Obesity related to antipsychotic liability and exposure (oracle): preliminary results

The NIMH-funded Obesity Related to Antipsychotic (AP) Risk and Exposure (ORAcLE) study leverages population-level data from large biobanking initiatives, epidemiological studies, and randomized clinical trials (RCTs) to identify genetic contributors to antipsychotic induced weight gain (AIWG). In the proposed presentation, we will describe the study design and characteristics of the data collected to date, as well as an examination of the relevance of polygenic risk scores (PRS) for cross-sectional body mass index (BMI) in predicting AIWG in the Geisinger MyCode study population, part of the Geisinger MyCode Community Health Initiative. Using the current interim genome wide association study (GWAS) of BMI by the Genetic Investigation of ANthropometric Traits (GIANT) consortium, which includes >1.1 million European participants, we derived PRS for cross-sectional BMI, examining change in BMI during treatment with an atypical antipsychotic. Individuals in the MyCode data set with > 2 BMI measures > 30 days apart while prescribed an atypical antipsychotic were included. A linear mixed model, where BMI was regressed with intercept and time (years) as both fixed and random effects, was employed to derive the slope of AIWG, stratified by sex and ancestry. Analyses were conducted for overall change in BMI in all eligible individuals, as well as for “gainers” only. PRS weights were estimated in SBayesR. Association analyses between standardized PRS and AIWG were done using generalized linear models while accounting for families using general estimating equation (GEE). To date, the ORAcLE consortium has identified >26,000 individuals across 4 cohort studies, 7 RCTs and 3 health systems. Specifically, we have relevant data on approximately 17,500 children, adults, and older adults from health systems data; 7,200 adults and older adults with relevant exposure from cohort studies; and approximately 1,500 children, adults, and older adults with relevant AP exposure from RCTs. In the PRS analysis, there were 8,493 participants (97% European ancestry, 68% female, 61% who gained weight while prescribed antipsychotic medication) with usable data who met criteria for inclusion in the analysis. There were 33 observations (SD=58) per individual participant, with a mean of 3.7 yrs (3.7) of follow-up, and AIWG of 0.18/yr (0.93). The final PRS included 1,080,060 variants and was significantly associated with baseline BMI (P=1e-305, β=4.03, R2=14.4%). Significant relationships between the cross-sectional BMI PRS and overall AIWG (R2=0.3%, P=3.9e-8, β=0.069 BMI units/year) and AIWG in gainers only (R2=0.3%, P=6.5e-6, β=0.051 BMI units/year), respectively, were detected. Finally, a small but significant difference (0.125 BMI/yr, P=2e-4) in adjusted mean AIWG between the lowest and upper quintiles of the PRS was also observed. These preliminary results demonstrate that the PRS derived from the GIANT Consortium's BMI GWAS can feasibly detect variation in genetic contributions to AIWG. This result extends observations from smaller studies that there is a clear heritable component to AIWG that is detectable using a PRS originally designed to detect genetic risk for obesity. The increase in percent variance explained in AIWG by the cross-sectional BMI PRS was small but significant, suggesting that genetic contributions to AIWG risk may involve variants not associated with underlying obesity risk.