|Title||Polygenic Risk Scores Predict Late-life Depressive Symptoms in Older Whites, but Not in Older African Americans: Stratified Analyses by Sex at Birth|
|Publication Type||Journal Article|
|Year of Publication||2023|
|Authors||Park, S, Zamil, K, Patrick, R, Forester, BP|
|Journal||The American Journal of Geriatric Psychiatry|
|Keywords||Depressive symptoms, Older Whites, Polygenic risk scores|
Introduction Given significant racial/ethnic disparities in the prevalence and treatment of late-life depression (LLD), investigating the roles of polygenic risk scores (PRS) will be vital in assessing genetic predispositions to LLD. However, empirical findings have been inconsistent across racial/ethnic groups. For example, PRS for depressive symptoms (PRS-ds) explain ∼0.5% of variance in adults of European ancestry (EA), while no genetic loci reached genome-wide significance in a sample of female African American (AA) adults (Okbay et al., 2016). Given the underrepresentation of AAs in genome-wide association studies (GWAS), the utility of established PRS-ds in explaining LLD symptom trajectories in older AAs remains unclear, highlighting the need for exploration. Moreover, sex differences in the effect of PRS-ds on LLD symptoms remain unknown in racial/ethnic groups, despite accumulated findings on the differential effect of sex on gene expression. Given this research gap, the current study aims to (1) examine the association between PRS-ds and LLD symptoms in older AAs compared to older EAs and (2) explore if there are sex differences in the association between PRS-ds and LLD symptoms by stratifying the analyses by sex at birth in each AA and EA group. Methods Growth curve modeling was conducted on a nationally representative sample of community-living AA (n=3,030) and EA adults (n=11,888) aged 50+, from 12 waves of the Health and Retirement Study (1996-2018). The AA sample (1,860 women and 1,170 men) provided an average of 6.2 observations per subject, while the EA sample (6,761 women and 5,127 men) provided 7.8 observations per subject. Depressive symptoms were measured by scores on the Center for Epidemiologic Studies Depression scale (CESD-8) at each biennial visit. PRS-ds was based on two genome-wide single nucleotide polymorphisms identified from a GWAS meta-analyses (Okbay et al., 2016). In the main analyses, we controlled for multifaceted covariates known to be associated with LLD. The first set of covariates consisted of demographic information, such as age, employment, marital status, education level, and household income. The second set of covariates included diagnoses of medical conditions (i.e., high blood pressure, diabetes, cancer, lung and heart problems, stroke, and arthritis), self-rated health status, health behaviors (i.e., Body Mass Index, drinking, smoking), and family characteristics (i.e., number of children and people in the household). The scores on the top five European or African ancestry-specific principal components (PCs) were included in all main analyses to account for the potential genetic effects of shared ancestry. Results When only ancestry-specific PCs were controlled (Model 1), PRS-ds was positively and significantly associated with LLD symptoms in both EAs (z = 10.48, p < .001, 95% CI: [0.11, 0.16]) and AAs (z = 3.14, p < .005, 95% CI: [0.03, 0.15]). In Model 2 where demographic covariates were added to Model 1, higher PRS-ds significantly predicted more LLD symptoms in both EAs (z = 8.74, p < .001, 95% CI: [0.08, 0.13]) and AAs (z = 2.80, p < 0.01, 95% CI: [0.02, 0.13]). In Model 3 where medical, family, and health-related covariates were added to Model 2, PRS-ds was significantly associated with LLD symptoms in EAs (z = 6.85, p < .001, 95% CI: [0.05, 0.09], but not in AAs (p > .05). Further stratification by sex analyses indicated that PRS-ds was a consistent and significant predictor of LLD symptoms in both EA women and men in all models (ps < .005), irrespective of covariates. However, in AA men, PRS-ds did not reach statistical significance in explaining LLD symptoms in any of the Models (ps > .05). In AA women, the effect of PRS-ds on LLD symptoms was significant when adjusting for PCs and demographic covariates (Model 1 and 2), but nonsignificant after adding more complex covariates in Model 3 (p > .05). Conclusions The current findings elucidate racial/ethnic and sex differences in the effects of PRS-ds in explaining LLD symptoms in older AAs and EAs. While PRS-ds was a robust predictor for LLD symptoms in both EA women and men, the predictive value of PRS-ds on AA's LLD symptoms largely fluctuated depending on the covariates, and importantly, differed by sex. More specifically, the effect of PRS-ds on LLD symptoms in AA women became nonsignificant after accounting for the effects of health and family-related covariates, in addition to demographics and PCs. Furthermore, the predictive value of PRS-ds in AA men was nonsignificant regardless of covariates. In sum, our findings highlight inconsistent prediction accuracy of PRS-ds in older AAs compared to EAs. This reinforces the urgent need for more representation of non-European samples in polygenic research to increase the clinical utility of PRS-ds in underserved racial groups to prevent further disparities in LLD. This research was funded by This study is funded by the 2022-2023 Pope-Hintz Fellowship (PI: Soohyun Park, Ph.D.) awarded at the McLean Hospital, Massachusetts.