To propose a percutaneous treatment for otherwise intractable pancreatic fistula (PF). From 2005 to 2011, 12 patients (9 men and 3 women, mean age 59 years, median 63 years, range 33-78 years) underwent radiological treatment for high-output PF associated with peripancreatic fluid collection. The percutaneous procedures were performed after at least 4 wk of unsuccessful conservative treatments. We chose either a one or two step procedure, depending on the size and characteristics of the fistula and the fluid collection (with an arbitrary cut-off of 2 cm). Initially, 2 to 6 pigtail drainages of variable size from 8.3 (8.3-Pig Duan Cook, Bloomington, Indiana, United States) to 14 Fr (Flexima, Boston Scientific, Natick, United States) were positioned inside the collection using a transgastric approach. In a second procedure, after 7-10 d, two or more endoprostheses (cystogastrostomic 8 Fr double-pigtail, Cook, Bloomington, Indiana, United States in 10 patients; covered Niti-S stent, TaeWoong Medical Co, Seoul, South Korea in 2 patients) were placed between the collection and the gastric lumen. In all cases the metal or plastic prostheses were removed within one year after positioning. Four out of 12 high-output fistulas fistulas were external while 8/12 were internal. The origin of the fistulous tract was visualised by computer tomography (CT) imaging studies: in 11 patients it was at the body, and in 1 patient at the tail of the pancreas. Single or multiple drainages were positioned under CT guidance. The catheters were left in place for a varying period (0 to 40 d - median 10 and 25(th)-75(th) percentile 0-14). In one case external transgastric drainages were left in place for a prolonged time (40 d) due to the presence of vancomycin-resistant bacteria (Staphylococcus) and fluconazole-resistant fungi (Candida) in the drained fluid. In this latter case systemic and local antibiotic therapy was administered. In both single and two-step techniques, when infection was present, we carried out additional washing with antibiotics to improve the likelihood of the procedure's success. In all cases the endoprostheses were left in situ for a few weeks and endoscopically removed after remission of collections, as ascertained by CT scan. Procedural success rate was 100% as the resolution of external PF was achieved in all cases. There were no peri-procedural complications in any of the patients. The minimum follow-up was 18 mo. In two cases the procedure was repeated after 1 year, due to the onset of new fluid collections and the development of pseudocysts. Indeed, this type of endoprosthesis is routinely employed for the treatment of pseudocysts. Endoscopy was adopted both for control of the positioning of the endoprosthesis in the stomach, and for its removal after resolution of the fistula and fluid collection. The resolution of the external fistula was assessed clinically and CT scan was employed to demonstrate the resolution of peripancreatic collections for both the internal and external fistulae. The percutaneous placement of cistogastrostomic endoprostheses can be used for the treatment of PF that cannot be treated with other procedures.