From June 1997 to May 2009, twenty-six patients receiving attempted embolization for acute LGIB in our institute were retrospectively reviewed. There were nineteen men and seven women with the mean age of 69.1 years (range, 19-95 years). The cases with tumor bleeding were excluded in this study. Major co-morbidities included hypertension (n = 13), diabetes mellitus (n = 9), coronary artery disease (n = 3), renal insufficiency (n = 7), history of malignancy (n = 4) and cerebrovascular accident (n = 2). Multidetector computed tomography (MDCT) has been used as the screening modality for detecting LGIB at our hospital since 2002, and positive findings on MDCT were considered as the only indication for angiography. Before 2002, angiography was used to detect the bleeding source in patients with massive LGIB, whereas a nuclear medicine bleeding scan was used to detect chronic or minor bleeding. The use of colonoscopy for the evaluation of LGIB was at the clinicians' discretion. Thirteen in-patients, including five referred from the intensive care unit, and thirteen out-patients referred from the emergency unit were included in this study. Informed consent was obtained from each patient or from his/her family. Retrospective reviews of the medical records and images were approved by the institutional review board of our hospital.

By using the Seldinger's technique, the common femoral artery was punctured, and a 4 Fr RC-1 catheter (Cordis, Miami, FL, USA) was used for superior mesenteric artery catheterization, or a 4.1 Fr RIM catheter (Cook, Bloomington, IN, USA) for inferior mesenteric artery and/or internal iliac artery if necessary. Once the bleeding source was localized angiographically, superselective catheterization was then attempted with either a 2.7 Fr Progreat microcatheter (Terumo, Tokyo, Japan) or a 2.8 Fr Renegade microcatheter (Boston Scientific, Cork, Ireland) with a 0.018-inch guide wire coaxially advanced to the bleeding site. Attempts were made to position the catheter as close to the bleeding site as possible. Embolization was performed only when the catheter could be advanced to or distal to the mesenteric border of the bowel (at least to the marginal artery or in most cases to the vasa recta). Microcoils measuring 2 × 3 mm in dimension or straight microcoils of 5 or 10 mm in length (Cook, Bloomington, IN, USA) were deployed until no further arterial extravasation was seen. In cases in which superselective embolization was not possible due to small vessel tortuosity, or any other reasons, embolization was not performed, and the microcatheter was left in the mesenteric branching artery for the purpose of guidance during later surgical resection.

All the embolization procedures were performed by three interventionalists with 6-20 years of vascular interventional experience. Intravenous prophylactic antibiotics (cephacin, 1 g) before the procedures, and analgesics (fentanyl citrate, 100 µg, Janssen Pharmaceutica N.V.; Beerse, Belgium) during the procedure were administered to the patients. Oxygen saturation, blood pressure, and heart rate and rhythm were monitored in all the patients by noninvasive means.

After embolotherapy, all the patients were monitored closely for clinical signs and symptoms suggestive of ischemic complications (abdominal pain and/or tenderness, peritoneal signs, nausea, diarrhea, or fever) or recurrent bleeding until discharge or death. These clinical findings were supplemented by laboratory studies (white blood count, lactate level, hemoglobin level). Post-embolization CT follow-up (1-3 days after embolization) was done routinely in our early patients, and later only in patients with clinical suspicion of recurrent LGIB or bowel ischemia. The long-term outcome of the patients, specifically, incidence of rebleeding, mortality and procedure related complications were determined by chart review and/or telephone interview where appropriate. Colonoscpic examination was not a routine practice to be performed following embolotherapy in our unit.

Primary endpoints of our study were technical success, recurrent bleeding rate and the rate of minor and major complications, which are reported according to the guidelines of the Society of Interventional Radiology (16). Technical success was defined as immediate cessation of extravasation on repeat angiography at the end of the embolization procedure. Recurrent bleeding was defined as the stabilization of hemoglobin levels that required more than two units of packed red blood cells within 30 days after the procedure. Any lower GI rebleeding occurring later than 30 days after embolization was defined as delayed. Ischemic complications after embolization were defined as bowel ischemia or infarction that required surgery or as late focal bowel strictures requiring surgery or endoscopic repair. Asymptomatic, radiologic, or endoscopic signs of ischemic sequelae after embolization were not considered as postembolization complications.

Secondary endpoints of our study were to evaluate the long-term clinical outcome, including late recurrent LGIB, bowel ischemia and the survival rate.

Kaplan-Meier survival estimates were used to construct a curve for the overall mortality. Date of embolization was used as start date for the follow-up times. The difference in survival time between in- and out-patient groups was tested using Log rank test. A p value < 0.05 was considered statistically significant. All analyses were performed with the Statistical Package for the Social Sciences 17.0 statistical package (Chicago, IL, USA).

Technical success with immediate cessation of extravasation was achieved in twenty-two patients (84.6%). The target vessel of embolization was vasa recta (Fig. 1) in seventeen patients and marginal artery in five. Technical failure was encountered in three patients because of small torturous vessels and subsequently they were referred to surgical resection. The pathological diagnosis was diverticular bleeding in two patients, and ulcer bleeding in one. Another patient had inadvertently induced vascular spasm during selective catheterization but with successful control of the bleeding. Early rebleeding after an initial successful embolotherapy occurred in two patients (7.7%). One was a 61 year-old female patient with corrosive injury after gastrectomy. Contrast extravasation at the proximal jejunum on superior mesenteric artery angiograms was noted. Successful embolization with straight microcoils was obtained initially, but rebleeding from another nearby vasa recta was noted a week later (Fig. 2). Subsequently, she underwent another successful embolization. No recurrent bleeding occurred in this patient, but unfortunately, this patient died 1.5 months later due to multi-organ failure. The other patient had a new focus of bleeding at hepatic flexure of the ascending colon as seen in MDCT images 5 days after the initial successful embolotherapy for the bleeding at the cecum. He received conservative management and died of sepsis 28 days after the embolotherapy.

Post-embolizatin abdominal CT images were followed in thirteen patients. One was for confirmation of a jejunal tumor growth (leiomyoma) adjacent to the embolized bleeding source, two for recurrent LGIB, four for clinical complaints of abdominal pain, and six for abdominal surveillance in our early patients. Normal or mild thickening of the bowel wall at the embolized bowel segments was found in nine patients, moderate bowel wall thickening in two patients, and CT evidence of bowel ischemia in two patients, of whom the embolized points were both at the marginal artery (Fig. 3).

Five patients eventually underwent surgical resection in this study, including the three patients, in whom embolotherapy was a failure. One patient had a hypervascular tumor stain at the adjacent bowel segment, which was incidentally found on the angiograms. He received elective surgery 11 days after the initial embolization with the pathological diagnosis of a leiomyoma of the jejunum. A healed shallow mucosal erosion was also noted at the previously embolized site. The other patient had the complication of bowel ischemia noted a day after the embolotherapy and thus underwent right hemicolectomy with uneventful clinical course.

Five of the twenty-six patients (19.2%) died within the first 30 days of intervention, three following surgery and two following embolization. All the three patients underwent urgent surgical resection after failed embolotherapy, but died at 1, 7 and 18 days, respectively after the operation. Of the two successfully embolized patients, one patient with complications of bowel ischemia had bowel perforation with localized abscess formation. No surgical intervention was done because of poor general condition. Consequently, the patient died of sepsis 46 days after the initial embolization. The other patient, an 87 years old man, died during sleep at midnight 2 days after successful embolotherapy without any clinical symptoms or signs of ischemic bowel or other clinical complaints.

More than 30 days clinical follow-up was available in 21 patients with an average follow-up time of 21.9 months (range, 1-72 months). Two patients (9.5%) experienced further bleeding, which was sufficient to warrant hospital admission. Both of them were angiodysplasia patients (Fig. 4) with re-admission to hospital at 5- and 8-months, respectively following the initial embolotherapy. They were haemodynamically stable and treated nonoperatively. None but one of the twenty-one patients had clinical symptoms or signs of bowel stricture or obstruction. Seven patients died with follow-up period ranging from 1.5-47 months. One was attributed to the ischemic complication. The others were considered not to be related to the LGIB. Five patients remained alive (follow up range, 10-72 months) and nine patients were lost to follow up (follow up range, 1-26 months). The overall survival rates in the twenty-six intended embolized patients and in the twenty-one patients with survival for more than 30 days were plotted in the Kaplan-Meier estimate graphic (Fig. 5). The estimated percentage survival (± standard error) after 1 year was 58.2 ± 10.3% and 72.0 ± 10.7%, respectively and after 5 years were 43.1 ± 12.1% and 53.4 ± 14.1%, respectively. As in the cases of in-patients and out-patients, the estimated survival time (± standard error) was 21.2 ± 7.6 months and 49.1 ± 9.5 months, respectively. Generally, there is a trend of better survival time for out-patient group, however, no statistical significance was observed (p = 0.133).