shock resulting from blood loss), coagulopathy, cause of bleeding (e.g., duodenal ulcer, tumour, trauma, iatrogenic causes), angiographic/embolization findings, and 30 day mortality. Patients were described as having coagulopathy if they had a partial thromboplastin time >40 s and/or platelet count of <80,000/μL or international normalized ratio or greater than 1.5 [3]. Characteristics of the patients are summarized in Table 1.

Table 1 Summary of the demographics and clinical characteristics of study patients ^a

Embolization Technique

All patients underwent an initial diagnostic angiography after they provided appropriate informed consent. Patients who had demonstrable active hemorrhage underwent direct selective targeted embolization of the culprit vessel (Fig. 1) with either microcoils alone or microcoils and either gelatin/gelfoam sponge or polyvinyl alcohol (PVA) particles (Table 1). Patients who demonstrated no active bleeding radiologically, but in whom endoscopic localization demonstrated or suggested an arterial territory (with or without placement of a metallic clip), underwent empiric embolization (Fig. 2), apart from a small number (group 2b) where no embolization was performed as a result of operator preference. The GDA was embolized for all empiric cases for presumed or known duodenal ulcers. We had no cases of empiric embolization of the left gastric artery for presumed gastric bleeding in our study cohort during the study period.

Fig. 1

Angiographic images demonstrating targeted embolization of the GDA. A Selective digital subtraction angiogram (DSA) of the GDA (white arrow) showing active contrast extravasation (arrowhead) from a vessel arising from the GDA (dotted arrow). B DSA after coil embolization of the distal and proximal GDA (black arrows) and gastroepiploic artery (dotted black arrow) demonstrating successful embolization with no further contrast extravasation of contrast medium

Fig. 2

Angiographic images demonstrating empiric embolization of the GDA. A Selective angiography of the GDA (white arrows) before embolization showing no evidence of active bleeding. B Selective angiography of the GDA after coil embolization of the distal and proximal GDA (black arrows), including the anterior and posterior superior pancreaticoduodenal arteries and the right gastroepiploic artery, to prevent retrograde flow (white arrowheads)

At our institution, we perform embolization with a selective 4F/5F catheter: SOS Omni catheter (Boston Scientific, France), Sidewinder catheter (Cordis, UK), or Cobra/Glide-Cobra catheter (Cook, Ireland) with or without the use of a 2.7F Progreat microcatheter (Terumo, Belgium). The GDA is selectively cannulated and then packed from distal to proximal with fibered 0.035 or 0.018 in. microcoils (either Nester, Cook, UK; Tornado, Cook, UK; or Azur Hydrocoils, Terumo, Japan) if the microcatheter is used, with or without the use of gelatin sponge. The number and selection of embolized arteries and the type of embolic agent used are at the discretion of the interventional radiologist performing the procedure. However, in the cases of empiric embolization, we routinely place the catheter within the right gastroepiploic artery, where the coils are introduced from this point as the catheter is withdrawn into the proximal GDA [2, 3]. If necessary, the anterior and posterior pancreaticoduodenal arteries are also embolized with coils (Fig. 2). Angiography of the superior mesenteric artery and IPDA is routinely performed after GDA embolization to check for further hemorrhage from these territories, unless the precise site of hemorrhage has been identified and arrested after initial access from the GDA.

Statistical Analysis

Fisher’s test was used for analysis for nonparametric data. Findings with a p value of <0.05 were considered to be significant.

Results

Patient Demographics

A total of 40 patients underwent a total of 41 catheter angiograms. There were 32 male and eight female patients, with a median age of 71 years (range 39–91 years). All patients had UGIH during their admission or immediately before angiography (defined as a decrease in blood pressure to <100 mmHg systolic, tachycardia >100 beats/min, and requiring blood transfusion of ≥4 U of red blood cells during the previous 24 h). However, 34 % (17 of 50) were considered hemodynamically stable immediately before angiography/embolization but had been previously unstable, with continued hemorrhage refractory to endoscopic control warranting embolization. Patients with multiorgan failure were defined as those who were acutely unwell with altered organ function (involving two or more organs) whose homeostasis could not be maintained without intervention. Those who had three or more of the following were assumed to have multiorgan failure: cardiovascular compromise, steroid-dependent chronic obstructive pulmonary disease, ventilator-dependent lung disease, dialysis-dependent renal failure, cirrhosis, and sepsis [2] (Table 1).

Sites of Hemorrhage

The majority of patients (30 of 40, 75 %) had an esophagogastroduodenoscopy (OGD) before angiography. The remaining 10 patients went on to catheter angiography because they either had a bleeding point demonstrated on CTA (8 of 10) or had previous upper gastrointestinal (GI) surgery and the bleeding point was presumed (2 of 10). CTA was performed rather than primary OGD in some patients (8 of 10) because no endoscopy was available at that time, and it was clinically decided to perform a CTA instead. In the three patients who had previous surgery and the bleeding site was presumed, all were too unstable for either OGD or CTA and therefore subsequently went on to have a catheter angiography. The causes of hemorrhage in the 40 patients were as follows: duodenal ulcer (n = 31), iatrogenic after upper GI surgery (n = 5), GDA pseudoaneurysm from pancreatitis presenting with hematemesis (n = 1), and unknown (n = 3) (Table 2).

Table 2 Cause of hemorrhage in the study groups

Endoscopic and Angiographic Data

In the 34 patients in whom embolization was attempted, it was successfully achieved in 33. All cases had embolization of the GDA (with or without embolization of the IPDA). The overall technical success rate for all procedures was 97 % (33 of 34), and the clinical success was 86 % (27 of 33) (Fig. 3; Tables 1, 3).

Fig. 3

Flow chart demonstrating the main findings of our study. CMA catheter mesenteric angiography, CTA computed tomography angiography

Table 3 Outcomes summary

Angiographically Targeted Embolization

Sixty-nine percent of the patients in group 1 who underwent targeted embolization (9 of 13) were hemodynamically unstable at the time of angiography. Nine of 13 patients had an OGD before angiography. Four patients demonstrated refractory hemorrhage after endoscopic treatment, and in five, endoscopy could not treat the lesion as a result of copious amounts of blood or an inability to identify the cause. In the remaining four patients who did not have endoscopy, two underwent initial CT angiography, and two went directly to catheter angiography because their condition was too unstable to permit OGD or CTA.

As defined by their inclusion in this group, contrast medium extravasation was demonstrated during angiography, and selective embolization was technically successful in all patients. Two of 13 patients experienced recurrent hemorrhage after embolization. One patient required a partial gastrectomy as a result of hemodynamic compromise, and the second patient who was stable was managed conservatively, with no further episodes of bleeding or need for intervention. The cause for recurrent hemorrhage after embolization was unknown. Clinical success in group 1 was 85 % (11 of 13).

Empiric Embolization Group

The majority of patients who underwent empiric embolization (group 2a) were hemodynamically unstable (17 of 20, 85 %) before embolization. Fourteen of 17 unstable patients had prior OGD. Nine of the 17 unstable patients had refractory hemorrhage after technically successful endoscopic treatment. In five of the 17 unstable patients, endoscopic treatment was not possible because of large amounts of blood or severe hemodynamic compromise during the procedure. The remaining three unstable patients had previous surgery for duodenal ulcer or underwent a Whipple procedure, and their condition was too unstable to permit OGD or CTA; therefore, they underwent empiric embolization because the bleeding site was presumed. The remaining three patients who had prior OGD and underwent prophylactic embolization were hemodynamically stable. Two experienced rebleeding after endoscopic treatment, and one could not be treated endoscopically because of the presence of blood. In total, 17 of the 20 patients who had empiric embolization had prior OGD.

In summary, the overall reasons for empiric embolization in the 20 cases were as follows: 11 had refractory hemorrhage after endoscopic treatment, six could not be treated endoscopically but a bleeding territory was described, and three had undergone previous upper gastrointestinal surgery and thus a bleeding site was assumed.

Four of 21 patients who were treated with empiric GDA embolization had refractory hemorrhage. Three had undergone prior endoscopic treatment (two continued to bleed despite endoscopic treatment, and one could not be managed endoscopically as a result of copious amounts of blood), and the condition of the fourth was too unstable to permit intervention; therefore, OGD was not performed. One patient whose condition was unstable underwent a secondary embolization procedure that demonstrated extravasation from the GDA stump and from the IPDA. This was successfully embolized with coils and PVA with no further hemorrhage. The initial embolization procedure had been performed with microcoils alone within the mid segment of the GDA; the IPDA had not been embolized.

In the remaining three patients who experienced refractory bleeding, the initial empiric embolization procedure had been performed by completely packing the GDA with coils alone from distal to proximal in two, and with coils and PVA in one. In all three cases, no embolization of the anterior or posterior pancreaticoduodenal arteries had been performed. Two patients were subsequently taken to surgery (for undersewing of their duodenal ulcers) because their condition was considered unstable; one died from multiorgan failure before further intervention was possible. Three of the four patients with refractory hemorrhage after empiric GDA embolization were unstable before embolization with abnormal clotting both before and after embolization. The clinical success in group 2a was 80 % (17 of 21).

No-Embolization Group

In seven patients in whom no embolization was performed (group 2b), one patient had a technically failed attempted empiric embolization as a result of a failure to cannulate the GDA. Two patients in group 2b had massive further episodes of hemorrhage and subsequently died. Both were hemodynamically unstable before and after angiography, and both deaths occurred relatively early in our experience. The single patient who had a failed embolization went on to successful undersewing of the duodenal ulcer.

There was no statistical significance in the recurrent hemorrhage rate between the targeted embolization (2 of 13) and empiric GDA embolization (4 of 20) groups (p = 1.0). The clinical success in both groups equates to the rebleeding rate within our study cohort because after patient discharge, there was no readmission of the patients for recurrent hemorrhage during our study period. There was a slightly higher reintervention rate in group 2a compared to group 1 (15 vs. 8 %), but this was not statistically significant. All-cause 30-day mortality was 20 % (8 of 40). Thirty-day mortality, excluding unrelated causes, was 18 % (7 of 40) (i.e., excluding a single cerebral hemorrhage that occurred after a fall). The mortality rate for groups 1 and 2a were similar, with no statistical significance (Table 3).

When we compared all embolization procedures (groups 1 and 2a) with patients who were not embolized, we found a higher recurrent hemorrhage rate and increased mortality in group 2b (18 vs. 29 %), although this was not found to be statistically significant, which may reflect the small sample size of this group.

Discussion

Endoscopic treatment is the established first-line treatment for active hemorrhage from duodenal ulcers. However, massive hemorrhage can be difficult to control by endoscopic treatment. Historically, patients have undergone surgery if bleeding could not be controlled endoscopically. These patients often have considerable comorbidities that make surgery a high-risk procedure, with mortality rates of 20–40 %. Angiographically targeted embolization has been demonstrated to have promising technical success and clinical success rates, with associated lower morbidity from the procedure itself with reported outcomes of 52–100 % technical success and 44–100 % clinical success [4, 5] (Table 4).

Table 4 Comparison of clinical success, rebleeding rates, need for surgery, and mortality in other studies

The limited data on empiric GDA embolization suggest similar outcomes between “blind” (endoscopic guided) embolization versus targeted embolization (Table 4) [3, 6–10]. Aina et al. [8] performed embolization on 75 patients with 29 undergoing empiric embolization (8 left gastric and 21 GDA embolizations). The study unfortunately did not separate empiric vs. targeted embolization, and therefore, the clinical and rebleeding rates for each group cannot be clearly assessed [8]. Another study by Padia et al. [9] embolized 36 patients (26 GDA and 10 left gastric) with contrast extravasation and 72 patients (64 GDA and 13 left gastric) without contrast extravasation (empiric) on angiography. They demonstrated no difference in outcomes in the targeted vs. empiric groups. Ichiro et al. [5] published a retrospective study in 59 patients with massive bleeding from duodenal ulcers treated with empiric GDA embolization or embolization guided by identification of a bleeding site angiographically. In their cohort, they used coils alone for empiric embolization and coils with or without gelatin sponge for their targeted embolizations. The results are similar to ours, with rebleeding rates of 14 % and a clinical success rate of 83 % compared to 15 and 86 %, respectively, in our cohort. Schenker et al. [2] reported a cohort of 163 patients with multiple causes for GI hemorrhage that also included 52 patients with duodenal ulcers; there was no difference in outcomes between patients in whom targeted embolization was performed vs. empiric embolization, including left gastric artery embolization (Table 4).

Our overall technical success rate (97 %) and clinical success rate (82 %) are comparable to the literature, which describes 52–100 % technical success and 44–100 % clinical success [4, 5]. The single technical failure in our study was due to our inability to catheterize the GDA as a result of a tortuous hepatic artery and a steeply angulated GDA origin. The most common documented cause for technical failure is a stenotic or occluded origin of the common hepatic artery or celiac trunk [4]. This can be predicted by CT angiography before the embolization procedure, and some patients will require surgery if there are anatomical constraints to embolization.

Compared to other studies, our refractory hemorrhage rate of 18 % is lower than most (Table 4). Loffroy et al. [10] had a rebleeding rate of 28 %, of which 15 of the 16 were after GDA embolization. They suggested that this may be due to the abundant collateral circulation supplying the duodenum. The most common causes of refractory bleeding after embolization are either the culprit vessel was not actually embolized, or the perfusion pressure has not been satisfactorily reduced by embolization of the major trunk from which it arises. This can be the result of antegrade flow through the coils (e.g., inadequate embolization due to poor packing of the coils) or collateral flow from adjacent vessels. In our study, one potential cause for refractory hemorrhage after empiric GDA embolization may be the maintenance of a normal perfusion pressure to the diseased duodenum via the supply from the vessels of the inferior pancreaticoduodenal arcade originating from the superior mesenteric artery, which will continue to perfuse the duodenum after complete occlusion of the main trunk of the GDA. Although angiography of these vessels was performed to try to identify a culprit vessel, these vessels were not routinely embolized in the empiric embolization group. We do, however, aim to embolize distally in the GDA to minimize retrograde flow into the GDA territory [3].

Six patients had refractory bleeding after embolization. Four had undergone empiric embolization (three with coils alone and one with coils and PVA), and two had undergone targeted embolization (both with coils alone). Four of the six patients with refractory hemorrhage were unstable and had marked coagulopathy before and after the procedure. All four of these patients had undergone coil placement alone. This suggests that instability of the patient, combined with their coagulopathic nature, may have prevented adequate thrombus and vessel occlusion [4, 8]. This is in agreement with the systematic review by Mirsadraee et al. [4] that notes that in the presence of coagulopathy, the use of coils alone for embolization and the presence of multisystem organ failure are associated with a threefold increased risk of rebleeding and poorer clinical success. Other studies have reported improved outcomes using a combination of gelatin sponge, gelfoam, and PVA, and coils or cyanoacrylate [3, 8]. We therefore recommend the use of particulate embolic material in addition to coils for embolization, particularly in patients with abnormal clotting.

The major complications that occur after transarterial embolization are ischemia, duodenal stenosis, and infarction, with rates varying from 0.04 to 9 % [4, 5, 11, 12]. Other suspected causes are the use of glue or gelatin sponge powder alone [4]. In our study cohort, we had no clinically overt postembolization ischemic complications, although there was limited interval endoscopic follow-up.

Our overall mortality rate (excluding unrelated causes) after targeted and empiric embolization is 18 %, which is comparable to the literature. The in-hospital mortality rate has been shown to be associated with the use of fresh frozen plasma before embolization, rebleeding, coagulopathy, the need for surgery after embolization, and multiorgan failure [2, 6, 13]. Patients with coagulopathy were approximately 3.4 to 10 times more likely to die compared with patients with normal coagulation [3, 4]. However, Schenker et al. [2] demonstrated that despite patients being unwell, those who underwent embolization were 13.3 times more likely to survive than those who did not have an intervention or who had unsuccessful intervention.

Limitations to our study include the retrospective collection of data, so the data were not randomized, and the relatively small number of patients assessed. However, group 1 and group 2a were well matched for age, coagulopathy, multiorgan failure, and instability at the time of embolization. Although there was no statistical difference between the rebleeding rate and the mortality of patients who underwent empiric embolization and those who did not undergo embolization at all, this is reflected by the small patient numbers, especially within the nonembolization group. We point out that the two patients in group 2b who experienced rebleeding both died and were hemodynamically unstable. Therefore, we consider that if no contrast medium extravasation is demonstrated at the time of catheter angiography and if the patient is unstable, then it would be sensible to perform empiric embolization. In patients who are hemodynamically stable, the decision to empirically embolize remains controversial. Larger multicenter randomized trials of maximal endoscopic treatment vs. maximal endoscopic and empiric GDA embolization in stable but high-risk patients appear feasible and may guide future practice.

The decision to carry out endoscopy-assisted/empiric embolization is not straightforward, and a multidisciplinary approach is required with consultant endoscopists, interventional radiologists, and intensive care and upper-GI surgeons. Endoscopic-guided embolization of the GDA in unstable patients with UGIH with refractory bleeding despite endoscopic treatment appears to be a safe and effective treatment with low reintervention rates comparable to those patients where the site of hemorrhage is identified and embolized.