Longitudinal Developmental Trajectories Do Not Follow Cross-Sectional Age Associations in Hippocampal Subfield and Memory Development

Abstract Many cross-sectional findings suggest that volumes of specific hippocampal subfields increase in middle childhood and early adolescence. In contrast, a small number of available longitudinal studies observed decreased volumes in most subfields over this age range. Further, it remains unknown whether structural changes in development are associated with corresponding gains in children’s memory. Here we report cross-sectional age differences in children’s hippocampal subfield volumes together with longitudinal developmental trajectories and their relationships with memory performance. In two waves, 109 healthy participants aged 6 to 10 years (wave 1: MAge=7.25, wave 2: MAge=9.27) underwent high-resolution magnetic resonance imaging to assess hippocampal subfield volumes, and completed cognitive tasks assessing hippocampus dependent memory processes. We found that cross-sectional age-associations and longitudinal developmental trends in hippocampal subfield volumes were highly discrepant, both by subfields and in direction. Further, volumetric changes were largely unrelated to changes in memory, with the exception that increase in subiculum volume was associated with gains in spatial memory. Importantly, the observed longitudinal patterns of brain-cognition coupling could not be inferred from cross-sectional findings. We discuss potential sources of these discrepancies. This study underscores that children’s structural brain development and its relationship to cognition cannot be inferred from cross-sectional age comparisons. HighlightsThe subiculum undergoes volumetric increase between 6-10 years of ageChange across two years in CA1-2 and DG-CA3 was not observed in this age windowChange across two years did not reflect age differences spanning two yearsCross-sectional and longitudinal slopes in stark contrast for hippocampal subfieldsLongitudinal brain-cognition coupling cannot be inferred from cross-sectional data