Polygenic Risk Scores and Twin Concordance for Schizophrenia and Bipolar Disorder

Abstract

Importance: Schizophrenia and bipolar disorder are highly heritable psychiatric disorders with strong genetic and phenotypic overlap. Twin and molecular methods can be leveraged to predict the shared genetic liability to these disorders.

Objective: To investigate whether twin concordance for psychosis depends on the level of polygenic risk score (PRS) for psychosis and zygosity and compare PRS from cases and controls from several large samples and estimate the twin heritability of psychosis.

Design, setting, and participants: In this case-control study, psychosis PRS were generated from a genome-wide association study (GWAS) combining schizophrenia and bipolar disorder into a single psychosis phenotype and compared between cases and controls from the Schizophrenia and Bipolar Twin Study in Sweden (STAR) project. Further tests were conducted to ascertain if twin concordance for psychosis depended on the mean PRS for psychosis. Structural equation modeling was used to estimate heritability. This study constituted an analysis of existing clinical and population datasets with genotype and/or twin data. Included were twins from the STAR cohort and from the Swedish Twin Registry. Data were collected during the 2006 to 2013 period and analyzed from March 2023 to June 2024.

Exposures: PRS for psychosis based on the most recent GWAS of combined schizophrenia/bipolar disorder.

Main outcomes and measures: Psychosis case status was assessed by clinical interviews and/or Swedish National Register data.

Results: The final cohort comprised 87 pairs of twins with 1 or both affected and 59 unaffected pairs from the STAR project (for a total of 292 twins) as well as 443 pairs with 1 or both affected and 20 913 unaffected pairs from the Swedish Twin Registry. Among the 292 twins (mean [SD] birth year, 1960 [10.8] years; 158 female [54.1%]; 134 male [45.9%]), 134 were monozygotic twins, and 158 were dyzygotic twins. PRS for psychosis was higher in cases than in controls and associated with twin concordance for psychosis (1-SD increase in PRS, odds ratio [OR], 2.12; 95% CI, 1.23-3.87 on case status in monozygotic twins and OR, 2.74; 95% CI, 1.56-5.30 in dizygotic twins). The association between PRS for psychosis and concordance was not modified by zygosity. The twin heritability was estimated at 0.73 (95% CI, 0.30-1.00), which overlapped with the estimate in the full Swedish Twin Registry (0.69; 95% CI, 0.43-0.85).

Conclusions and relevance: In this case-control study, using the natural experiment of twins, results suggest that twins with greater inherited liability for psychosis were more likely to have an affected co-twin. Results from twin and molecular designs largely aligned. Even as illness vulnerability is not solely genetic, PRS carried predictive power for psychosis even in a modest sample size.

Figure.. Distribution of the Polygenic Risk Score (PRS) for Psychosis in LifeGene

The shapes in the upper part of the figure show the mean PRS for psychosis in unrelated schizophrenia cases and controls (diamond symbol), monozygotic (MZ) twins (square symbol), and dizygotic (DZ) twins (circle symbol). The signs in the shapes represent disease status, “−“ for unaffected and “+” for affected and “−/−“ means concordant unaffected twin pair, “−/+” means discordant pair, and “+/+” means concordant affected pair. LifeGene controls were volunteers from the general population and anyone with a register diagnosis of a mood or psychotic disorder was removed. All PRS for psychosis were standardized using the mean and SD of PRS for psychosis from LifeGene controls (the LifeGene PRS for psychosis was standardized to a mean of 0 [red line] and SD of 1). Participant numbers in each group are as follows: 580 unrelated, 134 MZ twin pairs, and 158 DZ twin pairs.