Background Occult atheroma has often been suggested to be the potential aetiology of undetermined ischaemic stroke. If so, we hypothesised that the risk factor profile of undetermined stroke should resemble that of large-vessel stroke.

Methods We compared the risk factor profile of ischaemic stroke subtypes (TOAST classification) in a population-based study of TIA and stroke and included these data in a meta-analysis of associations in all published population—and hospital-based studies.

Results Risk factor associations in our study were broadly similar to those in nine other population-based studies. Although associations in 24 hospital-based studies were more heterogeneous, the pooled estimates were also similar. There were strong associations (all p<0.001) of large-vessel stroke with male sex (OR=1.5, 95% CI 1.3 to 1.8), hypertension (1.3, 1.2 to 1.5), diabetes (1.4, 1.1 to 1.6), smoking (1.8, 1.5 to 2.1) and previous TIA (1.7, 1.4 to 2.0), while hypertension (1.3, 1.1 to 1.4) and diabetes (1.2, 1.0 to 1.3) were consistently associated with small-vessel stroke. Stroke of undetermined aetiology, however, was negatively associated with conventional risk factors when compared with other subtypes for example, in comparison with large vessel stroke: male sex (0.5, 0.4 to 0.8), hypertension (0.7, 0.6 to 0.8), diabetes (0.7, 0.5 to 1.0), smoking (0.5, 0.3 to 0.8) and previous TIA (0.6, 0.4 to 0.7).

Conclusion Stroke of undetermined aetiology had a different risk factor profile from large-vessel stroke and had a lower prevalence of most conventional risk factors than all other subtypes. Occult atheroma is therefore unlikely to be the main aetiology for this subtype and more research on novel risk factors is required.