The patient-control sample was predominantly composed of individuals self-reporting European (EU) ancestry (n=4410), but did include a small number of individuals from Brazil, Mexico, and Costa Rica (n=78) to maximize the power to detect rare events. However, a sensitivity analysis restricted to genetically-defined EU ancestry (n=4276) via multidimensional scaling (Figures S2, S3, available online) confirmed that results were not biased by population stratification.
