Association of Cerebral Blood Flow With Longitudinal Changes in Cerebral Microstructural Integrity in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Key Points Question Is cerebral blood supply associated with cerebral tissue integrity? Findings In this cohort study of 732 healthy adults, at 5-year follow-up, significant associations between lower regional cerebral blood flow and decreased integrity of brain tissue were observed in critical brain regions. The association was most pronounced in white matter tissue. Meaning This finding suggests that cerebral blood flow might be an early risk factor of future change in cerebral tissue integrity among healthy adults.

performed on intermediate and final processing steps by visual inspection and by identifying outliers of calculated variable or parameter distributions. Following the QC protocol, the pseudocontinuous ASL (pCASL) (2) and DTI (3,4) MRI datasets were processed using automated pipelines for the calculation of the corresponding CBF or FA and MD maps, respectively. Briefly, the DTI and the ASL derived CBF maps were registered to the subject T1-weigthed images using FSL, from which the mean values of the different regions of interest were obtained. DTI was registered to the T1 space using FSL FLIRT, while the ASL images were registered to the T1weigthed images using the FSL boundary-based registration. The detailed description of the QC procedure and parameters mapping analyses can be found in (1,(5)(6)(7)(8). Ten regions of interest (ROIs) were defined corresponding to whole brain GM or WM, and frontal, parietal, temporal and occipital lobe WM or GM. These regional brain volumes were obtained using T1-weighted images.
The processing steps include reorientation, inhomogeneity correction with N4 (9) and multi-atlas skull-stripping (10). Thereafter, the anatomical regions of interest were segmented using a multiatlas segmentation method (6). Within each WM or GM ROI, the mean FA and MD values were calculated, while the mean CBF values were calculated in the GM ROIs only. Our main analysis was restricted to GM CBF due to the high sensitivity to noise of CBF values derived from ASL in WM ROIs in addition to the fact that the pCASL protocol used here was optimized for GM CBF determination (2).

Age, time, race, and sex-related differences in DTI indices
As expected, older individuals had higher MD and lower FA at baseline, indicating lower microstructural integrity at older ages (Tables. 2 and 3); these associations were statistically significant (p < 0.05) or close to significance (p < 0.1) in several ROIs studied. Here again, the anterior brain regions, especially the frontal and parietal lobes exhibited, overall, the highest slopes, while the posterior structures, especially the occipital lobes, exhibited the lowest slopes.
Furthermore, in almost all ROIs investigated, regional increases in MD or decreases in FA, that is, decreased microstructural tissue integrity, were associated with time ( Fig. 2, Tables 2 and 3).
Moreover, although women showed, overall, significantly lower MD values as compared to men, and Black participants showed, overall, higher MD values as compared to White participants, these associations were not consistent with FA (Table 2 and

Supplementary statistical analysis
Here, we investigated whether there was a selection bias in participants between Y25 and Y30.
For that, we conducted a complementary and similar analysis to the main analysis presented in the main document but restricted to only participants with DTI data obtained at both baseline (i.e., Y25) and follow-up (i.e., Y30) (eTable 1). The motivation for this sensitivity analysis was to eliminate possible confounding due to participants lost to follow up. We compared the results to those derived from the main analysis highlighted in the main document.
First, for each ROI, CBF values of participants with 1 (group 1) or 2 (group 2) visits were compared using the two-sample t-test and Kolmogorov-Smirnov test. Both tests indicated no significant differences in CBF distributions between the two groups in all ROIs. Furthermore, the results of the regression analysis are virtually identical to those of the main analysis. Indeed, although the statistical significance of the independent parameters on the changes in FA and MD was, overall, lower in this analysis, the regression coefficients of the longitudinal independent parameters, that is, time and time × CBF, were very similar between both analyses (eTables 2-3). Here, the mean diffusivity (MD) is the dependent variable.

Terms Incorporated in the Linear Mixed-Effects Model Given by Equation 1
Here, the fractional anisotropy (FA) is the dependent variable.