POWER, REPLICATION, AND EPIGENETIC SIGNAL IN PSYCHIATRIC DISORDERS: CONTRASTING METHYLOME-WIDE ASSOCIATION STUDIES OF MAJOR DEPRESSIVE DISORDER AND GENERALIZED ANXIETY DISORDER IN THE MILLION VETERAN PROGRAM

Epigenome-wide association studies (EWAS) of psychiatric disorders have historically been limited by modest sample sizes, contributing to inconsistent findings and poor replication. Leveraging DNA methylation data from the Million Veteran Program (MVP), we conducted large-scale multi-ancestry EWAS of both major depressive disorder (MDD; n=17,987 cases) and generalized anxiety disorder (GAD; n=2,586 cases) to evaluate how statistical power influences discovery and reproducibility. In MDD, we identified extensive methylome-wide significant (MWS) signal, with 649 CpG sites detected in European ancestry analyses, of which 102 remained significant after adjustment for smoking and body mass index. Cross-cohort replication demonstrated strong concordance with independent studies, including a large multi-cohort MDD EWAS and an antidepressant exposure proxy phenotype (effect size correlations r²:0.58–0.71), indicating robust and reproducible epigenetic signal. In contrast, GAD analyses yielded substantially fewer significant loci (33 CpG sites in European ancestry), with marked attenuation following adjustment for smoking and BMI, and more limited replication of methylation risk scores in an independent cohort. Across both traits, signal strength and locus discovery scaled with effective sample size, particularly within European ancestry analyses. However, replication performance diverged sharply between traits, with MDD demonstrating consistent cross-cohort reproducibility, while GAD results remained more sensitive to confounding and cohort-specific factors. These differences likely reflect both disparities in effective sample size and underlying phenotypic heterogeneity. Together, these findings highlight the central role of statistical power in EWAS discovery and replication, and suggest that, analogous to early genome-wide association studies, substantially larger and more harmonized samples will be required to achieve stable and generalizable epigenetic findings for anxiety-related phenotypes. Ongoing collaborative efforts within the Psychiatric Genomics Consortium will be critical for advancing this goal.