Depression is a complex and high-burden disorder, and predicting treatment response remains a major clinical challenge. This study aimed to develop a predictive algorithm for treatment response in depression using sociodemographic data, self-reports, ...
Depression is a complex and high-burden disorder, and predicting treatment response remains a major clinical challenge. This study aimed to develop a predictive algorithm for treatment response in depression using sociodemographic data, self-reports, and plasma proteomic profiles. Patients with moderate to severe depression were divided into two groups according to treatment response. No significant differences were observed between responders and non-responders in hospital, diagnosis, medication, clinical scales, medication adherence, or observed depressive and anxiety symptoms. However, non-responders reported higher levels of obsessive-compulsive symptoms, phobic anxiety, and interpersonal sensitivity, along with lower satisfaction in social relationships. The predictive algorithm for treatment response showed improved performance when subjective symptoms were incorporated, compared to models adjusted only for basic demographic and clinical variables. The model including all five self-reported symptoms demonstrated high predictive accuracy, with an AUROC exceeding 0.7.
Additionally, linear and logistic regression analyses were performed to identify plasma proteins significantly associated with treatment response. The baseline levels of IGLV9-49, MELTF, F12, SAA1, C4BPB, and SPARCL1, as well as the level changes of ACTN1, SOD3, SAA1, HSPA5, FLNA, and ADCY1, were statistically associated with treatment response. Proteins identified in both analyses were involved in platelet activation, complement, and coagulation pathways associated with immune and inflammatory processes. The integrative predictive model combining proteomic data with self-reports achieved a maximum AUROC of 0.855 (95% CI: 0.797–0.913), with baseline protein levels showing greater predictive power than longitudinal changes.
These findings suggest that combining baseline proteomic profiles with self-reported symptom data can substantially enhance the performance of predictive algorithms for treatment response in depression.