INTRODUCTION

Resilience, the ability to maintain cognition or brain integrity despite Alzheimer's disease (AD) pathology, is often quantified using the residual approach. However, the variability in methodology and correction methods for this approach raises concerns about the interpretability of findings across studies.

METHODS

We assessed brain resilience (BR) and cognitive resilience (CR) in a memory clinic population using the residual approach. We compared non-corrected and corrected residuals' associations with risk factors using linear regression models, and their impact on longitudinal cognition using linear mixed-effects models.

RESULTS

Corrected versus non-corrected BR yielded distinct, often opposing, associations. For example, glial fibrillary acidic protein (GFAP) was negatively associated with non-corrected BR (β = -0.33; p < 0.01) but positively with corrected BR (β = 0.5, p < 0.001). Only corrected CR measures yielded significant associations. Only corrected residuals predicted cognitive decline.

DISCUSSION

The observed discrepancies raise questions about the reliability of the residual approach in accurately capturing resilience.

HIGHLIGHTS

Corrected and non-corrected residuals show distinct associations with risk factors. Corrected and non-corrected residuals show different predictions of cognitive decline. These approaches may reflect general brain health rather than true resilience mechanisms.