ObjectiveMajor depressive disorder in older adults (late-life depression; LLD) is frequently associated with cognitive impairment, and some deficits (e.g., executive function) have been associated with a higher level of treatment resistance. However, the cognitive profile of treatment-resistant LLD (TR-LLD) has not been characterized. We hypothesized that patients with TR-LLD would show deficits in cognitive function, especially executive function, and that executive function deficits would predict poorer response to pharmacotherapy. DesignSecondary analysis of baseline cognitive data from OPTIMUM, a multicenter RCT evaluating pharmacotherapy strategies for TR-LLD. SettingFive outpatient academic medical centers (4 US, 1 Canada). ParticipantsAbout 369 participants aged 60 and older from the OPTIMUM study. MeasurementsBaseline scores on individual tasks and composite scores from the NIH Toolbox-Cognition Battery were transformed into demographically-adjusted T-scores and compared to published norms. Impairments in the set shifting and inhibitory control tasks were investigated as predictors of depressive symptom change following treatment using ANCOVA models. ResultsParticipants had low performance on tasks evaluating inhibitory control, processing speed, verbal/nonverbal memory, and the fluid composite, but normative performance on working memory and set shifting. Participants had high estimated premorbid IQ (superior Performance on oral reading recognition). Age and physical comorbidity negatively associated with processing speed. Impairments in set shifting predicted less improvement in depressive symptoms; impairments in inhibitory control did not. ConclusionsParticipants with TR-LLD presented with broad cognitive deficits relative to healthy norms. Given poorer outcomes following standard pharmacotherapy associated with impaired set shifting, future research needs to identify alternative treatment strategies.