Though color duplex ultrasonography (CDU) can identify threatened arterial bypass grafts, the natural history of grafts predicted to fail is not known. We examined patency of "failing grafts" followed by CDU for prolonged periods without intervention. A graft was defined as failing if there was elevation of the peak systolic flow velocity (PSFV) to a ratio of three times the PSFV in the adjacent graft, or if PSFV was less than 45 cm/sec throughout the graft. Only patients followed with CDU abnormalities without intervention for more than 2 months were included. Forty-six CDU abnormalities were noted after construction or revision of lower extremity bypass grafts in 34 patients. Grafts were autogenous in 25 cases, prosthetic in 16, and composite in 5. Focal abnormalities were noted in 35 grafts (76.1%), low PSFV throughout the graft in 6 (13.0%), while both findings were present in 5 grafts (10.7%). For various reasons no intervention was performed during follow-up ranging from 2 to 50 (mean 10) months, during which time patients had a mean of 3.6 CDU studies. Abnormalities regressed in 10 grafts (21.7%), progressed to 5 (10.9%), and were stable in the remainder. Fourteen grafts (30.4%) were ultimately revised with surgery or angioplasty at a mean of 5 months after the first abnormal CDU. Only 3 grafts (6.5%) occluded while being followed. Two of the 3 were among the 5 grafts with both focal elevated PSFV ratio and low PSFV throughout the remaining graft, while all 3 were among the 7 grafts with PSFV ratio in excess of 7.0. Compared to grafts without these features, occlusion was significantly more likely (p = 0.03 and p = 0.001, respectively). Currently defined threshold CDU criteria for prediction of graft failure may be excessively sensitive.