Characterizing mood disorders in the AFFECT study: a large, longitudinal, and phenotypically rich genetic cohort in the US

Abstract

There has recently been marked progress in identifying genetic risk factors for major depression (MD) and bipolar disorder (BD); however, few systematic efforts have been made to elucidate heterogeneity that exists within and across these diagnostic taxa. The Affective disorders, Environment, and Cognitive Trait (AFFECT) study presents an opportunity to identify and associate the structure of cognition and symptom-level domains across the mood disorder spectrum in a prospective study from a diverse US population.Participants were recruited from the 23andMe, Inc research participant database and through social media; self-reported diagnosis of MD or BD by a medical professional and medication status data were used to enrich for mood-disorder cases. Remote assessments were used to acquire an extensive range of phenotypes, including mood state, transdiagnostic symptom severity, task-based measures of cognition, environmental exposures, personality traits. In this paper we describe the study design, and the demographic and clinical characteristics of the cohort. In addition we report genetic ancestry, SNP heritability, and genetic correlations with other large cohorts of mood disorders.A total of 48,467 participants were enrolled: 14,768 with MD, 9864 with BD, and 23,835 controls. Upon enrollment, 47% of participants with MD and 27% with BD indicated being in an active mood episode. Cases reported early ages of onset (mean = 13.2 and 14.3 years for MD and BD, respectively), and high levels of recurrence (78.6% and 84.9% with >5 episodes), psychotherapy, and psychotropic medication use. SNP heritability on the liability scale for the ascertained MD participants (0.19-0.21) was consistent with the high level of disease severity in this cohort, while BD heritability estimates (0.16-0.22) were comparable to reports in other large scale genomic studies of mood disorders. Genetic correlations between the AFFECT cohort and other large-scale cohorts were high for MD but not for BD. By incorporating transdiagnostic symptom assessments, repeated measures, and genomic data, the AFFECT study represents a unique resource for dissecting the structure of mood disorders across multiple levels of analysis. In addition, the fully remote nature of the study provides valuable insights for future virtual and decentralized clinical trials within mood disorders.

Fig. 1. Flowchart illustrating the enrollment.

The procedural steps were: Informed consent, apply for enrollment and meet study inclusion and no exclusion criteria, return a saliva kit for genotyping (except for excisting costumers who purchased and returned a 23andMe kit prior to joining the study), and answer the baseline questionnaire. In the 9 months after enrolment, participants were asked to complete monthly surveys and cognitive tests.

Fig. 2. Baseline clinical features.

A Summary of key clinical features in cases reporting a diagnosis of MDD, BD subtype 1 (BD1) and BD subtype 2 (BD2), as per latest diagnosis recieved. Mood disorders cases in this 23andMe sub-cohort show high burdens of illness. Any medication class refers to medication received over the last 5 years and during the study. Percentage of those who answered one or several treatment questions in the medication survey. B Transdiagnostic symptoms. Radar plot of median score pr. symptom domain within controls (purple), MD (blue), and BD (orange) participants. Scores are based on the DSM-5 cross-cutting symptom measures, where max item score (ranging from 0 to 4) within each domain is reported and summarized.

Fig. 3. Genetic component.

A Liability-scale SNP-heritability of AFFECT BD and MD as a function of population prevalence, ranging from 0.001 to 0.03 for BD and 0.001–0.3 for MD with r_g estimates at every 0.001 step-wise increase. Dotted line represents s.e. B Estimated genetic correlations of European ancestry AFFECT BD and MD with PGC GWAS of MDD3 (excluding the 23andMe cohort), the 23andMe MD discovery cohort (Hyde et al, 2016), PGC-BD2, and of PGC-BD3, which is further divided into BD3 type I and type II. Correlations in AFFECT BD were performed with the full cohort (BD) and within BD type (BD1, BD2). All correlations were significant, circle size and values indicate r_g. P-values, Z-scores and s.e are reported in Supplementary Table 9.