BACKGROUND: Individuals with diabetes have a raised risk of stroke, but it is unclear whether sustained hyperglycaemia contributes to the development of cerebrovascular disease. Haemoglobin A1c (HbA(1c)), a measure of long-term glycaemia, is strongly related to retinopathy, nephropathy, and neuropathy in diabetes. We sought to assess the association between HbA(1c) and stroke in people with and without diabetes. METHODS: 10,886 participants without diabetes and 1635 participants with diabetes in the ARIC study, who did not have cardiovascular disease, were followed up for incident ischaemic stroke over 8-10 years. We assayed HbA(1c) for all 167 stroke cases and a sample of 680 non-cases in the adults without diabetes and for the full cohort of 1635 adults with diabetes (including 89 stroke cases). We assessed the relation between HbA(1c) concentrations (in tertiles specific for individuals with and without diabetes) and incident ischaemic stroke during follow-up using Cox proportional hazards models, controlling for risk factors for stroke. FINDINGS: The adjusted relative risks of stroke increased with increasing tertile of HbA1c in both adults without diabetes (p=0.02) and with diabetes (p<0.0001). Compared with adults without diabetes in the lowest tertile of HbA1c, the adjusted relative risks of stroke by HbA(1c) tertile were 1.18 (0.70-2.00) and 1.58 (0.94-2.66) in adults without diabetes and 1.75 (0.90-3.42), 2.29 (1.24-4.21), and 4.71 (2.69-8.25) in adults with diabetes. INTERPRETATION: Raised HbA(1c) could be an independent risk factor for stroke in people with and without diabetes, with relative risks similar to those previously reported for coronary heart disease.