Background: Despite several modifications to the original design, patency rates of radiocephalic arteriovenous fistulas have changed little since the first report in 1966. The use of non-penetrating clips for vascular anastomosis on the outcome of such fistulas was studied. Methods: Between January 2000 and August 2003, 107 primary radiocephalic fistulas were constructed in 98 patients. The vascular anastomoses were performed at random with either sutures (n = 56) or clips (n = 51). Results: Although there were trends for better primary and primary assisted patency of clipped fistulas, the differences were not statistically significant. The 6-month primary patency rate was 61 per cent with sutures and 69 per cent with clips (P = 0-393). The mean(s.d.) primary patency was 315(306) and 285(285) days for clipped and sutured fistulas respectively. With regard to secondary patency, clipped fistulas were better (P = 0-009). The mean(s.d.) secondary patency was 435(376) and 344(316) days for clipped and sutured fistulas, respectively. There were no significant differences in flow characteristics, number of revisions or other morbidity. Conclusion: This randomized clinical trial provided further evidence that the use of vascular clips may improve the patency rate of radiocephalic arteriovenous fistulas for haemodialysis.