How Selection Over Time Contributes to the Inconsistency of the Association between Sex/Gender and Cognitive Decline across Cognitive Aging Cohorts

The sex/gender and aging-related cognitive decline association remains poorly understood due to inconsistencies in findings. Such heterogeneity could be attributable to the cognitive functions studied and study population characteristics, but also to a differential selection by drop-out and death between men and women. This work aims to evaluate the impact of selection by drop-out and death on the association between sex/gender and cognitive decline. We first compared the most frequently used statistical methods for longitudinal data, targeting either population estimands (marginal models estimated by Generalized Estimating Equations) or subject-specific estimands (mixed/joint models estimated by likelihood maximization) on eight aging studies: six population-based (ACTIVE(1996-2009), Paquid(1988-2014), REGARDS(2003-2016), 3-City(1999-2016), WHICAP(1992-2017), Whitehall II(2007-2016)) and two clinic-based (ADNI(2004-2017), MEMENTO(2011-2016)) studies. We illustrated the differences in the estimands of the sex/gender association with cognitive decline in selected examples and highlighted the critical role of differential selection by drop-out and death. By using the same estimand, we then contrasted the sex/gender association across cohorts and cognitive measures suggesting residual differential sex/gender association depending on the targeted cognitive measure (memory or animal fluency) and the initial cohort selection. We recommend focusing on subject-specific estimands in the alive population for assessing sex/gender differences while handling differential selection over time.