studies seemed to account for the heterogeneity. Excluding the Finnish HBCS study (with the largest effect), heterogeneity was no longer significant (χ2 = 11.30, df = 8, P = 0.19) and the pooled p-value became 4.3 × 10−6. When excluding the NAG/IRPG study (a small opposite effect), heterogeneity was also no longer observed (χ2 = 11.20, df = 8, P = 0.19) and the pooled p-value became 2.2 × 10−9. The SNP was genotyped in seven of the studies and imputed with high quality in the SardiNIA, NTRNESDA and ERF studies (r-squared or proper_info > 0.98). The MAFs were very similar across studies and ranged between 0.37 and 0.46 and distributions were in HWE in all studies (P > 0.01).
