Polygenic and familial contributions to antidepressant continuation, switching, discontinuation and augmentation in the All of Us and Pharmlines cohorts

Polygenic scores for depression, schizophrenia and bipolar disorder are tested against treatment trajectories (continuation, switching, discontinuation, augmentation) to separate drug-specific non-response from a general propensity to modify treatment.

The authors analysed participants with at least one antidepressant monotherapy episode of 28 days or more in All of Us (n = 98,357) and Pharmlines (Lifelines linked to IADB.nl, n = 12,884), comparing continuation with switching, discontinuation and augmentation with an atypical antipsychotic or lithium. Among individuals with recorded major depressive disorder, switching, but not discontinuation, was associated with anxiety, higher depression symptom count and stress-related measures in both cohorts. Depression polygenic score was associated with switching in All of Us (OR = 1.16 per SD, 95% CI 1.12-1.20), with a concordant nominally significant estimate in Pharmlines (OR = 1.11, 1.01-1.22), and rose progressively from continuation to switching to augmentation (per-step OR = 1.18, 1.15-1.21), whereas schizophrenia and bipolar disorder scores were selectively associated with augmentation. No polygenic score showed drug-class-specific associations with switching. Familial aggregation was detectable for continuation, including SSRI and SNRI continuation, but not for switching or discontinuation.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The most useful result is a negative one: the absence of drug-class-specific associations suggests these scores capture severity and psychiatric complexity rather than molecule-specific pharmacokinetics or pharmacodynamics, which should temper commercial offerings marketing polygenic scores as antidepressant response predictors. Effect sizes remain around OR 1.16 per standard deviation, with no individual decision value in the clinic. This is a preprint, and treatment trajectories are reconstructed from dispensing and coding data, an imperfect proxy for the actual clinical decision.

Analysis by Dr Thibaut Benquey