¹¹C-PIB PET imaging in Alzheimer disease and frontotemporal lobar degeneration

Subject selection.

Patients were recruited from AD and FTLD research cohorts followed at the University of California San Francisco Memory and Aging Center (UCSF MAC). All patients had at least one clinical evaluation at the MAC, including a history and physical examination by a neurologist, a structured caregiver interview administered by a nurse, and a comprehensive battery of neuropsychological tests.⁹ Patients' functional status was measured using the Clinical Dementia Rating Scale (CDR).²³ MRI scans acquired clinically or through research were reviewed for all patients at the time of diagnosis. Clinical diagnoses were assigned at a multidisciplinary conference using standard research criteria for AD,²⁴ the FTLD clinical subtypes frontotemporal dementia (FTD), semantic dementia (SemD), and progressive aphasia (PA),²⁵ and amyotrophic lateral sclerosis (ALS).²⁶ Onset of symptoms was determined retrospectively based on the estimated date of the first symptom as identified by patients or caregivers and documented in medical records.

Patients were considered eligible for the study if they met research criteria for AD or FTLD, and did not have significant comorbid medical, neurologic, or psychiatric illness. Control subjects were recruited from the community by advertisement. All were free of significant medical illnesses and were not taking medications deemed to affect cognition. Control subjects were judged to be cognitively normal following an evaluation that included a medical history, functional assessment, neurologic examination, and neuropsychological assessment with a battery of tests similar to those performed by patients. Eligible subjects were recruited between April 2005 and May 2006, and all who consented were enrolled.

The final cohort included 7 patients with AD, 12 patients with FTLD, and 8 normal controls (table 1). At the time of this report, pathologic confirmation of the diagnosis was available for one patient with FTLD-FTD (autopsy diagnosis of Pick disease²⁷). In a second patient with familial FTD/ALS, a pathologic diagnosis of FTLD with motor-neuron (ubiquitin-positive, tau-negative) inclusions²⁷ was made in the patient's deceased brother.

The study was approved by the University of California at Berkeley Campus Committee for the Protection of Human Subjects.

Radiochemical synthesis.

¹¹C-PIB was synthesized at the Lawrence Berkeley National Laboratory's Biomedical Isotope Facility using a previously published protocol.²⁸ In brief, high specific activity ¹¹C-carbon dioxide produced on an 11 MeV CTI RDS-111 cyclotron was used to synthesize ¹¹C-CH₃I.^29,30 The PIB precursor 2-(4′-aminophenyl)-6-methoxymethoxybenzothiazole was prepared using previously described procedures, then methylated with ¹¹C-CH₃I prior to deprotection to afford the 6-hydroxy compound, ¹¹C-PIB.²⁸ The final compound was purified by semipreparative HPLC. Specific activity was determined for every batch, and averaged 4,648 ± 1,214 Ci/mmol. ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) was purchased from a commercial vendor (Eastern Isotopes, Sterling, VA).

Image acquisition.

PET scans were performed at Lawrence Berkeley National Laboratory using a Siemens ECAT EXACT HR PET scanner in three-dimensional acquisition mode. All 27 subjects (19 patients and 8 controls) underwent PET imaging with ¹¹C-PIB. An average of 16.4 mCi of ¹¹C-PIB (range 9.8 to 19.2 mCi) was injected as a bolus into an antecubital vein. Dynamic acquisition frames were obtained as follows: 4 × 15 seconds, 8 × 30 seconds, 9 × 60 seconds, 2 × 180 seconds, 8 × 300 seconds, and 3 × 600 seconds, for a total of 90 minutes. Most patients (17 of 19) also underwent PET imaging with ¹⁸F-FDG. At a minimum of 2 hours following ¹¹C-PIB injection (six Carbon-11 half-lives), patients were injected with an average of 9.4 mCi of ¹⁸F-FDG (range 8.8 to 10.0 mCi). Six emission frames of 5 minutes each were acquired starting 30 minutes after tracer injection, with the patient resting quietly in a dimly lit room with minimum ambient noise, and eyes and ears unoccluded during tracer uptake. Ten minute transmission scans for attenuation correction were obtained either immediately prior to or following each ¹¹C-PIB and ¹⁸F-FDG scan. PET data were reconstructed using an ordered subset expectation maximization algorithm with weighted attenuation. Images were smoothed with a 4 mm Gaussian kernel with scatter correction. All images were evaluated prior to analysis for patient motion and adequacy of statistical counts.

Most patients (18 of 19) underwent high resolution MRI scans on a 1.5-T Magnetom VISION system (Siemens Inc., Iselin, NJ) using a previously published protocol.³¹ In patients with multiple MRIs, the MRI closest to the date of the PET scan was used for analysis. Research protocol MRI scans were not available for controls and for one patient.

Image preprocessing and definition of regions of interest.

Structural scans (T1-weighted MRIs) were manually reoriented to the anterior-posterior commissure plane and origins were set to the anterior commissure (AC). Structural images were segmented in SPM2 (http://www.fil.ion.ucl.ac.uk/spm) and the resulting gray and white matter images were added to form a brain mask. This mask was then applied to the original MRI to effectively skull-strip the brain. The skull-stripped image was normalized (SPM2 defaults) to the Ch2bet template, a high-resolution skull-stripped single-subject T1-weighted scan in Montreal Neurological Institute (MNI) space.³² The resulting normalization parameters were then used to perform a reversed (backwards) normalization on a set of MNI space region of interest (ROI) templates that included lateral frontal, lateral temporal, medial temporal, precuneus and posterior cingulate cortices, subcortical white matter, and pons. All templates were a subset of the Automated Anatomic Labeling (AAL) Atlas³³ except the white matter and pons ROIs, which were manually drawn.

The reversed normalization warps MNI space ROIs into native-space. For each patient it involves extraction of a deformation field from the normalization parameters, inversion of the field, and finally application of the obtained inversion to the ROIs. In order to prevent bleeding into CSF areas, the native-space brain mask was applied to all ROIs. The resulting native-space ROIs were then coregistered and resliced to the mean ¹¹C-PIB PET image obtained during realignment of the ¹¹C-PIB frames (see below). Since the segmentation algorithm originally applied to the MRIs occasionally failed to distinguish dura mater or skull from gray matter, we applied an additional brain mask based on the subject's mean ¹¹C-PIB image. Finally, all ROIs were overlaid and visually inspected on the subject's coregistered MRI and ¹¹C-PIB mean image to ensure correct warping. Manual editing of ROIs was performed when necessary.

The ROI definition protocol had to be slightly altered for one patient and all eight controls who did not have a structural scan. The Ch2 template (not skull-stripped) and all AAL ROIs were coregistered and resliced to the SPM PET template image. The subject's native-space mean ¹¹C-PIB image was then normalized to the PET template, and reverse normalization was performed on the ROIs, analogous to the procedure described above.

Origins were manually set to AC in all PET frames. ¹¹C-PIB frames 6 through 34 were coregistered and resliced with SPM2 PET realignment parameters using a mid-scan frame visually judged to best represent the subject's anatomy (usually number 17) as the reference frame. Frames 1 through 5 were coregistered separately to the mean ¹¹C-PIB image obtained from realignment, since these frames typically contain a paucity of anatomic information due to low tracer counts. ¹⁸F-FDG scans were realigned with defaults, summed, and subsequently coregistered to the mean ¹¹C-PIB image.

Image analysis.

¹⁸F-FDG scans were normalized to mean activity in the pons ROI for each subject.³⁴ For ¹¹C-PIB, voxel-wise and ROI distribution volume ratios (DVRs) were calculated using Logan graphical analysis,³⁵ with the cerebellum ROI time-activity curve used as a reference tissue input function. The cerebellum was chosen as a reference region because it is relatively free of fibrillar plaques in AD,³⁶ and results obtained with this analysis are similar to those derived from arterial input functions.^18,20,21 Kinetic parameters (T = 35 to 90 minutes, k₂ = 0.15 minutes⁻¹) were based on previously reported values.²⁰ In one patient (clinical diagnosis of FTLD-FTD), the cerebellum was missing from many image frames due to patient motion and could not be used as a reference region. A DVR image for this patient was created using the pons ROI as a reference region, since ¹¹C-PIB binding in the pons does not differ between AD patients and controls,^18,20 and results obtained when pons is used as a reference region are qualitatively very similar to those obtained when cerebellum is used (unpublished observations). The DVR image from this patient was used for visual inspection only, and was excluded from quantitative analyses.

Visual inspection.

Voxel-wise ¹¹C-PIB DVR images from all subjects were qualitatively assessed by an experienced PET investigator (W.J.J.) blinded to clinical diagnosis. Scans were visually read as positive or negative for cortical ¹¹C-PIB. A positive scan was defined as a DVR image in which uptake was significantly greater in cortex than in white matter. Summed ¹⁸F-FDG frames from patients only (not including controls) were visually read by the same investigator in a separate session as consistent with the metabolic pattern of either AD or FTLD.^37–40 The AD pattern was defined as hypometabolism (which could be asymmetric) in parietal, posterior temporal, and posterior cingulate/precuneus cortical regions. Frontal hypometabolism was allowed if it was judged to be less severe than posterior hypometabolism. Conversely, an ¹⁸F-FDG pattern consistent with FTLD was defined as a more severe metabolic lesion in dorsolateral, dorsomedial, or ventromedial frontal or anterior temporal cortex. To assess inter-rater reliability, a second blinded investigator (H.J.R.) read all patient ¹¹C-PIB and ¹⁸F-FDG scans following a training session.

Statistical analysis.

Group differences in demographic and neuropsychological measures were examined using one-way analysis of variance (ANOVA) and Tukey post hoc contrasts (for comparisons involving three groups), or two-tailed independent sample t-tests (for comparisons involving two groups). When non-parametric tests were required, the Kruskal-Wallis equality-of-populations rank test was used to compare three groups, and Wilcoxon's rank sum test was used to compare two groups. Dichotomous variables were analyzed using Fisher exact tests. Inter-rater reliability for visual interpretations was calculated using Cohen's Kappa. Mean DVR ROI values were compared between and within groups using repeated measures ANOVA (due to the high correlation between ROI DVR values for a given subject), with planned post hoc comparisons between AD and controls, FTLD and controls, and AD and FTLD. Statistical analyses were implemented in SPSS 12.0 for Windows (SPSS Inc., Chicago, IL) and Statistica 6.0 software (StatSoft Inc., Tulsa, OK).

Demographics.

Study subjects were well matched for age, gender, and education (table 1). Patients had lower Mini-Mental State Examination (MMSE) scores than controls (p < 0.01). AD and FTLD patients were well matched for disease duration and for dementia severity, as measured by the MMSE and CDR. All six patients with AD genotyped for apolipoprotein E (ApoE) were carriers of the ApoE4 allele, in comparison to 2 of 10 genotyped FTLD patients (p < 0.01). Medication use differed significantly between groups, with patients with AD using cholinesterase inhibitors more frequently than both patients with FTLD and controls, and both AD and FTLD patients using memantine more frequently than controls.

Visual interpretation.

Results of PET visual reads by the primary reader are presented in table 2. All seven patients with AD demonstrated cortical ¹¹C-PIB retention on visual inspection (PIB-positive scan), while 8/12 patients with FTLD and 7/8 controls did not (PIB-negative scan) (figure E-1 on the Neurology Web site at www.neurology.org). The patient with pathologically confirmed Pick disease and the patient with familial FTD/ALS both had negative ¹¹C-PIB scans by visual inspection.

Visual reads of ¹⁸F-FDG scans agreed with the clinical diagnosis in 5/6 patients with AD and 8/11 patients with FTLD (table 2). ¹¹C-PIB and ¹⁸F-FDG suggested the same diagnosis (PIB-positive scan and FDG consistent with AD, or PIB-negative scan and FDG consistent with FTLD) in 13/17 patients. In 3/4 cases in which there was a discrepancy, the ¹⁸F-FDG scan suggested FTLD, while the ¹¹C-PIB scan was positive.

A second blinded reader agreed with the primary visual read on 19/19 patient ¹¹C-PIB scans (kappa = 1.00) and 15/17 ¹⁸F-FDG scans (kappa = 0.76). Both disparate ¹⁸F-FDG reads occurred in FTLD patients: one who was PIB-positive (read as AD by the primary reader and FTLD by the second reader) and one who was PIB-negative (primary read FTLD, second read AD).

Qualitative assessment of ¹¹C-PIB distribution.

Six of seven patients with AD showed a diffuse, symmetric pattern of ¹¹C-PIB uptake (see figure E-1 for illustration), with tracer retention throughout lateral (dorsolateral and peri-opercular), medial (anterior and mid-cingulate, supplementary motor area [SMA]), and orbital frontal cortex; lateral (angular and supramarginal gyri, superior and inferior parietal lobules) and medial (precuneus and posterior cingulate) parietal cortex; lateral temporal cortex (including superior, middle, and inferior temporal gyri); occipital visual association cortex; striatum; and thalamus. Four of seven patients with AD had elevated ¹¹C-PIB in calcarine cortex. One patient with AD (AD-4) demonstrated an atypical uptake pattern, with ¹¹C-PIB uptake restricted to dorsal medial (mid-cingulate and SMA) and lateral (superior and middle frontal gyri) frontal cortex, precentral gyrus, and dorsal precuneus (figure E-2). Medial temporal structures were relatively spared in all patients with AD.

The four PIB-positive patients with FTLD all showed a diffuse pattern of tracer uptake (figures E-3 and E-4), with involvement of all of the above-mentioned frontal, parietal, lateral temporal, and occipital regions (including calcarine cortex in three of four patients), striatum, and thalamus. In contrast, the one PIB-positive control had a restricted pattern of uptake, with moderately elevated ¹¹C-PIB in medial parietal cortex (posterior cingulate and precuneus), and mildly elevated ¹¹C-PIB in (predominantly medial) frontal cortex.

PIB-positive patients with FTLD.

Four of 12 clinically diagnosed patients with FTLD had positive ¹¹C-PIB scans. We retrospectively reviewed these patients' clinical data in light of the discrepancy between their clinical diagnosis and the ¹¹C-PIB result.

Two of the PIB-positive patients with FTLD had a clinical diagnosis of FTLD-FTD (figure E-3). At initial presentation, FTLD-4 was a 55-year-old left-handed man with 9 years of profound behavioral changes, including compulsive sorting, disinhibited and socially inappropriate behavior (e.g., stripping naked at a family beach outing, pretending to be blind so that his dog would be allowed on a train). On a neuropsychological battery,⁹ he scored 20/30 on the MMSE, and showed deficits in episodic memory (recalled 2/9 words after a 10-minute delay on the California Verbal Learning Test [CVLT]), visual-spatial tasks (could not correctly copy intersecting pentagons, and scored 7/17 on copy of the Modified Rey-Osterrieth design), and executive function (107 seconds to complete modified Trails-B, seven designs/minute on the Delis Kaplan Executive Function Scale [DKEFS] design fluency task). The visual-spatial deficits are unusual for FTD, and are more consistent with AD.^9,41 However, the degree of behavioral disturbance as measured by the Neuropsychiatric Inventory (NPI)⁴² (score of 46) was more typical of FTD. MRI demonstrated predominantly biparietal atrophy, and ¹⁸F-FDG PET (obtained 27 months after initial presentation as part of this study) showed biparietal hypometabolism, and was read by the blinded investigator as consistent with AD (figure E-3).

At presentation, FTLD-1 was a 53-year-old right-handed woman with a 5-year history of short-term memory loss, poor organization and planning skills, and behavioral changes including inappropriate jocularity, disinhibition, apathy, and poor insight. On neuropsychological testing, she scored 25/30 on the MMSE, and showed deficits in executive function (two correct lines and three errors in 120 seconds on modified Trails-B, one design in 1 minute on DKEFS design fluency), working memory (digit-span backwards of two), episodic memory (3/9 free recall at 10 minutes on the CVLT), naming (10/15 on the Boston Naming Test [BNT]), and verbal fluency (five animals and eight D words in 1 minute), a pattern consistent with either AD or FTD.⁹ However, her total NPI score (7) was low compared to other patients with FTD. MRI showed predominantly frontal atrophy, but ¹⁸F-FDG PET (obtained 1 year after initial presentation) showed prominent biparietal hypometabolism and was read by the blinded investigator as consistent with AD (figure E-3). Of note, on follow-up over 19 months, she demonstrated a significant decline in verbal and visual memory and confrontation naming, changes more consistent with AD than FTD.

The other two PIB-positive patients with FTLD had a clinical diagnosis of FTLD-SemD (figure E-4). At initial evaluation, FTLD-9 was a 58-year-old right-handed woman with a 10-year history of rheumatologic disease (diagnosed as either systemic lupus erythematosus [SLE] or Sjögren syndrome) and 3 years of progressive word-finding difficulties, compulsive video game and solitaire playing, and restrictive dieting. Her language was fluent with impoverished content, semantic paraphasic errors, nonspecific noun substitutions, and surface dyslexia and dysgraphia. Neuropsychological testing was notable for an MMSE of 29/30, with profound deficits in naming (3/15 on the BNT), and decreased semantic (five animals) relative to lexical (eight D words) fluency. Overall, her clinical and neuropsychological presentation was classic for SemD.⁹

FTLD-10 was a 63-year-old right-handed woman with 4.5 years of progressive word-finding difficulties, short-term memory loss, impaired judgment, loss of empathy, and disinhibited behavior. Her language was fluent with impoverished content, occasional semantic paraphasic errors, poor confrontation naming, and surface dysgraphia. Neuropsychological testing revealed an MMSE of 26/30, with deficits in episodic memory (0/9 10 minute free recall on the CVLT), verbal fluency (five animals and five D words per minute), and naming (10/15 on the BNT). This profile could be consistent with either SemD or AD, though naming performance was high for SemD.⁹ Both PIB-positive SemD patients showed left greater than right anterior temporal atrophy on MRI, and ¹⁸F-FDG PET (obtained 3.5 years [FTLD-9] and 3 months [FTLD-10] after presentation) demonstrated left greater than right anterior temporal hypometabolism (figure E-4). Both ¹⁸F-FDG scans were read by the blinded investigator as consistent with FTLD.

In summary, out of four patients clinically diagnosed with FTLD with positive ¹¹C-PIB scans, one patient had a neuropsychological profile more consistent with AD (albeit with profound behavioral disturbances), one had a classic profile for SemD, and two patients had cognitive profiles that could be consistent with either AD or FTLD. In both PIB-positive patients with FTD, ¹¹C-PIB and ¹⁸F-FDG agreed with each other in suggesting a diagnosis of AD, and disagreed with the clinical diagnosis of FTD. In contrast, both PIB-positive SemD patients had ¹⁸F-FDG scans consistent with FTLD.

Quantitative analysis of ¹¹C-PIB and ¹⁸F-FDG.

Scatterplots of ¹¹C-PIB whole brain and ROI DVR values are presented in figure 1. Differentiation of scans visually read as PIB-positive and PIB-negative provides a more informative representation of the data than simple clinical diagnosis. Patients with AD generally had higher DVR values than controls in the whole brain and in cortical ROIs. Patient AD-4, who qualitatively demonstrated restricted dorsal ¹¹C-PIB uptake (figure E-2), had an elevated DVR in lateral frontal cortex, whereas DVR values in other ROIs were similar to controls. There was greater overlap between groups in medial temporal cortex, where there is a relatively low burden of Aβ plaques in AD.⁴³ Patients with FTLD showed a bimodal distribution of DVRs in the whole brain and in cortical ROIs: patients with negative ¹¹C-PIB scans had DVRs that were comparable to controls, while patients with positive ¹¹C-PIB scans had DVRs that were similar to AD. As has been reported previously,¹⁸ we found intermediate levels of ¹¹C-PIB binding in both patients and controls in subcortical white matter and pons (latter not shown). This is thought to be due to nonspecific binding of ¹¹C-PIB to white matter,¹⁸ though the precise nature of this binding has not yet been determined.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Scatterplots of ¹¹C-PIB DVR values in specific regions of interest (ROIs), sorted by clinical diagnosis and visual interpretation. PIB(+) = visual read positive for cortical ¹¹C-PIB; PIB(−) = visual read negative for cortical ¹¹C-PIB; WM = white matter.

Comparison of mean DVR values using repeated measures ANOVA revealed main effects for both diagnosis [F(2,23) = 4.85, p < 0.05] and ROI [F(7,161) = 20.57, p < 0.001] with an interaction [F(14,161) = 3.88, p < 0.001]. Planned comparisons demonstrated elevated DVRs in AD vs FTLD in whole brain (p < 0.05), lateral frontal (p < 0.05), precuneus (p < 0.05), and lateral temporal (p < 0.05) ROIs. Similarly, DVRs in AD were higher than controls in whole brain (p < 0.01), lateral frontal (p < 0.01), precuneus (p < 0.01), and lateral temporal (p < 0.01) ROIs, with a non-significant trend in posterior cingulate (p = 0.08). Contrary to previous reports, we found higher DVRs in subcortical white matter in AD compared to controls (p < 0.05) and FTLD (p < 0.05), possibly due to cortical contamination of the white matter ROI as a result of the normalization procedure. DVRs in medial temporal cortex and pons were not significantly different between groups. FTLD DVR values did not significantly differ from controls in any of the ROIs.

Scatterplots of ¹⁸F-FDG normalized activity are presented in figure 2. Patients with FTLD are split by clinical subtype given the distinct patterns of ¹⁸F-FDG uptake typical of each FTLD variant.^40,44,45 As expected, patients with AD tended to have lower ¹⁸F-FDG uptake than patients with FTLD in the posterior cingulate and precuneus, and patients with SemD were most severely affected in the temporal lobes. There was considerable overlap between AD and FTLD in lateral frontal and medial temporal ROIs. Consistent with the visual interpretations, the two PIB-positive patients with FTD had low ¹⁸F-FDG uptake in posterior cingulate and precuneus. Interestingly, PIB-negative patients with SemD had lower ¹⁸F-FDG uptake in these AD-specific regions than PIB-positive SemD patients. Statistics were not performed on ¹⁸F-FDG normalized values given the small number of patients in each group once patients with FTLD were separated by clinical subtype.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Scatterplots of ¹⁸F-FDG activity (normalized to the pons) in specific regions of interest (ROIs), sorted by clinical diagnosis and ¹¹C-PIB visual interpretation. Frontotemporal lobar degeneration patients are divided into three clinical subtypes: frontotemporal dementia (FTD), semantic dementia (SemD), and progressive nonfluent aphasia (PA). PIB(+) = visual read positive for cortical ¹¹C-PIB; PIB(−) = visual read negative for cortical ¹¹C-PIB.