Abstract The management of bleeding gastric varices has not been standardized. Although transjugular intrahepatic portosystemic shunt (TIPS) is used in most centers, endoscopic treatment with N-butyl-2-cyanoacrylate (cyanoacrylate) glue has recently been shown to be effective. Cost-effectiveness analyses of these methods are lacking. Methods We performed a retrospective review of patients with bleeding gastric varices treated either by TIPS or cyanoacrylate glue injection. Economic analysis was based on direct costs for a fixed financial year. The two groups were compared for a period of 6 months follow-up, to liver transplantation, or death for each patient. Results Between January, 1995 and December, 1999, 20 patients with bleeding gastric varices had TIPS; 23 patients had cyanoacrylate glue injection from January, 2000 to October, 2001. There were no significant differences between the two groups in patient characteristics, transfusion requirement, and gastric variceal anatomy. In the TIPS group, 15/20 patients had the procedure performed within 24 h of hemorrhage, and 90% of stent insertions were successful. Complications consisted of two cases of pulmonary edema, two cases of severe encephalopathy, and a 15% stenosis rate at 6 months. In the glue group, there were 3 ± 1.5 endoscopies and 2 ± 1 injections per patient, with a 96% initial hemostasis. There was one case of (glue) pulmonary embolism and one blocked front endoscope lens, which required repair. The initial rebleed rate was significantly lower in patients who had TIPS (15% vs 30%, p = 0.005). The inpatient stay was shorter in the glue group (13 ± 1 vs 18 ± 2 days, p = 0.05), but there was no difference in the overall mortality rate. The median cost within 6 months of initial gastric variceal bleeding was $4,138 ($3,009–$8,290) for glue versus $11,906 ($8,200–$16,770) for TIPS ( p < 0.0001). Conclusion In this comparable group of patients, cyanoacrylate glue injection was more cost effective than TIPS in the management of acute gastric variceal bleeding. A prospective, randomized trial would be required to confirm our analysis.