Introduction
When used to describe older adults, the concept of “cognitively normal-for-age” suggests that decline is normal. That is, scores considered “normal” for an 80-year-old on many cognitive measures are markedly lower than normal scores for a 50-year-old, especially for tests of episodic memory (for a discussion on this concept, see Evans, Grodstein, Loewenstein, Kaye, & Weintraub, Reference Evans, Grodstein, Loewenstein, Kaye and Weintraub2011). What is poorly understood is whether age-related decline is inevitable beyond age 80 or whether there is an alternative trajectory that resists the cognitive and anatomic changes characteristic of normal aging.
This question was addressed in a prospective study of individuals called SuperAgers, aged 80 or older, selected for scores on tests of episodic memory that were above-average-for-age, at a level considered at least normal for individuals 20–30 years younger, and whose scores in other cognitive domains were at least average-for-age. While several studies of successful aging exist (for relevant reviews, see Depp & Jeste, Reference Depp and Jeste2006; Kaup, Mirzakhanian, Jeste, & Eyler, Reference Kaup, Mirzakhanian, Jeste and Eyler2011; Rowe & Kahn, Reference Rowe and Kahn1997), few incorporate cognitive function into their definition and none require memory performance to be at least as good as those of individuals two to three decades younger.
This study examined whether the SuperAgers’ cortical morphometry was distinct from typical age-related atrophy by quantitatively comparing their structural magnetic resonance imaging (MRI) scans to those of two cognitively normal reference groups: elderly controls and middle-aged controls.
Methods
Participants
Participants were identified based on chronologic age (SuperAgers: ≥ 80 years old, Middle-aged Controls: 50–65 years old, Elderly Controls: age-matched to SuperAger cohort), availability of a 3 Tesla (T) structural MRI scan and neuropsychological performance (described below). All participants were required to have preserved activities of daily living and lacked clinical or structural evidence of neurologic or psychiatric disease.
SuperAgers and middle-aged controls were community dwellers, recruited through Northwestern's Alzheimer's Disease Center (ADC) Clinical Core, community lectures, and word of mouth. Twelve SuperAgers and 14 middle-aged controls met our criteria and were included in the analysis.
Since Northwestern's ADC does not routinely perform structural MRI scans on healthy elderly participants, data were obtained through the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Cognitively normal and impaired individuals have been recruited into ADNI from over 50 sites across the United States and Canada, including Northwestern's ADC, and their imaging and neuropsychological data are made available to approved investigators (http://adni.loni.ucla.edu/). Eleven individuals met criteria; one participant was subsequently excluded because their MRI data contained fatal defects.
Written informed consent was obtained for all participants, which was approved by the IRB of each participating center.
The neuropsychological measures were chosen on the basis of their relevance for cognitive aging and for early detection of Alzheimer's disease (AD) (Weintraub, Wicklund, & Salmon, Reference Weintraub, Wicklund and Salmon2011). The delayed verbal recall score of the Rey Auditory Verbal Learning Test (RAVLT) was used to assess episodic memory. SuperAgers were required to perform at or above average normative values for individuals in their 50s and 60s (midpoint age = 61; RAVLT delayed-recall raw score ≥ 9) (Schmidt, Reference Schmidt2004), while healthy middle-aged controls and elderly controls were required to score within one standard deviation of the average range for their age and education according to published normative values (Figure 1A; Schmidt, Reference Schmidt2004). The 30-item Boston Naming Test (BNT; Kaplan, Goodglass, & Weintraub, Reference Kaplan, Goodglass and Weintraub1983), Trail Making Test Part B (Randolph, Reference Randolph1998), and the Category Fluency Test (Morris et al., Reference Morris, Heyman, Mohs, Hughes, van Belle, Fillenbaum and Clark1989) were used to assess cognitive function in non-memory domains. All participants were required to score within one standard deviation of the average range for their age and education on each of these measures according to published normative values (Heaton, Miller, Taylor, & Grant, Reference Heaton, Miller, Taylor and Grant2004; Randolph, Reference Randolph1998; Saxton et al., Reference Saxton, Ratcliff, Munro, Coffey, Becker, Fried and Kuller2000).
Episodic memory performance and differences in cortical thickness by group. (a) Group average delayed recall scores on the RAVLT word list show SuperAgers performing significantly better than elderly controls. There is no significant difference in episodic memory performance between SuperAgers and middle-aged controls. The solid and dotted lines represent the average normative values for a 60-year-old and an 80-year-old, respectively (Schmidt, Reference Schmidt2004). * represents significant differences at p < .05. (b) Red and yellow represent significantly thinner cortex in elderly controls compared to middle-aged controls. (c) There is no significant thinning in the SuperAging group in either hemisphere compared to middle-aged controls. A region within the anterior cingulate (blue) is significantly thicker in SuperAgers when compared to middle-aged controls. (d) Red and yellow represent significantly thinner cortex in elderly controls compared to SuperAgers. False discovery rate (FDR) was set at 0.05 for each MRI analysis. Color bars display significance using as −log(10) p value.
MRI Acquisition
All subjects received T1-weighted three-dimensional MP-RAGE sequences (TR = 2300 ms, TE = 2.86 ms, flip angle = 9°, FoV = 256 mm) (Jack et al., Reference Jack, Bernstein, Fox, Thompson, Alexander, Harvey and Weiner2008). Cortical thickness was calculated by measuring the distance between representations of the white-gray and pial-CSF boundaries across each point of the cortical surface using the image analysis suite FreeSurfer (version 4.5.0; http://surfer.nmr.mgh.harvard.edu/; Fischl & Dale, Reference Fischl and Dale2000). This sensitive and well-validated method uses both intensity and continuity information from the entire MR volume, and produces thickness maps capable of detecting submillimeter differences between groups (Kuperberg et al., Reference Kuperberg, Broome, McGuire, David, Eddy, Ozawa and Fischl2003).
Statistical Analysis
Statistical surface maps were generated using a general linear model that displays differences in cortical thickness between two groups for each vertex along the surface representations. A false discovery rate of 0.05 was used to adjust for multiple comparisons (Genovese, Lazar, & Nichols, Reference Genovese, Lazar and Nichols2002).
Cortical volume, a summary measure of cortical integrity (Desikan et al., Reference Desikan, Segonne, Fischl, Quinn, Dickerson, Blacker and Killiany2006), was calculated for each participant, normalized by intracranial volume (Buckner et al., Reference Buckner, Head, Parker, Fotenos, Marcus, Morris and Snyder2004) and compared between groups using an analysis of variance (ANOVA). Pearson correlations were used to examine the relationship between cortical morphometry measures and memory performance.
All statistical analyses were performed within FreeSurfer or with PASW 19.0 (SPSS).
Results
Twelve SuperAgers (average age = 83.5 ± 3.0 years), 10 elderly controls (average age = 83.1 ± 3.4 years), and 14 middle-aged controls (average age = 57.9 ± 4.3 years) were included in the analyses. There were no significant differences in education among the groups (average years of education for SuperAgers = 14.8 ± 2.4, elderly controls = 17.5 ± 2.2, middle-aged controls = 16.1 ± 2.9).
As defined by our experimental design, episodic memory performance for the elderly and middle-aged controls was within one standard deviation of published normative values (Schmidt, Reference Schmidt2004), while SuperAgers’ performed at least as well as middle-aged controls (p = .09) and significantly better than elderly controls (p < .001; Figure 1A). All participants scored at least within one standard deviation of the average range for their age and education on non-memory testing, confirming the absence of cognitive impairment. One-way ANOVAs showed significant differences between the three groups on Trails B but not on Category Fluency or the BNT (Trails B: F = 4.1, p = .027, Average Score SuperAgers = 96.2 ± 46.1, Middle Age Controls = 61.8 ± 22.0, Elderly Controls = 128.5 ± 92.7; Category Fluency: F = 3.0, p = .064, Average Score SuperAgers = 22.4 ± 6.0, Middle Age Controls = 23.7 ± 5.6, Elderly Controls = 18.4 ± 3.8; BNT: F = 0.9, p = .424, Average Score SuperAgers = 28.7 ± 1.1 Middle Age Controls = 28.9 ± 1.0, Elderly Controls = 28.0 ± 2.5). Post hoc tests revealed that the elderly controls were significantly slower than the middle-aged controls (p < .05) on Trails B but there were no other significant differences between the groups. Thus, as a group, the SuperAgers did not differ from the middle-aged controls on non-memory performance.
Middle-Aged Controls Versus Elderly Controls
The whole-brain cortical thickness comparison conducted between elderly controls and middle-aged controls revealed, as expected, significant atrophy in the older cohort in multiple regions across the frontal, parietal and occipital lobes, including medial temporal regions important for memory function (Figure 1B). This pattern of atrophy in the cognitively normal elderly controls is consistent with previous findings (e.g., Salat et al., Reference Salat, Buckner, Snyder, Greve, Desikan, Busa and Fischl2004), and presumably reflects the anatomical substrate of the age-related decline of cognitive function, including memory performance, as shown in Figure 1A.
In accordance with the whole-brain cortical thickness results, the average normalized cortical volume was significantly smaller in elderly controls compared to middle-aged controls (Average Volume: elderly controls = 244.13 mm3; middle-aged controls = 306.43 mm3; p < .001).
SuperAgers Versus Middle-Aged Controls
Surprisingly, the whole-brain cortical thickness analysis comparing SuperAgers and middle-aged controls did not reveal significant atrophy in the SuperAgers (Figure 1C). In addition, an area located in the left anterior cingulate (central estimated Talairach coordinates: −6, 10, 26) was thicker in SuperAgers than in middle-aged controls (http://surfer.nmr.mgh.harvard.edu/fswiki/CoordinateSystems). The average thickness of this region was quantified for each group and is shown in Figure 2.
Quantification of the left hemisphere anterior cingulate region by group. (a) Inflated medial surface of left hemisphere shows the anterior cingulate region, identified in the whole-brain group comparison (Figure 1B) as significantly thicker in SuperAgers than in middle-aged controls. (b) Boxplots of average left hemisphere anterior cingulate thickness data by group. * represents significant differences in average thickness at p < .05.
Results from the whole-brain normalized cortical volume analysis were similar to the whole-brain cortical thickness analysis, demonstrating no statistically significant differences between the SuperAgers and middle-aged controls (Average Volume: SuperAgers = 288.05 mm3; middle-aged controls = 306.43 mm3; p = .08).
SuperAgers Versus Elderly Controls
As expected from the results above, a comparison of healthy elderly controls and SuperAgers revealed areas of significant cortical thinning in the former group (Figure 1D). The cortical volume measures corroborated the thickness findings, showing SuperAgers had significantly larger normalized cortical volumes than their peer group of elderly controls (Average Volume: SuperAgers = 288.05 mm3; elderly controls = 244.13 mm3; p < .001).
Relationships Between Memory Performance and Brain Morphometry
Separate Pearson correlations were used to assess the relationship between episodic memory performance (i.e., RAVLT delay score) and cortical brain measures (i.e., normalized cortical volume and cingulate thickness) across the groups. Results showed a significant positive correlation between average normalized cortical volume and episodic memory performance (r = 0.532; p = .001) but no relationship between episodic memory performance and thickness of the cingulate.
Discussion
We identified octogenarians and nonagenarians with superior memory function compared to their age-matched cognitively normal peers and found anatomic features associated with SuperAging that deviated from those associated with normal cognitive aging. The SuperAgers showed significantly greater cortical thickness and volume than their cognitively normal age-matched peers and showed no significant cortical atrophy when compared to younger, cognitively intact individuals 20–30 years younger (50- to 65-year-olds). These findings are remarkable given the numerous reports that grey matter loss is a common, if not universal, part of normal aging (Drachman, Reference Drachman2006; Fjell et al., Reference Fjell, Walhovd, Reinvang, Lundervold, Salat, Quinn and Dale2006; Salat et al., Reference Salat, Buckner, Snyder, Greve, Desikan, Busa and Fischl2004). Although the relationship between structural MRI-based cortical thickness measures and cellular morphology is not entirely understood, the lack of atrophy in SuperAgers may be taken as a proxy measure of preserved cortical neuronal integrity.
In addition to the absence of age-related cortical atrophy, our data also suggest that SuperAging may be associated with an unusual prominence of the cingulate cortex (Figure 2). At this point it is unclear whether SuperAgers were born with a particularly thick cortex or whether they resisted cortical change over time. Multiple lines of evidence converge toward the conclusion that the cingulate constitutes a critical site of transmodal integration related to episodic memory, spatial attention, cognitive control, and motivational modulation (Mesulam, Reference Mesulam2009; Mesulam, Reference Mesulam1998). In our study, cingulate thickness was not directly correlated with memory performance, which is not surprising since the cingulate is not known as a primary component of the episodic memory circuitry. However, it is possible that the cingulate may mediate resistance of memory circuits to the deleterious processes of aging.
The posterior segment of the cingulate gyrus has recently received considerable attention in relationship to the pathophysiology and temporal evolution of AD. For example, cingulate hypometabolism as measured by metabolic positron emission tomography (PET) is one of the earliest correlates of neuronal dysfunction in AD (Minoshima et al., Reference Minoshima, Giordani, Berent, Frey, Foster and Kuhl1997) and in vivo amyloid imaging with PET shows early accumulation of amyloid in the cingulate gyrus of AD patients (Fripp et al., Reference Fripp, Bourgeat, Acosta, Raniga, Modat, Pike and Ourselin2008; Sperling et al., Reference Sperling, Laviolette, O'Keefe, O'Brien, Rentz, Pihlajamaki and Johnson2009). Likewise, functional MR data suggest that the cingulate cortex is an important component of a large-scale resting-state network, which shows abnormalities early in the course of AD (Buckner et al., Reference Buckner, Sepulcre, Talukdar, Krienen, Liu, Hedden and Johnson2009; Sperling et al., Reference Sperling, Laviolette, O'Keefe, O'Brien, Rentz, Pihlajamaki and Johnson2009). Thus, normal cingulate function appears to be important for the integrity of multiple cognitive domains, and the SuperAgers’ superior episodic memory function may reflect, at least in part, the presence of a more extensively developed cingulate region that is also physiologically more resilient to age- and AD-related pathology.
While recording systematic and detailed medical and demographic histories of the SuperAgers we found no reason to suspect that they had unusually superior memory abilities when younger. In fact, their level of education was not abnormally high, only 4 of the 12 SuperAgers obtained a college degree. The findings in this cohort of cognitive SuperAgers provide ‘proof of concept’ that maintenance of superior memory together with cortical integrity is a biological possibility. Identifying the underlying factors that promote this trajectory of unusually successful cognitive aging may lead to novel insights for preventing age-related cognitive impairments or strategies for evading the more severe changes associated with Alzheimer's disease.
Acknowlegments
This project was supported by a grant from The Davee Foundation and the Northwestern University Alzheimer's Disease Core Center grant from the National Institute on Aging (AG13854). We are grateful to Rebecca Gavett and Katherine Reiter for their role in the neuropsychological testing of participants in this project. The authors report no conflicts of interest.
A portion of the data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Abbott, AstraZeneca AB, Bayer Schering Pharma AG, Bristol-Myers Squibb, Eisai Global Clinical Development, Elan Corporation, Genentech, GE Healthcare, GlaxoSmithKline, Innogenetics, Johnson and Johnson, Eli Lilly and Co., Medpace, Inc., Merck and Co., Inc., Novartis AG, Pfizer Inc, F. Hoffman-La Roche, Schering-Plough, Synarc, Inc., as well as non-profit partners the Alzheimer's Association and Alzheimer's Drug Discovery Foundation, with participation from the U.S. Food and Drug Administration. Private sector contributions to ADNI are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of California, Los Angeles. This research was also supported by NIH grants P30 AG010129, K01 AG030514, and the Dana Foundation.
Some data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.ucla.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.ucla.edu/wp-content/uploads/how_to_apply/ADNI_Authorship_List.pdf
