192 Chapter 7 individual level (as can also be observed in Figure 2), potentially cancelling each other out, leading to absence of average effects or change on group level. Another explanation may lie in results of another study assessing the relationship between longitudinal perfusion measures and tau pathology (as measured with PET), as they found a lack of overlap between declining perfusion and increases in tau pathology, suggesting a lag phase between these two processes [49]. This may also be the case here, given that tau pathology did change over time, while R1 did not (yet) in our study. Lastly, an explanation might lie in the composition of our Aβ+ group. Some literature describes increases in CBF over time in individuals without dementia with low-, intermediate- or high amyloid load [23, 24]. This might suggest that the increase in CBF represents a compensatory mechanism in response to first-occurring pathology. The fact that our amyloid positive group also included non-demented individuals, in whom this compensatory increase in CBF potentially takes place, might have contributed to the positive association between increases in Braak III/ IV tau pathology and parietal CBF as found in our study. A strength of the current study is that fully dynamic [18F]flortaucipir PET data was used to obtain quantitatively accurate measures of both tau pathology and rCBF in a sample covering the whole AD spectrum (CN-dementia). Furthermore, relative to previous studies this study had a long follow-up period of 25 months. Some caveats are, however, to be considered in the interpretation of our results. First, three different atlases were used to process our data, i.e., Hammers (Braak III-VI) and Svarer (Braak I) for PET and Desikan-Killiany for MRI. We opted not to change our well-established PET pipeline and accept the inherent variations in ROI definitions introduced by this methodological decision. Second, the subset of individuals with longitudinal tau PET data available was relatively small. This may have reduced the statistical power to detect effects. Also, there was likely a bias in our follow-up sample, where older patients and patients with lower MMSE scores dropped out more frequently. This bias is unfortunately common in longitudinal AD studies and likely excludes more progressed AD patients at follow-up. Furthermore, our sample is relatively young and findings might not translate to older patient populations where co-pathologies (independently of tau pathology) associated with brain atrophy are more common. Last, some of the regions that yielded significant associations, like the temporal poles or orbitofrontal cortices, are known to be susceptible for Freesurfer segmentation errors. However, these regions are found repeatedly throughout literature and image segmentations were thoroughly checked prior to analyses. In conclusion, we assessed the association between i) baseline and ii) change in tau pathology with longitudinal atrophy and rCBF by using dynamic [18F]flortaucipir PET and structural MRI scans. We demonstrate that tau pathology accumulated in
RkJQdWJsaXNoZXIy MjY0ODMw