Amyloid imaging and cognitive decline in nondemented oldest‐old: The 90+ Study

The goal of this study was to examine cross‐sectional and longitudinal associations between cognitive performance and beta amyloid (Aβ) load determined by florbetapir F18 positron emission tomography (PET) in nondemented oldest‐old.


Introduction
The ability to image cerebral beta amyloid (Ab) deposition during life with positron emission tomography (PET) scanning [1,2] is a major advance in neuroscience and a powerful research tool for the investigation of Alzheimer's disease (AD) and cognition in aging. Previously, studies of amyloid deposits and cognition were dependent on clinical pathological investigations, with a single amyloid measurement performed at the end of life. Recent studies have demonstrated that PET scanning with amyloid-binding ligands correlates with the presence and density of Ab at autopsy [2]. It has been hypothesized that amyloid deposition is an early event in the pathogenesis of AD, increasing rapidly and reaching a plateau before the appearance of clinical symptoms [3]. In this scenario, normal individuals with amyloid deposition may be at higher risk of developing AD and may be experiencing subtle cognitive decline [3,4].
The oldest-old are the fastest growing segment of the population and have high rates of dementia [5] and cognitive decline. A high proportion of nondemented individuals older than age 90 have significant amyloid deposition on autopsy [6,7]. It is unknown whether these individuals are at higher risk of developing dementia, are experiencing cognitive decline, or perhaps are even protected from the development of clinical AD. We examined the crosssectional and longitudinal relationship between cognitive performance and amyloid load (florbetapir PET uptake) in 13 nondemented oldest-old individuals.

Methods
Participants were part of The 901 Study, a longitudinal, population-based investigation of dementia and aging in the oldest-old. Individuals were invited to participate in this imaging study as part of an investigation to examine the relationship between measurements of brain amyloid using florbetapir PET scanning and levels of amyloid burden as measured by postmortem histopathological assessment [2].
To meet inclusion criteria for our study, individuals had to be nondemented: normal or cognitively impaired nondemented (CIND) and agree to postmortem examination. Participants were followed every 6 months with procedures that include the Mini-Mental State Exam (MMSE), Modified MMSE (3MS), Animal Fluency, Boston Naming Test (BNT), and the California Verbal Learning Test (CVLT) short form [8]. At each visit, a trained neurological examiner determined cognitive status (normal, CIND, or dementia). Dementia was diagnosed using the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) criteria [9]. Participants with either cognitive or functional impairment resulting from cognition not severe enough to meet DSM-IV diagnostic criteria for dementia were classified as CIND [9]. The neurological examiner had access to the 3MS and MMSE, but was blinded to the remainder of the neuropsychological testing.
Each participant underwent a 10-minute PET scan at approximately 50 minutes after injection of 370 MBq of florbetapir F18. A semi-automated quantitative analysis was performed to calculate the standardized uptake value ratio (SUVr) using the mean of six predefined anatomically relevant cortical regions (frontal, temporal, parietal, anterior cingulate, posterior cingulate, and precuneus), relative to entire cerebellum. These SUVr values were used as our primary amyloid load variable for our cross-sectional analyses. Florbetapir-PET images were also assessed visually by three trained nuclear medicine physicians using a semiquantitative score ranging from 0 point (no amyloid) to 4 points (high levels of cortical amyloid). The median of the three visual scores was used to dichotomize participants into Ab-(score, 0-1 point) and Ab1 (score, 2-4 points) groups for our longitudinal analyses. All procedures were approved by the University of California at Irvine institutional review board.

Statistical methods
The visit closest to the PET scan was considered the baseline visit. For our cross-sectional analysis, Pearson correlation coefficients were calculated to assess the correlation between SUVr and neuropsychological scores at baseline. To analyze change in cognitive performance over time in Ab-and Ab1 participants, we took two different approaches. First, because of the small number of participants, we took a simple approach and estimated a slope for each participant, with a linear regression of neuropsychological test scores as a function of years from baseline. The average slope was then compared between Ab-and Ab1 participants using t-tests and Wilcoxon's rank tests. As a second approach, we compared cognitive decline in Ab-and Ab1 individuals using mixed-effects models, which accounts for the correlation between repeated measurements in the same individuals, but usually requires more observations. The covariates in the model were years from baseline, an indicator variable for Ab group, and the interaction between years from baseline and Ab group. A significant interaction indicates a difference in the rate of cognitive decline in the Ab-and Ab1 groups. The intercept was included as a random effect. Analyses were conducted separately for each neuropsychological test and were performed using SAS version 9.2. Table 1 shows characteristics of the 13 participants (nine women and four men) in the study. The baseline visit was within 90 days of the PET scan (median, 42 days). At baseline, the median age of the participants was 94.1 years, and eight participants had normal cognition and five participants had CIND.
All but one participant were followed for three or four total visits: four participants had two follow-up visits and eight participants had three follow-up visits after their baseline evaluation. The remaining participant died within 4 months of the baseline visit and was not included in the longitudinal analyses. Median follow-up between baseline and last evaluation was 1.5 years (Table 1). Three participants developed dementia during follow-up. The three participants who developed dementia during follow-up were cognitively impaired but not demented at baseline. Figure 2 shows longitudinal trajectories for the different neuropsychological tests by amyloid load group. All but one participant in the Ab-group appear to have stable scores across time on the 3MS, MMSE, and CVLT tests. In contrast, people in the Ab1 group show a decline in scores, particularly in the 3MS and MMSE tests. Some of these differences were confirmed when comparing the average slopes for the two groups ( Table 2). On all tests, cognitive decline was faster in the Ab1 group compared with the Ab-group and was significant for the 3MS, MMSE, and BNT tests. For example, people in the Ab-group declined, on average, 1.26 points per year on the 3MS, whereas people in the Ab1 group declined 12.35 points per year (P 5 .04). The results from the random effects models were very consistent with those of the average slope approach (Table 2).

Discussion
This investigation in nondemented oldest-old individuals found that amyloid load measured with florbetapir PET scanning was related to cognitive performance at baseline and was associated with greater cognitive decline over 1.5 years. The results of this study support a model in which amyloid deposition is an early event on a path that may lead to dementia, beginning insidiously in cognitively normal individuals and accompanied by subtle cognitive decline [10].
Previous studies have suggested that the relationship between cerebral amyloid deposition and cognitive performance in the oldest-old is extremely poor [11]. In large measure, these results arise because of the large percentage of nondemented oldest-old with significant amyloid deposition. Our investigation suggests that 90-year-old nondemented subjects with amyloid deposition have subtle cognitive changes compared with those without amyloid burden, and are experiencing measurable declines over 1.5 years of follow-up. Three participants (two with a high SUVr) appear to be in the early stages of dementia. We do not know whether the others will develop dementia before death, nor do we know how long they have carried this amyloid load or whether it is changing over time. Twoyear follow-up PET studies are planned for all participants as they continue in our longitudinal study. Most participants in the Ab1 group were CIND (3 of 4) compared to the Ab2 group (2 of 8). Indeed, baseline cognitive status may have predicted future cognitive decline independent of imaging. However, when random effects analyses were repeated adjusting for baseline cognitive diagnosis, the results were virtually unchanged (results not shown). These results suggest that the presence of amyloid can predict cognitive decline beyond baseline cognitive status. Moreover, it would be beneficial to establish the reasons for cognitive decline and identify people who may benefit from amyloid therapies if they become available.
In this investigation of the oldest-old, scans that were interpreted visually as Ab1 (score, 2-4 points) had SUVr's . 1.2. This value is slightly higher than the SUVr cut-point of 1.1 noted in the pivotal trial with subjects ranging in age from 48 to 104 years [2]. The pivotal trial had three Ab1 subjects older than the age of 90 and all had SUVr's . 1.2. Of interest, despite having among the highest SUVr's, two of these individuals exhibited only intermediate Several limitations of this study deserve mention. First, the sample size is very small. We were somewhat surprised to find this result with our modest number of participants and have plans to increase the sample size with additional 901 subjects to confirm our results. Second, four of the 12 participants were not followed for the full 1.5 years, thus their cognitive trajectories may have been somewhat different if they had been followed an additional 6 months. Third, some participants did not complete all cognitive tests at every visit; therefore, we are less certain about the estimates of cognitive decline for those tests. Fourth, all participants are white, well educated, and older than the age of 90, which limits our ability to generalize to other ethnic and demographic groups. Last, oldest-old subjects have high rates of sensory loss and physical frailty, which make it challenging to administer the cognitive battery and to determine cognitive status.
This study in 901-year-olds supports a growing body of research suggesting that amyloid deposition may be an early event in the development of cognitive decline associated with AD. With additional research, we will understand better the temporal events related to amyloid deposition and the potential role of these studies in the diagnosis, treatment, and, hopefully, eventual prevention of AD. y P value for the interaction between years from baseline and Ab group. A significant interaction indicates a difference in the rate of cognitive decline in the Ab-and Ab1 groups.