Leveraging transdiagnostic genetic liability to psychiatric disorders to dissect clinical outcomes of anorexia nervosa

Abstract

Anorexia nervosa (AN) has extensive genetic correlations with other psychiatric disorders, and genetic risk for different psychiatric disorders was associated with distinct clinical courses in AN. Uncovering associations between transdiagnostic psychiatric genetic liability and AN outcomes can facilitate its personalized treatment. In this study, we investigated the associations between transdiagnostic psychiatric genetic liability and outcomes of AN. Genomic structural equation models were fitted to genome-wide association data for AN and psychiatric disorders with high genetic correlations with AN (obsessive-compulsive symptoms [OCS], major depressive disorder [MDD], schizophrenia, and anxiety disorders) to extract one shared and five trait-specific genetic components. Next, we calculated the polygenic risk scores (PRS) for these components, including PRS_shared, PRS_AN-specific, PRS_OCS-specific, PRS_MDD-specific, PRS_SCZ-specific and PRS_ANX-specific, which index the shared genetic liability to all five psychiatric traits, and genetic liability specific to AN, OCS, MDD, SCZ and ANX, respectively. We then tested associations between these PRSs and clinical outcomes reported between 1997 and 2018 among AN cases from the Anorexia Nervosa Genetics Initiative (ANGI), linked to Swedish National Registers. The clinical outcomes included cumulative disease burden (i.e., number of diagnoses, medication prescriptions, and inpatient days), risks of psychiatric comorbidities, and AN symptomatology. Among 4028 included AN cases, the mean (SD) birth year was 1985 (9), and 3947 (98.0%) were female. Within AN, +1 SD increase of PRS_shared was associated with 9-39% excess risk of disease burden and psychiatric comorbidity, whereas the associations between PRS_AN-specific and most clinical outcomes were statistically non-significant. +1 SD increase of PRS_MDD-specific was associated with 3-29% increased risk of AN disease burden. Our findings show that shared psychiatric liability is associated with more adverse AN outcomes, whereas AN-specific liability is not a good indicator for its clinical course. This study provides a novel perspective on factors influencing heterogeneity in AN clinical course.

Fig. 1. Flow chart of the current study.

AN anorexia nervosa, SCZ schizophrenia, OCS obsessive-compulsive symptoms, OCD obsessive-compulsive disorder, MDD major depressive disorder, ANX anxiety disorders, gSEM genomic structural equation model, PRS polygenic risk scores, ASD autism spectrum disorder, ADHD attention-deficit hyperactivity disorder, SUD substance use disorder, BMI body mass index, CIA Clinical Impairment Assessment, CPRS Comprehensive Psychopathological Rating Scale, EDE-Q Eating Disorder Examination Questionnaire.

Fig. 2. Association between transdiagnostic genetic liabilities and cumulative disease burden from the quasi-Possion models.

Results are derived from quasi-Possion regression models based on 4028 AN cases with number of unique clinical diagnoses (any, psychiatric, somatic), prescriptions (any, antipsychotics, antidepressants) and inpatient days (any, due to EDs) as outcomes. For each of the eight outcomes, we constructed six models with PRS_shared (A), PRS_AN-specific (B), PRS_OCS-specific (C), PRS_MDD-specific (D), PRS_SCZ-specific (E) and PRS_ANX-specific (F) as exposure variable, respectively. Incidence rate ratios indicate the risk estimates for +1 SD increase of PRS. Sex, birth year and first 10 ancestry-informative principal components were adjusted for in all models. The points represent incidence rate ratio estimates, and the error bars indicate 95% confidence intervals. Blue points represent effect estimates for clinical diagnoses, red points represent effects for medication prescriptions and yellow points represent effects for inpatient days. “*” represents association that remained significant after Bonferroni correction. “#” represents trending association at P < 0.05 but was not significant after Bonferroni correction. PRS polygenic risk scores, EDs eating disorders, IRR incidence rate ratio, 95%CI 95% confidence interval, AN anorexia nervosa, SCZ schizophrenia, OCS obsessive-compulsive symptoms, MDD major depressive disorder, ANX anxiety disorders.

Fig. 3. Association between transdiagnostic genetic liabilities and psychiatric comorbidities from the Cox regression models.

Results are derived from quasi-Possion regression models based on 4028 AN cases with ADHD, ANX, ASD, MDD, OCD, SCZ and SUD as the outcome variables. For each of the seven outcomes, we constructed six models with PRS_shared (A), PRS_AN-specific (B), PRS_OCS-specific (C), PRS_MDD-specific (D), PRS_SCZ-specific (E) and PRS_ANX-specific (F) as exposure variable, respectively. Incidence rate ratios indicate the risk estimates for +1 SD increase of PRS. Sex, birth year and first 10 ancestry-informative principal components were adjusted for in all models. The points represent hazard ratio estimates, and the error bars indicate 95% confidence intervals. “*” represents association that remained significant after Bonferroni correction. “#” represents trending association at P < 0.05 but was not significant after Bonferroni correction. PRS polygenic risk scores, AN anorexia nervosa, OCS obsessive-compulsive symptoms, OCD obsessive-compulsive disorder, SCZ schizophrenia, MDD major depressive disorder, ANX anxiety disorders, ASD autism spectrum disorder, ADHD attention-deficit hyperactivity disorder, SUD substance use disorder.

Fig. 4. Association between transdiagnostic genetic liabilities and AN symptomatology from the linear regression models.

Results are derived from linear regression models with BMI, EDE-Q score, CPRS-depression score, CPRS-anxiety score and CIA score as the outcome variables. The analyses were based on 1934 ANGI-SE AN cases with available data on symptomatology in Stepwise Quality Register. For each of the five outcomes, we constructed six models with PRS_shared (A), PRS_AN-specific (B), PRS_OCS-specific (C), PRS_MDD-specific (D), PRS_SCZ-specific (E) and PRS_ANX-specific (F) as the exposure variable, respectively. Regression coefficients indicate the risk estimates for +1 SD increase of PRS. Sex, birth year and first 10 ancestry-informative principal components were adjusted for in all models. The points represent regression coefficients, and the error bars indicate 95% confidence intervals. “*” represents association that remained significant after Bonferroni correction. “#” represents trending association at P < 0.05 but was not significant after Bonferroni correction. PRS polygenic risk scores, AN anorexia nervosa, OCS obsessive-compulsive symptoms, SCZ schizophrenia, MDD major depressive disorder, ANX anxiety disorders, BMI body mass index, CIA Clinical Impairment Assessment, CPRS Comprehensive Psychopathological Rating Scale, CPRS-dep CPRS-depression score, CPRS-anx CPRS-anxiety score, EDE-Q Eating Disorder Examination Questionnaire.