Associations of patient‐specific resting state variability with CSF and clinical biomarkers: Data from Alzheimer’s disease clinical trials

Abstract Background Changes of brain functional connectivity assessed using MRI in patients with Alzheimer’s disease (AD) have been previously reported. However, factors contributing to the variability among patients have not been sufficiently investigated. Here, we investigated the association of patient‐specific resting state parameters with CSF markers and clinical measures. Method We analyzed the baseline rs‐fMRI of 158 patients (87 female, age 50‐89) with mild AD collected in the Marguerite RoAD trial of gantenerumab in mild AD (NCT02051608) and 132 cognitively unimpaired adults (88 female, age 56‐90) from ADNI. We investigated the within‐ and between‐subjects’ variability in five brain networks (Sensory Motor, Cerebellar, Visual, Default Mode and Cognitive Control), inferred from the two groups using a Bayesian probabilistic model (PROFUMO, Harrison‐Smith‐2015; Harrison‐Woolrich‐2020). We computed spatial maps, BOLD amplitude changes, and spatial (temporal) functional connectivity between‐ and within‐networks for each subject. Those metrics were compared between groups using T‐tests. In AD patients we used linear models to investigate associations of the extracted metrics with CSF markers (Amyloid Beta 42, p‐Tau217, pTAU/Abeta 1‐42 ratio and Neurogranin) and cognitive scores (ADAS‐Cog 13, MMSE and ADCS‐ADL), while controlling for age, sex and MRI‐scanner type. All results were corrected for multiplicity using FDR. Result On average the signal amplitude was lower in the Cognitive Control and the Sensory Motor (p < 0.0001) networks in AD patients compared to cognitively unimpaired adults. Interestingly, in AD patients the lower amplitude observed in the cognitive control network was associated with cognitive decline (β = 0.02, p = 0.01). Moreover, significant associations were observed in AD patients between lower signal amplitude and higher CSF levels of ABeta 42 (β = ‐0.02, p = 0.036) and cognitive decline (β = 0.02, p = 0.008) (Fig.2). AD group showed on average reduced functional connectivity compared to the healthy one. Furthermore, the within‐AD patients’ analysis showed weak between‐ and strong within‐networks connectivity (T‐test, q < 0.05, Fig.1). Conclusion In patients with mild AD, the signal amplitude and connectivity in the cognitive control and DMN networks were associated with both CSF level of ABeta42 and cognitive scores.