The aim of this study is to determine if early angiographic embolization can forestall recurrent bleeding in selected high risk ulcers after their initial endoscopic control; to validate prospectively the investigators proposed in selecting high risk ulcers for recurrent bleeding in spite of maximal endoscopic control and profound acid suppression using high dose intravenous infusion of proton pump inhibitor; to characterize the nature of bleeding arteries in severely bleeding peptic ulcers and determine the efficacy of angiographic embolization in the prevention of recurrent bleeding and to establish safety profile of angiographic embolization as an early elective treatment to bleeding peptic ulcers.

Endoscopic therapy is now the treatment of choice in patients with actively bleeding peptic ulcers and ulcers with non-bleeding visible vessels. Following endoscopic control of bleeding, we showed that the use of a high dose intravenous infusion of proton pump inhibitor (PPI) for 72 hours further reduced rate of recurrent bleeding [Lau NEJM 2000]. Recurrent bleeding still occurs in 8 to 10 percent of patients who receive the above treatment regime. The associated mortality following a rebleed is 4-10 fold higher when compared to those without recurrent bleeding. In a logistic regression model involving 1144 patients after successful endoscopic thermocoagulation to their bleeding peptic ulcers, we demonstrated that several factors independently predicted recurrent bleeding. They included hypotension, hemoglobin <10g/dl, fresh blood in the stomach, ulcer size > 2cm and active bleeding during endoscopy [Wong Gut 2003]. When we applied this model in a cohort of 945 patients who underwent endoscopic control of bleeding to their ulcers and adjunctive use of high dose intravenous PPI, 275 belonged to the high risk group. Of them, rebleeding leading to surgery or death occurred in 46 patients (16.7%)[Chiu DDW 2007]. Endoscopic treatment to bleeding peptic ulcers has its own limit. In an ex vivo bleeding model using canine mesenteric arteries, endoscopic thermocoagulation could only consistently seal arteries up to 2 mm in size [Johnson Gastro 1987]. Trans-arterial angiography allows clinicians to study and characterize bleeding arteries underneath peptic ulcers. In ulcers that erode into major arteries such as the gastro-duodenal artery complex and branches from left gastric artery, angiography complements endoscopic therapy in the form of selective coiling of the bleeding artery. Trans-arterial angiographic coiling can provide definitive control of bleeding from larger arteries i.e. > 2 mm in size. In cohort studies, trans-arterial angiographic coiling has been shown to compare favorably to surgery, and is less invasive in the control of severe bleeding in peptic

Inclusion Criteria:

• Actively bleeding peptic ulcers (Forrest I), NBVV or Forrest IIa ulcer,
• Successful endoscopic hemostasis by combination treatment of injected epinephrine followed by either 3.2mm heat probe 30J (4 continuous pulses) or hemo-clipping (at least 2 clips) And one of the followings
• Spurting hemorrhage during endoscopy;
• Ulcer >= 2 cm is determined by an opened biopsy forceps;
• Hb on admission of < 9 g/dl; or
• Hypotension prior to endoscopy defined by SBP of <90 mmHg AND HR of >110 bmp

Exclusion Criteria:

-