We compared the MDD2000+ results with other published studies. Muglia et al.11 listed 27 SNPs with P<1 × 10−5; 25 SNPs overlapped with our analysis but none had P<0.05 and association with the same allele. Of the top 200 SNPs in one or both of their study samples, ∼90% were analysed in our study but ∼5% had P<0.05, consistent with chance expectations. As the MDD2000+ controls were also used in GenRed13 and STARD,14 we did not make formal comparisons of our results, but we note that ∼5% of their top SNPs (P<1 × 10−5) had P<0.05 in MDD2000+. The SNP rs2251219 identified as a mood disorder risk factor in a GWAS meta-analysis (three bipolar studies plus the GAIN–MDD study)32 but was not associated in MDD2000+ (P=0.51, MAF=0.40, odds ratio (OR)=0.97, 95% CI 0.90–1.05). PCLO was the top finding in the GAIN–MDD study,12 replicated in Australian (QIMR)12 and Dutch33 samples, but we found no evidence for association with PCLO variants (for example, P=0.51 for rs2522833), despite partial overlap (562 cases and 264 controls) in the QIMR samples used here and in
