Results of genome-wide association studies (GWASs) for bipolar disorder (BD) have indicated ANK3 as one of the most promising candidates for a susceptibility gene. In this study, we performed genetic association analysis of two single-nucleotide polymorphisms (SNPs) in ANK3 (rs1938526 and rs10994336), whose genome-wide significant associations were reported in a previous meta-analysis of GWASs, using genotyping data of Korean and Japanese case–control samples and a part of data from a GWAS in Han-Chinese from Taiwan. The total number of participants was 2,212 cases (352 from Korea, 860 from Japan, and 1,000 from Taiwan) and 2,244 controls (349 from Korea, 895 from Japan, and 1,000 from Taiwan). We could not detect any significant difference of allele frequency in individual analyses using each of the three populations. However, when we combined the three data sets and performed a meta-analysis, rs1938526 showed nominally significant association (P = 0.048, odds ratio = 1.09). The over-represented allele in BD was same as that reported in Caucasian GWASs. On the other hand, any significant association was not detected in rs10994336. This discrepancy between two SNPs may be explained by the different degree of linkage disequilibrium between Asian and Caucasian. These findings further supported the association between ANK3 and BD, and also suggested the genomic region around rs1938526 as a common risk locus across ethnicities. © 2011 Wiley-Liss, Inc.