A small number of individuals (n = 45) had recorded haemorrhagic stroke before their primary IS event. We conducted two sensitivity analyses to assess the impact of this on our results: (i) excluding n = 45 individuals from the analysis; (ii) adjusting for haemorrhagic stroke status in the analysis. In both cases, there was essentially no difference in the association of the metaGRS with IS compared with the original analysis (HR = 1.27 per standard deviation of the metaGRS across the two analyses).
