Could tau-PET imaging contribute to a better understanding of the different patterns of clinical progression in Alzheimer’s disease? A 2-year longitudinal study

Alzheimer’s disease (AD) is a complex and heterogeneous disease, both in terms of clinical presentation [1] and evolution of cognitive disorders [2, 3]. Molecular imaging by positron emission tomography (PET) enables the in vivo detection of amyloid and tau pathologies, and the latter is closely associated with clinical symptoms in AD [4]. The prognostic value of tau PET imaging on the subsequent progression of brain atrophy and cognitive decline has been reported in recent studies [5‐7], and regional tau radiotracer binding was able to predict cognitive decline in domain-specific brain areas [7]. Given the topographical information it provides, the longitudinal progression of [¹⁸F]-flortaucipir PET imaging could also be of interest to better understand the spatial and temporal relationships between tauopathy, cortical atrophy progression, which reflects neurodegeneration, and the cognitive decline over the same time interval. Longitudinal tau PET imaging studies are still relatively few [8]. The results published thus far do not seem to be completely consistent with some studies reporting an increase in the tau radiotracer binding over time in the temporal [9, 10], frontal [10‐14], parietal [14], and occipital [13] regions or in all regions except the medial temporal lobes [15]. More surprisingly, some studies report a decrease in radiotracer uptake over time in some patients, which is occasionally considered a simple measurement error (processing artifacts, random statistical noise) [14‐17]. Some studies reported a greater increase in radiotracer uptake over time when the binding is higher at baseline [14, 18], while others suggested the opposite relationship [17]. A greater increase in tracer uptake has also been reported in women and in younger patients [18]. In addition, the relationship between the evolution of tau radiotracer binding and cognitive evolution by considering specific cognitive domains has been little studied so far.

In the present study, we aimed to explore the longitudinal progression of the tau SUVr values and of cortical atrophy after a mean time interval of 2 years in a cohort of well-characterized AD patients at the mild cognitive impairment/mild dementia stages compared with controls, and their relationships with baseline regional tau load, and clinical decline in specific cognitive domains. We hypothesized that (a) tau SUVr values increase to varying degrees in AD patients depending on the cortical region considered, (b) tau-PET progression could be correlated with cortical atrophy progression, and (c) tau-PET progression could be associated with cognitive decline in specific domains.

The present study contributes to our understanding of the relationship among the progression of tau pathology as measured by [¹⁸F]-flortaucipir PET; the progression of the regional cortical atrophy, which is considered a marker of nonspecific neurodegeneration and neuronal loss; and cognitive decline in AD. We found in AD patients an expected increase in the average SUVr values over time in the frontal and right inferomedial temporal cortices, contrasting with a surprising decrease in the average SUVr values in the lateral temporoparietal cortex. Individual analyses revealed two distinct trajectories of tau SUVr values that depended on the initial intensity of temporoparietal tau load, leading to the definition of two AD subgroups called low- and high-Tau1. High-Tau1 patients, who were younger than their counterparts with low-Tau1, demonstrated an increase in SUVr values over time in the frontal lobe but also a decrease in the temporoparietal cortex and rapid clinical decline. In contrast, low-Tau1 patients displayed an increase in SUVr values over time in all cortical regions and a slower clinical decline.

Plateauing or decreasing tau SUVr values in some AD patients have been reported in previous studies and are most often attributed to measurement errors or artifacts [14‐16]. However, it seems unlikely that our results are related to a measurement error, as we tried to minimize any source of error by using the most up-to-date techniques for the analysis of longitudinal tau PET data [29, 30]. We used a sophisticated method involving the construction of a nonlinear template of brain MRIs for registering the two PET images of each subject in MNI space and the VOIs defined from the segmentation performed by the FreeSurfer longitudinal pipeline to the PET images in the native space. This allowed us to optimize the quality of the VOI analyses and of the subtraction images used in the voxelwise analyses. In addition, we verified the consistency of our results regardless of the reference region (eroded supratentorial WM and cerebellar GM).

One can argue that the regional decrease in SUVr values could be an effect of regional cortical atrophy as brain atrophy progressed strongly in the regions where a decrease in SUVr values was observed, especially in patients with high Tau1. However, the following arguments go against this assumption: (a) we did not find a regional relationship between Tau2-Tau1 SUVr values and the progression of cortical atrophy when taking into account the effect of age, sex, ApoE genotype, and severity of the clinical alteration at baseline, which could have influenced the results; (b) we found an increase in SUVr values over time in regions where cortical atrophy also markedly progresses, such as the inferior temporal and frontal lobes; (c) we took into account the potential effect of atrophy in different ways, by verifying in subjects with a decrease in SUVr values over time that different grey matter segmentation (FreeSurfer and SPM toolbox) and partial volume effect correction methods (no partial volume correction, PSF correction within the reconstruction and GTM based method) did not modify the results obtained.

The decrease in SUVr values in temporoparietal regions could be related to a decrease of the tissue’s ability to retain pathologic aggregates over time or to the binding properties of the tau-PET radiotracer. It has been recently suggested that the decline in tau radiotracer binding may result from biological changes in the tau pathology that affect the affinity of the tracer [17]. Even if the time frame of this study is relatively short, this could be related to the transition to ghost tangles, which, contrary to mature tangles, contain predominantly 3R tau, for which the affinity of the radiotracer is lower, or to the occurrence of other modifications of the tau protein (post-translational modifications, truncation or conformational changes) which may affect the affinity of the tracer [33]. This decrease in SUVr values, regardless of whether it is related to brain atrophy, could therefore reflect the continuation of the disease process in the regions where it is most advanced and is ultimately more representative of the amplification of neuronal damage than the evolution of tau load.

Consistent with this view, we found a relationship between a decrease in SUVr values in the temporoparietal cortex and the severity of cognitive decline, particularly for the instrumental score. However, the opposite relation could have been expected in accordance with the well-known progression of tau pathology in AD. The finding of a significant association in the unexpected direction reinforces the idea that the decrease in SUVr values reflects the continuation of the disease process possibly due to the biological changes in tau pathology (transition to ghost tangles) such that the radiotracer is less able to detect at a more advanced stage of the disease.

We only found a weak relationship between the increase in SUVr values, mainly in the frontal regions, and cognitive decline, regardless of the cognitive domain considered. To ensure that this result was not biased by the regional decrease in SUVr values observed in some patients, we conducted the same analyses in the low Tau1 subgroup, in which SUVr values increase over time in all patients and found similar results. In addition, to account for the tendency of younger patients to have a higher Tau1 load and a biparietal clinical phenotype, age was included as a covariate in all analyses.

In contrast, we found a strong association between the progression of regional cortical atrophy and cognitive decline, depending on the cognitive component considered and congruent in anatomical-functional terms (memory and temporal lobes, instrumental functions and parietal cortex, and executive functions and fronto-parietal cortex), which is similar to what has been recently reported in the study of the predictive value of the baseline tau-PET [7]. The relationship between clinical progression and cortical atrophy progression seems, therefore, closer than that between cognitive decline and the evolution of tau-PET SUVr values, whereas the predictive value of baseline tau SUVr values on subsequent cognitive decline appears better than that of brain atrophy, as previously reported [7]. This finding is consistent with a previous study using the tau tracer THK-5317, which has the disadvantage of being strongly associated with MAO-B, especially in the basal ganglia. The results of this study suggested that changes in glucose metabolism over time, which is considered as a marker of neuronal injury, are more closely associated with clinical progression than changes in tau radiotracer binding [13].

The present study has some limitations. Its main weakness is the relatively limited number of AD patients included, which probably limits the statistical power and the generalizability of the results obtained, and could have influenced our voxelwise analyses. This limitation is inherent to longitudinal imaging studies in symptomatic AD patients. The relatively limited duration of follow-up, even if it is equal to or greater than most of the studies published to date, also constitutes a limitation of this study, in particular for the longer-term assessment of the impact of changes in tau protein, some of which may occur later.

Beyond the utility of tau-PET for determining AD diagnosis and prognosis [7, 34], its use to follow the evolution of the disease and monitor the effect of possible treatments still raises certain issues. In order to treat AD at the earliest clinical stage, it is tempting to include patients with a relatively low initial tau load in clinical trials. When using such tau-PET imaging criteria, there is a risk of selecting patients who will have a slower clinical progression and thus decrease the ability to detect beneficial effects of treatments. Nevertheless, the inclusion of prodromal AD patients with a higher risk of rapid clinical progression characterized by an higher initial temporoparietal tau load and younger age will require great caution in interpreting the evolution of the SUVr values, which could be biased by a paradoxical decrease in some regions (potentially explained by a decrease of the tissue’s ability to retain pathologic aggregates over time or by a rapid transition to ghost tangles, for which the affinity of the tracer is lower). To avoid such difficulties, the choice of neuroimaging outcome measures deserves to be discussed, for example, by focusing on the frontal lobes where the SUVr values increase in all AD patients, and by also considering the progression of the regional cortical atrophy assessed by MRI.