Although the associations of both CREB1 SNPs with each illness were not replicated in the new cohorts (BPD analysis in 871 cases and 1089 controls (rs2709370, P=0.0611; rs6785, P=0.0544); SCZ analysis in 1273 cases and 1072 controls (rs2709370, P=0.230; rs6785, P=0.661); and MDD analysis in 129 cases and 100 controls (rs2709370, P=0.114; rs6785, P=0.188)), an overall meta-analysis of all included samples suggested that both SNPs were significantly associated with increased risk of BPD (11 105 cases and 51 331 controls; rs2709370, P=2.33 × 10<sup>-4</sup>; rs6785, P=6.33 × 10<sup>-5</sup>), SCZ (34 913 cases and 44 528 controls; rs2709370, P=3.96 × 10<sup>-5</sup>; rs6785, P=2.44 × 10<sup>-5</sup>) and MDD (9369 cases and 9619 controls; rs2709370, P=0.0144; rs6785, P=0.0314), with the same direction of allelic effects across diagnostic categories.