The cluster analysis identified three significantly different patterns of FDDNP signal— a low global (LG) cluster with lower FDDNP signal in all ROIs; a cluster that had high signal in frontal and parietal regions (HF/PA); a HT/PC cluster, showing highest FDDNP signal in the lateral temporal and posterior cingulate regions, with similar medial temporal and relatively lower frontal and parietal signal compared to HF/PA.
To interpret these different signal patterns we examined the clusters according to diagnosis, demographics, APOE genetic risk, and cognition. The HT/PC was comprised almost entirely of clinically defined MCI patients (88%) with one cognitively normal subject, LG was mostly comprised (79%) of cognitively normal subjects. The HF/PA cluster was the most diagnostically diverse, with 71% MCI and 29% cognitively normal subjects. Neuropathological studies indicate a greater degree of amyloid and tau pathology in MCI compared to cognitively normal older persons6,7, 26
, which is consistent with the current findings of more MCI patients in the two clusters with relatively higher FDDNP signal. Others have found variable binding patterns in MCI and healthy controls using Pittsburg Compound B ([11C]PIB) 27
, which has been reported to have in vivo specificity for A-beta plaques, but is not known to label neurofibrillary tangle pathology. In that study, visual inspection of DVR images indicated that 5 of 9 MCI subjects showed greatest PIB binding in posterior cingulate/precuneus, frontal cortex, and caudate regions, followed by lateral temporal and parietal cortex (“AD-like” pathology), and the remainder showed no cortical or subcortical gray matter binding, similar to that of many cognitively normal persons. Of 27 cognitively normal subjects, 4 showed orbitofrontal, variable cingulate/precuneus and temporal binding, while 2 had either occipital or several focal areas of cortical binding. In comparing this study with ours, it can be observed that in vivo binding patterns of PIB and FDDNP were not identical. Most notable, is the higher signal using FDDNP in NFT rich areas in the MCI subjects such as the medial temporal region, and this discrepancy would be expected given that FDDNP is a marker of aggregates of tau and amyloid.
Most MCI patients were in the HT/PC and HF/PA clusters, but subjects with amnestic MCI or amnestic MCI plus other domains impaired were not differentially represented in those clusters. It should be recognized, however, that subtypes of MCI have not been consistently distinguished earlier from each other using other imaging methods..28
MRI measures of cerebral volumes29
in medial temporal and association cortex and cortical thickness30
in the precuneus have distinguished amnestic MCI from multiple domain MCI.
FDDNP signal clusters in our subject population differed on neuropsychological testing. The LG group, which included mostly cognitively intact individuals, had the best cognitive performances. Relative to LG, the other two clusters showed memory and visuospatial deficits; but HF/PA cluster had the most extensive cognitive deficits, while the HT/PC cluster had more variability in performances across the domains. The HT/PC and HF/PA clusters did not differ from each other. Having more extensive cognitive deficits, such as impairment in memory plus other domains, has been associated with more rapid progression to dementia than having memory impairment alone31
. Longitudinal follow-up will be important in determining not only diagnostic outcomes, but also whether the more extensive cognitive deficits in HF/PA has implications for rate of progression.
Eight cognitively normal persons were members of the higher binding clusters—seven were in HF/PA and one was in HT/PC. It would be expected that some normal subjects would have higher FDDNP signal because of the probability that they may eventually develop AD. Autopsy determinations have identified asymptomatic individuals with AD neuropathology.5,26,32,33,34
To understand the significance of higher FDDNP signal in the eight cognitively normal subjects, we examined their demographics, APOE genotype, and cognition. The subjects were similar in age, APOE-4 genotype, dementia family history, and education to the entire sample of cognitively normal subjects in the current study. Two subjects had no memory impairment; the remaining six performed more than 1 sd below the mean on the selective reminding test, but on no other memory tests. The significance of impairment on Selective Reminding is unclear; although it was earlier reported that this test is a significant predictor of dementia.35
Impairment on only
one memory test was not sufficient for an MCI diagnosis in this study, but this highlights the complexities of defining impairment for MCI.
Five MCI patients were in the LG cluster. These patients had less extensive cognitive deficits, as four were MCI-A and one was MCI-A+. No other variables set these MCI subjects apart from the rest.
As these data are part of an ongoing larger longitudinal study, follow-up cognitive data were available for 14 subjects (6 MCI and 8 cognitively normal) averaging 3.5 (range 1.5 to 7) years. Half of the MCI and cognitively normal subjects remained stable, and the remainder declined cognitively. Of the four cognitively normal subjects who declined, one developed a visuoconstruction deficit, and three developed MCI. Of the three MCI subjects who declined, two converted to dementia, and one developed more extensive deficits. Among the declining subjects, one was from HT/PC (the subject with worsening MCI), three were from LG (all cognitively normal), and three were from HF/PA (two MCI converted to dementia, and a cognitively normal developed MCI-A+ over two years). .
These findings may be useful for generating new hypotheses about the utility of these imaging clusters for predicting further cognitive decline and diagnostic outcomes, particularly as our group collects additional longitudinal data on these subjects. The HT/PC cluster, comprised almost entirely of MCI patients, had higher binding in temporal (particularly lateral temporal) and posterior cingulate regions with memory and visuospatial deficits. The high frontal-parietal signal with relatively lower signal in temporal and PC regions of the HF/PA cluster is not consistent with the pattern of FDDNP accumulation in AD patients,13
or in subjects with frontotemporal dementia,36
where we have observed elevated frontal and temporal FDDNP signals. The HF/PA cluster had the most extensive deficits. Overall, these results may have implications for using regional FDDNP signal patterns over global signal burden in further characterizing MCI, given the heterogeneous manifestations of this disorder. However, at this time, it is not clear whether these clusters represent different etiologies, variants, severities or different stages of disease (or some combination) related to underlying amyloid or tau pathologies. Longitudinal follow-up of these subjects will be important in further clarifying the significance of these clusters.
In conclusion, these results suggest that FDDNP can identify homogeneous groups of nondemented subjects with distinct signal patterns. Most MCI patients belonged to the two clusters with higher regional signal while most cognitively normal subjects had uniformly low FDDNP regional signal. Future longitudinal studies will help clarify whether any of these clusters predict an increase risk for AD. The major limitations of this study are the small number of subjects, and the limited follow-up data currently available. Increasing the number of recruits, adding a non-amnestic MCI group, and additional longitudinal follow-ups are underway to establish the prognostic implications of FDDNP.