Cross-sectional versus longitudinal estimates of age-related changes in the adult brain: overlaps and discrepancies

The healthy adult brain undergoes tissue volume decline with age, but contradictory findings abound regarding rate of change. To identify a source of this discrepancy, we present contrasting statistical approaches to estimate hippocampal volume change with age based on 200 longitudinally-acquired magnetic resonance imaging in 70 healthy adults, age 20-70 years, who had 2-5 magnetic resonance imaging collected over 6 months to 8 years. Linear mixed-effects modeling using volume trajectories over age for each subject revealed significantly negative slopes with aging after a linear decline with a suggestion of acceleration in older individuals. By contrast, general linear modeling using either the first observation only of each subject or all observations treated independently (thereby disregarding trajectories) indicated no significant correlation between volume and age. Entering a quadratic term into the linear model yielded a biologically plausible function that was not supported by longitudinal analysis. The results underscore the importance of analyses that incorporate the trajectory of individuals in the study of brain aging.