and takes into account the direction of effect at each SNP (33). To specify the direction of the effect, the PGC used the logistic regression coefficient beta and the CHARGE used z-scores. Weights were based on the number of the MDD cases in the PGC study (n=9,240) and the number of individuals in the CHARGE with clinically significant depressive symptoms (n=5,976) using population specific cutoff scores of the questionnaires were considered for weighting. To test whether the results are affected by different sample size weightings, equal weights per study, or no weight as suggested by Stouffer (34), we carried out a series of sensitivity analyses.
