Introduction: Mechanisms underlying the positive association between occupational mental demands and late-life cognition are poorly understood. The objective of this study was to assess whether the association between occupational complexity and cognition is related to and moderated by brain integrity in individuals at-risk for dementia. Brain integrity was appraised throughout structural measures (Magnetic Resonance Imaging, MRI) and amyloid accumulation (Pittsburgh Compound B (PiB)-positron emission tomography, PiB-PET).

Methods: Participants from the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) neuroimaging sample -MRI (N=126), PiB-PET (N=41)- were included in a post-hoc cross-sectional analysis. Neuroimaging parameters comprised the Alzheimer ' s Disease signature cortical thickness (ADS, Freesurfer 5.3), medial temporal atrophy (MTA), and amyloid accumulation (PiB-PET). Cognition was assessed using the Neuropsychological Test Battery. Occupational complexity with data, people, and substantive complexity were classified through the Dictionary of Occupational Titles. Linear regression models included cognition as dependent variable, occupational complexity, measures of brain integrity, and their interaction terms as predictors.

Results: Occupational complexity with data and substantive complexity were associated with better cognition (overall cognition, executive function) when adjusting for ADS and MTA (independent association). Significant interaction effects between occupational complexity and brain integrity were also found, indicating that, for some indicators of brain integrity and cognition (e.g., overall cognition, processing speed), the positive association between occupational complexity and cognition occurred only among persons with higher brain integrity (moderated association).

Conclusion: Among individuals at-risk for dementia, occupational complexity does not seem to contribute towards resilience against neuropathology. These exploratory findings require validation in larger populations.