In this study, we examined associations of 2 PET markers (florbetapir and FDG) with concurrent cognitive function and retrospective cognitive decline. First, amyloid deposition, hypometabolism, and cognition were associated cross-sectionally in early and late MCI. In normal subjects, amyloid deposition was only marginally associated with cognition, whereas in AD (which included 27 of 53 patients who had recently converted from LMCI), FDG (but not amyloid) was related to cognition. In our longitudinal analysis, amyloid deposition (but not hypometabolism) was associated with ongoing cognitive decline in normal subjects, whereas in LMCI, both amyloid and hypometabolism were related to decline, but the association was stronger for hypometabolism. These findings suggest that in normal older individuals, amyloid is closely linked to the earliest indications of clinical decline. Once individuals have sufficient clinical dysfunction to be diagnosed with MCI, however, hypometabolism (which represents synaptic and neuronal dysfunction) and cognitive loss progress together. These data are consistent with a model in which amyloid deposition is associated with cognitive dysfunction in the early and middle stages of decline, but at moderate and later stages of disease (LMCI/AD), synaptic dysfunction becomes more pronounced and more closely linked to ongoing cognitive decline.8
Almost a third of cognitively normal individuals in this study were florbetapir+
, a rate that is consistent with other reports of amyloid deposition in healthy older populations based on amyloid-PET imaging using Pittsburgh compound B (PiB),21–23
and rates of amyloid positivity increased with cognitive symptoms such that 43% of EMCI and 65% of LMCI participants were florbetapir+
. Seventy-seven percent of AD patients were florbetapir+
, which is similar to the percentages of AD amyloid positivity reported recently in different populations with florbetapir24
although AD amyloid positivity around 90% or above has also been observed.22,26,27
This variability could be due to different sample sizes, application of diagnostic criteria, the sensitivity of the imaging technique, or incorrect clinical diagnoses, which are difficult to estimate due to low sample sizes in autopsy studies but range from about 10%28,29
Although some studies have reported a correlation between age and amyloid deposition across diagnostic groups with florbetapir24
and with PiB in some diagnostic groups but not others,5,22
florbetapir uptake and age were related only in the EMCI group. However, the majority of individuals in the LMCI and normal groups were enrolled in ADNI >4 years prior to their florbetapir scans. The normal and LMCI subjects are therefore older and have a slightly reduced age range compared to the EMCI subjects, which may have biased associations between amyloid and age.
Evidence for cross-sectional associations between cognitive dysfunction and amyloid deposition is mixed and may depend on stage of disease,32
which is in agreement with our findings, because only the LMCI groups showed this relationship (the association in normal subjects was a nonsignificant trend). Some studies have reported significant cross-sectional associations,33,34
whereas others have reported no association22,23,35
between amyloid-PET measurements and cognitive function. Our findings are also consistent with recent longitudinal studies reporting that PiB+
normal subjects have greater retrospective,36
cognitive decline than PiB−
normal subjects. In addition, we previously observed similar low agreement between FDG and PiB measures in a subset of the same population,18
suggesting that there is only moderate shared variance between hypometabolism and amyloid measurements, particularly in the normal and AD groups.
An important limitation of this study is that the associations with longitudinal cognitive decline are retrospective rather than predictive, because the florbetapir and FDG measurements were collected at the end of the follow-up period. Amyloid deposition is likely to have fluctuated throughout the follow-up period, and change in amyloid deposition over time may differ between normal and LMCI subjects.5
Furthermore, there is evidence that metabolism changes over a longer time period than amyloid deposition does,9,11
both before and after the onset of cognitive symptoms. Hypometabolism has also been observed in young adult APOE4
several decades before the likely onset of amyloid deposition, suggesting that hypometabolism may reflect long-term patterns of brain and cognitive function that are independent of amyloid deposition. Prospective follow-up of these participants is ongoing and will be important for determining the predictive utility of PET imaging, particularly in the EMCI group, which may represent the earliest clinical stage of disease but did not have available longitudinal data at the time of this study. Repeated PET scans in this population will also be critical for determining the rate of amyloid and metabolic change (as well as measurement error) at different stages of disease. An additional limitation is that the distributions of FDG-PET and florbetapir differ. Florbetapir is more bimodal than FDG-PET, so the use of dichotomous predictor variables may more accurately reflect the underlying characteristics of the florbetapir distribution.
Our primary finding was that the relationship between amyloid deposition and cognitive decline differed between diagnostic groups. Positive amyloid status in both the normal and LMCI groups was associated with ongoing decline. However, in normal subjects, decline was more closely linked to amyloid status, whereas in LMCI, decline was more closely linked to hypometabolism. These data suggest that amyloid deposition precedes detectable metabolic dysfunction and that the consequences of high amyloid deposition can be observed at a stage prior to the onset of clinically recognized symptoms. After the onset of clinical symptoms, amyloid and cognitive dysfunction appear to become decoupled, so variability in cognitive decline is mediated by other factors such as synaptic dysfunction and atrophy, comorbidities, and factors related to cognitive reserve such as education or lifestyle.