Introduction
The use of positron emission tomography (PET) imaging with probes that bind specifically to β-amyloid and tau aggregates has received increased attention recently because this technique may provide an earlier diagnosis of Alzheimer’s disease (AD). Currently, AD [
1,
2] must reach the dementia stage, in which cognitive and noncognitive symptoms significantly alter activities of daily living, to be clinically diagnosed. However, disease symptoms are considered a consequence of the cumulative burden of brain alterations that may begin to appear years before initial clinical manifestations [
3,
4]. Consequently, the new AD diagnostic criteria suggest that the diagnosis of "prodromal AD" (also called the AD predementia stage) [
5] or "MCI due to AD pathology" [
6] should rely on in-vivo biomarkers of amyloid pathology, such as PET imaging that uses ligands of amyloid plaques and degenerative neurofibrillary tangles. Among the candidate probes,
11C-labelled tracers have been extensively studied. The PIB compound (
N-methyl-[
11C]2-(4′-methylaminophenyl)-6-hydroxybenzothiazole), a derivative of thioflavin T, was the first to demonstrate the ability to clearly distinguish AD patients from healthy control (HC) subjects [
7]. PET imaging with
11C-PIB has also been used to efficiently differentiate patients with mild cognitive impairment (MCI) who have converted to AD from nonconverters [
8], and to estimate the decreased amyloid load in patients treated with bapineuzumab [
9]. Nevertheless,
11C-PIB uptake is also increased in nondemented elderly subjects [
10] and shows a variable level of agreement with other biomarkers, such as cerebrospinal fluid (CSF) dosage of Aβ [
11]. Another
11C-labelled, benzoxazole-derived compound,
11C-BF-227, has good neocortical uptake in AD patients [
12] and appears to be able to discriminate MCI converters from nonconverters better than voxel-based morphometry using magnetic resonance imaging (MRI) [
13,
14]. However, the 20-min radioactive half-life of both
11C-PIB and
11C-BF-227 is a serious barrier to increasing the accessibility of biomarkers for routine clinical purposes, as the use of these markers is limited to centres with an on-site cyclotron.
Consequently,
18F-labelled amyloid ligands appear to be the best alternative, as the 110-min half-life of
18F allows the centralized production and locoregional delivery of compounds, similar to other PET radiotracers, such as
18F-FDG. Several clinical studies have been conducted with a new naphthalene family compound, 2-(1-{6-[(2-
18F-fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile (
18F-FDDNP). PET with
18F-FDDNP has demonstrated the ability to discriminate between HC subjects, MCI patients and AD patients [
15]. However,
18F-FDDNP has limitations because of its low metabolic stability and its high white matter binding [
16]. More recently, the standard uptake value ratio (SUVr) of
18F-labelled PIB (
18F-flutemetamol) was observed to be highly correlated with the SUVr of its parent molecule,
11C-PIB [
17]. However, another family of
18F-labelled ligands derived from stilbene has been developed, and two compounds belonging to this family (AV-1 and AV-45) display favourable properties for brain amyloid PET imaging.
18F-AV1 (BAY-94-9172, florbetaben) has brain kinetic characteristics appropriate for clinical use. When studied 90 to 120 min after injection, the level of cortical uptake clearly distinguished 15 AD patients from 15 healthy elderly subjects [
18]. A recent phase II study has shown that the visual rating of PET scans with florbetaben has sufficient sensitivity (80%) and specificity (91%) to distinguish AD patients from HC subjects [
19].
The last stilbene family derivative,
18F-AV-45 (florbetapir) [
20], also has strong affinity for amyloid proteins in AD brain homogenates (
ki 10 ± 3.3 nM) and faster in vivo kinetics [
21]. The use of florbetapir in amyloid imaging has recently been validated in an autopsy study [
22], and its safety profile allows clinical applications in brain imaging [
23]. A pilot study comparing 16 AD patients with 16 HC subjects confirmed that florbetapir binds selectively to amyloid proteins enabling AD patients to be distinguished from HC subjects, and has kinetics that allow imaging sessions as short as 10 min 50 to 60 min or 60 to 70 min after injection [
24]. More recently, a multicentre study conducted at 31 US research sites that included 82 elderly HC subjects, 62 AD patients and 62 MCI patients reported that both qualitative visual ratings and a continuous mean cortical SUVr quantification show high sensitivity and specificity for distinguishing AD patients from MCI patients and elderly HC subjects [
25]. These findings suggest that florbetapir is a useful biomarker for distinguishing AD and MCI patients from HC subjects. However, these results were derived from premarketing studies, and little is known about their possible replication under "real life" conditions and everyday practice.
In the present study, our aim was to assess the feasibility of using PET imaging with florbetapir in third-level clinical settings to differentiate patients with mild to moderate AD or MCI patients from normal HC subjects in three PET centres. In addition, we assessed the safety of a florbetapir injection immediately after injection and during the follow-up period.
Discussion
Our results indicate that PET with florbetapir is suitable for routine use to improve the accuracy of AD diagnosis in clinical settings, such as memory clinics, with acceptable tolerability and sufficient reliability. In particular, the quantitative analyses showed a higher global SUVr and SUVr in several cortical regions (precuneus, anterior and posterior cingulate, frontal median, temporal, parietal and occipital cortex) in AD patients than in HC subjects. Additionally, the SUVr in the posterior cingulate and frontal median regions was significantly higher in AD patients than in MCI patients. To the best of our knowledge, the results of only five clinical trials with
18F-labelled tracers that specifically bind to brain amyloid plaques have been published to date [
15,
17,
19,
22,
25]. In all studies PET imaging with
18F-labelled tracers was able to distinguish AD patients from HC subjects, but only three [
15,
17,
25] included MCI patients, and the differences in SUVr between MCI patients and HC subjects were only reported with an FDDNP tracer that binds to both amyloid and tau proteins [
15], which does not appear to be the case with florbetapir [
25]. The pattern of florbetapir cortical uptake found in the present study is remarkably similar to that found in previous studies conducted by Wong et al. [
24] and Clark et al. [
22]. The pattern also appears to be similar to those found with other amyloid-labelling compounds, such as
11C-PIB [
10] and its
18F-flutemetamol-derived molecule [
17],
11C -BF-227 [
13],
18F-FDDNP [
15] and
18F-BAY94-9172 [
18,
19]. These patterns closely match the neuropathological stages of AD progression, which was strengthened by the high correlation found between florbetapir PET imaging and autopsy results [
22]. Our results also indicate that there is high uptake of florbetapir in the white matter (centrum semiovale) and the striatum region (in particular, the putamen) to a similar degree as with other
11C- [
39,
40] and
18F-labelled compounds [
17,
41,
42]. A high rate of amyloid deposition in the striatum has been suggested to be more frequently associated with dementia with Lewy bodies [
43] or some early-onset variant of AD associated with spastic paraparesis and a presenilin I (PSI) gene mutation [
44]; however, this finding has also been consistently reported in PET imaging studies using amyloid ligands in AD patients [
17,
39‐
42], and this high uptake level in the striatum does not appear to be associated with decreased glucose metabolism (as assessed by FDG PET), as it is in other cortical areas [
45].
The sensitivity (92,3%) and specificity (90,5%) values provided a good quantitative assessment for the global cortex ROI, with a cut-off value (1.12) lower than that calculated by Fleisher et al. (SUVr ≥1.17), but the mean global SUVr (mean ± SD) values were also slightly lower in the present series (AD patients 1.26 ± 0.15, MCI patients 1.12 ± 0.05, HC subjects 1.07 ± 0.09) than in the study by Fleisher et al. (AD 1.39 ± 0.24, MCI 1.17 ± 0.27, HC 1.05 ± 0.16) [
25]. However, the visual assessment less accurately distinguished AD patients from HC subjects, demonstrating a specificity of 38.1%. There are at least two possible explanations for this low specificity. First, the three different cameras between the three participating sites required adaptation of the reconstruction parameters to those of the tomograph with the lowest spatial resolution (Dual Philips GEMINI) and this could have led to some borderline positive scans being rated as negative or conversely borderline negative scans being rated as positive. Second, at least one HC subject was a clear outlier, with high SUVr values and a positive visual rating. Interestingly, this HC subject had a family history of AD and was a heterozygotic ApoE4 gene carrier.
The association of the ApoE4 genotype with the likelihood of having a positive florbetapir PET scan on visual rating is another remarkable finding of this study. The association of the ApoE genotype with a higher brain amyloid load has previously been documented in several PET imaging studies with PIB [
46,
47], florbetaben [
19] and florbetapir [
25]. Strikingly, the ApoE genotype appears to increase florbetapir uptake levels in several cortical regions in AD and MCI patients but not in HC subjects, suggesting that the presence of ApoE could strengthen or accelerate the level of amyloid deposition. Here the HC subjects with a positive scan could have been at an early preclinical stage of the disease in which criteria for MCI or AD were not yet reached, and only the follow-up will confirm its progression towards a full symptomatic feature of disease.
This study has several limitations. First, the sample size was limited, but this work was initiated and conducted in only three academic clinical settings independent of any premarketing phase II studies now being conducted by industrial sponsors. Another limitation was a possible selection bias, expressed by the significantly older age in the MCI group than in the AD and HC groups. Nevertheless, age did not differ between subjects with positive and negative scans on visual rating (Table
2). Moreover the MCI patients had a significantly older mean age than the AD patients and HC subjects. It cannot be excluded that this could have increased the mean values of florbetapir uptake in favour of showing a significant difference in relation to the HC subjects. Despite these limitations, the preliminary results indicate that imaging techniques using
18 F-labelled compounds, such as florbetapir, can be easily conducted in demented and predemented patients in clinical settings, such as third-level memory clinics, and could become a routine clinical procedure for patients with suspected AD. The low specificity of the visual PET scan assessment is problematic, but this could be significantly increased by improvements in the spatial resolution of tomographs as well as by the use of appropriate training programmes for raters, and by the development of semiautomatic or automatic quantifications method or software.
Conclusion
These preliminary results agree with previously reported results obtained using other PET tracers that bind to brain amyloid plaques, and suggest that PET with florbetapir could become a routine clinical procedure to improve the reliability of AD diagnosis and the detection of typical or atypical forms of predementia stages, such as amnestic MCI and MCI associated with multidomain deficits or neuropsychiatric symptoms (e.g. depression). More studies testing the feasibility and tolerability of consecutive scans with florbetapir are needed to better document the accuracy of PET imaging with florbetapir in the AD diagnostic process at the dementia or predementia stages. Comparisons (or combinations) with other biomarkers, such as FDG PET, MRI and CSF dosages of tau and protein, are also needed. The clinical relevance of changes in quantity of brain amyloid over time, especially when disease-modifying treatments are available, should also be assessed. Finally, the reliability of visual rating of PET scans remains critical and should be urgently addressed through the improvement in spatial resolution of images, the use of semiquantitative methods and the support of appropriate training programmes for nuclear medicine physicians.
Acknowledgments
This work was funded by the French Ministry of Health with grant no. PHRC-N 2008-1004 and in part by the EC-FP6-project DiMI, LSHB-CT-2005-512146, the Région Centre and FEDER: “Radex” programmes. This work was conducted with the support of the GIS Radiopharmaceutiques Caen, Toulouse, and Tours centres with Laboratoires Cyclopharma. The authors are grateful to D Skovronsky (AVID Radiopharmaceutucals), to JB Deloye (Laboratoires Cyclopharma) and to the technical staff involved in florbetapir production. Authors are also grateful to Drs. N. Daluzeau and F. Bouvier (CH Lisieux), Dr. B. Dupuy (CH Cherbourg), A. Abbas and A. Pèlerin (Inserm U923), Dr. A. Manrique (GIP Cyceron), C. Roussel, PhD, C. Baringthon, PhD, A. Matysiack, H. Bansard and F. Teasdale (CIC/CIC-IT 202) for their contributions to the clinical investigation.