Biomarkers.

Background: The brain's ability to perform a cognitive task is a dynamic process and requires small blood vessels to dilate or constrict in real time to adjust blood flow in a region-specific manner. Cerebrovascular Reactivity (CVR) measures the ability of vessels to react to vasoactive challenges. In this work, we investigated the role of CVR as a possible biomarker in small vessel disease related vascular contributions to cognitive impairment and dementia (VCID), as part of the NINDS-funded MarkVCID study.

Methods: This work consisted of 4 studies. CVR was measured by MRI signal changes in response to inhalation of 5% carbon dioxide (CO2) (Figure 1). Study 1 was a single-site study of 72 participants (aged 69±8 years, 28 normal cognition and 44 cognitive impairment), in which we examined the relationship between CVR and cognitive function, CSF-measured AD pathological markers, and CDR-sum-of-boxes. Study 2 conducted an instrumental validation to benchmark the test-retest reproducibility of the CVR measure. Study 3 developed an automatic, cloud-based tool for automatic and standardized processing of CVR MRI data. Study 4 was a multi-site biological validation study in which CVR was measured across 3 sites and data from each site were separately analyzed to investigate the relationship between CVR and cognitive function.

Results: Study 1 revealed that whole-brain CVR was associated with MoCA (β=29.64, 95% CI, 9.94 to 49.34) and global composite scores (β=4.32, 95% CI, 0.05 to 8.58), and these associations were independent of CSF-measured amyloid and tau markers. CVR was also associated with CDR-SB score. Study 2 revealed that inter-scanner CoV of CVR was 6.90 ± 5.08% and had an ICC of 0.8498. Study 3 showed that CVR processing can be standardized and the processing pipeline can be accessed by any researchers around the world through an installation-free cloud platform referred to as MRICloud (https://braingps.mricloud.org/cvr.v5). Study 4 showed that CVR was significantly associated with MoCA in data analyzed separately in each of the three sites through a pre-defined analysis (p=0.003, 0.002, 0.046 for the three sites, respectively; when pooling data together p=3.7*10^-6) (Figure 2).

Conclusion: CVR measured with MRI is a promising candidate biomarker in VCID.