We further conducted GWAS meta-analyses on four subgroups, defined by differences in their sample ascertainment, either (i) clinical-OCD cases diagnosed by a health care professional in a clinical setting (ncases = 9,089, ncontrol = 21,077; including IOCDF, IOCDF_trio, EPOC, NORDiC-nor, NORDiC-swe, EGOS, OCGAS, OCGAS-ab, OCGAS-gh, OCGAS-nes, Psych_Broad, WWF, MVP, Michigan/Toronto IGS, YalePenn, Chop, CoGa), (ii) comorbidindividuals that were primarily ascertained for another comorbid psychiatric disorder (ncases = 5,266, ncontrol = 43,760; AGDS, IPSYCH), (iii) biobank data from large-scale biobanks or registries with ICD or DSM codes (ncases = 9,138, ncontrol = 1,049,776; BioVU, EstBB, FinnGen, HUNT, MoBa, UKBB) or (iv) 23andMe data (ncases = 30,167, ncontrol = 929,804). While these groups are not exclusive (e.g., diagnoses in health records were originally given in a clinical setting, or comorbid cases were also assessed in a clinical setting or derived from health records), we defined these groups by the cohort’s primary characteristic. We also conducted one meta-analysis including all clinical, comorbid, and biobank subgroups, while excluding the 23andMe data, resulting in 23,493 cases and 1,114,613 controls. As 23andMe is the only consumer-based dataset, we intended to compare this dataset to all others.
