Longitudinal fMRI reveals multi-network reconfiguration and a recurrent frontoparietal shift across rest and task states during adult aging

Abstract

Functional connectivity changes across adult aging, but most evidence is cross-sectional, and covariance-based connectivity depends on blood oxygenation level dependent (BOLD) fluctuation amplitude. Covariance decline can reflect changing fluctuation amplitude, and its interpretation may depend on whether normalization uses signal amplitude or includes estimated noise. It is also unclear which longitudinal changes recur across brain states. We analyzed 852 adults (baseline age 36-89 years) from the Aging Adult Brain Connectome with repeated functional MRI (fMRI) during rest, inhibitory control, associative memory, and visual-motor processing (2,220 sessions; median follow-up 5.5 years). Baseline age and elapsed follow-up were modeled separately. Ten cortical networks were tested jointly across states, and covariance was normalized either by released signal amplitude or by a candidate total amplitude reconstructed from released signal and noise amplitudes. Native common covariance declined by 7-11% per 5 years, while signal pair amplitude declined by 9% on average across states, significantly more than estimated noise pair amplitude. Under signal normalization, the relative balance among networks reconfigured over follow-up and differed across states, and the frontoparietal network shifted positively relative to the other nine networks in all four states, by 5-8% per 5 years. The estimated common trajectory was near zero under signal normalization but declined by 4-7% per 5 years under candidate total normalization. The network reconfiguration and the positive frontoparietal-relative direction persisted under both. Within-person coupling between the frontoparietal shift and fluid cognition was not detected. Adult aging thus combines normalization-dependent overall change with a state-dependent redistribution among networks that includes a recurrent frontoparietal shift.