Amyloid imaging in Alzheimer's disease: A literature review

Abstract

Therapies targeting amyloid-β peptide currently represent approximately 50% of drugs now being developed for Alzheimer's disease. Some, including active and passive anti-Aβ immunotherapy, directly target the amyloid plaques. The new amyloid tracers are increasingly being included in the proposed updated diagnostic criteria, and may allow earlier diagnosis. Those targeting amyloid-β peptide allow identification of amyloid plaques in vivo. We need to gain insight into all aspects of their application. As florbetapir (Amyvid™) and flutemetamol (Vizamyl™) have received marketing authorization, clinicians require deeper knowledge to be rationally used in diagnosis. In this paper, we review both completed and ongoing observational, longitudinal and interventional studies of these tracers, our main objective being to show the performance of the four most commonly used tracers and their validation.

Introduction

The number of patients with a dementia disorder will increase exponentially in the coming years, mainly because of aging of the population. From 25 million in 2000, it has been estimated that their number will reach 63 million in 2030 and 114 million in 2050 (1). Among the causes of neurodegenerative “dementia”, Alzheimer's disease (AD) is the most frequent and accounts for about 80%. In the US, it has been estimated that number of people with AD dementia will be 13.8 million in 2050 with 7 million aged 85 years and older (2).

Several hypotheses relating to molecular mechanisms have been described in the pathophysiological process of AD, but the most commonly accepted is the amyloid hypothesis of Hardy et al. 3, 4. Amyloid deposits form at the cortical level, and correspond to extracellular accumulation of amyloid-β (Aβ) peptide (1-42), formed after cleavage of amyloid protein precursor (APP) by β and γ secretases. These extracellular deposits lead to an inflammatory reaction with microglial activation and neuronal death. The other pathological process is neurofibrillary tangles (NFTs), related to hyperphosphorylation of tau protein, a normal component and stabilizer of microtubules. The last stage of the disease process is neuronal death, leading to cortical atrophy. Knowledge of this pathophysiological process has led to the development of specific markers of amyloid plaque, the main aim being early diagnosis. Pittsburgh compound B (PIB), the first amyloid marker to be developed, was studied in subjects with moderate AD versus controls (three young and six old) by Klunk et al. in 2004 (5), visualizing amyloid plaques in vivo for the first time. New amyloid tracers are currently being studied, usually labeled with fluorine 18 (18F).

AD is diagnosed as part of an exclusion process, and is established according to the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA criteria) of 1984 (6). Diagnosis is based on evaluation of cognitive functions by neuropsychological tests, magnetic resonance imaging (MRI) and laboratory tests. The recent update to the NINCDS-ADRDA criteria of the working group of the Alzheimer's Association and the National Institute on Aging, presented in 2010 at the Alzheimer's Association International Conference on Alzheimer's Disease (AAICAD), define mild cognitive impairment (MCI) as a clinical entity which requires demonstration of cognitive impairment that does not affect the activities of daily living. The prodromal stage of AD, or MCI related to AD, is defined by memory impairment, changes in thought processes or behavior, associated to the presence of AD biomarkers. These biomarkers, in particular those that mark amyloid disease, are henceforward included in the newly proposed diagnostic criteria. They also make it possible to define the “preclinical” stages of the disease by revealing the amyloid process in patients with few clinical signs of cognitive disorder. We review the literature on molecular imaging markers of the amyloid plaque, by bringing together known observational and longitudinal data as well as ongoing interventional and observational studies. The aim is to characterize the precision of these pathophysiological markers for early (or even preclinical) diagnosis of AD, and also for providing supplementary data on change in amyloid load during follow-up.

Methodology of the literature review

We carried out a PubMed search of the literature on amyloid imaging markers from April 2009 to September 2013. We selected cross-sectional and longitudinal studies of the four principal amyloid markers: Pittsburgh compound B (PIB™), flutemetamol (Vizamyl™), florbetapir (Amyvid™) and florbetaben (Neuraceq™). The last part of our literature review deals with interventional studies using amyloid imaging. Our methodology is shown on Figure 1, Figure 5.

Figure 1.

Flowchart of the PubMed search “amyloid imaging AND alzheimer's disease” from April 2009 to September 2013

Figure 5.

Methodology of search for current clinical interventional studies using amyloid biomarkers

The principal radiopharmaceutical markers of amyloid plaques

Introduction to amyloid theory and definition of amyloid status

Despite the progress in understanding the pathogenesis of AD in recent decades, the precise mechanism of AD remains uncertain. Several independent hypotheses have been proposed, but none of them is sufficient to explain the diversity of biochemical and pathological abnormalities of AD. Amyloid metabolic cascade and modifications of tau protein are considered the most important assumptions in AD. The histopathological lesions described consist of aggregation of β-amyloid substance and NFTs related to hyperphosphorylation of tau protein (a stabilizing component of neuron microtubules). Although direct causal link between Aβ accumulation and cognitive disorders is unclear, it is certain that amyloid deposits play a central role in the neuropathology of AD. It is this theory of amyloid cascade that is the basis for the development of amyloid tracers, reason why we will discuss mainly this theory in our literature review.

The theory of the amyloid cascade has been described for the first time by Hardy in 1992 (4). The amyloid peptide originates from a precursor: amyloid beta A4 precursor protein (APP), cleaved by β-secretases and then by γ-secretases. The two forms of amyloid peptide Aβ (1-40) (short form) and Aβ (1-42) (long form) produce the principal neurotoxic effects. These amyloid peptides exist in mono- or oligomeric forms, whose prolonged accumulation (time- and concentration-dependent) finally results in the aggregated (insoluble) form of amyloid-β peptide (Aβ peptide). It is the oligomeric form of Aβ (1-42) that is thought to be the principal cause of neurotoxicity (7). These deposits of amyloid substance, in association with NFTs, may lead to neuronal death.

During AD, amyloid binding occurs in the frontal, temporal, parietal and occipital cortices, and is inversely correlated with brain metabolism on positron emission tomography with 2-[18F]fluoro-2-deoxy-d-glucose (FDG PET) (8), which reflects neuronal dysfunctioning. The most specific regions for analysis of amyloid deposits using florbetapir appear to be the frontal regions and to a lesser extent the posterior association areas, while with brain atrophy on MRI regions of interest are the hippocampal regions and with FDG PET mostly posterior association areas (9).

The markers that we are now studying were developed on the basis of the amyloid theory. The aim is to demonstrate amyloid plaques in the brain regions of interest which reflect the AD disease process: the frontal, temporal, parietal and occipital cortices. The cerebellum and the brain stem regions bind weakly to amyloid markers, and for this reason they are often used as control regions in quantitative analyses. Among the techniques for analysis of amyloid status, calculation of the automated standardized uptake value ratio (SUVr) is the most commonly used in the literature. The signal obtained in the regions of interest is compared with the signal of a control region (cerebellum or pons). To limit noise, the top and bottom 10% of values within each region of interest are excluded from the computation of the mean (10). One of other methods of analysis of amyloid imaging is visual, such as that proposed by Clark in 2011 (11). The aim is to classify the images as “no significant cortical retention” or “presence of significant cortical retention” (Figure 2).

Figure 2.

Visual classification scale of amyloid load using florbetapir. Reproduced with permission of Prof. Payoux, Nuclear Imaging Department, Purpan University Hospital, CHU Toulouse, France

A study with Florbetapir (12) including AD at dementia stage, MCI and healthy controls, compares the SUVr (cut-off value of 1.122, with the cerebellum defined as control region) and the visual method. Automated SUVr performed considerably better (sensitivity 92%, specificity 90%) than visual rating (sensitivity 80%, specificity 38%) in discriminating demented from healthy subjects.

The amyloid cascade theory is the basis of the development of amyloid markers. “Amyloid (+)” status is defined by several methods: automated, such as SUVr (ratio of the signal compared with that of a control region), and visual (Clark's visual scale), which seems to perform less well. In AD, the typical uptake pattern lies in the frontal, temporal, parietal and occipital cortices.

Characteristics of the principal radiopharmaceuticals

The oldest radiopharmaceutical, and the one that has received the most attention in the literature, is PIB. Its main drawback is its short half-life (labeling with carbon-11 (11C)), with a major imperative: the radioactive ligand must be produced in proximity to the imaging device. The three other radiopharmaceuticals (florbetapir, flutemetamol and florbetaben) are labeled with fluorine-18 (18F), which has a longer half-life than 11C. Florbetapir and florbetaben bind to the aggregated form of amyloid peptide, whereas PIB mainly binds to an oligomeric form of Αβ peptide carrying an N3 terminal segment (13). Flutemetamol, a PIB derivative, also binds to a soluble form of Αβ peptide (1-42). Table 1 shows the characteristics of the four principal radiopharmaceuticals studied: half-life, SUVr, dose injected and acquisition time. The inhibition (Ki) and binding (Kd) affinities are the means of those found in the literature, as are the examples of SuVr values.

Table 1.

Characteristics of the four principal radiopharmaceuticals

Radiopharmaceutical	PIB	Florbetapir	Flutemetamol	Florbetaben
C or F labeling	11C	18F	18F	18F
Half-life (min)	20	110	110	110
Binding	Aβ (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40) peptide (fibrillar oligomer)	Aβ peptide (aggregated form)	Aβ peptide (soluble form 1-42)	Aβ peptide (aggregated form)
Chemical group	Benzothiazole	Styrylpyridine	Benzothiazole	Stilbene
Dose injected (MBq)	240 – 450	300 – 380	185	300
Affinity (Ki, nM)	0,9	2,2	0,7	2.4
Affinity (Kd, nM)	1.4 – 2.4	3.7 ± 0.3	2.4	6.7
Examples of SUVr
AD	2.5	1.7	2.2	1.6
C	1.5	1.3	1.4	1.3
Acquisition time (min)	40 – 90	50 – 70	80 – 100	70 – 130
References	(42, 105, 106, 107, 108, 109)	(11, 108, 110, 111, 112, 113, 114, 115)	22, 2, 108	21, 2, 116, 117

PIB: Pittsburgh compound B, 11C: labeling with carbon 11, 18F: labeling with fluorine 18, SUVr: standard uptake value ratio, AD: Alzheimer's disease, C: controls, Ki: inhibition affinity, Kd: binding affinity, units: minutes (min), megabecquerels (MBq), nanomoles (nM).

Sensitivity and specificity of the different markers for the diagnosis of positive AD

The sensitivity and specificity of the four principal amyloid markers are summarized in Table 2, and compared with those of Αβ measurement in cerebrospinal fluid (CSF). The results show some variation between studies, in particular according to the number of subjects included. Rowe et al. (14) performed PIB scans in MCI, demented subjects and healthy controls, in order to differentiate healthy subjects from AD demented subjects. These authors found 98% sensitivity and 66% specificity (78% and 80%, respectively, for decrease of hippocampal volume on MRI). In another study (15), PIB PET scans distinguished 75% of MCI patients from healthy controls (54% with FDG PET), whereas precision reached 90% when FDG PET and PIB PET were used in combination. This suggests the value of using an amyloid marker in synergy with a marker of neuronal death. However, positive PIB PET with clinical criteria for probable Alzheimer disease can be secondary to cerebral amyloid angiopathy like demonstrated in an autopsy case report of 83 year old men (16). In this case, in spite of clinical criteria and typical binding with PIB, the autopsy revealed no amyloid plaques or neurofibrillary tangles and only cerebral amyloid angioapathy. This could mean that amyloid PET cannot distinguish lesions of amyloid angiopathie and cortical amyloid due to AD.

Table 2.

Examples of sensitivity and specificity of amyloid biomarkers for AD diagnosis

Biomarker	Sensitivity (%)	Specificity (%)	Sample (n) AD/MCI/controls	References
PIB	98	66	53/57/177	Rowe et al. (14)
Florbetapir	95	95	19/0/21	Newberg et al. (18)
Florbetaben	80	91	81/0/69	Barthel et al. (116)
Flutemetamol	93	93	27/20/25	Vandenberghe et al. (22)
CSF/Aβ (1-42)	86	90	600/0/450	Blennow et al. (24)

AD: Alzheimer's disease, MCI: mild cognitive impairment, PIB: Pittsburgh compound B, CSF/Aβ: measurement of Aβ in cerebrospinal fluid.

In a study (17) with demented subjects and young healthy subjects, 100% of both patients and healthy subjects were correctly classified by florbetapir. In another study (18) with AD demented subjects and healthy elderly controls, the sensitivity and specificity of florbetapir were 95% for diagnosis of AD demented subjects (compared with 89% and 86%, respectively, for FDG PET). Analysis of the findings of the Alzheimer's Disease Neuroimaging Initiative (ADNI) (19) in 32 subjects, who had undergone at least one imaging study with PIB and one with florbetapir, found a strong correlation between the results obtained with the two radioligands (p = 0.86-0.95), which appears to confirm that their performances are equivalent.

Sensitivity of florbetaben was 80% for a specificity of 91% (20) and respectively 100% and 90% (21) to distinguish AD patients and elderly controls.

Vandenberge et al. (22) carried out flutemetamol scans in demented subjects, MCI, healthy elderly and healthy young subjects, and found a sensitivity of 93% and a specificity of 93% for diagnosis of AD. The same authors found 100% agreement between the results of flutemetamol and those of PIB.

In comparison, the data of the literature on FDG PET generally find equivalent sensitivity and specificity: In a meta-analysis by Bloudek et al. (23), sensitivity of FDG PET was 90% and specificity 89% for diagnosis of AD compared with healthy controls. With regard to the sensitivity and specificity scores of cerebrospinal fluid biomarkers, they were variable depending on whether total tau protein, phosphorylated tau protein or Aβ (1-42) levels were used. For example, a meta-analysis by Blennow and Hampel [24] found that decreased Aβ (1-42) levels had a sensitivity of 86% and a specificity of 90%.

Overall, the performances of the various amyloid markers are similar for AD diagnosis compared with control subjects, with satisfactory sensitivity and specificity at around 90%. They are also similar to those of other markers (carbohydrate metabolism using FDG PET and amyloid biomarkers in CSF), even if data are variable across studies.

Autopsy validation

The in vivo results for 3 of the principal radiopharmaceuticals have been validated by autopsy. One study (11) compared the results of florbetapir amyloid imaging, performed a mean of 99 days before death, and post-mortem histological findings. In half of the autopsies performed, autopsy criteria of AD were found. In 96% of cases, findings of florbetapir amyloid imaging and autopsy were correlated (consistent or not consistent with a diagnosis of AD). In the same study, florbetapir imaging in healthy young controls yielded only negative results. Autopsy validation thus showed 100% specificity (amyloid (–), autopsy (–)) and 93% sensitivity (amyloid (+), autopsy (+)). Florbetapir imaging followed by post-mortem examination in demented subjects (25) confirmed a strong correlation between amyloid binding in vivo and the majority of plaques found on autopsy, but florbetapir did not bind to neurofibrillary tangles.

For flutemetamol, the literature shows full agreement of in vivo amyloid imaging with the histological findings of frontal cerebral cortex biopsy in AD demented patients (26).

With PIB scan, correlation between in vivo and autopsy findings is good for patients who had “elevated or negligible”, but less good for those who had “intermediate” amyloid binding (27). This direct correlation with Aβ plaque load at autopsy doesn't exist with neurofibrillary pathology (28). Autopsy validation found a closer correspondence for certain regions after histopathological analysis: anterior and posterior cingulate gyrus and precuneus (29).

Cross-sectional studies

Amyloid load according to cognitive status

We present the data on amyloid (+) or amyloid (–) status according to cognitive status, whatever the amyloid label used. Table 3 summarizes the percentage of patients considered amyloid (+) and their cognitive status.

Table 3.

Percentage of amyloid (+) patients in three population subgroups in function of cognitive status, % (n+/n total)

Controls	MCI	AD	References	Biomarker
% (n+/n total)	% (n+/n total)	% (n+/n total)
23 (11/48)	NA	NA	Perrotin et al. [31]	PIB
18 (17/91)	48 (28/58)	100 (56/56)	Hatashita et al. [42]	PIB
33 (58/177)	NA	NA	Pike et al. [50]	PIB
31 (33/106)	69 (45/65)	97 (34/35)	Villemagne et al. [70]	PIB
NA	55 (17/31)	NA	Okello et al. [85]	PIB
29 (6/21)	59 (19/31)	88 (7/8)	Jack et al. [41]	PIB
33 (32/95)	60 (20/33)	100 (36/36)	Bourgeat et al. [10]	PIB
7 (1/15)	50 (10/20)	93 (25/27)	Duara et al. [118]	Flutemetamol

MCI: mild cognitive impairment, AD: Alzheimer's disease, NA: Not applicable

Normal subjects and MCI

Amyloid plaques are not specific to subjects with AD at the dementia stage (using the NINCDS-ADRDA criteria). According to the literature, 20 to 50% of healthy elderly subjects have amyloid deposits 30, 31. These findings are supported by numerous studies that revealed similar proportions of healthy elderly subjects who were amyloid (+): overall, a previous review of the literature (32) found that 20 to 30% of cognitively normal subjects were amyloid (+) 33, 2, 35, 36, in agreement with autopsy findings (37).

According to some authors, there is a dichotomy in healthy subjects between those with high amyloid binding and those who do not bind amyloid. This finding was confirmed by Rowe et al. (14) who compared PIB findings in healthy, MCI and demented subjects (definition of amyloid (+) status by SUVr, cut-off value 1.5 compared with the neocortical standard value). In addition to the greater proportion of Apolipoprotein E4 (APOE4) carriers in demented patients, in healthy subjects a non-linear distribution of binding was found, with a separation between weak and strong binders. This distribution in two groups suggests heterogeneity among “healthy” subjects, where some may present the first neuropathological manifestations of AD in advance of the clinical signs. These data are similar to those reported by Jack et al. (38) where changes in markers, and particularly in amyloid deposits, preceded the onset of cognitive impairment in the disease cascade of AD.

In subjects with MCI, Jack et al. (38) observed that the proportion who were amyloid (+) was intermediate to that of cognitively normal individuals and demented subjects, and was about 50% to 60% 33, 2, 40.

AD demented subjects

The proportion of amyloid (+) demented subjects, clinically due to AD, ranges from 88% (41) to 100% (42). In the latter study, 74% of amyloid (+) subjects were classified as AD at dementia stage according to the Clinical Dementia Rating scale (CDR) criteria.

Amyloid accumulation follows a sigmoid curve, as described by Jack et al. (38), and precedes the onset of clinical signs. Thus, amyloid binding is not specific to AD at a clinical dementia stage according to the NIA-AA criteria, and about 20% of healthy subjects are amyloid (+). Those could be preclinical stage of AD. About 50% of MCI subjects are amyloid (+), whereas over 90% of AD demented subjects are amyloid (+).

Other factors affecting amyloid load

Cerebral atrophy, cognitive performances and amyloid status

In an analysis carried out by the Australian Imaging Biomarkers & Lifestyle Flagship Study of Ageing (AIBL), Ellis et al. (43) found that hippocampal atrophy was linked with amyloid binding only in the lower temporal lobe region, which suggests that Aβ accumulation may interrupt the connections of this region with the hippocampus. This is confirmed in a cross-sectional study (44) that found a significant association between cortical atrophy in the precuneus and hippocampus and memory impairment in healthy amyloid (+) subjects but not in amyloid (-). In those asymptomatic amyloid (+) subjects, Aβ deposition is associated with cerebral atrophy and memory impairment (44).

It is in the early or even preclinical stages that the association between global cognitive performance and amyloid status is the strongest (45). In a study (46) that included 57 non-demented participants of whom 6 had MCI, increased PIB retention was associated with poorer performances in verbal memory but not in visual memory. In women, amyloid burden was linked to episodic memory and visuospatial performances, but not to other cognitive components. In these preclinical stages of AD, a relation has been found between episodic memory deficits and neocortical temporal PIB retention, independently of hippocampal atrophy (47). This was confirmed in an analysis by Mormino et al. (48) of data from ADNI, wich found in healthy elderly subjects an inverse relationship between cognitive loss in episodic memory and hippocampal atrophy, as well as between amyloid deposition and hippocampal volume. Similar results were found in a meta-analysis (49) in healthy elderly subjects, confirming the possible relationship between episodic memory loss and amyloid burden with PIB, although association with other cognitive domains (executive function, working memory, processing speed) is not found here. Moreover, the elderly amyloid (+) subjects, who were cognitively healthy, had less confidence in their memory than other subjects of the same age, and their episodic memory performance was less good 31, 50. Examination of amyloid plaque deposition, in cognitively healthy subjects, showed that elevated amyloid was associated with poorer cognitive performance in processing speed, working memory and reasoning in the subgroup aged 60 and over (51) and with premorbid intellect (52). In a community duelling non demented older subject, amyloid uptake and premorbid intellect are strongly correlated. Furthermore, the association between cognition and amyloid uptake is stronger when controlling for premorbid intellect. This suggests that cognitive reserve could partly compensate the symptoms of amyloid deposition in non demented elderly (52).

By contrast, another analysis of ADNI data (14) found no association between amyloid retention on PIB scans and cognitive status, hippocampal volume or age in AD at dementia stage. Amyloid load therefore does not appear to be correlated with the severity of cognitive impairment in demented patients, as shown in a study (18) with florbetapir carried out in demented patients and healthy aged controls. Amyloid binding and Mini Mental State Examination score (MMSE) were correlated only in the overall group of subjects, but not in the subgroup of demented patients. Conversely, hypometabolism on FDG PET (reflecting neuronal dysfunctioning) was significantly associated with MMSE score in the global study population as well as in the subgroup of demented patients alone.

These findings appear to confirm that the presence of amyloid plaques in healthy subjects may reflect a preclinical stage of AD. Amyloid deposits are more correlated with memory performance in early stage of cognitive decline. These elderly cognitively healthy patients with amyloid plaques have poorer episodic memory performances than amyloid (–) patients of the same age. On the other hand, in subjects with a diagnosis of AD at dementia stage, cognitive performances are more correlated with brain hypometabolism which reflects neuronal dysfunctioning than amyloid binding.

Amyloid load according to other clinical factors

Amyloid deposition is influenced by factors other than cognitive status, such as age, family history or presence of the APOE4 allele. These data are summarized in Table 4, where only p significance values are given since odds ratios are found in only a minority of studies.

Table 4.

Factors influencing imaging of amyloid load

Biomarker	Sensitivity (%)	Specificity (%)	Sample (n) AD/MCI/controls	References
	Cognitive status	Significance	References
Hippocampal atrophy	Healthy		p=0.0016	Mormino et al. (48)
	Healthy		p=0.018	Bourgeat et al. (10)
	MCI		NS	Bourgeat et al. (10)
	AD		NS	Rowe et al. (14)
Dementia syndrome	AD (moderate)	p<0.01	Hatashita et al. (42)
Disease severity	AD		NS	Rowe et al. (14)
Decreased episodic memory	MCI		p<0.001	Pike et al. (34)
	Healthy		p<0.034
	AD		NS
APOE4	MCI		p<0.0001	Rowe et al. (14)
	Healthy		p=0.0002
Family history of AD	Healthy		p<0.05	Mosconi et al. (57)
Amyloid angiopathy	AD		p=0.002	Ly et al. (59)
Advanced age	Healthy		p<0.0001	Rowe et al. (14)

AD: Alzheimer's disease, MCI: mild cognitive impairment, NS: non-significant

The main factor that influences amyloid deposition is APOE4 status. The majority of asymptomatic APOE4 carriers (53) had increased PIB retention compared with controls. APOE4 allele (+) and amyloid (+) status were found to be associated in a population of clinically healthy elderly subjects: those classified amyloid (+) (33%) were twice as likely to carry the APOE4 allele in the Australian Imaging Biomarkers and Lifestyle study (AIBL) of healthy elderly patients (50). In cognitively healthy subjects aged 30 to 89 years, amyloid deposition was more elevated and more patients were APOE4+ (38%) in the subgroup aged over 60 years (51). Another AIBL study (14) participating in ADNI investigated amyloid binding in healthy controls, MCI subjects and patients with mild AD. High binding was present in 33% of healthy subjects, and in 49% of healthy APOE4+ carriers. The association of the APOE4 allele and of amyloid load is also quantitative, as found in a study which included healthy subjects with a family history of AD and a variable number of APOE4 alleles: amyloid load was statistically correlated with the number of duplications of the APOE4 gene (53). These findings all confirm that, in healthy subjects, amyloid deposition increases with the presence of the APOE4 allele, independently of cognitive performances. The association of the APOE4 allele and amyloid deposition exists in cognitively healthy subjects, but also in demented subjects as shown by a study (54) in AD demented, MCI, healthy elderly subjects and healthy young controls. This found that APOE4 carriers had higher mean deposition and were more often amyloid (+) in all three subgroups. A study with ADNI-2 data confirms the association of APOE4 statut and amyloid deposits in preclinical stages of AD (amyloid (+) MCI and cognitively healthy subjects) (55). APOE genotype has a differential effect on the distribution of amyloid plaques: it seems that amyloid binding is higher in frontal cortex of APOE4 noncarriers compared with APOE4 carriers (56).

With regard to family history, a PIB PET study (14 subjects with negative family history, 14 with maternal family history, and 14 with a paternal history of AD) analysed the risk of amyloid deposition in relation to family history of late-onset AD (57). Healthy children with a parent with late-onset AD (particularly those with an affected mother) had increased amyloid load compared with controls. The association of family history for AD and amyloid binding varied according to age (58): before the age of 55 years, the APOE4 allele was associated with PIB binding but family history was not, whereas after 55 years, both family history and APOE4 were associated with PIB binding.

A diagnosis of amyloid angiopathy (AA) influences amyloid load, as found in a study (59) in which demented subjects with AA had moderately increased amyloid binding compared with healthy subjects, with a slightly different pattern from AD (demented subjects) (lower in the cortical and frontal regions and higher in the occipital regions).

Lastly, amyloid deposition in non demented patients (MCI and healthy subjects) increases with aging 47, 60. For example, the AIBL study (14) showed that amyloid deposition increased with age, in particular in APOE4+ subjects: 18% of healthy subjects were amyloid (+) between 60 and 69 years, compared with 65% over the age of 80. The association of amyloid deposition and age seems to exist mostly in non demented subjects (54). This finding was confirmed in another study (61) of cognitively healthy elderly subjects, where age and APOE4 status were found to be independent risk factors for elevated amyloid accumulation.

To a large extent, accumulation of amyloid substance precedes the onset of clinical signs of AD. A substantial proportion of healthy subjects are considered amyloid (+), and this may correspond to preclinical stages of AD. These Amyloid positive healthy subjects have poorer episodic memory performances, correlated with hippocampal atrophy. Several factors influence amyloid deposition in non-demented subjects: age, family history (maternal in particular), and the presence of the APOE4 allele, which may be risk factors for future degenerative disease. In demented subjects, cognitive performances are more correlated with brain hypometabolism which reflects neuronal dysfunctioning than amyloid binding.

Longitudinal data

The second part of our literature review deals with longitudinal data on amyloid plaque markers, changes in amyloid binding during follow-up, and conversion rates of subjects according to their amyloid status.

Change in amyloid load and cognitive status of healthy subjects

The data of the literature show that 20 to 30% of healthy elderly subjects have increased amyloid deposition. In these subjects amyloid load progresses during follow-up, as shown in a study (62) that followed non-demented elderly subjects for 1.5 years and found that amyloid retention increased by 0.9% yearly. The increase was greater in subjects who had high amyloid deposition at baseline, mainly in the prefrontal, parietal, lateral temporal, occipital, and anterior and posterior cingulate cortices. Moreover, amyloid deposition in healthy subjects appears to be irreversible. A 2.5-year follow-up of clinically healthy patients (63) observed initially elevated amyloid deposition in 14% of subjects, and most of them had plaque growth during follow-up (about 8% per year); none reverted to a state of no Aβ deposits.

Unlike non-healthy subjects, accelerated cortical atrophy (64) and cognitive decline in healthy subjects are related to amyloid load in longitudinal studies. This finding was underlined in a study (65) that included healthy elderly subjects, where 21% had elevated amyloid levels on PIB scans (sometimes comparable to AD demented subjects, classified PIB+), and smaller volumes in the hippocampus, temporal neocortex, anterior cingulate and posterior cingulate regions. However, cognitive decline (in episodic and working memory and visuospatial abilities) during follow-up was related to amyloid levels and decreased hippocampal volume. A study of non-demented subjects (66) confirmed that higher baseline amyloid binding was associated with decline in most cognitive functions (language, attention, visuospatial abilities and executive functions) and that progression of amyloid binding was associated with memory impairment during the 2-year follow-up. In analysis of AIBL data (67), during an 18-month follow-up of cognitively healthy elderly subjects, elevated amyloid load was found to be associated with decrease in working and episodic memory. Lastly, amyloid (+) status and future cognitive decline were also found to be associated in the oldest-old, as in a study including subjects aged 90 years and over, non-demented at inclusion, in whom an 18-month follow-up showed decline mainly in global cognitive measures (68).

Moreover, a longitudinal data study with cognitively normal adults showed that abnormal levels of amyloid biomarkers (CSF or PET) were associated with faster time to cognitive impairment (69), without difference in predictive value between each biomarkers.

In healthy elderly amyloid (+) subjects, amyloid load progresses irreversibly during follow-up, especially in subjects with initial high deposition in the prefrontal, parietal, lateral temporal and occipital regions. The proportion of cognitively healthy amyloid (+) subjects is 20 to 30%, depending on the studies, and presence of the APOE4 allele is a risk factor for amyloid (+) status. Cognitive decline in subjects initially considered healthy is related to amyloid load and to decreased hippocampal volume.

Progression of amyloid load in “non-healthy” subjects

In longitudinal studies, amyloid load in MCI and demented subjects increases during follow-up. A study (70) which included demented, MCI and healthy subjects, with a 20-month follow-up, found that amyloid load increased in all three populations (5.7% in AD, 2.5% in MCI and 1.5% in healthy controls amyloid (+) at baseline). Increase was, in this study, associated with the number of APOE4 alleles. The same progression rate of 2.5% in the MCI population was found in a 2-year follow-up study (71). Although amyloid load increase is confirmed in most of studies, some do not confirm the difference of rate increase. As shown in a longitudinal study (72) over 2 years, which included AD demented subjects and controls, change in amyloid uptake did not (or only little) increase during follow up in AD demented subjects. However, brain volume change in the hippocampal region, temporal cortex and precuneus was significantly greater in AD demented subjects, as was decline in cognitive functions (MMSE and neuropsychological tests). Increased amyloid binding in longitudinal studies in MCI patients occurs mainly in the lateral temporal lobe, as shown in a study with a follow-up of 2.5 years in MCI, demented subjects and healthy controls (73). The same study confirmed that increase in amyloid load was not greater in demented patients.

Change in amyloid load is therefore weakly associated with progression of cognitive disorders in “non-healthy” patients. On the other hand, hippocampal atrophy, reflecting neuronal death, is indeed associated with AD progression. This was found in the study (41) of Jack et al., that included cognitively normal, MCI and demented subjects, with a one-year follow-up. It reveals no difference in change in amyloid load between the three groups and weak correlation of change on PIB scans with MMSE score but not with CDR-SB score, whereas both scores were correlated with ventricular expansion on MRI. These authors confirmed that amyloid load is not necessarily correlated with progression of cognitive decline in “non-healthy” patients; in particular, the rate of amyloid deposition is constant during the course of the disease. Different analyses of the ADNI cohort have examined change in amyloid load in “non-healthy” subjects, like Apostolova et al. (74) who found a significant association between precuneal amyloid binding and hippocampal volume.

With regard to conversion of MCI to AD at dementia stage, Jack et al. (75) found a linear relationship with hippocampal atrophy, whereas amyloid load reached a ceiling earlier in the progression of cognitive impairment. This confirms the hypothesis that amyloid deposits appear in the first stages of the disease, are weakly correlated with stages of severity at dementia stage, and later become stable, whereas the neurodegenerative process accelerates as the disease progresses, together with cognitive decline. Two analyses by Caroli et al. (76) and Beckett et al. (77) confirmed this hypothesis by showing that measures associated with early stages of the disease (Aβ 1-42) underwent more significant change in MCI than demented patients, whereas measures associated with later changes (FDG PET) were more marked in demented subjects.

Jack et al. (38) described serial change in the different markers (Aβ, tau, cerebral atrophy) during AD, with progression which followed a sigmoid curve. Amyloid deposits exist prior to clinical diagnosis of AD, and therefore prior to memory impairment. When the dementia stage is reached, progression of amyloid load is low or nil, and it is progression of NFTs and atrophy that mark the stages of the disease. The chronological sequence of change in AD markers was investigated in a follow-up study (78) of patients with autosomal dominant AD: Aβ (1-42) in CSF showed a decrease 25 years before symptom onset, PIB imaging revealed amyloid deposits 15 years before symptom onset (as well as cerebral atrophy and increased CSF tau levels), and brain hypometabolism and impaired episodic memory 10 years beforehand. These data are similar to those found in a prospective cohort study with 200 participants followed up for 18 months (79). Authors estimate that it will take 19.2 years for the amyloid levels in healthy subjects reach those of AD. Subsequently amyloid deposition reached a plateau and authors suggest a prolonged preclinical stage of AD, and Aβ deposition plateaued positivity after 17 years, while hippocampal atrophy and cognitive impairment appear respectively 4.2 and 3.3 years before the onset of dementia stage.

Some changes were made in the model adjusted by Jack in 2013 (80): first, interindividual variations in the input of pathophysiological process, and different growth depending on whether dealing with a subject at high risk or low risk of cognitive impairment. This risk varies depending mainly lifestyle, genetic profile, cognitive reserve, and comorbidities including cerebro-vascular comorbidities. Moreover, some changes have been made regarding the pathophysiological sequence in AD, such as changes initially independent of the NFTs and the amyloid process.

Figure 3 is adapted from Jack CR et al. (38). We show here only the progression of amyloid deposits, memory impairment and loss of independence, which are the theme of this paper. Percentages indicate amyloid (+) subjects in each subgroup.

Figure 3.

Adapted from Jack CR et al. (38). Dynamic change during the AD cascade

In “non-healthy” subjects there is a constant progression of amyloid load during follow-up, which is greater in presence of the APOE4 allele. Progression is not more marked in demented subjects, in whom amyloid load even becomes stable. At the dementia stage, progression of cognitive disorder is morely strongly related to NFTs and neuronal death (shown by atrophy on morphological examination) than to amyloid deposition.

Conversion rate according to amyloid status

In longitudinal studies, the percentages of high amyloid deposition are the same as those found in the preceding observational studies: 25% of healthy subjects, 50% of MCI subjects.

The conversion rate of clinically normal to AD demented subjects is higher when they are amyloid (+). The relative risk of conversion from cognitive normality to AD dementia increased five-fold when baseline amyloid binding was high (81). Of cognitively normal amyloid (+) subjects, 25% developed MCI or AD at dementia stage at 3 years (70). The same authors also showed that amyloid binding increased in amyloid (+) cognitively normal controls during follow-up, resulting in progression of amyloid accumulation in all cases in these healthy patients with high retention at baseline. Amyloid load thus predicts progression from cognitive normality to dementia, as found in a study (81) which included elderly healthy patients (CDR 0): 15% progressed to CDR 0.5 (MCI), and 40% of these were diagnosed with dementia. Age and mean cortical amyloid retention on amyloid PET scans were predictive of progression to CDR 0.5. Subjects who had converted to dementia stage showed decline in three cognitive domains (episodic memory, semantic memory and visuospatial performance) and volume loss in the parahippocampal gyrus compared with subjects who remained CDR 0.

The conversion rate of MCI to dementia stage was also markedly higher in amyloid (+) MCI patients. Moreover, no amyloid (–) MCI subject converted to dementia stage (82), suggesting that high amyloid retention may predict risk of conversion to dementia in this MCI population, and that amyloid (+) MCI and demented subjects represent different stages of severity of the same disease. In an analysis of the ADNI cohort, Petersen et al. (83) examined data at inclusion and after 12 months in healthy controls, MCI and moderate AD subjects: the conversion rate of MCI to dementia was 16.5%.

Table 5 shows the results of conversion of healthy or MCI subjects to dementia stage according to their amyloid status in five studies. Globally and according to duration of follow-up (1 to 3 years), 30% to 80% of MCI who were amyloid (+) at baseline converted to dementia, compared with 0% to 12.5% of amyloid (–) MCI patients.

Table 5.

Conversion to AD of MCI subjects and controls according to their amyloid status on PIB scans in five longitudinal studies

Villemagne et al.(70)	Ossenkoppele et al. (71) Okello et al. (85)	Wolk et al. (82)	Grimmer et al. (104)	References
Sample size (n)
35	NA	NA	NA	NA	AD
65	29	31	26 (1)	28	MCI
106	NA	NA	NA	NA	Controls
20	24	12 to 36	24	24	Duration of follow-up (months) Amyloid+ status (%)
97	NA	NA	NA	NA	AD
69	72	55	54	61	MCI
31	NA	NA	NA	NA	Controls Conversion (%)
67	71	82	31	53	MCI+ to AD
5	12.5	70	0	MCI– to AD
16	NA	NA	NA	NA	C+ to MCI or AD

AD: Alzheimer's disease, MCI: mild cognitive impairment, C: controls, (1): including 13 amnestic MCI, 6 multi-domain MCI, 7 non-amnestic MCI, NA: not applicable.

High amyloid deposition is therefore an independent risk factor for conversion to dementia of healthy or MCI subjects 70, 81. Some brain regions appear to be more involved in the amyloid process responsible for conversion: the posterior cingulate, lateral frontal and temporal cortices, putamen and caudate nucleus, as found in a two-year follow-up (84) of amnestic MCI subjects. In this study, hippocampal atrophy at baseline was greater in patients who converted to dementia, but it increased significantly in both groups. On the other hand, in nonconverters there was an increase in amyloid binding in the anterior and posterior cingulate cortices, putamen and temporal and parietal cortices.

Figure 4 shows conversion of amyloid (+) MCI to AD dementia in longitudinal studies. At baseline, between 50% and 70% of MCI subjects were amyloid (+). Their number decreased during follow-up when the amyloid (+) MCI subjects converted to dementia. Such a representation shows the ineluctable course of amyloid (+) MCI subjects to dementia, confirming the hypothesis that amyloid (+) status is a marker of AD at a pre-dementia stage. The horizontal axis shows follow-up at 6-month intervals. The vertical axis shows number (n) of amyloid (+) MCI patients. Depending on the studies, the conversion rate of amyloid (+) MCI subjects to dementia is variable, according to number of subjects included, psychometric tests used and duration of follow-up. If we consider that progression of the conversion rate is linear over time, we find a conversion rate ranging from 20% per year (82) to 46% per year (84) in amyloid (+) MCI subjects.

Figure 4.

Conversion of amyloid (+) MCI subjects during follow-up This figure shows conversion of amyloid (+) MCI to AD dementia in longitudinal studies. The horizontal axis shows follow-up at 6-month intervals. The vertical axis shows number (n) of amyloid (+) MCI patients. At baseline, between 50% and 70% of MCI subjects were amyloid (+). Their number decreased during follow-up when the amyloid (+) MCI subjects converted to dementia. Such a representation shows the ineluctable course of amyloid (+) MCI subjects to dementia, confirming the hypothesis that amyloid (+) status is a marker of AD at a pre-dementia stage

It thus appears that the majority of amyloid deposits are formed during the preclinical stages. Amyloid binding on PET appears to be a risk factor for conversion to a clinical diagnosis of AD in both healthy and MCI subjects. The rapidity of conversion of amyloid (+) subjects to dementia stage was greater in APOE4 carriers, as found in numerous studies. For example, in a longitudinal study (85) of MCI patients, followed between 1 and 3 years, at baseline 55% of MCI patients were amyloid (+). Of these subjects, 82% converted to clinical AD at dementia stage after the follow-up period, whereas only 7% of amyloid (–) MCI patients converted to dementia. Among converters, half were described as rapid converters (less than one year), and these had high amyloid retention levels at baseline in the anterior cingulate and frontal cortex. The APOE4 (+) genotype was also associated with rapidity of conversion to dementia in this study. Other factors were associated with rapidity of conversion of MCI to dementia: in a longitudinal study over 2 years (75) using ADNI database in MCI patients, those who were amyloid (+) (in CSF or on PIB scans) were more likely to progress to dementia (50% vs 19%). Among these amyloid (+) MCI patients, hippocampal atrophy was predictive of conversion time, but Aβ (1-42) load was not. On the other hand, if the data of all MCI patients (amyloid (+) and amyloid (–)) were analysed, rapidity of conversion was linked with hippocampal atrophy and with amyloid load.

Amyloid load assessed on PET scan may be a marker of risk of progression of healthy and MCI subjects to dementia. Factors such as presence of the APOE4 allele and hippocampal atrophy are predictive of rapidity of conversion of MCI to dementia, and certain regions of amyloid deposition (prefrontal, posterior cingular, lateral temporal cortices) appear to be more predictive of risk of progression toward dementia. When the stage of dementia is reached, a “ceiling effect” of Aβ deposition appears, and progression of cognitive disturbance (and therefore disease severity) is more related to cerebral atrophy (hippocampal cortex, ventricular expansion) and neuronal loss (hypometabolism on FDG PET) than amyloid process.

Interventional studies

Amyloid plaque markers are of considerable interest in interventional studies of AD. They make it possible to select study populations according to a pathophysiological marker. By labelling of amyloid plaques in MCI or dementia subjects, AD can be selected whatever the stage. This is illustrated in the analysis of the distribution of amyloid status according to APOE4 in the phase III trials of bapineuzumab presented to the Clinical Trials in Alzheimer's Disease (86). Nearly 36.1% of APOE4 (–) subjects were considered amyloid (–). They could present a non-AD cognitive disorder, which supports the need to use a pathophysiological marker of AD to identify subjects in therapeutic trial cohorts. These markers also allow validation of the mechanisms of action of AD treatments targeting the amyloid plaque 87, 88, as in immunotherapy (proof of concept studies) in phase I, II and III clinical studies. In the study with bapineuzumab (86), pooled analysis showed a significant decrease in the outcome measure “change in amyloid load on PIB scan” in the global study population and in mild stages of AD at the end of the duration of the trial.

In our PubMed search of the literature, we found only two published interventional studies that used PIB as amyloid marker, and their findings are summarized in Table 6. In both studies, change in amyloid load on PIB scan was the main outcome measure, and results showed a significant decrease in amyloid load compared with placebo. Other ongoing interventional studies were sought among the clinical trials and are presented in the next section. We excluded studies that used measurement Aβ (1-42) in CSF.

Table 6.

Published interventional studies using an amyloid marker with PET imaging as main outcome measure

	Bapineuzumab	Gantenerumab
Mode of action	Humanized monoclonal antibody anti-N terminal segment (Aβ 1-5)	Human monoclonal antibody anti-N terminal and central segment (Aβ 40-42)
Tracer	PIB	PIB
Study population	20 bapineuzumab, 8 controls (19/7 in ITT)	12 gantenerumab (1), 4 controls
Stage of severity	Mild to moderate AD	Mild to moderate AD
Duration of follow-up	78 weeks	2 to 7 infusions every 4 weeks, max. 28 weeks
Efficacy in reducing amyloid	dose 0.5 mg/kg -24%	dose 60 mg -15.6% (CI -42.7/+11.6)
progression (2)	1 mg/kg -18%	200 mg -35.7% (CI -63.5/-7.9)
	2 mg/kg -29%
References	Rinne et al. (119)	Ostrowitzki et al. (120)

PIB: Pittsburgh compound B, ITT: intention-to-treat analysis, AD: Alzheimer's disease, CI: 95% confidence interval, (1) 6 = dose 60 mg, 6 = dose 200 mg, (2) vs placebo

Ongoing studies

We carried out a search for ongoing clinical trials as of September 2013 using the methodology shown in Figure 5. We excluded observational studies, and those using amyloid markers other than PET, such as serum or CSF. The majority are phase II studies, still recruiting or undergoing analysis, and generally using PIB or florbetapir PET.

Interventional studies

Selected interventional studies were sought using “clinical trials” database as of September 2013. Table 7 summarizes the characteristics of current interventional studies using amyloid labeling. In some of these studies, change in amyloid load is the principal outcome measure: these are marked with an asterisk (*). In the other studies change in amyloid binding is a secondary outcome measure, while the principal outcome measure is change in cognitive function (MMSE, Alzheimer's Disease Assessment Scale (ADAS)) or in independence (Alzheimer's Disease Cooperative Study -

Table 7.

Interventional clinical trials using amyloid PET

Amyloid biomarker	Intervention	Type of intervention	Phase	Subjects included	Subjects (n)	Cognitive inclusion criteria	Follow-up duration (months)	References
Amyloid PET*	ACC-001	Active immunization (fragment Aβ 1-6)	II	Pre-dementia AD	108	CDR 0.5 MMSE 25 – 30	24	NCT01227564
Amyloid PET*	ACC-001	Active immunization active (fragment Aβ 1-6)	II	Mild to moderate AD	108	MMSE 18 – 26	24	NCT01284387
Amyloid PET*	bapineuzumab	Humanized monoclonal antibody anti-segment N terminal of Aβ	II	Mild to moderate AD	120	MMSE 18 – 26	24	NCT01254773
Amyloid PET* Aβ (1-42) in CSF	ponezumab	Humanized monoclonal Ab anti-fragment C terminal of amyloid peptide 1-40 (IgG2deltaA)	II	Mild to moderate AD	36	DSM-IV NINCDS-ADRDA MMSE 18 – 26	18	NCT00945672
Amyloid PET* Aβ (1-42) in CSF	MABT5102A	Passive immunotherapy, humanized anti-Aβ monoclonal antibody NINCDS-ADRDA	II	Mild to moderate AD	72	MMSE 18 – 26	18	NCT01397578
Amyloid PET Aβ (1-42) in CSF	solanezumab	Anti-Aβ antibody	III	AD	1275	NINCDS-ADRDA	24	NCT01127633
PIB PET	IV Ig	Intravenous polyvalent immunoglobulins	II	Mild to moderate AD	24	MMSE 14 – 26	18	NCT00299988
FDDNP PET*	Theracurmin	Dietary supplementation of curcumin	II	MCI and age-related cognitive decline Healthy subjects	132	Standard criteria	18	NCT01383161
Flutemetamol PET*	Physical exercise	Mild to intense physical exercise	II		24	DSM-IV	6	NCT01202994
Florbetapir PET	PF-04494700	Oral inhibitor of receptor for advanced glycation	II	Mild to moderate AD	402	MMSE 14 to 26	18	NCT00566397
Florbetapir PET	Semagacestat	end products LY450139, gamma-secretase inhibitor	III	AD	189	NINCDS-ADRDA	24	NCT01035138
Florbetapir PET	Semagacestat	LY450139, gamma-secretase inhibitor	III	Mild to moderate AD	1100	MMSE 16 to 26	18	NCT00762411
Florbetapir PET	BIIB037	Fully human monoclonal antibody	I	Prodromal and mild AD	160	MMSE 20 to 30	24	NCT01677572
Florbetapir PET	Bexarotene	Agonist's retinoid X receptor	II	Mild to moderate AD	20	MMSE 10 to 20	8	NCT01782742
Florbetapir PET Flutemetamol PET	solanezumab	Anti-Aβ antibody	III	Mild AD	2100	MMSE 20 to 26	18	NCT01900665
	MK-8931	Béta secretase inhibitor	III	Prodromal AD	1500	CDR-SB	104	NCT01953601
Amyloid PET	BAN 2401	Humanized conformational specific antibody targeting the Aβ protofibrils	II	Prodromal and mild AD	800	CDR 0.5 to 1; MMSE 22 to 30	18	NCT01767311

^*Amyloid biomarker studies in which change in amyoid load was the main outcome measure. Abbreviations: amyloid PET: positron emission tomography with amyloid biomarkers, MMSE: Mini Mental State Examination, CDR: Clinical Dementia Rating scale, AD: Alzheimer's disease, CSF: cerebrospinal fluid, NINCDS-ADRDA: criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association, DSM-IV: criteria of the Diagnostic and Statistical Manual of Mental Disorders, fourth edition

Activities of Daily Living (ADCS ADL)). In both phase III solanezumab and bapineuzumab trials, PET imaging revealed that about a quarter of patients lacked fibrillar amyloid pathology at baseline, suggesting that they did not have Alzheimer's disease in the first place. So a new third Phase 3 clinical trial for solanezumab, called EXPEDITION 3, in patients with mild Alzheimer's disease and evidence of amyloid burden has been started.

ADNI 2

The ADNI 2 study, that is currently recruiting, is a longitudinal observational study of 550 patients [89]. They are divided into four groups: cognitively healthy, early MCI, late MCI and AD dementia subjects, with a 4-year follow-up. Although the main outcome measure is volumetric change on MRI of various brain structures (overall brain volume, hippocampal volume), the progression of amyloid load as shown by florbetapir during the follow-up period is one of the secondary criteria. In the ADNI 2 study, diagnosis of AD demented subjects is made using the CDR, the MMSE, the Wechsler Memory Scale and independence rating scales.

MAPT-AV45

The MAPT [90, 91] study is a multicenter, multidomain interventional trial. The main objective is to evaluate the efficacy of omega-3 fatty acid supplementation alone, of a multidomain intervention (nutrition, physical activity, cognitive stimulation), or of their combination, on change in cognitive function of frail elderly persons aged 70 years and over. It is a phase III study that has included 1680 subjects aged 70 years and over, with at least one criterion of frailty (walking speed, instrumental activities, memory complaint) and no dementia syndrome apparent on inclusion. Although the main outcome measure is memory evaluation on the Grober and Bushke scale, an ancillary MAPT-AV45 study aims to evaluate the prevalence of brain amyloid plaques in this population of frail elderly subjects using florbetapir. The final results should be available in 2014.

Towards new diagnostic criteria for AD and recommendations for use of amyloid imaging

Revision of diagnostic criteria for AD by Dubois et al. in 2007 (92) and 2009 (93) included amyloid markers. According to these criteria, “probable” AD is defined by the association of memory impairment of hippocampal type and the presence of a marker of NFTs or amyloid process in the CSF or by molecular imaging.

Figure 6 shows the new definitions presented in the new lexicon (94). In addition to AD at the dementia stage, they introduce AD at the prodromal or predementia stage (presence of biomarkers and isolated characteristic memory loss) as well as presymptomatic disease (absence of AD symptoms and presence of a known autosomal dominant genetic mutation). The term “asymptomatic subjects at risk of AD” is defined by the absence of symptoms in presence of markers of AD. They also address atypical presentations such as logopenic or progressive primary aphasia, logopenic aphasia, posterior cortical atrophy or frontal variants which are recognized as atypical AD with in vivo demonstration of the amyloid process (in CSF or by molecular imaging). In this context, they would be used in preference to topographic or metabolic markers, reflecting dysfunctional areas.

Figure 6.

New definitions of AD according to the new lexicon (94)

The National Institute on Aging and the Alzheimer's Association entrusted a working group with the task of updating the 1984 AD criteria. Definitions were proposed for AD at the dementia stage, MCI and preclinical AD. AD biomarkers are included in these proposals, and are divided into biomarkers of amyloid (Aβ in CSF or amyloid PET) and of neuronal lesions (tau in CSF, FDG PET or brain atrophy on MRI).

Definition of preclinical stages of AD by Sperling et al. (95)is based on distinguishing the “pathophysiological process of AD” which precedes “clinical AD”. Its main interest lies in identifying subjects with asymptomatic stages of AD for therapeutic protocols (96). Use of biomarkers of the AD process contributes to define three stages of preclinical AD: stage 1, asymptomatic with abnormal amyloid markers; stage 2, asymptomatic with positive amyloid and neuronal lesions markers; and stage 3, that associates the features of stage 2 with impaired cognitive performances but not meeting criteria for MCI. A one-year follow-up study (97) showed increased conversion of healthy subjects to MCI or AD according to classification stages.

Albert et al. (98) defined the clinical criteria of MCI by the onset of involvement of one or several cognitive domains (memory, executive or visuospatial functions) which do not affect basic activities of daily living. The biomarkers provide an “intermediate” probability that MCI may be linked with an AD process when a biomarker (of either amyloid or neuronal lesions) is positive, and a “high” probability when both types of biomarkers are positive. When markers of amyloid and of neuronal lesions are negative, they suggest an MCI that is not due to AD process.

Lastly, McKhann et al. (99) propose the terms of “possible and “probable” AD according to cognitive history, symptoms or the association of criteria in favor of non-AD dementia. Using biomarkers, probable AD can be defined with evidence of an AD pathophysiological process (positive biomarkers), and possible AD with the presence of biomarkers (clinical criteria of non-AD dementia and positive biomarkers), even if a mixed condition cannot be excluded. Finally, non-AD dementia may be considered in presence of criteria for possible AD with negative biomarkers.

On the other hand, the recommendations do not propose the use of biomarkers in routine practice, for several reasons. Clinical criteria have very good efficacy for the majority of patients, whereas biomarkers have several drawbacks (limited accessibility, need for further research to identify and standardize normal values). Here use of biomarkers is recommended only in clinical research or investigational studies, or in practice in some rare cases left to the clinician's judgment.

The highlight of in vivo amyloid plaques allowed by PET scan permits the evaluation, in routine clinical practice, of pathophysiological process of AD. Even if opportunities in current practice as in research activities are important, experts have proposed criteria of good use of these examinations, to optimize their use to the most relevant indications. Appropriate criteria of the Amyloid Imaging Task Force propose that the use of PET either reserved for “experts” (100) psychiatrists, neurologists or geriatrics, whose activities evaluation patients with cognitive disorders or AD are frequent (at least 25% of time), with formations by learned societies (SNMMI). Certain information must be included at patient medical record to allow the practice of a PET Amyloid (100) (results of clinical and paraclinical evaluations, reasons of the examination) (100). Thus are defined criteria of good use. When the diagnosis of AD is possible but not certain despite expert assessments (101), and provided that the detection of amyloid process increases the likelihood of diagnosis and directs medical support (100)(anticipation of needs, followed up, planning financial aid,…). MCI potentially eligible are those of uncertain etiology (associated with AD or other cause), MCI of potentially mixed etiology or with confounding situations (psychiatric pathology,…). Finally, progressive dementias of young subject (before age 65) are included in these criteria 100, 101. It is specified that these examinations have no indication for assess stage of severity, for asymptomatic patients, or for diagnosis of body disease Lewy because of the possible positivity 100, 101. These appropriate criteria also assert the interest of an early cognitive diagnosis, notably for the establishment of aids, legal protection, participation in therapeutic studies, or differential diagnosis of curable pathology (101).

These criteria have the interest to introduce certain principles regarding to the use of amyloid PET. The purpose is the selection patients who may benefit most from this investigation, whose cout remains high, and that have, despite good results in terms of specificity and sensitivity, a non negligible proportion of “uncertain” results. Only patients for whom an increase in the diagnostic certainty is expected will benefit, and because of the absence at present of taken into preventive burden of disease Alzheimer, asymptomatic patients are excluded.

In contrast, the question of the place of amyloid PET in diagnostic approach is not mentioned: in these recommendations, others amyloid biomarkers (research of Abeta 42 in the LCR), either biomarkers of neuronal lesions (atrophy at MRI, functional imaging of metabolism at PET FDG, tau in the LCR) are not mentioned. This poses a question to view the new criteria for AD at clinical stage by Sperling & al (95), and Albert & al (98) for the MCI, for whom the probability of AD increases with the complementarity of amyloid biomarkers and neuronal damage biomarkers. Thus, these criteria for using amyloid PET seem to be a working basis useful for the markers marketing. They adduce limits recognized use, expertise necessary, quality of study before marketing.

Conclusion

According to our review of the literature, there is no difference between the amyloid imaging markers in terms of sensitivity and specificity, close to 90%. All have been validated by autopsy, although florbetapir and PIB appear better documented.

The findings of cross-sectional studies show that 20% of “healthy” elderly subjects and 50% of MCI subjects are amyloid (+). To a large extent, accumulation of amyloid substance precedes the clinical diagnosis of AD, in accordance to the dynamic theory of Jack et al. (38). This subgroup of healthy amyloid (+) subjects may represent preclinical stages of AD. Several factors influence amyloid binding, such as age, family history of AD, or presence of the APOE4 allele.

Longitudinal studies show that there exists, in healthy amyloid (+) subjects, an irreversible progression of amyloid load. This progression is greater if at baseline there is increased uptake in the prefrontal and cingulate regions, or in presence of the APOE4 allele. Hippocampal atrophy and amyloid load contribute to the conversion from “healthy” to “non-healthy” subjects, that confirms the importance of amyloid plaque deposition in cognitive decline at the predementia stages of the disease. Moreover, study of conversion rates of MCI patients to AD at dementia stage shows that amyloid (+) status is an independent predictive factor, and very few amyloid (–) MCI subjects become demented during follow-up. On the other hand, during the dementia stage of the disease, the severity of cognitive impairment is related only to the degree of hippocampal atrophy. A “ceiling effect” occurs in progression of amyloid load, and NFTs and neuronal death gradually become the principal markers of the course of the disease. These findings support the amyloid process in the neuropathological cascade of AD as a necessary but not sufficient condition for dementia disease. It underscores the factors influencing the onset of dementia, the principal factor being the presence of the APOE4 allele.

The value of amyloid markers for interventional studies lies principally in the selection of AD populations according to a neuropathological marker and in the detection of patients at the predementia stage of AD. Numerous studies with these new markers are currently under way: autopsy validation, analysis of diagnostic accurancies, follow-up of specific populations (conversion rates, predementia stages …). In 2011, new diagnostic criteria for AD at the dementia stage were published, including amyloid biomarkers: their aim is principally to reveal the amyloid process when the diagnosis of AD is “possible”, whereas they are not useful when the diagnosis is “probable”.

In view of the recent marketing authorization in the United States by Food and Drug Administration (FDA) of florbetapir (Amyvid™) and flutemetamol (Vizamyl™) and by European Medicines Agency (EMA) of Florbetapir (Amyvid™), for use in the diagnostic process of cognitive disturbances, the attitude of practitioners will come to change with in vivo demonstration of amyloid plaques. A recent study (102) has shown that in over 50% of cases the results of functional imaging with florbetapir lead to a change in clinical diagnosis, and in 20% of cases it strengthens diagnostic certainty. Future studies will complete these findings on AD, and better understanding of the pathophysiological process will surely lead to the development of new, effective treatments.