Polygenic liability for major depression, schizophrenia and bipolar disorder and treatment outcome across major psychiatric disorders

Polygenic scores for major depression, schizophrenia and bipolar disorder have been related to treatment outcome. We analysed 4,007 patients of European ancestry from 7 cohorts: three with major depression (STAR*D, GSRD, PANDORA), two with bipolar disorder (STEP-BD, Barcelona) and two with schizophrenia (CATIE, HSRRm). The three scores were computed with PRS-CS from the latest Psychiatric Genomics Consortium studies. The primary outcome was good response or good course, harmonized from each cohort's instrument. Firth logistic regression within cohorts, adjusted for age, sex and ancestry components, was followed by random-effects pooling. A higher depression score was associated with lower odds of a good outcome (odds ratio per standard deviation 0.88, 95% confidence interval 0.79-0.99, p = .04; false discovery rate q = .11). The estimate remained below one in all 8 sensitivity analyses, including adjustment for baseline severity, and was negative in all 5 cohorts with a continuous improvement measure. The schizophrenia score pointed the same way (odds ratio 0.93, 95% confidence interval 0.85-1.01) with no heterogeneity. The bipolar score showed no pooled effect (odds ratio 0.97), and its direction differed by diagnosis. In pooled data spanning three diagnoses, higher polygenic liability to depression was nominally associated with a less favourable treatment outcome, although the association did not survive correction for multiple testing; schizophrenia liability showed a similar directional trend. These findings support further evaluation of disorder liability in multivariable models of treatment outcome.