Twenty-seven vascular prosthetic graft infections in 25 patients were managed from 1981 through 1990 using the principles of extraanatomic bypass through uncontaminated fields and removal of the infected prosthesis. This experience included 18 aortic, three femoral-femoral, three femoral-popliteal, two axillofemoral grafts, and one popliteal endarterectomy patch. The predominant organism was Staphylococcus epidermidis (26%). Mean follow-up was 36 months. There was only one early and one late death, with two late amputations. Extraanatomic bypass grafts were placed in 21 of 25 patients including all 18 infected aortic grafts. Of these 21 patients, 11 (52%) experienced at least one extraanatomic bypass failure within 15 months, resulting directly in two major lower extremity amputations and two graft reinfections. Axillounifemoral bypass had a higher incidence of failure than axillobifemoral bypass (54% versus 29%). More importantly, however, extraanatomic graft failure was also associated with the presence of groin sepsis. The revision rate was 63% when the graft required circuitous tunneling to avoid groin sepsis, in contrast to a revision rate of only 17% when the graft could be anastomosed directly to the common femoral artery. Of extraanatomic bypass grafts that failed once, 63% had multiple failures. Graft removal and extraanatomic revascularization produced excellent overall results when not involving the groin. Late complications occurred more frequently when groin sepsis was present. These results suggest that, to reduce the incidence of late graft failure and amputation, more aggressive early direct reconstruction should be performed in lieu of atypical graft tunneling, especially if the extraanatomic graft has failed once and the causative organism is Staphylococcus epidermidis.