A retrospective study was carried out of patients from a single institution over a 30-year period. Thirty-one patients presented with 33 fistulas, four non-enteric and 27 enteric. In 25 of 27 patients with a prosthesis-related enteric fistula gastrointestinal bleeding was present. Angiography revealed the fistula in five patients endoscopy in three, and barium studies, echography and computed tomography each revealed one fistula. Six patients died before and five died during operation. In 20 patients various techniques were used for treatment. In-hospital mortality decreased from six of eight patients before 1970, to seven of ten between 1971 and 1980, and to four of 13 after 1981. In the long term, patients treated with an extra-anatomic reconstruction had a poorer prognosis than those treated by in situ reconstruction. This experience shows that diagnostic tests often fail to reveal a prosthesis-related fistula and that mortality can be substantially reduced by early exploration in patients with negative diagnostic studies.