There are several limitations to this study. Firstly, the sample size for MDD and rMDD groups are very similar (n = 43,062 and n = 39,556, respectively), therefore perhaps rMDD is not the best stratifier in this sample. The sample size of all depressed cases could have been increased by including individuals with mild depressive/manic symptoms (n = 26,847), however as case classification was based on very few items (two symptoms and help-seeking behavior), it wasn’t possible to determine whether mild symptoms should be classified as cases or controls17, therefore including these individuals could introduce further phenotypic heterogeneity. Whilst the differential model selection in GS:SFHS and UKB used to adequately account for differential family and population structure introduces analytical heterogeneity, the genetic correlation of MDD GWAS summary statistics from the two samples was rG(SE) = 0.997(0.26). Lastly, the higher prevalence of females in GS:SFHS caused a gender imbalance in the sample sizes, resulting in lower statistical power for the male only analysis.
