Phenotype-by-phenome-wide association study of treatment resistant depression

Abstract

AbstractObjectiveTreatment-resistant depression (TRD), defined as inadequate response to at least one or at least two antidepressant (AD) trials, is common in major depressive disorder (MDD). In this study, electronic health records (EHR) were used to identify clinical associations with TRD.MethodsUsing two biobanks, phenomes of patients with at least one MDD-related diagnostic code and one AD prescription (N=17,049) were generated using aggregated diagnostic codes (phecodes) from EHRs. Phenotype-by-phenome-wide association analyses were performed for two binary definitions of TRD, based on either one or more, or two or more, AD switches after at least 30 days but within 14 weeks, and a quantitative measure defined as the number of unique ADs prescribed for at least 30 days.ResultsOf the 17,049 patients with MDD, 1624 (9.5%) had at least one switch, 422 (2.5%) had at least two switches, and the number of unique antidepressant prescriptions ranged from one to twelve. After accounting for multiple testing, 142, 18, and 7 phecodes were significantly associated with the quantitative definition and the two binary definitions (≥1 AD switch or ≥2 AD switches), respectively. All three outcomes were significantly associated with known TRD risk factors including anxiety disorders, insomnia, and suicidal ideation. The quantitative measure was uniquely associated with other conditions including irritable bowel syndrome and decreased white blood cell count.ConclusionsIn addition to identifying known clinical associations, the quantitative measure of treatment resistance uncovered new factors potentially associated with TRD. This measure may also facilitate discovery of genetic correlates of TRD in future analyses.