Gastrointestinal bleeding after failed endoscopic hemostasis: diagnostic efficacy of angiography compared with computed tomography and treatment outcomes of transcatheter arterial embolization.

Abstract

This study compared the detection sensitivity of catheter angiography to that of contrast-enhanced multi-detector computed tomography (MDCT) for detecting acute non-variceal gastrointestinal bleeding (GIB) to elucidate the diagnostic efficacy of catheter angiography. We also determined GIB outcomes of transcatheter arterial embolization after failed endoscopic hemostasis. Data were collected retrospectively from 63 patients managed at four institutions who met the following criteria during a 3-year period: (1) ongoing non-variceal GIB confirmed during endoscopy; (2) failed endoscopic hemostasis; and (3) endoscopy, MDCT, and catheter angiography performed within 24h. The diagnostic efficacies of MDCT, selective angiography with a 5-Fr catheter (5Fr-angiography), and super-selective angiography with a microcatheter (micro-angiography) were compared using endoscopic diagnosis as the reference method. The rates of technical success, clinical success, and complications were analyzed when arterial embolization was performed. All transcatheter angiographies were performed after MDCT. Micro-angiography had a significantly higher GIB detection rate (73.0%) than MDCT (57.1%) and 5Fr-angiography (39.7%) (micro-angiography vs. MDCT, P = 0.021; MDCT vs. 5Fr-angiography, P = 0.043). Arterial embolization was attempted in 55 of 63 patients, with technical success achieved in 53 of 55 patients (96.4%) and clinical success in 38 of 42 patients (90.5%). Eleven patients were lost to follow-up. Three patients experienced complications, including bowel infarction (two patients) and common bile duct stricture (one patient). In cases of endoscopic hemostasis failure, angiography can be performed even if MDCT yields negative results but should include micro-angiography; moreover, embolization can be performed safely and effectively. None.