The effect of socioeconomic status on schizophrenia comorbidities: a genetically informed causal inference analysis

Schizophrenia (SCZ) is a severe psychiatric disorder associated with large health and societal costs. Affected individuals suffer from psychiatric and somatic comorbidities that substantially reduce their life expectancy. Several negative health outcomes in SCZ patients are consistently associated with socioeconomic inequalities. However, the ways in which socioeconomic status (SES) affect psychiatric and somatic comorbidities of SCZ remains unclear. In this study, we use genetically informed causal inference methods to estimate whether SES partially contributes to the association of schizophrenia with a range of negative health outcomes. Genetic correlation analysis was performed using LD Score Regression analysis between the Psychiatric Genomic Consortium SCZ genome-wide association summary statistics results and a range of psychiatric and somatic traits. Information regarding psychiatric and somatic traits and diseases was derived from genome-wide association statistics available from the UK Biobank (7,222 traits and diseases assessed in up to 484,267 participants), and FinnGen (2,803 traits and diseases assessed in up to 218,792 participants). To account for the potential causal effect of SES on schizophrenia comorbidities, we used the multi-trait conditioning and joint analysis (mtCOJO) that permitted us to correct the SCZ per-SNP effects with respect to SES. Specifically, the genetic effects were corrected with respect to the causal relationship of household income and Townsend deprivation index on SCZ inferred by Mendelian randomization (MR). After Bonferroni correction accounting for the number of traits tested, SCZ was genetically correlated with a wide range of health outcomes in both UKB (N=344) and FinnGen (N=100). These included both psychiatric comorbidities such as depression (rg=0.43, p=4.66e-63, anxiety (rg=0.49, p=2.81e-79, mood disorders (rg=0.51, p=6.49e-99), and addiction (rg=0.4, p=4.45e-11), and somatic illnesses such as gastrointestinal (e.g., viral hepatitis: rg=0.34, p=9.82e-7), hematological (e.g., anemia: rg=0.2, p= 10e-7), respiratory (e.g., COPD: rg=0.33, p=6e-3) and urinogenital (e.g., cancer of urinary organs: rg=0.23, p=3.92e-17). MR analyses showed a putative causal effect of household income variables on SCZ (odds ratio=0.93, 95% confidence interval=0.87-0.99). However, conditioning genome-wide association statistics of SCZ for SES variables did not have a substantial impact on the genetic correlation of SCZ with comorbid conditions. Although SES could have a causal effect on SCZ, this relationship does not appear to significantly affect the pleiotropy of SCZ with psychiatric and somatic comorbidities. Further multivariable analyses will be needed to confirm these initial findings.