Alzheimer's disease (AD) presents clinically and evolutionarily heterogeneous characteristics. Positron emission tomography (PET) offers in vivo detection of amyloid and tau pathologies, with tau strongly linked to clinical symptoms. Prior research has shown the prognostic value of tau PET imaging in predicting brain atrophy and cognitive decline; however, the spatial and temporal relationships between tauopathy, cortical atrophy progression (reflecting neurodegeneration), and cognitive decline remain incompletely understood. Longitudinal tau PET studies are limited, with inconsistent findings regarding tau radiotracer binding changes over time. Some studies report increases in various cortical regions, while others show decreases, sometimes attributed to measurement errors. The relationship between tau radiotracer binding evolution and cognitive changes in specific domains is also understudied. This study aimed to explore the longitudinal progression of tau SUVr values and cortical atrophy in AD patients at mild cognitive impairment/mild dementia stages compared to controls, and their relationships with baseline tau load and clinical decline in specific cognitive domains. The hypotheses were that tau SUVr values would increase variably across cortical regions in AD patients, that tau-PET progression would correlate with cortical atrophy progression, and that tau-PET progression would associate with cognitive decline in specific domains.

The literature on longitudinal tau PET studies in AD is mixed. Some studies reported an increase in tau radiotracer binding over time in temporal, frontal, parietal, and occipital regions, while others reported a decrease, sometimes attributed to measurement errors. The relationship between the baseline tau load and the rate of increase in tau load is also debated, with some studies showing a positive correlation and others showing the opposite. Furthermore, the relationship between the evolution of tau radiotracer binding and cognitive evolution by considering specific cognitive domains has received little attention.

This prospective longitudinal study (Shatau7-Imatau, NCT02576821-EudraCT2015-000) enrolled 27 AD patients (mild cognitive impairment/mild dementia) and 12 healthy controls. Inclusion criteria for AD patients included a temporoparietal cognitive deficit, CSF AD profile (t-tau/Aβ42 > 0.52), and [11C]-PiB PET GCI > 1.45. Controls had MMSE ≥ 27, normal neuropsychological assessments, CDR = 0, and negative PiB-PET imaging. All participants underwent neuropsychological assessments (MMSE, CDR, cognitive battery assessing verbal episodic memory, executive functions, etc.), 3T brain MRI, [11C]-PiB PET, and [18F]-flortaucipir PET (Tau1) at baseline. Annual follow-ups for two years included the same assessments and a second [18F]-flortaucipir PET (Tau2). Data analysis involved mixed effects models, ANCOVA, multiple regression, voxelwise analyses using VoxelStats 1.1, and RFT-based multiple comparison correction. Tau SUVr images were generated using BrainVisa software, with both cerebellar grey matter and eroded supratentorial white matter used as reference regions. Longitudinal tau-PET voxelwise analyses used annualized subtraction images (Tau2 - Tau1) divided by the time interval, after longitudinal registration using ANTs software and nonlinear template creation from MRI scans. Cortical thickness was measured using FreeSurfer 6.0.0. Relationships between Tau2-Tau1 SUVr or cortical atrophy progression (MRI2-MRI1) and cognitive decline were assessed using mixed effects models.

In AD patients, average SUVr values increased in most regions over two years (e.g., 6% in inferomedial temporal, 6.7% in frontal), except for a decrease in the left temporoparietal region. Controls showed minimal changes. Compared to controls, AD patients had higher Tau2-Tau1 SUVr values in frontal regions, but this was not significant after multiple comparisons correction. Voxelwise analysis in AD patients revealed positive Tau2-Tau1 values in frontal and right inferomedial temporal regions, contrasting with negative values in left parieto-temporal regions. A significant negative correlation existed between Tau1 and Tau2-Tau1 in temporal and lateral parietal cortices. Individual analysis revealed two subgroups based on temporoparietal Tau1 uptake: high-Tau1 (n=10) and low-Tau1 (n=17). High-Tau1 patients, significantly younger, showed frontal SUVr increases, temporoparietal decreases, and rapid clinical decline, while low-Tau1 patients showed increases across all regions and slower decline. Cortical atrophy progressed in temporoparietal and frontal cortex in AD patients, significantly more in high-Tau1 compared to low-Tau1 patients. No significant correlation was found between tau-PET and MRI progression. Cognitive decline strongly associated with cortical atrophy progression, but weakly with tau SUVr progression, particularly in frontal regions and left temporoparietal decrease. In the low-Tau1 subgroup, a modest association was found between increased frontal SUVr and memory decline.

This study demonstrates heterogeneous patterns of tau progression in AD, characterized by contrasting changes in different brain regions. The decrease in temporoparietal SUVr in high-Tau1 patients, despite apparent disease progression, may reflect a shift to ghost tangles with lower tracer affinity or other tau protein modifications. This decrease in SUVr might represent a continuation of the disease process in advanced regions, possibly indicative of amplified neuronal damage rather than evolving tau load. The weak association between tau SUVr increase and cognitive decline, even in the low-Tau1 subgroup, contrasts with the strong association between cortical atrophy and cognitive decline. This suggests that cortical atrophy progression might be a more direct marker of clinical progression than changes in tau-PET SUVr, although baseline tau SUVr shows greater predictive value than baseline atrophy for subsequent cognitive decline. These findings are consistent with other studies suggesting that neuronal injury markers (such as glucose metabolism changes) are more closely linked to clinical progression than changes in tau binding.

Tau-PET imaging offers valuable insights beyond diagnosis and prognosis, assisting in disease monitoring and treatment evaluation. Careful consideration of neuroimaging outcome measures is crucial when selecting patients for clinical trials, particularly regarding the paradoxical decrease in temporoparietal SUVr. Future studies should focus on frontal regions (consistent SUVr increases) and cortical atrophy progression measured by MRI to better understand AD progression. The limited sample size warrants future research with larger cohorts to validate these findings.

The main limitation of this study is the relatively small sample size of AD patients, potentially impacting statistical power and generalizability of results. The relatively short two-year follow-up period might also limit the assessment of long-term effects of changes in tau protein. The findings may not be fully generalizable to other populations.