Most arteriovenous grafts fail due to irreversible thrombosis, and most clotted grafts have an underlying stenotic lesion. These observations raise the plausible hypothesis that early detection of graft stenosis with preemptive angioplasty will reduce the likelihood of graft thrombosis. A number of noninvasive methods can be used to detect hemodynamically significant graft stenosis with a high positive predictive value. These tests include clinical monitoring, as well as surveillance by static dialysis venous pressures, flow monitoring, or duplex ultrasound. However, these surveillance tests have a much lower positive predictive value for graft thrombosis in the absence of preemptive angioplasty. In other words, none of the currently available surveillance tests can reliably distinguish between stenosed grafts destined to clot, and those that will remain patent without intervention. As a consequence, any program of graft surveillance necessarily results in a substantial proportion of unnecessary angioplasties. Moreover, a substantial proportion of grafts thrombose despite a normal antecedent surveillance test. Numerous observational studies have found an impressive reduction of graft thrombosis after implementation of a stenosis surveillance program. In contrast, 5 of 6 randomized clinical trials failed to show a reduction of graft thrombosis in patients undergoing graft surveillance, as compared with those receiving only clinical monitoring. The lack of benefit of surveillance is largely attributable to the rapid recurrence of stenosis after angioplasty. Thus, routine surveillance for graft stenosis, with preemptive angioplasty, cannot be recommended for reduction of graft thrombosis. Future research should be directed at pharmacologic interventions to prevent graft stenosis.