64 - ANOREXIA NERVOSA AND OBSESSIVE COMPULSIVE POLYGENIC RISK SCORE: ASSOCIATIONS WITH ADOLESCENT EATING DISORDER PHENOTYPES

Abstract

Background Anorexia Nervosa (AN) and Obsessive Compulsive Disorder (OCD) are highly heritable (~50–60%) and positively significantly genetically correlated (rg=0.48–0.55), suggesting that the same markers confer risk for both disorders. We hypothesized that an AN and OCD polygenic risk scores (PRS) could explain individual differences in adolescent eating behaviors in a large population-based cohort (~4,500 participants), the Avon Longitudinal Study of Parents and Children (ALSPAC). Methods We calculated PRS in our dataset of 3,495 female AN cases from the Psychiatric Genomics Consortium-Eating Disorders workgroup and 2,688 male and female OCD cases from the Psychiatric Genomics Consortium-Obsessive Compulsive Disorders workgroup. We calculated Principal Components (PCs) using Eigensoft and identified case-control significant PCs in R. PRS calculation was performed using PRSice (http://prsice.info/) with case-control significant PCs as covariates. We correlated AN and OCD PRS with parent- and child-reported continuous and categorical eating disorder phenotypes for boys and girls in the ALSPAC cohort and controlled for multiple comparisons using Benjamini-Hochberg false discovery rate. Results Preliminary results show that, for the AN PRS and OCD PRS, the best fitting p-value thresholds were at pT=.01 and pT= 0.05 respectively. The AN PRS was significantly correlated with adolescent-report of greater thin ideal internalization (r=0.04, p=0.018), dieting (r=0.04, p=0.013), and pressure to lose weight (r=0.023, p=0.05) at age 14. The OCD PRS was significantly positively correlated with any parent or adolescent-reported eating disorder diagnosis by age 18 (r=0.05, p=0.006). Discussion Our analyses provide evidence that AN and OCD PRS can explain a small but statistically significant percentage of the variance in eating disorder phenotypes in adolescence in the general population. In addition, generating a multi-trait genomic best linear unbiased prediction (mtGBLUP) PRS of genetically-correlated disorders has been shown to increase predictive ability equivalent to 34–76% increase in sample size. Thus, we plan to construct a multi-trait AN-OCD PRS and continue to examine its association with adolescent eating disorder cognitions and behaviors in this large population-based cohort.