Age‐Related Variations in Cerebrovascular Reactivity Measured With Resting‐State BOLD MRI

ABSTRACT Cerebrovascular reactivity (CVR) is a crucial physiological marker of vascular health and has been linked to aging‐related cerebrovascular decline. Resting‐state BOLD MRI‐based relative CVR mapping (RS‐rCVR) offers a noninvasive and compliance‐friendly alternative to gas‐challenge methods, making it suitable for lifespan studies. This study aimed to examine age‐related differences in RS‐rCVR among healthy adults using voxel‐wise and region‐of‐interest (ROI) analyses. We prospectively recruited 54 healthy adults, including 27 younger (20–28 years, mean = 23.3 ± 3.4; 16 females) and 27 older (57–75 years, mean = 66.5 ± 5.3; 16 females) participants. Resting‐state fMRI data were acquired using a T2*‐weighted gradient‐echo EPI sequence at 3T. RS‐rCVR maps were generated by linear regression of the voxel‐wise BOLD time series against the global BOLD signal and normalized to cerebellar gray matter. Group‐level analyses included voxel‐wise comparisons, histogram analyses, and ROI‐based statistical tests. The distribution of RS‐rCVR values significantly differed between age groups (Kolmogorov–Smirnov test: KS = 0.039, p < 0.001). Voxel‐wise comparisons revealed age‐related reductions in RS‐rCVR in the medial frontal cortex, precuneus, and cuneus in older adults. In contrast, ROI‐averaged RS‐rCVR values showed no statistically significant group differences across frontal, temporal, and occipital cortices ( p > 0.05). Effect size analysis indicated small to moderate differences in specific regions (e.g., occipital cortex: d = 0.441; parietal cortex: d = 0.430; frontal middle gyrus: d = 0.275), but negligible effects in others (e.g., cingulate cortex: d = 0.006). While voxel‐wise RS‐rCVR mapping detects spatially localized age‐related reductions in cerebrovascular reactivity, ROI‐based analysis may obscure these effects due to anatomical averaging. These findings underscore the spatial heterogeneity of cerebrovascular aging and support the utility of voxel‐level RS‐rCVR approaches in lifespan research.