Age-related change in task-evoked amygdala-prefrontal circuitry: a multiverse approach with an accelerated longitudinal cohort aged 4-22 years

There has been considerable interest in the development of the amygdala and its connections with medial prefrontal cortex (mPFC) given the central role of these brain regions in emotional processes. While several studies have suggested that this circuitry exhibits functional changes across the first two decades of life, they have typically employed cross-sectional designs, and findings have been mixed. Additionally, analytic choices may contribute to discrepancies across studies. Here we used an accelerated longitudinal design to examine task-evoked changes in amygdala-mPFC circuitry from 4-22 years of age (N=98; 183 total scans; 1-3 scans per participant). Participants were recruited from the greater Los Angeles area, and completed an event-related emotional face (fear, neutral) task designed to be appropriate for the wide age range. 9Multiverse9 analyses examined the robustness of our findings to fMRI analysis choices. 2808 parallel analyses varying in preprocessing and modeling choices found evidence for average age-related decreases in amygdala reactivity to faces. Greater amygdala reactivity at younger ages was attributable to elevated responses during the first few trials relative to later trials. Within-participant changes in amygdala reactivity with age could not be differentiated from between-participant differences, however. Across analysis decision points, we did not find consistent evidence of age-related change in amygdala-mPFC connectivity through generalized psychophysiological interaction (gPPI) or beta-series correlation (BSC) methods. We also did not find evidence for associations between separation anxiety behaviors and amygdala reactivity or amygdala-mPFC connectivity. Within the context of this faces task and age range, age-related changes in amygdala reactivity were more robust to processing pipeline than were task-evoked functional connectivity measures, particularly those using gPPI. These findings highlight both the challenges in estimating developmental change in longitudinal cohorts and the value of multiverse approaches in developmental neuroimaging for assessing robustness of results.