Abstract

Abstract Diffusion MRI derived free-water (FW) metrics show promise in predicting cognitive impairment and decline in aging and Alzheimer’s disease (AD). FW is sensitive to subtle changes in brain microstructure, so it is possible these measures may be more sensitive than traditional structural neuroimaging biomarkers. In this study, we examined the associations among FW metrics (measured in the hippocampus and two AD signature meta-ROIs) with cognitive performance, and compared FW findings to those from more traditional neuroimaging biomarkers of AD. We leveraged data from a longitudinal cohort (nparticipants=296, nobservations=870, age at baseline: 73 ± 7 years, 40% mild cognitive impairment [MCI]) of older adults who underwent serial neuropsychological assessment (episodic memory, information processing speed, executive function, language, and visuospatial skills) and brain MRI over a maximum of four time points including baseline (n=284), 18-month (n=246), 3-year (n=215), and 5-year (n=125) visits. The mean follow-up period was 2.8 ± 1.3 years. Structural MRI was used to quantify hippocampal volume in addition to Schwarz and McEvoy AD Signatures. FW and FW-corrected fractional anisotropy (FAFWcorr) were quantified in the hippocampus (hippocampal FW) and the AD signature areas (SchwarzFW, McEvoyFW) from diffusion-weighted (dMRI) images using bi-tensor modeling (FW elimination and mapping method). Linear regression assessed the association of each biomarker with baseline cognitive performance. Additionally, linear mixed-effects regression assessed the association between baseline biomarker values and longitudinal cognitive performance. A subsequent competitive model analysis was conducted on both baseline and longitudinal data to determine how much additional variance in cognitive performance was explained by each biomarker compared to the covariate only model, which included age, sex, race/ethnicity, apolipoprotein-e4 status, cognitive status, and modified Framingham Stroke Risk Profile scores. All analyses were corrected for multiple comparisons using an FDR procedure. Cross-sectional results indicate that hippocampal volume, hippocampal FW, Schwarz and McEvoy AD Signatures, and the SchwarzFW and McEvoyFW metrics are all significantly associated with memory performance. Baseline competitive model analyses showed that the McEvoy AD Signature and SchwarzFW explain the most unique variance beyond covariates for memory (ΔRadj2 = 3.47 ± 1.65%) and executive function (ΔRadj2 = 2.43 ± 1.63%), respectively. Longitudinal models revealed that hippocampal FW explained substantial unique variance for memory performance (ΔRadj2 = 8.13 ± 1.25%), and outperformed all other biomarkers examined in predicting memory decline (pFDR=1.95x10-11). This study shows that hippocampal FW is a sensitive biomarker for cognitive impairment and decline, and provides strong evidence for further exploration of this measure in aging and AD.