While much is known about structural MRI correlates of longitudinal memory performance in AD [28
], analogous fMRI correlates are poorly delineated. This study explores the hypothesis that, in subjects with mild AD, candidate brain regions supporting associative-memory processes may be identified in which changes in encoding-related fMRI activity over 24 weeks vary with changes in performance on behavioral tests and clinical-trial memory measures. The rationale underlying this hypothesis is that individuals who show increased brain modulation during the memory task at follow-up show better performance at follow-up and individuals who show decreased brain modulation during the memory task at follow-up may also show worse task performance at follow-up. Using whole-brain statistical parametric maps, changes in fMRI activity during encoding of face-name pairs for a group of 22 subjects with mild AD significantly correlated with changes in ADAS-Cog total and word recognition, FCSRT-free, MMSE, and postscan name recall and postscan forced-choice name recognition for faces in several brain areas. These areas included parietal, medial temporal lobe (MTL) and frontal regions linked to retrieval of episodic memory and attentional control; these areas all undergo early structural changes in AD. The results presented here extend findings from cross-sectional clinical and cognitive studies [12
] to identify candidate functional measures that may provide additional support in proof-of-concept human clinical trials to gauge effects for promising AD drugs.
This study design and implementation have several important strengths that make results relevant to early-phase clinical researchers. These are inclusion of well-characterized mild-AD subjects with clinical, demographic and medication usage profiles typical of AD clinical-trial populations, administration of several standard clinical and neuropsychological measures commonly used in AD clinical trials, application of a well-characterized and robust fMRI associative-encoding paradigm, implementation of fMRI within the structure of a short-term, 24-week, randomized, controlled, blinded, AD clinical-trial format similar to phase 2 studies of symptomatic efficacy in AD, achievement of high subject follow-up and data integrity, and utilization of standard and widely available MRI processing and analysis streams and software. Important to the feasibility of using cognitive tests and fMRI as reliable measures in clinical trials is that there be no learning effect over time. The present paradigm has been shown to have high reliability across time and the stimuli used at baseline and the week-24 visits were different [1
]. Additionally, the neuropsychological measures utilized in this study have both high validity and reliability across time in mild AD dementia [20
The relationships between performance changes and fMRI changes in brain regions reported here are also supported by previous work that implicates these regions as supporting aspects of visual and verbal memory-related processes, as well as data from our previous study that examined a cross-sectional behavioral-fMRI link using the same paradigm [12
]. In that study, the N>R contrast was correlated with ADAS-Cog scores, postscan memory performance, MMSE and FCSRT scores.
Effects Related to FCSRT-Free Recall Change Scores
Activity changes in the left inferior parietal and left precuneus correlated significantly with changes in performance on FCSRT-free recall. Inferior parietal regions are associated with correctly recalled items during cued recall [36
]. The left inferior parietal cortex, in particular, is involved in procedural and declarative memory processes [37
] and encoding of unfamiliar faces [39
] – conditions akin to encoding of novel name-face pairs in this fMRI paradigm.
Episodic retrieval has been implicated in the activity of the lateral posterior parietal cortex and the precuneus [40
], which are important components of the default mode network (DMN) [43
]. The DMN is characteristically active during task-independent, introspective thought processes, and its function may be particularly affected in cognitive aging and disrupted in patients with mild cognitive impairment and AD [5
]. The precuneus, considered a major hub in the DMN, is also strongly connected with the inferior and superior parietal lobules [50
] and is important for successful encoding [4
The regional investigation of activity in the DMN yields further insight into how the DMN is affected by AD [3
]. Over the 24-week trial window, activity related to novel face-name encoding increased in all regions and groups; whether and how such an increase may represent a signal of potential drug effect requires additional studies. Future analyses will investigate how the DMN and other intrinsic brain network patterns change over time and address associations with potential signals of antidementia drug effects.
Effects Related to ADAS-Cog Change Scores
Changes in activation were observed in frontal and parietal regions, including the inferior parietal cortex and middle temporal gyrus that significantly correlated with ADAS-Cog change scores. Lateral and inferior parietal regions are implicated in several memory processes including in old/new recognition [53
] and in visual recognition of words [54
Effects Related to FR and FCR Change Scores
Changes in activation that significantly correlated with postscan memory test scores were observed in frontal, temporal and parietal regions, including the inferior frontal gyrus, middle temporal gyrus and inferior parietal cortex; regions that may serve as hubs in important cognitive networks, including the DMN [43
], in the perceptual decision-making circuit, and in attentional control [58
]. The findings of significant correlations between changes in memory tests outside the scanner with changes in magnitude of encoding-related activity inside the scanner for MTL structures supports the utility of this block design, paired-associate, encoding paradigm for detecting and potentially tracking fMRI signal changes even in AD patients with significant MTL atrophy.
Summary of Effects Related to Cognitive Test Score Changes and Candidate Brain Regions
These results support activity changes in several regions previously implicated in cross-sectional studies to have encoding-related activity associated with memory performance outside the scanner (e.g. inferior parietal lobule, superior temporal gyrus, precuneus). However, we did not find extensive similar associations with more general measures of cognition that have low memory load (e.g. MMSE) [12
]. Other than a true lack of association, other plausible explanations for these null results are that such global measures may have relatively smaller memory-related signals that would be otherwise detectable in isolation (e.g. via ADAS-Cog recognition) and in higher-powered studies, and that may be obfuscated in the context of a multidomain cognitive scale (i.e. ADAS-Cog total). Also, associations for encoding-related fMRI activity with FCSRT-free and FCR may have been more readily detected since a greater range in performance was observed on these tests compared to the other memory tests.
The present study focused on probing the functional neuroanatomical specificity of memory changes in mild AD, that generally occur earlier in the disease process relative to semantic cognition, apraxia and aphasia [60
], rather than probing more general changes in cognition. As such, paradigms that strongly recruit memory-sensitive regions to complete the task will yield the best results. In comparison, tasks probing motor function, such as studies of swallowing in AD [61
], are likely to be better suited than the present study to investigate functional neuroanatomical specificity for apraxia. Likewise, language tasks would be better suited than the present task for probing the functional neuroanatomical specificity of aphasia in neurodegenerative diseases and AD [63
]. Uncovering the functional neuroanatomical specificity of cognitive changes in AD requires varied studies that probe different cognitive systems at different stages of the disease. Here we make a start by probing the longitudinal functional neuroanatomy of a brain system that supports encoding of highly ecologically important explicit associations, those of faces and names, and one that this is also a relatively early casualty of the destruction wrought by AD.
Identification of candidate brain regions that track cognitive performance provides researchers with locations to inspect for functional changes in early-phase or proof-of-concept clinical trials; these regions could be monitored to assess for signals of drug effect and to track such effects over time. Ultimately, this information may be combined in integrative models with other biomarker data to better inform go-no go decisions for experimental drugs or dose-response relationships, to elucidate early trends in efficacy, or to predict individual treatment responses.
The utilization of imaging biomarkers is not a replacement for neuropsychological testing, but rather a complement to them. If the correlation between imaging and behavior were 1, then there would not be a need for imaging biomarkers; however, this is not the case in the present results. High correlations provide evidence that we are capturing variance related to changes in memory performance, but also provide evidence that imaging biomarkers may have unique variance that could be leveraged in early-phase trials. Thus, the goal is not to replace neuropsychological tests, but to develop imaging biomarkers that, when combined in integrated models with other variables (including neuropsychological measures), yield greater explanatory power, particularly for clinical changes over time. In particular, fMRI potentially manifests signals of effects at the synaptic/local field potential levels, and these signals are likely to be detectable much earlier than changes in behavior in order to guide drug discovery, selection and therapeutic trial efforts. While the current results do not support accuracy at the level of the individual, they are nonetheless encouraging as they support the detection of a potential signal at the group level with N's that are modest enough to support their investigation in small early-phase/proof-of-concept AD trials [1
The main limitations of this study and the caveats in the results are as follows: First, there is an implicit assumption that chronic administration of antidementia medications does not significantly alter the function-behavioral performance relationship of the network(s) utilized during the encoding task. In other words, there is an invariance in the nature, not necessarily of the degree of any fMRI signal modulation when, as in this analysis, the aim is to assess whether fMRI activity may track clinical and cognitive measures, and to delineate regions where such changes may occur (i.e. that chronically stable antidementia drugs between week 0/12 and 24 do not fundamentally alter the shared regions involved in the networks supporting these tasks). Second, in the present analysis, we do not study the potential modulatory effects of memantine on fMRI-BOLD signal or on function-behavioral performance relationships because effect sizes on cognitive measures in mild AD are unlikely to be detectable in short time frames and with only 24 subjects [64
] using such analysis – and in fact, as would be expected, the clinical measures did not show significant differences attributable to drug over the 24-week course of this study (i.e. accounting for multiple comparisons, there were no behavioral differences between the group that was on drug for 12 weeks and the group that was on drug for 24 weeks). Also, the relative homogeneity of our subjects and methods (e.g. single-site study, well-characterized subjects with high education, excellent adherence to protocol and follow-up) yields high internal reliability but may potentially limit generalizability to multisite studies. Finally, these results should be viewed as preliminary and will require replication in larger confirmatory studies, especially to assess whether performance changes on more global measures of cognition correlate with regional fMRI signals and whether the results are generalizable.
In conclusion, these results suggest that functional changes in several MTL and parietal brain regions may be potentially useful in tracking signals for change in memory performance over short periods in patients with mild AD. Our findings also provide further support for the potential of encoding task-fMRI as a complementary biomarker of AD that warrants further study, particularly in the context of early-phase clinical trials.