Haemodialysis is a predominant modality for the treatment of end-stage renal failure, but is dependent on high quality access to the circulation to allow effective removal, treatment and return of blood through an extracorporeal circuit. The preferred method is via an established arteriovenous fistula, created surgically by anastomosing a patient’s own artery and vein. This technique is, however, hampered by a high failure rate (up to 40% before first use) (1) caused by early neointimal hyperplasia, in the genesis of which haemodynamic factors appear to play an important role (2). Given the seeming importance of the local flow, including its physiological three-dimensionality (3), we have attempted to determine in model studies whether geometric modification at the anastomosis can reduce the burden of neointimal hyperplasia. Perspex models of a vascular anastomosis were created with planar orientation of the vessels at a range of anastomotic angles. A further, otherwise identical set of models, was created with ‘offset’ junctions, novel non-planar anastomoses, which mimicked the geometry at the origin of arterial branches. Comparative observations were made under non-pulsatile (but otherwise physiological) flow conditions, using flow visualisation and computational fluid dynamics (CFD). Similar results were observed in the flow visualisation and CFD work (see Figure). Flow separation occurred in both planar and offset (non-planar) models within both vessels. It is suggested that flow disturbance (including separation, wall shear abnormality and instability) occur in regions associated with neo-intimal hyperplasia development in humans and animal models. Flow visualisation and CFD allowed complementary assessment of vascular anastomosis models. Preliminary results suggest that modification of anastomosis configuration can alter the peri-anastomotic and downstream flow fields. Planned observational studies in human patients will increase understanding of how current practice, and resulting flow patterns, correlate with clinical outcomes. Concurrent work, including with animal models, will allow examination of the hypothesis that these changes affect the burden of neointimal hyperplasia, possibly mediated through wall shear stress and/or mass transport mechanisms.