Short-and long-term results of transcatheter embolization for massive arterial hemorrhage from gastroduodenal ulcers not controlled by endoscopic hemostasis

1Department of Interventional Radiology and Endovascular Therapy; 2Department of Gastroenterology and Hepatology; 3Department of Abdominal and Oncological Surgery, University of Dijon School of Medicine, Bocage Teaching Hospital, Dijon, France Correspondence: Dr Romaric Loffroy, Department of Interventional Radiology and Endovascular Therapy, University of Dijon School of Medicine, Bocage Teaching Hospital, 2 bd Maréchal de Lattre de Tassigny, BP 77908, 21079 Dijon cedex, France. Telephone 33-380-293-686, fax 33-380-293-243, e-mail romaric.loffroy@chu-dijon.fr Received for publication July 23, 2008. Accepted September 13, 2008 A bleeding is the leading complication of peptic ulcer disease, which contributes to approximately one-half of the cases of upper gastrointestinal bleeding (1,2). Bleeding from peptic ulcers is fatal in 5% to 10% of patients, a proportion that has not changed substantially over the past two decades (3-5). The bleeding persists or recurs in approximately 20% of cases requiring emergency medical treatment or endoscopic intervention (1,5). When these methods fail, either surgery or transcatheter arterial embolization must be performed (6). Bleeding from peptic ulcers occurs selectively in older patients with serious underlying conditions, in whom surgery is associated with a high risk of mortality (40%) and original article

morbidity (45%) (7).Embolization is used at the Bocage Teaching Hospital, Dijon, France, in patients with failed endoscopic treatment for severe peptic ulcer bleeding.
The objective of the present study was to report the safety, efficacy, and short-and long-term outcomes of selective embolization in 60 consecutive patients with failed endoscopic treatment for severe peptic ulcer bleeding between October 1999 and January 2008.

Patients
A retrospective chart review was conducted to identify patients admitted to the Department of Interventional Radiology and Endovascular Therapy for transcatheter arterial embolization to treat massive peptic ulcer bleeding (defined as a need for more than four units of blood/24 h), causing hemodynamic instability (defined as clinical hypovolemic shock requiring volume replacement).The present study was approved by the institutional review board.Given the retrospective design, informed consent was not required.
Sixty patients were identified, 41 men and 19 women, with a mean age of 69.4 years (range 29 to 95 years).The operative risk was high in most of the patients as a result of advanced age (older than 70 years in 61.7% of patients and older than 80 years in 23.3%) and serious comorbid conditions (one or more in 90% of patients; two or more in 65% of patients, respectively).Furthermore, 58.3% of patients were on anticoagulant and/or anti-inflammatory treatment at hospital admission.
Emergency endoscopy showed gastric or duodenal peptic ulcers (n=55) or Dieulafoy lesions (n=5) with active bleeding (type 1; n=33), recent bleeding (type 2; n=22), or no bleeding (type 3; n=5) according to Forrest's classification (8).Endoscopic hemostasis involved noradrenaline injection or placement of metal clips around the bleeding site.Mean time from hypovolemic shock onset to referral was 2.3 days.At the time of embolization, the patients had received a mean of 11 packed red blood cell units (range two to 40 packed red blood cell units).Mean hemoglobin level at admission was 65 g/L (range 30 g/L to 107 g/L).

Technique
All procedures were performed by one of three experienced interventional radiologists.Selective angiography of the celiac trunk, gastroduodenal artery (GDA) and superior mesenteric artery was performed using a 5-Fr Simmons-type 1 catheter (Cook Inc, USA) inserted through a 6-Fr sheath placed in the common femoral artery.In three patients, bleeding was from the right (n=2; extravasation) or left (n=1; no extravasation) gastric artery, which was too thin and tortuous to allow catheterization; the 57 other patients were treated with embolization.Extravasation of contrast medium or a false aneurysm-like lesion was noted at the bleeding site in 38 patients, of whom 17 were managed with embolization after superselective arterial catheterization using a 2.9-Fr coaxial microcatheter (Terumo Medical, Belgium).The 19 remaining patients underwent 'sandwich' embolization of the GDA.The 22 patients who had no identifiable bleeding site underwent blind embolization based on the endoscopic findings, using the technique chosen by the interventional radiologist.The left gastroepiploic artery was embolized through the splenic artery in two patients, and the left gastric artery was selectively occluded in six patients.
Overall, embolization involved the GDA and its branches or the pancreaticoduodenal arches in 49 patients, of whom 15 underwent selective embolization of the bleeding branch (Figure 1) and 34 underwent sandwich embolization (Figure 2) as described elsewhere (occlusion of the GDA trunk on either side of the bleeding site) (9).When the sandwich technique was used, the anterior and posterior superior pancreaticoduodenal arteries were routinely embolized; in three of these 34 patients, persistent extravasation prompted selective embolization of the anterior and posterior inferior pancreaticoduodenal arteries through the superior mesenteric artery.Embolization was usually achieved using 0.089 cm steel coils for the sandwich method or 0.046 cm soft platinum multiplecurled microcoils for the superselective method (Cook, USA), mechanically disrupted gelatin powder sheet (CuraMedical, Netherlands), or cyanoacrylate surgical glue (GEM SRL, Italy) mixed with ultrafluid lipiodol (Therapex, E-Z-EM, Canada) in a 1:3 ratio.Microspheres ranging in size from 500 µm to 700 µm in diameter (Biosphere Medical, France) were used in a few patients.In 28 patients, a single material was used (coils, n=15; glue, n=8; gelatin, n=3; microspheres, n=2).In 13 patients, both coils and gelatin sponge were used to occlude the GDA.The coils were released distally to prevent retrograde filling through the right gastroepiploic artery.A gelatin sponge was placed in the arterial trunk, and coils were then released in the proximal part of the GDA.Coils and glue were used in nine patients, gelatin sponge and microspheres in three patients, glue and gelatin in one patient, and coils and microspheres in one patient.In two patients, three embolic agents were used in combination.All agents were released near the bleeding site until angiographic extravasation stopped and/or occlusion of the targeted vessel was achieved, as shown by fluoroscopic guidance.The mean volume of contrast medium was 190 mL per patient, and the mean total procedure time was 70 min.

Follow-up
Follow-up information was available for all patients.The mean length of hospital stay was 18 days (range one to 98 days) after embolization.Mean follow-up was 22 months (range one day to 103 months).A physical examination was performed during the hospital stay and one month after hospital discharge as part of routine patient care.Data on subsequent events were collected during telephone interviews of patients during follow-up by attending physicians.
Procedural success was defined as the absence of extravasation on postembolization arteriography; early clinical success as absence of rebleeding within one month after embolization; and long-term clinical success as absence of rebleeding throughout follow-up.Rebleeding was defined as bleeding with a greater than 20 g/L decrease in the hemoglobin level and/or failure of conservative medical treatment; early rebleeding was defined as bleeding occurring within 30 days after embolization.Complications were classified as major complications if they required surgery and/or prolonged hospitalization and as minor complications otherwise.
Table 1 provides a summary of patient management according to angiographic findings.In 22 patients, no extravasation was identified by angiography but the artery supplying the endoscopically identified bleeding site was embolized in 21 patients.In three patients, the left gastric artery was occluded using resorbable or nonresorbable particles, and in 18 patients, treatment of the GDA was achieved using the sandwich technique (n=15) or by selective embolization (n=3).Selective embolization of the bleeding vessel was performed in 17 additional patients.The 19 remaining patients with angiographic extravasation underwent sandwich embolization of the GDA.
The main results are summarized in Tables 2 and 3. Embolization was consistently effective in stopping the bleeding.However, 16 patients required further treatment within 72 h for recurrent bleeding: eight were managed endoscopically; three, all of whom initially underwent sandwich occlusion of the GDA, were treated with embolization of the inferior pancreaticoduodenal artery; and five underwent duodenotomy.The second treatment was initially successful in all 16 patients but three rapidly experienced fatal rebleeding from the same site.In total, gastrointestinal surgery was needed in eight of the 60 patients (13.3%).Initial embolizations were unsuccessful in 13 of 36 patients (36.1%) who were taking anticoagulant and/or anti-inflammatory drugs at admission and in three of 21 patients (14.3%) without these medications (Table 4).
No serious embolization-related complications occurred.Four patients experienced minor complications.Transient liver enzyme elevation occurred in a patient with proximal celiac trunk occlusion and retrograde filling of the GDA through the inferior pancreaticoduodenal artery, in whom catheterization beyond the bleeding site was not feasible; after verification of portal venous flow, resorbable particles were used to achieve retrograde occlusion of the inferior pancreaticoduodenal artery, GDA and common hepatic artery.A transient increase in serum amylase levels without symptoms was noted in another patient.In two patients, a microcoil migrated into the common hepatic artery during embolization, with no detectable effects.Angiography was responsible for two major complications.A hematoma in the groin area responsible for hemodynamic instability developed in one patient, who required vascular surgery.In the other patient, who was obese (160 kg), a false aneurysm of the femoral artery was identified a few days after the procedure but produced no symptoms; thrombosis was achieved by mechanical compression for 20 min under Doppler sonography guidance.No ischemic gastrointestinal complications were identified by clinical examination.
Of the 57 patients treated with embolization, 16 (28.1%)died within one month after the procedure from multiple organ failure (n=7); hypovolemic shock due to massive rebleeding

DISCUSSION
Endoscopy is the first-line method for diagnosing and treating actively bleeding peptic ulcers because its success rate is high.Endoscopic hemostasis is effective in most patients, thereby obviating the need for surgery, a valuable benefit in patients at high risk for surgical complications as a result of advanced age or comorbid conditions.When endoscopy fails, surgery is associated with high mortality rates of 20% to 40% (7,10).Endovascular embolization is an alternative to surgery in highrisk patients who fail endoscopic treatment.However, most studies of embolization were conducted in small patient populations who had a wide variety of diseases including peptic ulcers, malignant tumours, vascular malformations, traumainduced lesions and post-inflammatory lesions (11).We focused on patients with bleeding from gastric or duodenal peptic ulcers (n=55) or Dieulafoy lesions (n=5).The bleeding site was identified endoscopically in all of our patients.In a study of 28 patients with bleeding duodenal ulcers (12), celiac arteriography showed extravasation in only 11 patients (39%).We visualized the extravasation site in a higher proportion of patients (38 of 60 [63.3%]), probably because we injected both the celiac trunk and the GDA, followed by the superior mesenteric artery.These results are in concordance with most published series (13)(14)(15) reporting a 60% to 80% sensitivity rate of arteriography in the detection of active upper bleeding.Furthermore, our study establishes that embolization can be performed successfully even when angiography fails to visualize the extravasation site.Previous reports (16)(17)(18) have found empirical embolization based on endoscopic findings, in the absence of contrast extravasation, to be helpful in controlling hemorrhage, with no difference between patients with negative and positive angiography results, confirming the practice of endoscopy-directed blind embolization.All but one of the 16 rebleeding events in our study occurred after embolization of the GDA and its branches; three occurred after selective embolization of a pancreaticoduodenal branch and 12 after sandwich embolization of the GDA not including the inferior pancreaticoduodenal artery.The extensive collateral circulation in the duodenum probably increases the risk of rebleeding after embolization in this territory.In addition, rebleeding after selective embolization of a branch of the GDA or a pancreaticoduodenal artery alone, or sandwich embolization of the GDA alone not including the inferior pancreaticoduodenal artery, may be ascribable to retrograde flow into the pancreaticoduodenal arches and right gastroepiploic artery.Complete sandwich occlusion of the GDA including the superior and inferior pancreaticoduodenal arteries has been suggested to decrease the rebleeding rate, even when extravasation is not visualized (19,20).Furthermore, it seems that embolization was more likely to be unsuccessful in patients under anticoagulant and/or anti-inflammatory treatment (36.1%) than in those without (14.3%),confirming that all efforts should be made to stop it early in the course of hemodynamically unstable bleeding patients.
In our study, six of the 15 deaths (40%) recorded within one month after embolization occurred among the 15 patients who underwent selective embolization of a bleeding branch of the GDA.The clinical success rate in this subgroup was higher than in the subgroup of 34 patients treated with sandwich GDA embolization (80% versus 64.7%, respectively).Of these 34 patients, seven (20.6%) died.Two factors may contribute to explain the high mortality rate (six of 15 patients) after selective embolization.First, patients who underwent selective embolization had angiographic evidence of contrast-medium extravasation that is associated with active bleeding, severe blood loss and hypovolemic shock.Second, patients treated with selective embolization were slightly older (mean age 73.1 years) than those treated with sandwich embolization (mean age 68.8 years).Therefore, our data cannot be used to determine whether one of these embolization techniques is superior over the other in terms of 30-day survival.
Finally, three patients with acute bleeding from the pyloric artery (right gastric artery) (n=2) or left gastric artery (n=1) were treated surgically because the bleeding vessel was too slender and tortuous to allow catheterization.
Although rates of procedural success (95%) and early clinical success (71.9%) were high in our study, 26.7% of patients died within the first month.Severe hypovolemic shock at admission, advanced age and comorbid conditions probably contributed to this high mortality rate, which simply reflected the high surgical risk in our population.The impact of medications associated with increased bleeding on the one-month mortality rate was not clear in our study.Only three of the 16 deaths were due to early rebleeding.In the only previously published retrospective study (21) comparing outcomes of embolization (31 patients) and surgery (39 patients) after failed endoscopic treatment of bleeding peptic ulcers, no differences were found regarding rates of rebleeding, surgery or mortality.Unfortunately, the postprocedural morbidity rate was not compared between the two techniques.Few data are available regarding postsurgical morbidity, most notably complications related to the surgical method and infectious complications.In a retrospective study of 49 patients (7), the morbidity rate was 45%.No cases of bowel ischemia were noted in our study.In addition, major complications developed in only 3.3% of patients and were unrelated to vascular occlusion.Puncture-site injuries were the main complications.Using small introducers and catheters (4-Fr), delaying introducer retrieval until the coagulation parameters are normalized, or using vascular closure devices might help to decrease the rate of puncture-site injuries.The low morbidity rate is a major advantage of embolization over surgery.In addition, embolization is less invasive and less aggressive.The most troublesome long-term complication was duodenal stenosis (25%) in a study of 28 patients ( 22) followed for at least five years after embolization for a bleeding duodenal ulcer.Symptomatic duodenal stenosis was not reported in any of our patients, even those treated with surgical glue, probably because the follow-up was too short.Routine investigations may have identified cases of asymptomatic duodenal stenosis.The potential role for embolization in the occurrence of duodenal stenosis is unclear because duodenal stenosis can occur as a complication of peptic ulcer disease.After the first month, none of our patients experienced severe rebleeding requiring endoscopy or invasive treatment during follow-up, which lasted 22 months on average.Although the results of the present retrospective study should be interpreted with caution, they indicate that arterial embolization performed by experienced interventional radiologists is effective in controlling bleeding from gastroduodenal ulcers, even when extravasation is not visualized by angiography and does not cause ischemia.At our hospital, this technique is the treatment of choice after failure of endoscopic hemostasis.Embolization usually obviates the need for surgery.

Figure 1 )
Figure 1) Arteriogram images of bleeding from a bulbar duodenal ulcer in a 76-year-old man.A, B Arteriogram showing contrast medium extravasated from a slender branch of the gastroduodenal artery into the duodenum (arrows).C, D After microcatheterization, selective glue embolization (radiopaque because of associated lipiodol (arrow) preserving the gastroduodenal artery ensured control of the bleeding, with no early or late recurrences

Figure 2 )
Figure 2) Typical sandwich embolization in a 75-year-old man with bleeding from a postbulbar duodenal ulcer at endoscopy.A,B Angiography before embolization: No evidence of active bleeding; C Image after coil embolization of the distal and proximal gastroduodenal artery (with gelatin sponge placed in the arterial trunk), including the anterior and posterior superior pancreaticoduodenal arteries and the right gastroepiploic artery, to prevent retrograde flow (arrows).No ischemic complications were reported