Abstract
Background: Antidepressant selection in major depressive disorder remains trial-and-error, and no biomarker identifies which drug will work for an individual patient. Because neuroplasticity is a convergent mechanism of antidepressant response, we tested whether it could serve as a drug-specific biomarker in patient-derived neurons. Methods: We developed the Neuroplasticity Score (NPS), integrating structural and synaptic features from high-content imaging of patient-derived cortical neurons, and validated it against graded brain-derived neurotrophic factor (BDNF) exposure in neurons from 16 healthy donors. NPS was then applied unmodified to 75 antidepressant treatment episodes from 44 patients across two independent retrospective cohorts: 40 episodes from 20 STAR*D participants (Cohort 1; USA multi-site) and 35 episodes from 24 participants at Sheba Medical Center and Geha Mental Health Center (Cohort 2; Israel). Results: NPS increased with escalating BDNF dose (Spearman rho=0.83; P<.001). Response rate rose monotonically across NPS tertiles in both cohorts (trend P=.001 and P<.001). Across all 75 episodes, 84.6% of episodes in the highest NPS tertile responded, compared to 48.0% overall. NPS discriminated responding from non-responding episodes with an area under the receiver operating characteristic curve of 0.81 in both cohorts. The pooled odds ratio per 1-SD increase was 6.05 (95% CI, 2.80-13.05; I-squared=26%). Within patients contributing both a responding and a non-responding episode, the highest-scoring drug was the one that worked in 86% of patients (P<.001). Conclusions: Higher antidepressant-induced neuroplasticity in a patient's own neurons corresponded to a higher likelihood of response to that specific drug across two independent cohorts.