This longitudinal fMRI study demonstrates that clinical decline in older individuals is accompanied by significant loss of hippocampal activation over time. Consistent with our previous findings, subjects with the highest activation at baseline demonstrated the greatest clinical decline, supporting the hypothesis that there is a period of hyperactivation in the earliest stages of prodromal AD. These subjects also demonstrated the greatest loss of hippocampal activation over 2 years, consistent with the hypothesis that hyperactivity may be an indicator of impending hippocampal failure. These longitudinal fMRI findings were not explained by age, APOE status, or hippocampal volume loss. Our results suggest that longitudinal fMRI is capable of detecting early brain dysfunction predictive of subsequent decline, and tracking clinically meaningful changes over the course of early cognitive impairment.
Previous work from our group using a scene-encoding task in a separate group of older subjects found that fMRI activation in the posterior MTL was predictive of subsequent cognitive decline,2,3
but these studies did not include longitudinal imaging. The current study replicates our finding that increased baseline MTL activation is associated with greater subsequent clinical decline, in a separate cohort of subjects and using a different paradigm. The face-name paradigm engages more anterior hippocampal regions at baseline and was associated with loss of anterior hippocampal activation over time, suggesting that the hyperactivity predictive of decline may be regionally specific to the demands of the cognitive task. It is possible that this finding represents a regression to the mean phenomenon; however, we did not observe any longitudinal differences in the cognitively stable normal older individuals, and the loss of hippocampal activation paralleled clinical decline on both the CDR and neuropsychological measures of episodic memory.
In addition to our own studies that have reported evidence of increased hippocampal activation in the earliest stages of MCI, relative to normal controls,2,3,20,21
several other groups have reported hyperactivation in MCI, in particular for successful memory formation.8,9,22
This paradoxical increase in activation has also been observed in carriers of the APOE
even among young carriers,13
and in asymptomatic offspring of individuals with confirmed familial late-onset AD.23
We did not observe a significant effect of APOE
carrier status in this study, perhaps because of the more advanced age of our subjects. We have previously speculated that hyperactivity might reflect an increase in neuronal recruitment to maintain memory performance in the setting of early AD pathology,10
which is supported by the finding of regionally specific hyperactivity according to memory task demands. The finding that increased activity is a predictor of impending hippocampal failure, however, suggests that the hyperactivity may be an indicator of neuronal stress or be directly related to the pathophysiologic process of AD. Hyperactivity could be related to aberrant sprouting of cholinergic fibers,24
excitotoxicity from the excessive stimulation of NMDA receptors,25
or stress from Aβ-induced neuronal alterations (e.g., excitatory activity triggers compensatory inhibitory mechanisms, which jointly contribute to network dysfunction).26
It should be noted that blood oxygen level dependent (BOLD) is an indirect measure of neuronal activity, and may reflect other physiologic properties, including cerebrovascular changes.27
The BOLD signal may depend on baseline perfusion, cerebral blood volume, vascular compliance, and interactions between these variables.28
The increase in BOLD activation observed in our fast decliner group could represent a decreased resting oxygenation state, increased perfusion, or some other disruption in neurovascular coupling.
One other potential explanation for apparent decreases in functional activation over time could be loss of hippocampal volume. Controlling for volume both within ROI analyses and in a multivariate model, we still observed loss of hippocampal activation. We did not observe significant hippocampal atrophy over the 2 years using automated volumetric analyses, perhaps because many of our subjects were still at relatively early stages of prodromal AD, prior to significant neuronal loss. It has been postulated that hippocampal atrophy may occur later in the pathophysiologic sequence of AD.29,30
We hypothesize that those subjects who demonstrated the greatest decrease in function will subsequently manifest hippocampal atrophy, and are actively pursuing this hypothesis in ongoing longitudinal studies.
Over the course of the study, we modified our postscan memory test to include more test items at the follow-up scan. Unfortunately, this limited our ability to determine if loss of hippocampal signal accompanies worsening performance on postscan memory testing. Longitudinal fMRI studies now underway in our laboratory include more detailed postscan testing, and we will be able to analyze successful vs failed encoding using event-related analyses.
If fMRI is to be useful as a tool for evaluating longitudinal change in brain function or the efficacy of pharmacologic interventions, it is essential to demonstrate the stability of fMRI measures over extended periods of time in the absence of clinical change. Two recent studies suggest that fMRI reproducibility is reasonable over short time frames in older populations.31,32
Our finding of reproducible patterns of activation over 2 years in clinically stable normal subjects is encouraging that fMRI can provide a reliable estimate of hippocampal activation over time, and thus may prove a useful tool for detecting clinically meaningful change in longitudinal pharmacologic studies in early prodromal AD.
Our results in subjects with very mild impairment demonstrate that clinical decline is related to the loss of hippocampal activation, and support previous findings that paradoxically increased activation may be predictive of impending clinical decline. These findings suggest that fMRI may prove valuable in tracking very early progression of brain dysfunction, prior to the point of irreversible neuronal loss, and at a time when disease-modifying therapies are likely to be most efficacious.