OBJECTIVE Duplex ultrasound (DUS) imaging for vein bypass graft (VBG) surveillance is confounded by technical and physiologic factors that reduce the sensitivity for detecting impending graft failure. In contrast, three-dimensional computed tomography angiography (CTA) offers high-fidelity anatomic characterization of VBGs, but its utility in detecting at risk grafts is unknown. The current study analyzed the correlation between DUS and CTA for detection of vein graft stenosis and evaluated the relationship of the observed abnormalities to VBG failure. METHODS Consecutive lower extremity VBG patients underwent surveillance with concurrent DUS imaging and CTA at 1 week and at 1, 6, and 12 months postoperatively. A standardized algorithm was used for CT reconstruction and extraction of the lumen geometries at 1-mm intervals. At each interval, CT-derived cross-sectional areas were coregistered and correlated to DUS peak systolic velocities (PSVs) within six predesignated anatomic zones and then analyzed for outcome association. Vein graft failure was defined as pathologic change within a given anatomic zone resulting in thrombosis, amputation, or reintervention within the 6-month period after the observed time point. RESULTS The study recruited 54 patients, and 10 (18%) experienced failure ≤18 months of implantation. The expected inverse relationship between cross-sectional area and PSV was only weakly correlated (Spearman rank coefficient = -0.19). Moderate elevations in the PSV ratio (PSVr; 2-3.5) were frequently transient, with 14 of 18 grafts (78%) demonstrating ratio reduction on subsequent imaging. A PSVr ≥3.5 was associated with a 67% failure rate. CT stenosis <50% was highly correlated with success (0 failures); however, high-grade (>80%) CT stenosis was more likely to succeed than to fail (25%). Significant discordance between CT and DUS was found in 18 patients. Although 14 of these patients had CT stenosis >70% with a PSVr <3.5, subsequent failure occurred in only two. Conversely, graft failure occurred in three of four patients with CT stenosis <70% but PSVr >3.5. Focused analysis of these patients using computational fluid dynamic modeling demonstrated that vein side branches, local tortuosity, regional diameter variations, and venovenostomies were the drivers of these discrepancies. CONCLUSIONS This analysis demonstrated that a PSVr ≥3.5 is strongly correlated with VBG failure, whereas the natural history of moderately elevated PSVr (2-3.5) is largely clinically benign. Although minimum stenosis on the CT scan was highly predictive of success, high-grade CT stenosis was infrequently associated with failure. The interaction of anatomic features with the local flow dynamics was identified as the primary confounder for a direct correlation between CT and DUS imaging.