< 0.001; a minor allele count in cases and controls of < 20; imputation INFO score < 0.1; or that had alleles that were inconsistent with the reference sample. For each set of summary statistics we calculated the median odds ratio (OR) and standard error (SE) of the association statistics for SNPs with MAF > 0.01. We checked that median ORs were close to 1 and plotted median SEs against effective sample size to detect potential effect size scaling errors. We estimated pairwise LDSC genetic correlations44 among all cohorts and inspected genetic covariance intercepts for evidence of sample overlap. Pairs of studies with covariance intercepts > 0.025 were returned to cohort analysts for scrutiny to resolve potential sample overlap or close relatedness between sub-cohorts. For the diverse ancestry cohorts, only variants with an imputation information score of 0.7 or higher were deemed eligible. Furthermore, for studies with a sample size smaller than 10,000, a minor allele frequency (MAF) of at least 5% was required; for larger studies, we required a minimum effective sample size (Neff) of 50, calculated as Neff = 2 × MAF × (1-MAF) × N × Info, where ‘Info’ represents the imputation quality score, ‘MAF’ is the
