¹⁸F-Florbetapir PET in Patients with Frontotemporal Dementia and Alzheimer Disease

DISCUSSION

We report the first study, to our knowledge, of ¹⁸F-florbetapir PET imaging in patients with FTD, compared with those with AD. Consistent with previous studies, neocortical and subcortical gray matter ¹⁸F-florbetapir retention was increased in patients with AD, compared with healthy controls (21,22). In line with the primary study hypothesis, ¹⁸F-florbetapir binding was significantly increased in the cortical and subcortical gray matter of patients with AD relative to those with FTD.

The visual interpretation data in our study with respect to AD patients and controls were comparable to those reported in previous studies (21,22) and in line with the overlap seen on regional analysis (Fig. 1) between patients with AD and healthy controls. However, rating FTD patients was challenging, with only 2 of 8 FTD scans unanimously rated as negative, and the high disagreement rate indicated considerable uncertainty, whereas quantitative regional analysis identified increased global cortical binding comparable to AD in only 1 FTD patient. Retrospective unmasked inspection of scans in conjunction with MR scans revealed that severe cortical atrophy had confounded visual rating of some scans by mimicking high cortical uptake in some brain areas (Fig. 2B). Two FTD patients (patients 12 and 15) and 1 control subject (subject 22), who had received positive ratings, also had low cerebral white matter SUVR (1.91, 1.84, and 1.86, respectively, compared with a median of 2.46; interquartile range, 1.98–2.72, in entire sample). This low cerebral white matter may have reduced the contrast between gray and white matter that was guiding visual interpretation. Scans with discordant ratings did not show the typical full AD pattern but had reduced gray/white matter contrast in some lobes to a degree that fulfilled formal criteria for scan positivity (Fig. 3). Possible reasons of that appearance included regional cortical atrophy, low tracer uptake in white matter, and minor head movement. As demonstrated in Figure 3, MR image coregistration and image fusion could have clarified the actual gray/white matter borders, demonstrating the actual contrast and thus, perhaps, preventing misclassification. This somewhat atypical appearance contrasts with the unequivocally increased cortical ¹⁸F-florbetapir uptake seen in subjects with elevated ¹⁸F-florbetapir SUVRs (Figs. 2A and 2C). These observations suggest that visual interpretation of ¹⁸F-florbetapir images may be challenging in patients with FTD, and guidance should be refined on the basis of further studies in FTD using standard clinical scanners.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Gray-scale ¹⁸F-florbetapir scans as used for visual interpretation. (A) Patient 4 in Supplemental Table 1, patient 12 (B), patient 18 (C), patient 11 (D).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Transaxial coregistered images from FTD subject with discordant visual interpretations (subject 15 in Supplemental Table 1) of gray-scale ¹⁸F-florbetapir PET (A), T1-weighted MR brain (B), and color-scale ¹⁸F-florbetapir PET fused to MR brain (C). Arrows indicate apparent areas of high ¹⁸F-florbetapir uptake as a result of loss of contrast due to low white matter uptake or cortical atrophy leading to inconsistent visual interpretations.

Limitations of our study include the relatively small sample size and the use of clinical diagnosis as a standard of truth, in the absence of pathologic verification. However, clinical diagnosis was based on the assessment of experienced clinicians working in a tertiary cognitive neurology unit with a good record for diagnostic accuracy in AD and FTLD as confirmed by clinicopathologic studies (23). In addition, none of the FTD patients had patterns of ¹⁸F-FDG hypometabolism typical for AD. However, as only FTD patients underwent ¹⁸F-FDG PET, the diagnostic utility of ¹⁸F-florbetapir, compared with ¹⁸F-FDG PET, cannot be estimated. Given that our FTD cohort met research criteria for FTD, the generalizability of our findings to clinical situations in which there is a genuine diagnostic equipoise between AD and FTD is potentially limited. Previous ¹¹C-PiB data from patients with pathologically confirmed FTLD showed that all were visually rated as negative for amyloid (10). A strength of our study is that study groups were well matched in terms of age, relatively short disease duration, and dementia severity. Although the groups were not balanced in terms of sex, previous data suggest that PiB binding is greater in men with AD (24), and therefore the greater proportion of men in the FTD group in this study is unlikely to have influenced the outcome.

Our findings in FTD are consistent with other published studies using ¹¹C-PiB, in which lower neocortical uptake has been observed relative to AD (8–10). We used a different method to delineate ROIs for our analysis, namely an atlas-based approach after segmentation of anatomic MR imaging scans, in contrast to previous papers (8,9). Together with motion correction of the PET data, MR-based segmentation should improve the accuracy of ROI definition and minimize the effect of spill-over from adjacent regions. Although we did not perform partial-volume correction on these data, this is comparable to existing studies reporting amyloid imaging in FTD, including a recent report of another ¹⁸F-labeled tracer, florbetaben (25). Studies on partial-volume correction of ¹¹C-PiB data suggest that our approach may underestimate group differences between AD and other conditions, so the absence of partial-volume correction is unlikely to invalidate our findings (26). In most previous studies, ¹⁸F-florbetapir SUVRs were calculated with reference to the entire cerebellum (21,22). We used the cerebellar cortex to avoid contamination of the reference values by the variation of nonspecific cerebellar white matter uptake, whereas SUVRs calculated with either method were highly correlated (R² = 0.89).

Previous studies have suggested a significant overlap between AD and FTD patients in binding of ¹¹C-PiB, with 2–4 positive subjects in cohorts of a size comparable to that reported here (8,9). However, several of these reported cases had neuropsychologic features, a subsequent clinical course, or ¹⁸F-FDG PET findings to suggest an alternative diagnosis of AD (8,9). Patients with pathologically confirmed AD may have behavioral features of FTD at presentation, although they tend to develop neuropsychologic features, such as visuospatial impairment, characteristic of AD during the disease course (4,27). However, the clinical features, subsequent clinical course, and ¹⁸F-FDG imaging of the FTD patients with high regional uptake were all in keeping with the clinical diagnosis of FTD. One FTD patient was homozygous for the APOE ε4 allele and showed elevated neocortical ¹⁸F-florbetapir retention, in line with previous studies demonstrating that APOE ε4 is associated with higher amyloid burden (13). Deposition of Aβ has also been described in pathologically confirmed FTLD (28), with a greater Aβ burden in APOE ε4 carriers and extensive neuritic plaque deposition in an APOE ε4 homozygote. A positive ¹⁸F-florbetapir PET scan has also been reported recently in a patient with frequent cortical neuritic plaques and frontotemporal lobar degeneration with TDP43-positive inclusions (29). Most likely, this positive scan represents FTD–AD comorbidity, with the FTD component dominating the clinical presentation.

Previous studies on the topography of white matter binding of amyloid tracers have produced variable findings. Several ¹¹C-PiB PET studies showed no significant differences in binding to subcortical white matter between AD and controls, whereas PiB did not show specific binding to white matter homogenates or postmortem specimens from AD patients (7,30). Other studies have reported elevated ¹¹C-PiB retention in the subcortical white matter of patients with AD, compared with FTD (8), indicating the importance of resolving this issue. It is possible that white matter pathology, seen prominently in some molecular subtypes of FTD, might lead to a reduction in amyloid binding, compared with AD. When using eroded white matter masks to minimize the possibility of contamination of adjacent white matter regions by the differences in specific gray matter binding between AD and FTD due to partial-volume effects, we could not find significant differences in white matter binding between groups, whereas there was considerable interindividual variation.

Our observations underline that referring physicians should evaluate amyloid PET scan results in the context of clinical findings. Moreover, it may be useful for interpreters to indicate when the presentation is atypical and classification is uncertain. In our series, we would not have changed the clinical diagnosis in any of the FTD patients, but the scans made us aware of additional amyloid pathology in at least one of them.