Patients with treatment-resistant unipolar depression (TRD) are treated with antidepressant combinations (ADs) or with second-generation antipsychotics plus AD (SGA+AD) augmentation; however, the clinical characteristics, the factors associated independently with response to SGA+AD, and the outcome trajectories have not yet been characterized. We performed a naturalistic study on the latest stable trial (medication unchanged for about 3 months) in 86 TRD patients with resistance to at least two ADs trials, who received ADs (n=36) or SGA+AD (n=50) treatments. Montgomery–Asberg Depression Rating Scale (MADRS), Hamilton-Depression Rating Scale (HAM-D17), and other scales were administered before (T0) and after the latest 3-month stable trial (T3). Compared to ADs, the SGA+AD group showed increased percentage of depression with psychotic features, comorbidity for personality disorders and substance use disorders (SUD), higher number of failed ADs pharmacotherapies and depressive symptoms at T0 on all scales (P<0.001). Compared to T0, both treatments significantly decreased depressive symptoms on MADRS and HAM-D17 at T3 (P<0.001); however, the SGA+AD augmentation produced a greater decline in mean score. Logistic regression analysis indicated that psychotic features, personality disorders, and SUD were independently associated with SGA+AD treatment. Given the greater improvement in depression following SGA+AD augmentation, SGA augmentation should be indicated as a first-line treatment in severe TRD with psychotic features, SUD, and personality disorders.