Calibration of the metaGRS risk score was evaluated by fitting logistic regression models of the metaGRS (adjusting for sex, chip, and 10 genetic PCs) in the derivation set, predicting the absolute risk of event in the test set (allowing for the 9.38-fold lower observed baseline rate of events between the testing set compared with the derivation set), and evaluating the proportion of test set individuals with stroke events within each decile of the predicted risks (Supplementary Fig. 9). Pointwise confidence intervals were obtained via the binomial test for proportions.
