Amyloid imaging in Alzheimer's disease: comparison of Florbetapir and Pittsburgh Compound-B PET

Abstract

Amyloid imaging provides in vivo detection of the fibrillar amyloid-β (Aβ) plaques of Alzheimer's Disease (AD). The PET ligand Pittsburgh Compound-B (PiB-C11) is the most well studied amyloid imaging agent, but the short half-life of carbon-11 limits its clinical viability. Florbetapir-F18 recently demonstrated in vivo correlation with post-mortem Aβ histopathology, but has not been directly compared with PiB-C11. Fourteen cognitively normal adults and 12 AD patients underwent PiB-C11 and florbetapir-F18 PET scans within a 28 day period. Both ligands displayed highly significant group discrimination and correlation of regional uptake, supporting the hypothesis that florbetapir-F18 provides comparable information with PiB-C11.

Introduction

With the increasing potential of disease-specific therapeutic interventions, the search for non-invasive biomarkers of Alzheimer's Disease (AD) pathology remains an active area of pursuit. Amyloid imaging, which provides an in vivo measurement of one of the hallmark pathological features of Alzheimer's Disease (AD) – fibrillar amyloid-β (Aβ) plaques – holds great promise in filling this role and has already contributed significantly to our understanding of disease pathophysiology [1].

The positron emission tomography (PET) ligand ¹¹C-labeled Pittsburgh Compound-B (PiB-C11) is by far the most well studied amyloid imaging agent. In addition to reliably differentiating patients with AD from healthy controls and predicting the likelihood of progression to AD in patients with mild cognitive impairment (MCI) [2-8], strong correlations with histological measures of Aβ aggregates have been observed [9-12]. However, because of the short half-life of carbon-11 (∼20 minutes), PiB-C11 PET has limited potential for use outside of the research setting.

Fluorine-18 labeled PET ligands allow for more general clinical use due to their longer half-life (∼110 minutes). Several fluorine-18 amyloid imaging agents are in varying degrees of development [13-15]. Similar to PiB-C11 PET, florbetapir-F18 has demonstrated in vivo correlation with post-mortem histopathology, is currently being utilized in the Alzheimer's Disease Neuroimaging Initiative (ADNI), as well as other clinical intervention trials, and was recently approved by the United Stated Food and Drug Administration (FDA) for clinical use in the United States. Given the wealth of experience with PiB-C11 PET, an improved characterization of the correspondence between florbetapir-F18 and PiB-C11 PET would assist with interpretation of these studies. To examine the overlap of the two ligands, we obtained PET scans with both agents in patients with AD and cognitively normal controls.

Methods

Participants

Fourteen patients with AD [age: 69 ± 10 (SD) years; MMSE: 22 ± 4.6 (SD); 12 female;] and 15 cognitively normal (CN) adults [age: 71 ± 9 (SD) years; MMSE: 30 ± 0.5 (SD); 8 female] participated in the study. All participants were recruited from the cohort of the University of Pennsylvania's Alzheimer's Disease Center (ADC). As part of enrollment in the ADC, an extensive annual evaluation is performed, including medical history and physical examination, neurological history and examination, semi-structured psychiatric evaluation, and neuropsychological assessment; including all components of the National Alzheimer's Coordinating Center's (NACC) Uniform Data Set [UDS;(24-26)]. Clinical diagnosis is determined at a consensus conference attended by neurologists, neuropsychologists, geriatricians, and psychiatrists.

Diagnosis of AD was made according to the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer's Disease and Related Disorders Association criteria [16]. CN adults were defined as an absence of significant cognitive complaints, normal performance on age-adjusted cognitive measures, and consensus conference designation as ‘normal.’ Inclusion criteria included age between 55-90, fluency in English, and Mini-Mental State Examination (MMSE) score between 18-26 for AD and ≥ 27 for CN adults. Participants were excluded if they had a history of another significant neurological condition, such as clinical stroke, alcohol or drug abuse/dependence within 2 years of enrollment, any significant medical/psychiatric condition that would impact compliance with the study protocol, current participation in any amyloid specific therapeutic trial, or prior involvement in an immuno-based clinical trial for AD. The study was approved by the Institutional Review Board of the University of Pennsylvania.

Neuroimaging Acquisition and Analysis

PET imaging was performed on an Allegro whole-body PET scanner (Phillips). PiB-C11 PET and florbetapir-F18 PET were acquired on either the same day or two separate days separated by no more than 28 days. If performed on the same day, at least 120 minutes (∼6 half-lives of PiB-C11) must have passed between injection of PIB-C11 and florbetapir-F18. PiB-C11 PET imaging followed the ADNI protocol. Dynamic 3-D imaging for 20 minutes (4 × 5 minute frames) began 50 minutes following bolus intravenous injection of 555 MBq (15 ± 1.5 mCi) PiB-C11. Florbetapir-F18 PET imaging was conducted as previously described [14]. Dynamic 3-D imaging for 10 minutes (2 × 5 minute frames) began 50 minutes following bolus intravenous injection of 370 MBq (10 ± 1.0 mCi) florbetapir-F18. Due to technical failures, one AD patient and CN adult did not complete the PiB-C11 PET scan while one AD patient did not complete the florbetapir-F18 scan.

The multiple frames of each image were averaged and normalized to a tracer (PiB-C11 or Florbetapir-F18) specific PET template in Talairach space using SPM 2 (these were standard templates, previously derived by averaging scans from both CN and AD subjects from other studies with the respective tracers). Standardized anatomical regions-of-interest (ROI) were chosen based on typical association with amyloid plaque deposition [2, 17] and applied to the PET image in template space [14]. These ROI's included: anterior cingulate, posterior cingulate/precuneus, frontal cortex, parietal cortex, lateral temporal cortex, pons, and whole cerebellum. Given the general symmetry of PiB-C11 uptake [18], ROI's were averaged across right and left hemispheres. A composite ROI was also generated from the above cortical regions. Standardized uptake values (SUVs), calculated as the integrated activity over a given time period per unit of injected dose and body weight, were determined. ROIs were then referenced to the whole cerebellum to calculate standardized uptake value ratios [SUVRs; [6, 14]].

Statistical Analysis

Pearson correlation coefficient (r) was used to determine the correlation between the florbetapir-F18 and PiB-C11 SUVR values. The receiver operating characteristic (ROC) curve of cortical SUVR for classification of subjects as AD or NL was derived for both ligands. In addition, two-sample t tests were used to compare SUVR values between the two cohorts for each ligand separately.

Results

SUVRs for individual and the composite ROIs are presented in Table 1. Both ligands displayed significantly higher uptake in the AD relative to CN group in ROIs that are generally associated with amyloid plaque deposition in AD (p's < 0.05) [2, 17]. A composite measure of these regions was highly significant in group discrimination for both florbetapir-F18 [AUC=0.90, 95% CI 0.77 to 1.00] and PiB-C11 [AUC = 1.00]. Neither ligand displayed a group difference in the pons (p's > 0.1), a region in which uptake would be expected to be non-specific (Figure A).

Table 1. florbetapir-F18 and PiB-C11 PET regional SUVR's, discrimination of AD versus CN adults, and inter-tracer correlation.

	florbetapir-F18 PET				PiB-C11 PET
	CN n=14 (SD)	AD n=12 (SD)	AUC (95% CI)	Cohen d	CN n=14 (SD)	AD n=12 (SD)	AUC (95% CI)	Cohen d	Pearson Correlation n=26
Anterior Cingulate	1.12 (0.23)	1.52 (0.19)	0.88 (0.73,1.00)	1.90	1.19 (0.27)	2.35 (0.49)	1.00 (0.99,1.00)	2.93	0.81 (p<0.0001)
Posterior Cingulate/Precuneus	1.10 (0.23)	1.48 (0.15)	0.89 (0.74,1.00)	1.96	1.15 (0.29)	2.26 (0.42)	0.99 (0.98,1.00)	3.08	0.79 (p<0.0001)
Frontal	1.00 (0.18)	1.33 (0.20)	0.88 (0.74,1.00)	1.73	1.00 (0.26)	1.90 (0.41)	0.98 (0.94,1.00)	2.62	0.81 (p<0.0001)
Parietal	0.99 (0.16)	1.15 (0.20)	0.75 (0.54,0.96)	0.88	0.82 (0.28)	1.25 (0.30)	0.85 (0.67,1.00)	1.48	0.58 (p=0.002)
Lateral Temporal	1.11 (0.17)	1.43 (0.18)	0.90 (0.78,1.00)	1.83	1.19 (0.18)	1.92 (0.37)	0.97 (0.90,1.00)	2.51	0.68 (p=0.0001)
Pons	1.63 (0.18)	1.57 (0.12)	0.41 (0.19,0.64)	0.39	1.71 (0.27)	1.72 (0.29)	0.57 (0.34,0.79)	0.04	0.38 (p=0.06)
Composite	1.06 (0.17)	1.38 (0.15)	0.90 (0.77,1.00)	2.00	1.07 (0.23)	1.94 (0.36)	1.00 (1.00,1.00)	2.88	0.78 (p<0.0001)

Figure.

A. Standardized uptake value ratio (SUVR) for both florbetapir-F18 and PiB-C11 in the composite cortical region of interest (ROI) and pons for the cognitively normal adults (grey boxes) and AD patients (white boxes). ‘Boxes’ are drawn between lower and upper quartiles; ‘whiskers’ indicate minimum and maximum values, minus the outliers, indicated by squares; the bold line represents the median and the plus-sign the mean. B. Correlation between florbetapir-F18 and PiB-C11 standardized uptake value ratios (SUVRs) in the composite cortical region of interest (ROI) from 14 cognitively normal (CN; blue) older adults and 12 patients with Alzheimer disease (AD; green). ‘+’ symbols represent cases with AD-range CSF Aβ (< 192 pg/ml) and circles represent normal Aβ (≥ 192 pg/ml). The equation of the best linear fit is given.

SUVRs for all of the cortical ROIs were highly correlated between the two methodologies (r's=.58-.81; see Table 1). Interestingly, the ROI with the weakest correlation, parietal cortex (r=.58), was also the cortical ROI that displayed the smallest group difference. Consistent with the individual ROI's, the composite cortical ROI was highly correlated across the two ligands (r=.78, p<0.001), but the range of SUVRs was greater with PiB-C11 PET (Figure B).

Discussion

The development of fluorine-18 PET amyloid imaging ligands that reliably detect AD-related amyloid plaque pathology has the potential to greatly expand the ability to use these methodologies in research and clinical practice. The current study examined the relationship of one of the most developed of these compounds, florbetapir-F18. In vivo imaging with this compound has already demonstrated significant correlation with histological findings on autopsy [14]. This is the first published study directly comparing this agent to PiB-C11 in the same patients within a short temporal interval (< 28 days).

We found very similar and robustly correlated binding characteristics for the two compounds. Both PET ligands displayed excellent ability to discriminate CN adults from those with mild AD. Further, there was no evidence of group differences in the degree of non-specific binding in the pons, a region not associated with fibrillar amyloid pathology. While fluorine-18 labeled amyloid imaging agents such as florbetapir, and other newer ligands (e.g. flutemetamol-F18), generally have been associated with higher white matter uptake [15, 19], SUVRs for the pons were similar for the two ligands. However, florbetapir-F18 uptake in the pons was relatively higher in relation to cortical regions compared to that observed for PiB-C11. Nonetheless, all regions displayed highly significant correlation across the two methodologies, providing additional support that these ligands have very similar binding characteristics.

Despite this overlap, there were some differences. Although there was a strong correlation between the methods, the range of SUVRs between the lowest and highest values was larger for the PiB-C11 PET scans (see Figure B). Second, while both compounds clearly distinguished the two populations, there was less overlap with PiB-C11 PET, based on the ROC analysis (see Table 1). One potential explanation for this is the greater relative non-specific white matter uptake of florbetapir-F18 relative to cortical uptake. Particularly in the context of the grey matter atrophy associated with AD, inclusion of some white matter signal in cortical ROI's is likely and could result in reduced specificity of the findings.

However, comparison of the two compounds to discriminate based on clinical status should be viewed with caution given the frequent finding of amyloid plaque pathology in cognitively normal individuals [20]. Since we do not have the “gold standard” of histological confirmation, it is possible that at least some of the overlap between CN adults and those with AD of florbetapir-F18 PET uptake may reflect sensitivity to this “preclinical” pathology. Indeed, Joshi et al. suggested an SUVR cutoff of 1.10 for establishing a positive florbetapir scan [21]; the sensitivity and specificity of this cutoff for detecting patients with moderate to frequent neuritic plaques at autopsy has been recently confirmed [22]. If the regression line (Figure B) is used to impute the comparable cutoff point for PIB (SUVR = 1.15) all of the CN subjects identified as amyloid positive by florbetapir would also be identified as amyloid positive by PIB.

While the present data do not allow us to directly determine which of the present subjects were truly amyloid positive, it is possible to compare the PET results for some of these subjects to an independent amyloid biomarker; eight individuals in this study (3 CN; 5 AD) also had cerebrospinal fluid (CSF) levels of Aβ_1-42 obtained within one year of their scan (mean: 72 days). Previous work at the University of Pennsylvania, including an autopsy series, has established a cutoff for Aβ_1-42 (< 192 pg/ml) with 96% sensitivity to underlying AD-related amyloid pathology [23]. Based on this cutoff, all of the AD subjects with low (amyloid positive) CSF Aβ_1-42 values (green crosses in Figure B) were clearly elevated in PiB uptake. While two of these AD patients had florbetapir uptake in the CN range, both were above the above noted threshold of an SUVR of 1.10 (1.12, 1.17). Two CN adults (blue crosses in Figure B) displayed evidence of amyloid pathology by CSF. Interestingly, one of these adults was within the low range of PET uptake for both compounds, inconsistent with the CSF Aβ_1-42 result. The other CN adult with an abnormal CSF had the second highest composite PiB SUVR value of the CN participants and was in the AD-range for florbetapir-F18, suggesting that at least some of the overlap in the clinical groups may be due to CN adults harboring amyloid pathology.

Finally, there was also some variability in the correlation of the two measures in particular regions, such as the parietal cortex. The reason for this is unclear, but could be due to the fact that this is a region with significant atrophy in AD and, thus, more susceptibility to local non-specific white matter uptake. Alternatively, differential binding of the two compounds to different species of fibrillar amyloid is possible and merits further investigation.

Overall, we found a high correlation between the binding properties of florbetapir-F18 and PiB-C11 PET supporting the general hypothesis that these agents provide generally analogous information. The present results echoes findings with flutemetamol-F18 PET, essentially fluorine-18 labeled PiB, which displayed similar correlations with PiB-C11 PET [19]. Taken together, these findings suggest a role for the more accessible fluorine-18 amyloid imaging ligands in clinical research studies. Furthermore, research data with PiB-C11 PET can be used to inform clinical experience now that florbetapir-F18 PET has been approved for use by the FDA in the United States.