The results of this study demonstrate that when cognition is fully characterized over a sufficient period of time during the phase when older persons are considered cognitively healthy, substantial individual variability in slopes of cognitive change is observed and a faster rate of cognitive decline, particularly in working memory, can be linked to hippocampal atrophy, a well-established biomarker of risk for MCI and/or AD.
Our findings are consistent with a number of studies that report cognitive decline in the healthy years preceding a clinical diagnosis of MCI and/or AD (e.g., Amieva et al., 2008
; Johnson et al., 2009
; Wilson et al., 2011
; Rosano et al., 2012
) and extend these results by underscoring the substantial variability in cognitive function that occurs within the normal range during these years. It is clear that some persons decline, some stay stable and others improve, and this heterogeneity may be one explanation for mixed findings regarding the relationship of cognitive decline to measures of brain integrity in the cognitively healthy years. In this study, those persons who were cognitively healthy at the time of scan, but who declined cognitively in the years preceding the scan, had smaller hippocampal volumes.
We measured cognitive decline globally, but also in five different domains, and found that the association with smaller hippocampal volume was driven most strongly by decline in working memory. This finding is in line with studies that have connected the soundness of working memory in aging to the integrity of the hippocampal region (reviewed in Salthouse, 2011
). However, the association between episodic memory and hippocampal atrophy was weaker, a finding that is often noted in studies of cognitively healthy older persons (reviewed in Van Petten, 2004
). When we separated our sample into domain-related decliners and maintainers, the percentage of working memory-decliners was quite high (97%) and the association with hippocampal atrophy was strong, whereas the percentage of episodic memory-decliners was quite low (14%) and the association with hippocampal atrophy was marginal. Again, these findings emphasize the importance of addressing sample composition in longitudinal studies of cognition and brain integrity in cognitively healthy older persons. Most importantly, however, they suggest that older persons who are considered cognitively healthy but have evidence of cognitive decline, particularly in working memory, may be amid a pathological cascade and on a protracted trajectory toward neuronal injury, episodic memory impairment and eventually a clinical diagnosis of MCI or AD.
It has been established in many studies that maintaining cognitive function in older age lowers the risk of adverse cognitive and functional outcomes (reviewed in Hertzog et al., 2009
), however, the association of cognitive maintenance with brain integrity is not well-studied. Only one study that we know of has examined the relationship of cognitive maintenance to brain integrity in cognitively healthy older persons in the years prior to imaging. Rosano et al. (2012
) reported that 59% of persons in their sample maintained global cognitive function, based on the Modified Mini-Mental State Examination (3 MS; Teng and Chui, 1987
), over 4 time points in the decade prior to time of scan. Cognitive maintainers had larger medial temporal lobe (hippocampus, parahippocampus, entorhinal cortex) gray matter volumes. The results of the current study support this finding in that global cognitive maintainers (42.2% of our sample) had larger hippocampal volumes compared to cognitive decliners. However, the possibility that cognitive maintenance reflects susceptibility to practice effects needs to be addressed. We examined this possibility in secondary analyses using scores from the cognitive domain that generated the largest percentage of cognitive maintainers, episodic memory (86%). We added additional terms to the linear mixed models representing the number of follow-up years of cognitive testing, as previously reported (Wilson et al., 2006
). We found some evidence of a practice effect on episodic memory, however, the percentage of episodic memory maintainers still reached 60% after adjusting for this practice effect. This suggests that these individuals are genuinely maintaining or improving their episodic memory. A number of lifestyle behaviors that can potentially protect cognition have been examined (reviewed in Hertzog et al., 2009
). For example, it has been shown that frequent mental stimulation leads to better cognitive function (Wilson et al., 2012
), particularly in episodic memory. More studies are needed to further understand the brain basis of this phenomenon.
This study has important strengths. The data were sampled from a large, longitudinal clinical-pathological study in which subjects have participated in up to 14 annual assessments using well-established clinical and cognitive measures. The study also has limitations. The period of time over which cognitive change was measured in this study cannot be considered preclinical. All subjects were cognitively healthy at time of scanning and we await clinical outcomes. Although the Rush Memory and Aging Project was designed to closely represent the general population of persons aged 65 and over, the sample in this study was selected. Multiple years of cognitive data allowed the examination of cognitive change, but the volume data are from one time point. Thus, these data cannot address simultaneous change in pre-scan cognition and hippocampal volume. However, participants of the Rush Memory and Aging Study agree to bi-annual scanning until death so it will be possible to examine the associations between cognitive change, transition to clinical diagnoses, and macrostructural change in the future. Hippocampal volume in the elderly may be influenced by the presence of not only AD pathology, but other pathologies such as hippocampal sclerosis (Dawe et al., 2011
), Lewy bodies (Burton et al., 2012
), and amyloid angiopathy (Jagust et al., 2008
; Erten-Lyons et al., 2013
). Although we cannot address the neuropathology of reduced volume in this study, histopathologic information will be available for these subjects in the future. Finally, a stronger magnet would have allowed a closer examination of associations with hippocampal volumes in specific subfields. Using a 4T magnet, the CA1 subfield has been shown to be most strongly affected by age, particularly in the seventh decade (Mueller et al., 2006
) and hippocampal deformation was primarily attributable to CA1 volume loss in a post-mortem imaging shape analysis of elderly persons over the age of 65 (Dawe et al., 2011
). Post-mortem imaging will also be available for these subjects in the future. These limitations notwithstanding, the results of this study are important for at least two reasons. First, they emphasize the need to deeply characterize cognition, brain structure and their relation during the years in which older persons are considered cognitively healthy. Second, the findings are clinically relevant. Whereas clinicians often use imaging biomarkers such as hippocampal volume to predict subsequent cognitive decline, these findings show that hippocampal volume can inform on the trajectory of cognitive change during the period of time preceding the patient's first presentation to the clinic. This information would help the clinician elucidate the patient's cognitive history, identify risk of developing a clinical diagnosis of MCI or dementia due to AD and optimize treatment.
Conflict of interest statement
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.