|Home | About | Journals | Submit | Contact Us | Français|
Alzheimer’s disease (AD) is a progressive age-related neurodegenerative disease. At the time of clinical manifestation of dementia, significant irreversible brain damage is already present, rendering the diagnosis of AD at early stages of the disease an urgent prerequisite for therapeutic treatment to halt, or at least slow, disease progression. In this Review, we discuss various neuroimaging measures that are proving to have potential value as biomarkers of AD pathology for the detection and prediction of AD before the onset of dementia. Recent studies that have identified AD-like structural and functional brain changes in elderly people who are cognitively within the normal range or who have mild cognitive impairment (MCI) are discussed. A dynamic sequence model of changes that occur in neuroimaging markers during the different disease stages is presented and the predictive value of multimodal neuroimaging for AD dementia is considered.
About 35.6 million people are affected world-wide today by a mild to severe clinical dementia syndrome, associated with costs of about US$ 604 billion (World Alzheimer Report 2010, www.alz.org). The number of affected people is predicted to dramatically increase to 115 million by 2050 (World Alzheimer Report 2010, www.alz.org). Dementia of the Alzheimer’s disease (AD) type is the most frequent form of age-related dementia [1–2].
Clinically, initial progressive memory deficits that are eventually accompanied by more global cognitive and attention deficits are typical in AD dementia. Major pathologies in the brain associated with AD dementia have been identified, but the initial causes of such pathological changes are largely unknown. Causative factors include autosomal-dominant inheritable mutations in the genes of presenilin 1 (PS1) and presenilin 2 (PS2) and the amyloid precursor protein (APP). Familial AD (an inheritable form of AD) is typically associated with an early onset of the disease before the age of 65 years, but accounts only for about 1% of all AD cases . For AD without the presence of such known genetic causes (sporadic AD), a late onset of dementia (age > 65 yrs) is typical. The most important risk factor for sporadic late-onset AD is age, with the annual incidence of clinically diagnosed AD dementia being < 1% of adults between 65–69 years old, but > 8% in adults 85 years and older . Presence of the apolipoprotein E (ApoE) ε4 allele is the strongest genetic risk factor of sporadic AD . Presence of at least one ApoE ε4 allele advances the age of clinical onset of AD dementia significantly (from age 84 years to 68 years) and increases the risk of AD dementia by a factor of 4 (for regularly updated meta-analyses, see http://www.alzgene.org/, last update Jan 29, 2010).
Core neuropathologies in AD include the accumulation of the protein amyloid-beta (Aβ) and the development of neurofibrillary tangles, which have been associated with neuronal degeneration and clinical symptoms of dementia  Such brain changes occur decades before the onset of dementia. The detection of Aβ in the brain of living subjects has been made possible by recent developments in positron emission tomography (PET) using radiotracers such as [C-11]-labeled Pittsburgh Compound-B (PiB), which label Aβ deposits . Such studies have shown that substantial levels of Aβ deposits are present in subjects before the onset of dementia or even before any overt signs of cognitive impairment (as discussed later in the review). Other early brain changes include a decline in synaptic function, as assessed by [F-18]-fluorodeoxyglucose positron emission tomography (FDG-PET), gross neuronal loss causing atrophy, as measured by volumetric magnetic resonance imaging (MRI), and white matter changes within axonal projections as detected by diffusion tensor imaging (DTI)(Box 1). These different types of changes may evolve sequentially and relate to the development of cognitive impairment within the clinical course of AD .
Mild cognitive impairment (MCI, see Glossary) is a well defined clinical syndrome which includes deficits in memory or other cognitive abilities that may herald AD dementia . Subjects with MCI have a higher risk to progress to AD dementia  but a substantial proportion of MCI subjects remain stable for years or revert to normal, indicating that clinical MCI symptoms can also stem from non-AD related etiologies. This notion is substantiated by the finding that only a subset of of patients with MCI display measurable amounts of Aβ on PiB-PET scans (for a review, see ). Therefore, the combination of cognitive deficits (as present at the MCI stage) with abnormal values for biomarkers indicative of AD pathology (e.g. Aβ levels as measured by PiB-PET) has been proposed as the research diagnostic criteria of predementia AD, called “MCI due to AD” or “prodromal AD” [3–4, 13]. Arevision of the clinical diagnostic criteria of AD by the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) and the Alzheimer's Disease and Related Disorders Association (ADRDA)  has been recently proposed [3, 13, 15]. The National Institute of Aging (NIA) and Alzheimer’s Association (AA) work group- still recommended purely clinical and neuropsychological criteria for routine clinical diagnosis of AD dementia and MCI [13, 15]. For the use as research diagnostic criteria of AD and MCI, however, the dual clinico-pathological definition based on clinical testing and biomarker measurements has been adopted [13, 15]. . In this Review, we use the term “MCI due to AD” in reference to subjects who show abnormalities in neuroimaging biomarkers of AD in the presence of progressive cognitive deficits such as the presence of MCI. It should be mentioned that MCI due to AD is a diagnostic label in view of clinical and biomarker evidence, but neither does this diagnosis mean that AD pathology has necessarily caused the clinical symptoms nor that subjects with MCI due to AD necessarily progress to AD dementia. Rather it is a research concept that has not yet been fully validated.
The question of the clinical fate of subjects with AD-like brain changes becomes even more urgent at an earlier time point, i.e. the preclinical phase of AD , when such brain changes are present without any overt cognitive deficits. The extent to which subjects with preclinical AD progress to dementia is currently not known. Of note, it has not been established that all subjects with AD-like brain changes inevitably progress to AD dementia, and many subjects with preclinical AD die of natural age-related causes before any signs of dementia are ever observed.
In the current article, we review recent imaging studies that have examined brain changes in both the preclinical and MCI phase, and evaluate the utility of neuroimaging markers for the prediction of clinical progression in these two distinct phases. The association between different imaging modalities is shown for each of the stages, and stage-specific changes in neuroimaging markers are discussed with respect to the predictive value for cognitive decline and the progression from MCI to AD dementia.
Before discussing neuroimaging changes of relevance for the preclinical and MCI due to AD stages, it is helpful to begin by briefly overviewing the typical neuroimaging findings that have been observed over the past several years in patients with clinically manifest AD dementia. Numerous studies have determined that the pattern of Aβ deposits detected in the brains of living subjects diagnosed with AD dementia using [C-11]-PiB PET closely matches the pattern predicted by the histochemical detection of Aβ in postmortem brain tissue from subjects with clinical AD [8, 12] (Figure 1). Other amyloid PET imaging tracers that have been recently developed, such as the [F-18]-labeled radioligand florbetapir (also known as [F-18]-AV-45) have shown similar results of a good correspondence between the distribution of amyloid PET and histochemically detected Aβ in the brain [11–13]. Clinical utility is expected, especially from these[F-18]-labeled ligands, which possess a radiotracer half-life that is sufficiently long enough to be produced off-site and shipped to clinics. However, compared to [C-11]PiB-PET, such novel amyloid PET ligands are currently less well characterized in clinical populations of MCI and AD patients .
During the pre-clinical stages of AD, the pattern of Aβ in the brain, as derived from histochemical post-mortem brain examinations , shows an initial diffuse distribution within the neocortex and progresses towards the temporal allocortex, including the hippocampus, as well as subcortical brain structures during the course of the disease (Figure 1A). In the clinical stages of cognitive impairment, Aβ deposits are increased in both the amount and spatial extent throughout the brain  (Figure 1A). At the stage of AD, Aβ deposits are most frequent in neocortical brain regions, hippocampus, amygdala, and subcortical brain regions (e.g. striatum and basal ganglia), but only weakly distributed within the cerebellum and some brain stem areas . . Consistent with the histochemical post-mortem findings on Aβ deposition in the brain, a distribution of amyloid PET uptake can be already seen widely within the brain in the preclinical stage . A significant increase in Aβ deposition in predilection brain areas of AD can be observed in AD dementia patients (Figure 1B). Based on global brain PiB-PET uptake, subjects are often dichotomized into groups with high and low PiB-PET values, i.e. PiB-PET(+) and PiB-PET(−) groups . Data from 15 research groups have shown that 96% of 341 clinically diagnosed AD patients and 24% of 651 cognitively normal elderly controls were PiB-PET(+) [18–27]. This finding points out the diagnostic sensitivity of PiB-PET imaging, but also demonstrates that Aβ pathology is not specific for the dementia phase of AD, similar to what has been known from brain autopsy studies .
Longitudinal studies including serial scans provide an assessment of individual trajectories of brain changes and give insight into variability of brain changes between different persons. A recent longitudinal study with serial PiB-PET scans has reported a mean annual increase of 4% in global PiB-PET uptake over a period of 2 years in elderly subjects with clinically diagnosed AD . However, such changes are relatively small, and show substantial variability between AD patients . Furthermore, some studies do not report a significant increase in global PiB-PET uptake with time, at least when assessed over a few years [31–33], suggesting that Aβ deposition may increase only slowly or has already reached a plateau in patients with AD dementia . However, it should be noted that most longitudinal studies have so far included only small numbers of subjects (ie. usually <20 subjects), and due to the substantial variability between subjects, such studies may have lacked sufficient statistical power to detect a significant annual increase in PiB-PET [31–32, 34].
FDG-PET-derived measures of brain glucose metabolism and cerebral blood flow are markers of synaptic dysfunction, typically obtained during resting state (Box 1). A characteristic pattern of hypometabolism in the temporo-parietal region of the cortex, which is involved in episodic memory function, is present at the AD dementia stage [35–36]. Joint assessment of FDG-PET and PiB-PET shows the expected inverse association between both modalities within the temporoparietal region, although not frontal regions [37–38], suggesting that hypometabolism in core brain regions is associated with Aβ pathology.
A well-established finding using functional MRI (fMRI; Box 1) to examine AD-associated differences in brain activation is decreased activation in the hippocampus during episodic memory tasks [39–40]. Such results are consistent with clinical findings of early episodic memory deficits in AD subjects [Ref]. Abnormal fMRI-assessed brain activation can also be observed in AD subjects without engaging subjects in a cognitive task. In healthy subjects, the default network of brain regions is active during resting periods (probably a reflection of cognitive processes such as introspection) but becomes deactivated during cognitive processes that are focused on external stimulation, such as performance of a cognitive task . The intrinsic functional connectivity (i.e. the coordinated co-activation of the default network’s brain regions measured with resting state fMRI) is impaired in AD dementia subjects during the resting state as compared to cognitively healthy elderly subjects [43–44]. Furthermore, the default mode network of AD subjects do not show the beneficial deactivations that healthy subjects show when assessed during memory tasks .The abnormal fMRI activation and connectivity within the default model network brain regions overlaps spatially with the temporo-parietal regions of FDG-PET hypometabolism in AD subjects, suggesting that both modalities exhibit a converging pattern of functional brain impairment in AD.
Structural MRI is another imaging technique that has been used to assess brain changes in AD subjects, specifically with respect to grey matter volume changes. A recent meta-analysis that included 826 patients with AD and 1027 elderly cognitively normal subjects found that the medial temporal lobe showed the strongest changes in AD dementia , with a volume loss of 20% in the hippocampus already present at a mild stage of AD dementia. Other brain areas including the lateral temporal lobe, parietal and prefrontal lobes were found to be negatively affected as well, but changes in brain regions outside of the medial temporal lobe were smaller and more variable across studies .
An assessment of white matter fibers using the imaging technique of DTI has revealed that the fibers connecting the hippocampus and posterior cingulate gyrus are impaired in AD subjects to a significantly greater degree as compared to control subjects . This suggests that white matter damage may relate to grey matter atrophy within the temporo-parietal brain network in AD. A recent meta-analysis of DTI studies of AD subjects confirmed a large effect size of white matter damage in the posterior cingulum, but also within major white matter bundles connecting the prefrontal cortex with the medial temporal lobe or the parietal cortex , suggesting that white matter damage affects large-scale networks in AD.
Taking such diverse imaging modality findings together, it is clear that gross changes in the living brain of AD dementia subjects can be visualized and measured. A hypothetical model of sequential neuroimaging changes in AD has recently been proposed and Box 2 provides an overview of disease-stage specific changes for different neuroimaging markers throughout the course of the disease. The neuroimaging changes in the early stages of disease preceding dementia are discussed in the following sections.
There are a variety of neuroimaging biomarkers which are in use for the prediction and validation of AD. Some of these techniques detect core AD pathology such as amyloid deposition (as measured by amyloid PET), while other imaging modalities detect neurodegeneration such as FDG-PET and MRI assessed functional and structural changes that may occur in a specific temporal order as recently proposed , Some neuroimaging detected changes occur already at the preclinical stage, others at the MCI stage and all of them are useful for detecting changes at the AD dementia stage (Figure I). The dynamic changes in neuroimaging markers are known to occur non-linearly throughout the stages of the disease [111–112]. However, the exact temporal sequence of structural and functional brain changes, and how changes in these various different imaging modalities are related, remains to be determined. The model sequence model displayed presents a hypothetic framework based upon currently available neuroimaging data.
Another major category of primary biomarker candidates for AD consists of CSF biomarkers (for review, see ). CSF samples can be obtained from subjects by lumbar puncture and analyzed in standard laboratory tests. Major CSF-derived markers include:
In this section, we will provide an overview of neuroimaging studies that have assessed changes in subjects with preclinical AD, i.e. in elderly subjects who show normal cognitive abilities but have already AD-like brain abnormalities such as discussed above. Findings from amyloid PET studies will be discussed first and compared to findings on amyloid PET in AD dementia. Functional and structural brain changes (as assessed by FDG-PET and MRI methods) will be discussed subsequently, especially in relation to early Aβ pathology, taking into account the influence of ApoE genotype as a potential modulating factor of such brain changes at this early disease stage.
As discussed above, PET imaging studies using the PiB ligand have revealed that about 10–30% of elderly cognitively normal subjects are PiB(+) . ApoE genotype is known to modulate the level of Aβ deposition in the brain , and homozygous ApoE ε4 carriers show the highest proportion of PIB(+) . A recent study that imaged amyloid in the brains of subjects within the age range of 18–50 years using [F-18]-AV-45-labeled PET revealed an absence of Aβ deposits, regardless of ApoE genotype . This suggests that substantial elevations in Aβ occur only after 50 years of age . An increase in Aβ levels, as assessed by PiB-PET, has been associated with more rapid memory and global cognitive decline ; whether cognitively normal subjects with increased PiB-PET uptake will show higher likelihood to progress to dementia is, however, unknown and studies have so far not included sufficient follow-up durations to adequately assess this question. Vulnerability factors (such as sub-clinical cerebrovascular disease) and brain/cognitive reserve factors are likely to play a modulatory role in the process of Aβ deposition [26, 53–55]. However, the strongest determining factor at this stage remains the ApoE4 allele, therefore this genetic information is important to take into account when examining functional and structural brain changes in subjects.
A recent study of cognitively normal subjects who were characterized as PiB-PET(+) did not find any evidence of FDG-PET hypometabolism in the prefrontal cortex and parietal lobe during resting state . However, additional studies are needed to replicate this finding and to assess whether such a conclusion is common across different experimental groups. Other studies which have assessed FDG-PET hypometabolism in a subset of cognitively normal subjects that carry the ApoE ε4 allelle have detected changes in a subset of regions affected in AD [57–58], and these changes predict progression to MCI . Since ApoE ε4 increases the deposition of Aβ already in the preclinical stage , it is possible that such genetic effects on FDG-PET are related to the presence of Aβ pathology. However, results from FDG-PET data obtained in ApoE ε4 carriers within the 20–39 year age range suggest that a fundamental metabolic abnormality is present already at an age when substantial fibrillar Aβ deposition is unlikely to have developed . This suggests that ApoE genotype may exert an influence on FDG-PET metabolism beyond its influence on Aβ in the preclinical phase.
FMRI studies have uncovered several AD-like alterations in brain activity that are associated with Aβ or ApoE ε4 genotype at an early preclinical phase. Although memory task-induced brain activation within the medial temporal memory system does not appear to be strongly affected in cognitively normal PiB-PET(+)subjects, the default-network shows an AD-characteristic lack of deactivation, and even reverts into paradoxical increases in some of the network’s brain areas [62–63]. In further parallel to the findings in AD dementia, the functional connectivity between brain regions of the default mode network is disrupted in elderly PiB-PET(+) adults [22, 64–65]. These findings suggest that Aβ is associated with abnormal regulation of large-scale networks, especially within the default mode brain regions, even at this early preclinical stage . However, disruption of default mode network connectivity has recently been reported in elderly adults who carry the ApoE ε4 allele but who are PiB-PET(−) and in young ApoE ε4 carriers who are unlikely to have substantial Aβ deposition . Furthermore, the presence of the ApoE e4 allele has also been associated with increased cortical activation during a visual learning task  and disruption of functional connectivity of default mode network . Thus, similar to the findings of an ApoE effect on FDG-PET at the preclinical stage, ApoE ε4 genotype and elevation of Aβ may both influence - to some extent independently - brain activation .
There is now accumulating evidence that Aβ deposition in the brain is associated with grey matter atrophy in the preclinical stage of AD. In the so far largest combined cross-sectional MRI & PiB-PET study that included 135 elderly cognitively normal subjects, PiB(+) subjects showed higher hippocampus and cingulate cortex atrophy compared to PiB(−) subjects . Furthermore, clinical longitudinal assessment revealed that PiB(+) subjects, but not PiB(−) subjects, showed decline in episodic memory and working memory over a time span of 16 years that preceded or succeeded the PET scan . Consistent with these findings, other cross-sectional studies have shown an association between PiB-PET and hippocampus volume loss [70–71] and posterior cingulate gyrus atrophy . A major question is what is the threshold level of Aβ deposition that is associated with atrophy? Some researchers argue that even small levels of elevation in Aβ, including at levels that are currently defined as PIB(−), may already be associated with increased brain atrophy . On the other hand, the observation in PiB-PET studies that cognitively normal subjects show a bimodal segregation into groups of high and low levels of PIB binding [i.e. PIB(+) and PIB(−)]  suggests the existence of a distinct group of elderly adults with abnormal and potentially disease-specific elevations of Aβ. At this stage, any differences in the sensitivity of using absolute levels of Ab-labeled, versus the identification of two distinct groups of PIB(+) vs. PIB(−), for the detection of brain atrophy and prediction of dementia have yet to be validated. Furthermore, other pathological events such as the formation of soluble Aβ oligomers  ,or the development of tau-related neuropathologies , are likely to be critical factors in Aβ-related neurodegeneration, and will need to be taken into consideration in future assessments of Aβ-associated brain atrophy.
ApoE ε4 genotype may also be an important determinant of brain atrophy at the preclinical stage. Similar to cross-sectional findings in AD dementia subjects , elderly cognitively normal subjects who carry the ApoE ε4 allele have greater hippocampus volume loss when compared to elderly cognitive normal adults without that allele . A large longitudinal population-based study with serial MRI scans in 1186 healthy subjects, showed that cognitively normal subjects with two ApoE ε4 alleles (i.e. homozygous) had a faster atrophy rate for the hippocampus and whole brain over a 4 year interval as compared to heterozyguous ApoE ε4 carriers and non-carriers . . Thus, both elevated Aβ levels and ApoE genetic risk are associated with grey matter atrophy already in subjects without cognitive impairment.
Microstructural white matter changes, such as detected by DTI, have not been assessed in preclinical subjects in association with PiB-PET so far, and only few studies have investigated the impact of ApoE genotype on white matter changes in this subject group. cognitively normal adults with at least one copy of ApoE ε4 show reduced white matter integrity (as measured by a reduced FA value, Box 1) in the posterior cingulum, corpus callosum, and other major white matter bundles that connect different lobes of the brain and support the default mode network . These changes were observed in young adults (20–35 years) , middle aged (49–65 years)  and elderly cognitively normal subjects (> 65 years) . Such changes are reminiscent of the early functional brain differences that were observed in ApoE ε4 carriers at both a young and old age, suggesting that these changes may occur independently of Aβ, as discussed above.
In summary, the amyloid PET findings are in accordance with histochemical autopsy studies demonstrating substantial deposition in the neocortex in elderly subjects without cognitive deficits, who are in the preclinical stages of AD. These PiB-PET(+) subjects already already a number of structural and functional brain alterations including reduced functional connectivity between brain regions of the default mode network and accelerated medial temporal lobe atrophy that is associated with episodic memory impairment. Such functional and structural brain changes in PiB-PET(+) subjects are reminiscent of brain changes seen in AD dementia. Whether these early brain changes at the preclinical stage herald clinical progression to AD dementia is, however, currently known. In the next section, we review neuroimaging findings in subjects who are at a clinically more progressed stage, i.e. MCI.
In the following section, we review neuroimaging findings in subjects with clinically manifest MCI without full-blown dementia. The same imaging modalities as discussed previously are reviewed to allow for a direct comparison with findings in the preclinical AD and AD dementia stages discussed above.
About 33 – 61% of subjects diagnosed with MCI show AD pathophysiology by virtue of being PiB-PET(+) . The proportion of PiB-PET(+) subjects is intermediate compared to the proportion of PiB-PET(+) subjects found at the preclinical stage and the AD dementia stage (see above), suggesting an increase in the frequency of PiB-PET(+) from preclinical to clinical AD dementia stages. Clinical longitudinal studies show that subjects with MCI who convert to AD dementia within 2–3 years have higher baseline PiB-PET levels than MCI subjects who do not clinically worsen within that timeframe [80–81]. The evidence for the utility of PiB-PET in identifying AD pathophysiology in the setting of clinical MCI is becoming increasingly convincing. When pooling data from nine PiB-PET studies that examined the PIB-PET status in MCI [20, 23, 27, 81–82], overall 161 of 272 MCI subjects were amyloid-positive (ie. 59%). Five of these studies (155 MCI patients) included longitudinal clinical follow-up assessments, which showed that between 20–50% of subjects progressed to clinical AD over 1–3 years; the large majority of these MCI-AD converters were PiB-PET(+) at baseline [20, 23, 81–83]. This number will almost certainly increase as the follow-up time is increased to at least five years for all subjects. In contrast, only a relatively small proportion of PiB-PET(−) MCI patients (~8–15%) progressed to clinical AD in these studies [20, 23, 81–83]. Amongst PiB-PET(+) subjects with MCI, it is known that the ApoE ε4 genotype accelerates the time to progression from MCI to AD . Thus, the combination of amyloid-imaging together with genetic determination of ApoE alleles is valuable for predicting higher risk of progressing to AD at the MCI stage.
The AD-like hypometabolic pattern (e.g., decreased signals in the precuneus and temporo-parietal cortex) has likewise been observed in MCI subjects (for a review, see ; Figure 2). Indeed, several studies have shown that abnormalities in FDG PET predict progression from MCI to AD [85–87]. Based upon FDG-PET metabolism within the temporo-parietal brain regions and posterior cingulate gyrus, the test accuracy to predict conversion from MCI to AD dementia within 1 – 1.5 years ranged between 80–90% among these three studies, and both the sensitivity and specificity to predict AD dementia were above 80% in each study [85–87]. The predictive accuracy was even higher in ApoE ε4 carriers , suggesting that ApoE genotype may provide additional clinically relevant information for the prediction of AD dementia by FDG-PET hypometabolism. Whether FDG-PET hypometabolism in MCI is linked to deposition of Aβ is not clear at this stage. Only few studies have examined the association between FDG-PET and amyloid PET at the MCI stage, reporting weak negative associations between FDG-PET and PIB-PE , which is consistent with the observation that FDG-PET and PIB-PET provide complimentary information for the prediction of conversion from MCI to AD . Factors such as cognitive reserve may modulate the association between FDG-PET assessed brain metabolism and amyloid PET uptake .l conversion to AD.
During resting state, disrupted connectivity of the default mode network is present in MCI , similar to what was reported in the preclinical stage in PiB-PET(+) subjects and subjects with AD dementia (as discussed above). Reduced default mode network connectivity has been associated with a higher risk of converting to AD dementia within a 3–4 year follow up interval , suggesting that impaired default mode activity is predictive of impending clinical progression to dementia. Cognitive–task associated changes in fMRI have also shown to be predictive of cognitive decline in MCI . Specifically, hyperactivation within the medial temporal lobe has been observed in the mild stages of MCI [45, 93] in association with subsequent cognitive decline . Such hippocampus hyperactivation during memory performance is already present in PIB-PET(+) subjects with a clinical dementia rating (CDR) of 0.5, who are in the transition phase between preclinical AD and MCI due to AD . This suggests that memory-related hyperactivation of specific brain regions becomes abnormal early in the course of the disease. Interestingly, at clinically more advanced stages of MCI and in AD dementia subjects, this memory task-induced hyperactivation of the temporal-parietal memory network, however, turns into an activation deficit . Specifically what this initial hyperactivation is due to and how that turns into an activation deficit remains unknown at this time, however, it may represent a transient phase of impending breakdown of neuronal networks. Possibly, the hyperactivation of particular brain areas reflects mechanisms of compensation within neural circuits. Alternatively, it may be more directly related to Aβ-pathology, either as a consequence of Aβ-induced hyperexcitability of neurons [94–95] or by increasing vulnerability to Aβ toxicity , but the exact mechanism remains to be determined in future studies. Whatever the underlying reason, initial evidence suggests that hyperactivity may reflect impending neuronal dysfunction and clinical decline .
Several studies have shown that increased brain levels of Aβ (as assessed by PiB-PET), or decreased cerebral spinal fluid (CSF)-Aβ levels, are associated with atrophy of the hippocampus and other brain structures [70, 97] in subjects with MCI, as assessed by volumetric MRI. These findings are consistent with those observed in preclinical AD. Multivariate statistical analyses have shown that an AD-typical pattern of MRI-assessed brain atrophy is present in a subgroup of MCI subjects [98–99]. Specifically, an atrophy pattern including the medial temporal lobe, posterior cingulate and orbitofrontal cortex is predictive of the development of AD dementia in MCI .
Apart from grey matter volume changes, increased white matter abnormalities have been observed in subjects with increased risk of AD. DTI abnormalities within the medial temporo-parietal network associated with episodic memory impairment have been repeatedly reported in the literature (for a review, see ) and are among the brain regions that show the strongest predictive value for the discrimination between MCI and cognitively normal subjects [48, 101–102]. Currently, there are only a few clinical longitudinal studies available for DTI [103–105], however, a recent study has shown that DTI-assessed diffusivity of the left hippocampus is predictive of the conversion from MCI to AD within 1.5 years .
In summary, AD-typical structural and functional changes at the MCI stage are reflected by changes in a variety of measures. Such measures include significant levels of amyloid PET binding in the brain, fMRI-assessed task-related medial temporal lobe hyperactivation and default network dysfunction, FDG-PET assessed temporo-parietal hypometabolism, MRI-assessed medial temporal lobe atrophy and diffusion changes. In the next section, we review data in MCI subjects of whether the combination of such markers is beneficial for predict the risk of developing AD dementia within a few years.
There is now accumulating evidence that a combination of neuroimaging markers shows additive effects of different modalities for the prediction of progression from MCI to AD dementia. Cross-sectional studies suggest that PiB-PET and hippocampus volume provide complementary information for the diagnostic classification of AD dementia . Within MCI subjects, the combination of structural MRI to assess hippocampus volume and DTI in the posterior parietal lobe contributes to the prediction of the severity of the memory deficits . Neuroimaging studies with longitudinal clinical follow-up have shown that the combination of MRI-assessed grey matter atrophy and FDG-PET hypometabolism within the posterior cingulate is associated with increased risk of clinical progression within 2 years when compared to either predictor alone . When neuropsychological measures are included as predictors as well, MCI subjects who are abnormal on both FDG-PET and episodic memory are ~12 times more likely to progress to AD than subjects who were normal on these measures, which is significantly higher when compared to an increase in risk by ~4 times based on episodic memory deficits alone . In that study, the hippocampus volume did not add to the predictive accuracy, suggesting redundancies between the different predictors tested. In fact, the prediction accuracy of some of the best single predictors including right entorhinal cortical thickness, right hippocampus volume, or trail making test B (TMT-B) may only be marginally improved by combining multiple markers . Thus, although the risk prediction of AD may be increased by combining markers, case-by-case decisions of classifying subjects into MCI to AD converters may not necessarily be improved by the use multiple neuroimaging markers. For the assessment of the utility of neuropsychological predictors as discussed above, it should be taken into account that neuropsychological performance is part of the definition of AD dementia, and thus neuropsychological predictors may correlate with progression to dementia, especially in studies that use only short clinical follow-up time intervals.
Eventually, any gain in predictive accuracy of a biomarker or test for the detection of AD needs to be weighed against costs in terms of side effects (e.g. radioactivity of PET tracers), potential invasiveness of methods (eg. lumbar puncture in the case of CSF-based tests; Box 2), and availability of the technology (e.g. the imaging scanners and/or radiotracers may only be available at major hospitals and research centers). Notably, there are also psychological costs for the individual consequent to learning about the presence of AD-like brain pathology and possible clinical prognosis based on such biomarker test results. Such considerations are especially important to carefully weigh in light of the fact that no cure for AD or treatment to effectively slow or halt disease progression currently exists. At this stage, the main utility of biomarker-based early detection of AD is in identifying subjects with high likelihood of progression to AD dementia in clinical trials for the testing of novel drugs for disease prevention or treatment. Large-scale cost effectiveness studies are needed to evaluate the utility of biomarker-aided prediction of AD in clinical practice, and additional studies to better understand the underlying neurobiological changes that are occurring during disease progression in AD are needed (Box 3).
Neuroimaging methods are capable of detecting substantial brain changes not only in subjects with AD dementia, but also in subjects in the mildly symptomatic MCI due to AD stage and even in cognitively normal subjects who may be in the preclinical stage of AD (Box 2). There are imaging modality-specific changes within these clinical/preclinical stages. In the preclinical stage, Aβ deposits are already present in a substantial number of subjects. Such changes are associated with increased grey matter brain atrophy, especially within the hippocampus, a key region affected in AD dementia. FMRI-assessed resting state functional connectivity and reduced deactivation of the default network is already impaired in association with Aβ deposition in the preclinical phase. Hyperactivation within the hippocampus memory network during memory performance occurs early in the MCI phase but reduced hippocampus activation is visible shortly before progressing to dementia, suggesting that such hippocampus hyperactivation is a transient sign impending clinical worsening. In contrast to fMRI detected changes in brain activity, FDG-PET abnormalities have not yet been associated with Aβ deposition in the preclinical phase and only become detectable in the MCI phase. Therefore, resting state fMRI may be a more sensitive measure than FDG PET to detect early changes associated with AD pathology. DTI lacks a clear characterisation at the preclinical stage so far, but similarly to the other neuroimaging markers shows specific patterns of alteration in association with ApoE genotype. Across the various imaging modalities, ApoE genotype is an important modulating factor. The presence of the ApoE ε4 allele is associated with accelerating Aβ deposition, hippocampus atrophy and default mode network dysfunction, beginning in the preclinical stage.
At the stage of MCI, AD-like patterns of brain changes are observed. These include a high proportion of amyloid deposits, medial temporal and parietal brain atrophy, and temporo-parietal FDG-PET hypometabolism, all of which are predictive of short-term conversion to AD. Although imaging modalities show independent contribution towards the prediction of AD, there are considerable redundancies among the neuroimaging modalities for the prediction of AD, and additional costs of combining different imaging acquisitions need to be weighed against the actual gain in prediction accuracy. Accessibility to imaging equipment can also be a limiting factor. The [11-C]PiB-PET radioligand, with a relatively short half-life of decay, can be manufactured most likely only at major clinical centers where the imaging equipment is located on-site. However, the development of other 18F-labelled tracers of amyloid with properties more amenable for wider distribution may supplement PiB-PET for clinical use. Of note, the Food and Drug Administration (FDA) has given given conditional support for the use of the [F-18]-AV-45 PET radiotracer, which should significantly enhance the use of amyloid-imaging during the clinical and preclinical stages of AD.
Already many in the research community appear ready to accept biomarker-aided diagnostic criteria for research purposes, such as been proposed for the revision of the NINCDS-ADRDA criteria, in part because they provide useful hypotheses for future verification. However, in view of the current evidence, further studies concerning the specific approaches to implement these criteria need to be conducted, including the establishment of quantitative criteria for different biomarkers and their extensive validation (by observing conversion to dementia of the AD type and pathological verification). Only after the clinical utility of such biomarker-aided criteria for the prediction of dementia are firmly established, can such measures be fully adopted for standardized clinical trials and routine clinical use. Finally, almost all results are reported in carefully selected cohorts, not epidemiological community-based populations; thus generalizability of these results to the population as a whole needs to be established. Such a validation process will likely take many years. This is especially true when one considers that new biomarkers, including blood and CSF assays of compounds not currently known (ie. evolving from exploratory proteomic approaches), and new imaging radiotracers and techniques (such as PET measures of tau or inflammation, and high-field MRI methods to detect brain amyloid) will likely emerge. Nevertheless, such efforts to develop effective, reliable and robust biomarkers for AD are critical given the significant number of people who will continue to be diagnosed with AD dementia in the coming decades.
This work was supported by National Institutes of Health (NIH) grants R01AG10897, P41RR023953, and U01AG024904 (to MW), which were administered by the Northern California Institute for Research and Education, and with resources of the Veterans Affairs (VA) Medical Center, San Francisco, California. Additional support came from by NIH grants: P50 AG005133, R37 AG025516, P01 AG025204 (to BK), NIA: P01 AG036694; R01AG027435 (to RS), the Science Foundation Ireland (SFI) investigator neuroimaging program award 08/IN.1/B1846 (to HH), the “Landesoffensive zur Entwicklung wissenschaftlich-ökonomischer Exzellenz” (LOEWE) neuroimaging-neurophysiology research program grant „Neuronale Koordination Forschungsschwerpunkt Frankfurt“ (NeFF), Neuronal Coordination, Neurodegeneration & Alzheimer’s disease project (to HH).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
GE Healthcare holds a license agreement with the University of Pittsburgh based on the PiB-PET imaging technology described in this manuscript. W.E.K. is a co-inventor of PiB and, as such, has a financial interest in this license agreement. GE Healthcare provided no grant support for this study and had no role in the design or interpretation of results or preparation of this manuscript. R.A.S. has served as a consultant for Elan, Janssen, Pfizer, Link and Bristol-Myers Squibb (BMS), and receives research support from Elan, Janssen, and BMS. H.H. has received research support from Novartis and Janssen-Cilag. He has served as a consultant for BMS, Pfizer, Novartis, Janssen-Cilag and Merz Pharmaceuticals. M.W.W. has been on Advisory boards for Lilly, Araclon and Institut Catala de Neurociencies Aplicades, Gulf War Veterans Illnesses Advisory Committee, Veterans Affairs Central Office, Biogen Idec, Elan/Wyeth Alzheimer’s Immunotherapy Program North American Advisory Board, Novartis Misfolded Protein Scientific Advisory Board Meeting, and the Banner Alzheimer’s Institute Alzheimer’s Prevention Initiative Advisory Board Meeting. He has been a consultant for Elan/Wyeth, Novartis, Forest, Ipsen, Daiichi Sankyo, Inc., Astra Zeneca, Araclon, Medivation/Pfizer, TauRx Therapeutics LTD, Bayer Healthcare, Biogen Idec, Exonhit Therapeutics, Servier, and Synarc. He has received research/ travel support from Merck, Avid, Elan/Wyeth, Forest, Innogenetics, NeuroVigil, Inc., Siemens, AstraZeneca, Lilly, Pfizer, and Novartis, and holds stock options with Synarc and Elan.