Although arteriovenous fistulas are considered superior to grafts, it is unknown whether that is true in the subset of patients with a previous failed fistula. For investigation of this question, a prospective vascular access database was queried retrospectively to compare the outcomes of 59 fistulas and 51 grafts that were placed in the upper arm after primary failure of an initial forearm fistula. Primary access failure was higher for subsequent fistulas than for subsequent grafts (44 versus 20%; P = 0.006). Fistulas required more interventions than grafts before their successful use (0.42 versus 0.16 per patient; P = 0.04). The time to catheter-free dialysis was longer for fistulas than for grafts (131 versus 34 d; P < 0.0001) and was associated with more episodes of bacteremia before permanent access use (1.3 versus 0.4 per patient; P = 0.003). Cumulative survival (from placement to permanent failure) was higher for fistulas than for grafts when primary failures were excluded (hazard ratio 0.51; 95% confidence interval 0.27 to 0.94; P = 0.03), but similar when primary failures were included (hazard ratio 0.99; 95% confidence interval 0.61 to 1.62; P = 0.97). Fistulas required fewer interventions to maintain long-term patency for dialysis after maturation (0.73 versus 2.38 per year; P < 0.001). In conclusion, as compared with grafts, subsequent upper arm fistulas are associated with a higher primary failure rate, more interventions to achieve maturation, longer catheter dependence, and more frequent catheter-related bacteremia. However, once the access is usable for dialysis, fistulas have superior cumulative patency than do grafts and require fewer interventions to maintain patency. Copyright © 2007 by the American Society of Nephrology.