| Abstract: | Background :Major depressive disorder (MDD) is a prevalent mental illness with a significant burden, affecting 5% of adults worldwide. About one-third of MDD patients suffer from treatment resistance despite adequate dosage and duration of multiple anti-depressants (ADs). Previous pharmacogenetic studies suggest that the discrepancy in medication response may partially arise from genetic variability. However, these studies are often limited by sample size and exhibit heterogeneity in study designs and definitions of treatment-resistant depression (TRD). To conduct a well-powered genetic investigation considering a range of TRD phenotypes within a single source, we leveraged the real-world primary care data from the UK Biobank. Methods: We identified 15,528 MDD patients of White British descent with no prior psychiatric comorbidities. By examining AD prescription patterns, we constructed various proxy TRD definitions (TRDp) to identify patients who had ever changed ≥2 ADs (Change_AD: = 1,879); received ≥3 types of ADs (Receive_AD: 4,671); used augmentation alongside AD (AUG_Antipsychotics: 3,110; Mood stabilizers: 2,717; Valproate: 305; Lithium: 287); or undergone electroconvulsive therapy (ECT: 85). To increase the confidence of phenotyping, we considered symptom severity and additionally identified hospitalized TRDp patients (HOSP-TRDp). We then performed genome-wide association studies (GWAS) for each TRDp definition vs. non-TRDp or control group and estimated SNP-heritability (h2) using LD Score Regression. Finally, we assessed the genetic burden of nine common psychiatric diseases in TRDp through polygenic risk score (PRS) analysis. Results :MDD patients classified with TRDp tended to be females and have a higher rate of unemployment, lower education and income levels, and a higher BMI. Hospitalization was twice as common in TRDp as in non-TRDp across definitions, highest for ECT (65% vs. 26%). Although no credible single variant discoveries emerged from GWAS, heritability analysis revealed a significant genetic influence on TRDp, even larger for hospitalized patients (liability h2 = 9%, 10-20%, 15-25% for non-TRDp, TRDp, and HOSP-TRDp vs. control; 18-23%, 28-32% for TRDp and HOSP-TRDp vs. non-TRDp). TRDp identified as Change_AD, Receive_AD, AUG_A, or AUG_M carried a stronger polygenic burden of MDD, ADHD, BD, SCZ, and alcohol dependence than non-TRDp (odds ratio [OR] per SD = 1.05-1.14), with PRS of BD and SCZ more enriched in AUG_A than in other definitions. More profoundly, MDD patients with a higher BD PRS were more likely to receive intensive treatments like ECT or lithium, valproate augmentation (ORs = 1.73, 1.52, and 1.35, p <0.001). Lastly, compared to the average PRS group, those in the highest 10% were at a 15-20% elevated risk of experiencing TRDp, and the effects were more pronounced in the HOSP-TRDp group (e.g., ORBD = 1.53 [95% CI: 1.3-1.8] for AUG_A; ORADHD = 1.43 [1.2-1.7] for AUG_M). Discussion : Our findings support a significant role of common genetic variations in the etiology of TRD, underscoring a polygenic basis. While challenging to accurately ascertain treatment response from medical records, our results highlight distinction between TRDp vs. non-TRDp both genetically and phenotypically. By employing broader to more extreme phenotyping (e.g., ECT), we also showed how various TRDp definitions may capture the underlying genetic liability to a spectrum of mental illnesses, together advancing our understanding of the genetic architecture of TRD. |
