Longitudinal association between structural brain change and episodic memory change

1087 

Abstract Submission 

Marie Strømstad¹, Didac Vidal-Piñeiro¹, Øystein Sørensen¹, Anders Fjell¹, Kristine Beate Walhovd¹

¹Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo, Oslo, Norway

Marie Strømstad  
Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo
Oslo, Norway

Didac Vidal-Piñeiro  
Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo
Oslo, Norway

Øystein Sørensen  
Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo
Oslo, Norway

Anders Fjell  
Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo
Oslo, Norway

Kristine Beate Walhovd  
Department of Psychology, Center for Lifespan Changes in Brain and Cognition, University of Oslo
Oslo, Norway

Longitudinal studies are crucial for understanding the intra-individual variability in structural brain changes and their associations with cognitive functions. While correlations between episodic memory decline and gray-matter atrophy have been found (Gorbach et al., 2020, Sele et al., 2021, Oschwald et al., 2019), extensive investigation into these associations are limited, with studies often focusing on specific regions (e.g. hippocampus) and relying on relatively small sample sizes. Hence, here we have harmonized several longitudinal datasets to study the relationship between age-related episodic memory change and regional structural brain changes (cortical thickness and subcortical volumes). We use a mega-analytical to aggregate the different datasets and maintain statistical power (Eisenhauer, 2021). The main aim of the study is to explore the regional brain-memory change associations and their interaction with age.

Longitudinal neuroimaging and cognitive data from 13 datasets (Lifebrain cohorts and open-sharing datasets, see Figure 1), including 3,763 cognitively healthy participants (1,744 females; mean age = 62.5, age range 16.8-93 years), were analyzed. To harmonize cognitive data, we fitted – in each dataset - memory scores with age, sex, and retest effects as covariates using generalized additive mixed models (GAMM). Structural MRI data was processed with longitudinal FreeSurfer (version 7.1.0, http://surfer.nmr.mgh.harvard.edu), parcellated using the Destrieux and aseg atlases, and fed into a normative modeling pipeline (Rutherford et al., 2022). The slopes of change for both memory and brain data were estimated using linear models per subject using follow-up time as a predictor. The mega-analysis included linear mixed-effects models on each ROI, with memory change as a function of brain change. To account for varying reliability of measurements across datasets, weights derived from squared intraclass correlation coefficient were used in the models (longitudinal reliability estimated as in Fitzmaurice et al., 2012). For the age interaction we applied a GAMM with a tensor interaction of brain change and age on memory change. We also conducted a supplementary analysis with a meta-analytical approach, using the R metafor package (Viechtbauer, 2010). All models were corrected for multiple comparisons using False Discovery Rate (FDR).

57 memory - brain change associations were significant across several regions (p < 0.05, FDR corrected, see Figure 2a). Specifically, positive associations were found in the medial temporal regions, including the hippocampus, amygdala and parahippocampal gyrus, as well as the medial frontal sulcus, posterior cingulate gyrus, inferior frontal gyrus, and precentral sulcus. We also found significant age interaction effects in many of these regions, including the left hippocampus, amygdala and parahippocampal gyrus, where the effect increased with higher age (see Figure 2b). The meta-analysis results were similar, with highly correlated estimates across regions between the methods (r = .72), although fewer regions survived multiple comparison corrections due to higher error estimates.

The results reveal significant change-change correlations between brain and episodic memory in distributed cortical and subcortical regions with known links to memory processing. The strength of these relationships tended to increase with age.

Long-Term Memory (Episodic and Semantic) ¹

Aging

Databasing and Data Sharing

Anatomical MRI ²

Aging

Cognition

Cortex

Meta- Analysis

MRI

Open Data

STRUCTURAL MRI

Other - Memory