Aging
Adjustment of age for the association of neuroimaging biomarkers and cognition in a diverse cohort of older adults Yi Lor* Yi Lor Alexander Ivan B. Posis Charles DeCarli Paola Gilsanz Elizabeth Rose Mayeda M. Maria Glymour Rachel A. Whitmer Kristen M. George Rachel L. Peterson
Introduction: Age and brain volumes are highly correlated, thus there are disciplinary disagreements on whether age-adjustment is appropriate in models estimating effects of brain volumes on late-life cognition. We examined how different age-adjustment approaches affect associations of MRI-based brain markers and cognitive change.
Methods: Using two harmonized older adult cohorts (Kaiser Healthy Aging and Diverse Life Experience, and Study of Healthy Aging in African Americans), we used linear mixed models to evaluate the associations of volumetric MRI brain regions of interest (ROI) with annual rate of change in executive function and verbal episodic memory over 3 waves (~16 months apart) using five age-adjustment approaches: 1) no adjustment, 2) residual age at MRI estimated from regression models on intracranial volume, within model adjustment for 3) age at baseline cognitive interview, 4) age at MRI scan, and 5) age at both baseline and MRI. All models adjusted for practice effects and interview mode.
Results: Of 484 participants (mean baseline age=74.5; SD=6.3, mean MRI age=87.7; SD=8.9), older age was associated with worse brain ROI volumes and lower baseline cognitive scores. Multiple ROI were associated with lower baseline cognition and faster rate of change. The ROI-cognition association was smaller in all age-adjusted models versus models without age adjustment (Figure 1). However, associations with cognitive change did not differ after age adjustment (Figure 1), regardless of adjustment approach. Model fit statistics (AIC, BIC, and log-likelihood) suggest that adjustment for both baseline age and age at MRI yield the best fit.
Discussion: Age adjustment altered the estimated association of MRI markers and baseline cognition but not cognitive change. Due to the relatively short follow-up time, age-adjustment may not have affected cognitive change. Studies that adjust for age may be underestimating the association of MRI brain markers with cognition.
