BackgroundAge at first onset of depression as a clinical factor affecting cognitive improvement in late life depression was investigated. MethodsThis is a secondary analysis of an eight-week randomized controlled trial involving 452 elderly patients treated by vortioxetine, duloxetine or placebo (1:1:1). Patients were subcategorized into early-onset (LLD-EO) and late-onset (LLD-LO) groups divided by onset age of 50. Cognitive performance was assessed by composite score of Digit Symbol Substitution Test (DSST) and the Rey Auditory Verbal Learning Test (RAVLT) tasks, while depressive symptoms were assessed by Montgomery–Åsberg Depression Rating Scale (MADRS). ResultsVortioxetine and duloxetine exhibited advantages versus placebo in improving cognitive performance in the LLD-LO group, yet not in the LLD-EO group after eight weeks. Patients in the LLD-EO group showed overall advantage to placebo in depressive symptoms before endpoint (week 8) of treatment, while patients in the LLO-LO group showed no advantage until endpoint. Path analysis suggested a direct effect of vortioxetine (B = 0.656, p = .036) and duloxetine (B = 0.726, p = .028) on improving cognition in the LLD-LO group, yet in all-patients treated set both medications improved cognition indirectly through changes of depressive symptoms. LimitationReliability of clinical history could raise caution as it was collected by subjective recall of patients. ConclusionAge at first onset might affect cognitive improvement as well as change in depressive symptoms and its mediation towards cognitive improvement in late life depression treated with vortioxetine and duloxetine.