The role of the microcirculation in the pathogenesis of osteoarthritis (OA) remains unclear. This prospective cohort study examined the association between retinal vascular calibre and incidence of knee replacement for OA.1838 participants of the Australian Diabetes, Obesity and Lifestyle (AusDiab) Study had retinal vascular calibre measured using a nonmydriatic digital fundus camera in 1999-2000 and were aged ≥ 40 years at joint replacement data collection commencement. The incidence of knee replacement for OA during 2002-2011 was determined by linking cohort records to the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR).77 participants underwent knee replacement for OA. They had narrower retinal arteriolar calibre compared with those without knee replacement (166.1 ± 24.8 μm vs 174.3 ± 24.5 μm, P = 0.004). For every one standard deviation reduction in retinal arteriolar calibre, the incidence of knee replacement increased by 25% (HR 1.25, 95% confidence interval (CI) 1.00-1.56). Participants in the narrower two-thirds of arteriolar calibre had twice the risk of knee replacement compared with those in the widest one-third (HR 2.00, 95% CI 1.07-3.74, P = 0.03) after adjustment for sex, body mass index (BMI), physical activity and HbA1c. There was no association for retinal venular calibre.Retinal arteriolar narrowing is associated with increased risk of knee replacement for OA suggesting that further work is warranted to determine the role of the microcirculation in the pathogenesis of knee OA.