Arteriovenous graft thrombosis is a frequent event in hemodialysis patients, and usually occurs in grafts with significant underlying stenosis. Regular surveillance for graft stenosis, with pre-emptive angioplasty of significant lesions, may improve graft outcomes. This prospective, randomized, clinical trial allocated 126 hemodialysis patients with grafts to either clinical monitoring alone (control group) or to regular ultrasound surveillance for graft stenosis every 4 months in addition to clinical monitoring (ultrasound group). The two randomized groups were closely matched with respect to demographic, clinical, and graft characteristics, with the exception of a lower frequency of diabetes in the ultrasound group. The primary outcome was graft survival, and the secondary outcome was thrombosis-free graft survival. The frequency of pre-emptive graft angioplasty was 64% higher in the ultrasound group than in the control group (1.05 vs 0.64 events per patient-year, P < 0.001), whereas the frequency of thrombosis was not different (0.67 vs 0.78 per patient-year, P = 0.37). The median time to permanent graft failure was similar between the two groups (38 vs 37 months, P = 0.93). Likewise, the median time to graft thrombosis or failure did not differ (22 vs 25 months, P = 0.33). There was no significant association between diabetes and time to graft failure (P = 0.93) or time to graft thrombosis or failure (P = 0.88). In conclusion, the addition of regular ultrasound surveillance for graft stenosis to clinical monitoring increases the frequency of pre-emptive angioplasty, but may not decrease the likelihood of graft failure or thrombosis. © 2006 International Society of Nephrology.