Head-to-head comparison of relative cerebral blood flow derived from dynamic [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET in subjects with subjective cognitive decline

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by accumulation of intracellular neurofibrillary tangles (NFTs) and extracellular amyloid-β (Aβ) plaques in the brain [1]. The accumulation causes neuropathological changes, such as synaptic and neuronal cell death and decreased cognitive function. Aβ plaques are believed to be one of the hallmarks driving AD pathogenesis and are used for diagnosis; however, there is evidence that vascular and metabolic factors are involved in the development and progression of AD as well [2]. Previous imaging studies have shown that reduced cerebral blood (CBF) and cerebral metabolic rates of glucose in the brain could serve as possible biomarkers of AD [3, 4]. More importantly, these biomarkers are associated with cognitive decline and conversion to AD; therefore, they can be used to identify clinical drug trial participants who are in a milder and earlier phase of the disease [5, 6].

Radiolabelled water, also called [¹⁵O]H₂O, has a linear relationship with CBF [7]. Hence, the gold standard for measurement of CBF is [¹⁵O]H₂O positron emission tomography (PET). However, the short half-life of ¹⁵O makes it difficult to use it in clinical practice. Moreover, the test–retest repeatability (TRT) of K₁ obtained for [¹⁵O]H₂O is quite low (9.7 ± 10.4%) [8], which is inconvenient for longitudinal or drug intervention studies. A way to measure CBF is arterial spin labelling of magnetic resonance imaging (MRI) [9] which has a good correspondence with [¹⁵O]H₂O PET, but also moderate reliability over time (ICC = 0.63–0.74) [10, 11]. However, previous studies have presented R₁ obtained from a dynamic PET scan as a surrogate of relative cerebral blood flow (rCBF) [12‐17]. R₁ represents the ratio of tracer influx in target regions relative to the reference region and can be obtained from the same PET (dynamic) scan without requirement of any additional scans. Several studies have shown that R₁ estimates obtained with some amyloid and tau tracers such as [¹¹C]PIB, [¹⁸F]THK5317 and [¹⁸F]AV45 are strongly correlated with CBF obtained from [¹⁵O]H₂O PET and metabolic activity derived from [¹⁸F]FDG PET [12, 13, 16, 17]. Furthermore, the TRT of [¹⁸F]florbetapir and [¹⁸F]flortaucipir R₁ is higher when compared to the TRT of [¹⁵O]H₂O, 2.1 ± 1.1% and 1.8 ± 1.3%, respectively [8, 18]. This suggests that dynamic [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET scans do not only provide quantitative information of amyloid-β and tau pathology but also yield estimates on rCBF and ultimately circumvent the need for an additional [¹⁵O]H₂O, [¹⁸F]FDG scan or MRI for AD patients. This approach would be highly useful in the clinic, since it results in decreased costs, lower radiation dose and increased patient comfort.

A study by Rodriguez-Vieitez [13] investigated the comparability of [¹⁸F]THK5317- and [¹¹C]PIB-derived rCBF in 11 mild cognitive impairment subjects and 8 AD patients. The researchers showed a high correlation (r = 0.90) in R₁ between the two tracers. Moreover, another recent study found [19] a strong correlation between [¹⁸F]MK6240 and [¹¹C]PiB R₁ (r = 0.93). These findings suggest that R₁ obtained from different tracers could be used interchangeably and is irrespective of target tissue. Although a few studies have indicated that R₁ can be used as a surrogate of rCBF, the similarities or differences of R₁ estimates obtained from different PET tracers still require validation. It is of importance to determine whether the outcome of R₁ is similar between different tracers, since rCBF should not depend on the choice of the tracer. Therefore, the aim of the current study was to compare the R₁ estimates derived from dynamic [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET scans and assess whether R₁ can serve as a tracer-independent surrogate of rCBF, in case of these specific tracers. To this end, a head-to-head comparison was made using [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET scans obtained from the same subjects with subjective cognitive decline (SCD).

The current data did now show significant differences in age (p = 0.053) or MMSE (p = 0.40) between the Aβ-positive and Aβ-negative subject groups. Furthermore, no significant differences were observed in R₁ between the Aβ-positive and Aβ-negative subjects in whole brain for [¹⁸F]flortaucipir (p = 0.15) and [¹⁸F]florbetapir (p = 0.51).

The current study made a head-to-head comparison between R₁ obtained from two different dynamic PET tracers. In line with previous studies [13, 19], the results of the current study showed that RPM R₁ derived from dynamic [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET scans were well correlated with each other irrespective of the small clusters (range 0.2–2.6% of the whole brain) with significant differences observed during voxel-wise comparisons. In a previous head-to-head comparison study, it has been demonstrated that R₁ estimates from a [¹⁸F]florbetapir PET scan strongly correlated with CBF obtained from [¹⁵O]H₂O PET [17]. Based on the results of the current study and the study by Ottoy et al. (2019) [17], one could state that R₁ may serve as a surrogate for rCBF, for both [¹⁸F]flortaucipir and [¹⁸F]florbetapir (particularly for cortical regions). The R₁ obtained with [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET are of clinical interest as amyloid and tau tracers benefit from the ability to provide information on both rCBF and Aβ pathology or tau pathology in one imaging session, respectively. Most of the cortical regions had no significant differences in R₁ between the tracers for both Aβ-negative and Aβ-positive SCD subjects. Despite the significant clusters, the R₁ obtained from [¹⁸F]florbetapir and [¹⁸F]flortaucipir were well correlated (r² > 0.5). These findings were supported by the results of the regional analysis in native space.

Significantly higher R₁ values in the hippocampus for [¹⁸F]flortaucipir were observed when compared to [¹⁸F]florbetapir in both Aβ-negative and Aβ-positive SCD subjects. Hippocampus is a region that is crucial in AD with respect to tau studies, since the involvement of the hippocampus is thought to occur at a critical stage of tau pathology progression [29]. [¹⁸F]flortaucipir has been characterized by off-target binding in the basal ganglia, thalamus and choroid plexus [30‐32]. The off-target binding in the choroid plexus is of particular interest since it may cause spill-in to the anatomically near hippocampus. This spill-in effect may lead to an artificially increased [¹⁸F]flortaucipir activity in the hippocampus which leads to an inaccurate quantification of hippocampal tau load. The choroid plexus consists of a dense collection of capillaries in an ependymal stroma surrounded by a layer of epithelium [33]. The choroid plexus of the lateral ventricles may have several structures that could bind to [¹⁸F]flortaucipir, including melanin [34], calcification [35], Biondi bodies [36, 37] and iron deposits [38]. Off-target binding is problematic as the hippocampus is among the earliest regions affected by tau pathology and accurate assessment of tau accumulation in this region is important in the understanding of the natural time course of AD [29]. Earlier studies have investigated methods to reduce the spill-in effect of the choroid plexus, such as eroding voxels of the hippocampus [39], various partial volume correction methods [39‐41] and linear regression approaches [39, 40]. These techniques led to decreased correlation between hippocampus and choroid plexus tracer binding, presumably due to decreasing the spill-in effects [39‐42]. The elevated [¹⁸F]flortaucipir signal in the hippocampus when compared to [¹⁸F]florbetapir could be partly explained by spill-in from the off-target binding in the choroid plexus.

Higher R₁ values were observed for [¹⁸F]florbetapir in part of the thalamus (bilateral) when compared to [¹⁸F]flortaucipir in both Aβ-negative and Aβ-positive SCD subjects. The thalamus mainly consists of GM; however, it also contains two thin layers of white matter (WM), including the stratum zonale that covers the dorsal surface, and the external and internal medullary laminae [43]. [¹⁸F]florbetapir has previously shown non-specific binding in WM [44]. WM mainly consists of myelin, which is highly lipidic. The lipophilic character of [¹⁸F]florbetapir presumably explains the non-specific binding in this region [45, 46]. In our study, the number of voxels that were significant in the thalamus for WM was ± 863 voxels for Aβ-negative SCD subjects, accounting for 55.4% of total voxels in the WM thalamus and ± 500 voxels (31.5% of total voxels in the WM thalamus) for Aβ-positive SCD subjects, while the significant number of voxels in the GM for this region was much lower (Aβ−: ± 211 voxels: 24.1% of total GM voxels, Aβ+: ± 155 voxels; 17.7% of total GM voxels). This may be a possible explanation for the observed significant differences in R₁ between [¹⁸F]florbetapir and [¹⁸F]flortaucipir in the thalamus in this study.

A decrease in rCBF in the thalamus in case of Aβ-positive SCD subjects (as shown in Fig. 6, subplot associated with Thalamus) could be because of the involvement of the Papez circuit in the memory function. The Papez circuit is an anatomical circuit which starts and ends in the hippocampus [47]. Earlier studies have shown that lesions in any part of this circuit can cause memory dysfunction [48‐50]. It is not uncommon that there is decreased cerebral blood flow in the thalamus in Aβ-positive SCD subjects [51]. Kobayashi et al. [51] also observed decreased thalamic blood flow in mild AD patients using quantitative brain perfusion SPECT. The researchers concluded that lower rCBF in the thalamus might be the result of a remote metabolic effect in the Papez circuit.

Almost all other GM cortical regions showed similar R₁ values for both tracers. Besides that, the R₁ values obtained from [¹⁸F]flortaucipir and [¹⁸F]florbetapir showed good correspondence in the voxel-wise analysis and in the regional analysis (Figs. 4, 6), indicating that R₁ may be considered as a reliable measure of cortical rCBF irrespective of tracer choice. The smaller brain regions, such as the insula and hippocampus, had weaker correlations in R₁ between the two tracers, and this could be explained by the fact that smaller regions are more prone to noise. Furthermore, some ROIs showed significant differences in the Aβ-negative subjects that were not present in the Aβ-positive group. No clarification could be found for these results. Therefore, further research, aimed to address this issue, is necessary to determine the clinical implications. Despite different kinetics, the present study showed that R₁ obtained from [¹⁸F]florbetapir was comparable with [¹⁸F]flortaucipir R₁ in the cortical regions. Furthermore, the TRT of R₁ is better for both tracers, 2.1 ± 1.1% and 1.8 ± 1.3% [8, 18], respectively, when compared to the TRT of perfusion as measured with [¹⁵O]H₂O PET (9.7 ± 10.4%) [8]. K₁ estimates obtained from [¹⁵O]H₂O PET presented a lower test–retest repeatability, possibly due to day to day intrasubject variabilities (that are not related to any underlying pathology). Since the intrasubject differences (non-pathological) remain more or less constant throughout the brain, normalizing the K₁ to a reference region K₁’ (i.e. R₁) could correct for these differences and improve the TRT.

One of the limitations of the current study is that it was performed only in SCD subjects. It should preferably be reevaluated in patient groups, especially if pathological alterations in rCBF are expected such as in neurodegenerative disorders. For this purpose, the gold standard, [¹⁵O]H₂O PET can be considered to evaluate direct CBF in first place. Furthermore, no partial volume correction method was applied on the data, which can eliminate the spill-in effects from the choroid plexus and the WM. Yet, as both tracer studies were collected on the same PET system and used the same reconstruction methods and settings, both datasets have a matched spatial resolution. It is important to acknowledge that partial volume effects (PVE) depend on the contrast between regions, influenced by both tracer and uptake time. Consequently, PVE may not be the same across the tracers under examination. However, R₁ is primarily determined by the early uptake phase of the tracer (< 100 s). Therefore, we anticipate that the influence of PVE on R₁ may be similar for both tracers. Nonetheless, certain distinctions, such as possible spill-in from the choroid plexus and white matter, cannot be definitively dismissed. Additionally, validation of the use of R₁ as cortical rCBF surrogate was evaluated for [¹⁸F]florbetapir and [¹⁸F]flortaucipir and it is essential to note that the extension of these results to other tracers requires further assessments.

The present study showed that [¹⁸F]flortaucipir and [¹⁸F]florbetapir R₁ values were well correlated in GM cortical regions. Put together with previous findings by Ottoy et al. (2019) [17], a cautious claim can be made that R₁ values of both tracers are a valid marker for cortical rCBF. However, the use of R₁ as a tracer-independent rCBF surrogate should be performed with caution in case of some non-cortical areas, as shown in this study. Furthermore, generalization to other tracers is not directly implied by this study, since each new PET tracer needs thorough validation.