OUP user menu

Two-year follow-up of amyloid deposition in patients with Alzheimer's disease

Henry Engler, Anton Forsberg, Ove Almkvist, Gunnar Blomquist, Emma Larsson, Irina Savitcheva, Anders Wall, Anna Ringheim, Bengt Långström, Agneta Nordberg
DOI: http://dx.doi.org/10.1093/brain/awl178 2856-2866 First published online: 19 July 2006


Beta amyloid is one of the major histopathological hallmarks of Alzheimer's disease. We recently reported in vivo imaging of amyloid in 16 Alzheimer patients, using the PET ligand N-methyl[11C]2-(4′-methylaminophenyl)-6-hydroxy-benzothiazole (PIB). In the present study we rescanned these 16 Alzheimer patients after 2.0 ± 0.5 years and have described the interval change in amyloid deposition and regional cerebral metabolic rate for glucose (rCMRGlc) at follow-up. Sixteen patients with Alzheimer's disease were re-examined by means of PET, using PIB and 2-[18F]fluoro-2-deoxy-d-glucose (FDG) after 2.0 ± 0.5 years. The patients were all on cholinesterase inhibitor treatment and five also on treatment with the N-methyl-d-aspartate (NMDA) antagonist memantine. In order to estimate the accuracy of the PET PIB measurements, four additional Alzheimer patients underwent repeated examinations with PIB within 20 days (test–retest). Relative PIB retention in cortical regions differed by 3–7% in the test–retest study. No significant difference in PIB retention was observed between baseline and follow-up while a significant (P < 0.01) 20% decrease in rCMRGlc was observed in cortical brain regions. A significant negative correlation between rCMRGlc and PIB retention was observed in the parietal cortex in the Alzheimer patients at follow-up (r = 0.67, P = 0.009). A non-significant decline in Mini-Mental State Examination (MMSE) score from 24.3 ± 3.7 (mean ± standard deviation) to 22.7 ± 6.1 was measured at follow-up. Five of the Alzheimer patients showed a significant decline in MMSE score of >3 (21.4 ± 3.5 to 15.6 ± 3.9, P < 0.01) (AD-progressive) while the rest of the patients were cognitively more stable (MMSE score = 25.6 ± 3.1 to 25.9 ± 3.7) (AD-stable) compared with baseline. A positive correlation (P = 0.001) was observed in the parietal cortex between Rey Auditory Verbal Learning (RAVL) test score and rCMRGlc at follow-up while a negative correlation (P = 0.018) was observed between RAVL test and PIB retention in the parietal at follow-up. Relatively stable PIB retention after 2 years of follow-up in patients with mild Alzheimer's disease suggests that amyloid deposition in the brain reaches a plateau by the early clinical stages of Alzheimer's disease and therefore may precede a decline in rCMRGlc and cognition. It appears that anti-amyloid therapies will need to induce a significant decrease in amyloid load in order for PIB PET images to detect a drug effect in Alzheimer patients. FDG imaging may be able to detect a stabilization of cerebral metabolism caused by therapy administered to patients with a clinical diagnosis of Alzheimer's disease.

  • Alzheimer's disease
  • amyloid
  • PET
  • PIB
  • FDG
  • follow-up


The pathology of Alzheimer's disease has been coupled to an abnormal deposition of amyloid in the brain (the amyloid cascade hypothesis) (Hardy and Higgins, 1992). The regional evolution of Alzheimer's disease pathology in terms of amyloid deposition (amyloid plaques) and neurofibrillary tangles at several stages of the disease was described by Braak and Braak (1991), using post-mortem brain tissue. The distribution of amyloid plaques shows high variability in patients and has been difficult to correlate with the stage of the disease (Braak and Braak, 1991). Owing to the limitations of post-mortem brain studies, several attempts have been made to develop tracers for in vivo imaging of amyloid, and three compounds have been applied in PET studies in Alzheimer patients (Shoghi-Jadid et al., 2002; Klunk et al., 2004; Verhoeff et al., 2004). The finding that certain benzothiazol derivatives bind to amyloid with high affinity and cross the brain–blood barrier (BBB) led to the development of N-methyl[11C]2-(4′-methylaminophenyl)-6-hydroxy-benzothiazole (PIB) (Klunk et al., 2001; Mathis et al., 2002; Wang et al., 2002; Mathis et al., 2003). PIB binds amyloid with sufficiently high affinity to be detected in PET examinations. Animal studies have shown that PIB crosses the BBB and is cleared from normal brain tissue (Bacskai et al., 2003). The results of binding studies in autopsy-derived human brain tissue suggested that PIB might be a relevant ligand for the measurement of amyloid deposition but not for the detection of neurofibrillary tangles, since PIB mainly binds to aggregated fibrillar Aβ deposits (Klunk et al., 2003). In 2002–2003, the first study on measurement of amyloid in humans, based on the retention of PIB, was performed in Uppsala, Sweden, among 16 patients with mild Alzheimer's disease from Karolinska University Hospital Huddinge, Stockholm, Sweden (Klunk et al., 2004). PIB showed high retention in the frontal and temporal–parietal association cortices and striatum in these patients versus controls. The retention of PIB was similar in Alzheimer patients and healthy controls (HC) in brain regions known to be relatively unaffected by amyloid deposition, such as the pons, cerebellum and subcortical white matter (Klunk et al., 2004). A strong negative correlation was found between cerebral glucose metabolism [measured using 2-[18F]fluoro-2-deoxy-d-glucose (FDG)] and PIB retention in the parietal areas, suggesting a relationship between deficits in neural function and amyloid deposition (Klunk et al., 2004). In order to explore PIB further as a tracer, we have now performed a follow-up study with PIB, and FDG, in these 16 Alzheimer patients 1.5–2.5 years after the baseline study.



Follow-up study

The 16 patients with mild Alzheimer's disease previously recruited at the Department of Geriatric Medicine, Karolinska University Hospital Huddinge, Stockholm (Klunk et al., 2004), and fulfilling the diagnosis of probable Alzheimer's disease according to the criteria of the National Institute of Neurological and Communication Disorders, Alzheimer's Disease and Related Disorders Association (NINCDS–ADRDA) were clinically followed for a time period of 2.0 ± 0.5 years (1.5–2.5 years) and then re-examined, using PIB and FDG, at Uppsala Imanet, Uppsala, Sweden. The demography of the Alzheimer patients is presented in Table 1.

View this table:
Table 1


Age at baseline66.4 (±10.2)69.8 (±10.0)58.9 (±6.4)*
Age range at baseline51–8155–8151–66
Education (year)12.3 ± 3.713.6 ± 2.59.6 ± 4.8
MMSE at baseline24.3 ± 3.725.6 ± 3.121.4 ± 3.5
MMSE at follow-up22.7 ± 6.125.9 ± 3.715.6 ± 3.9#,**
RAVL at follow-up (Z-score)−2.60 ± 1.45−1.55 ± 1.36−3.35 ± 0.68***
Duration of disease (year)4.3 ± 1.54.4 ± 1.54.2 ± 1.6
Time from diagnosis to baseline (year)2.4 ± 1.82.4 ± 1.82.3 ± 1.9
APOE ɛ4-carrier853
CSF Aβ (pg/ml) at baseline372 ± 129401 ± 131326 ± 125
CSF tau (pmol/ml) at baseline599 ± 311441 ± 223883 ± 238**
Between scans (m)23.9 ± 5.623.2 ± 6.125.6 ± 4.4
Treatment at baseline11 on ChEls7 on ChEls3 on ChEls
Treatment at follow-up16 on ChEls 5 on memantine11 on ChEls 3 on memantine5 on ChEls 2 on memantine
  • AD = Alzheimer's disease; SD = standard deviation; MMSE = Mini-Mental State Examination; ChEI = cholinesterase inhibitor; AD-S = cognitively stable patients (MMSE decline <3); AD-P = patients with disease progression (MMSE decline >3); RAVL = Rey Auditory Verbal Learning test; *significant difference in age at baseline between AD-S and AD-P (Student's t-test; P < 0.05); #significant difference in MMSE score between baseline and follow-up (paired t-test; P < 0.01); **significant difference between AD-S and AD-P (Student's t-test; P< 0.05); ***significant difference between AD-S and AD-P (Student's t-test, P = 0.01).

Mini-Mental State Examination (MMSE) (Folstein et al., 1975) was used as measurement of global cognitive status and Rey Auditory Verbal Learning (RAVL) (Lezak, 2004) as test of episodic memory. For statistical analysis Z-scores of RAVL were generated on the basis of the raw scores, by comparing data with the mean value and standard deviation (SD) of a large control group.

Data from six healthy, age-matched controls (HC), studied at baseline, were utilized for comparisons (Klunk et al., 2004). The oldest healthy control (OHC) from the baseline study, who showed high PIB retention but normal cognitive performance in neuropsychological tests (Klunk et al., 2004), was also reinvestigated in the follow-up study.

The Alzheimer patients and the OHC gave written consent to participate in the study. The Ethics Committees of Uppsala University, the Karolinska Institute and the Isotope Committee at Uppsala Academic Hospital approved the study.


In order to study the variability in PIB retention, four additional Alzheimer patients (T1–T4) (3 females and 1 male) (58–79 years of age) (MMSE 9–28) were recruited from the Department of Geriatric Medicine, Karolinska University Hospital Huddinge, Stockholm, and they underwent repeated PET investigations with PIB. For three of the Alzheimer patients the two PIB studies were performed within 12 h and for the fourth patient, after a 20-day interval. The studies were performed under a separate study protocol with corresponding authorization as regards ethics. The changes in PIB retention were calculated for each patient by means of the formula %Difference = [(R−T)/(R+T)] × 200; (T = test, R = retest). The mean absolute percentage difference [Mean abs (%Diff)] was calculated for all four patients, yielding an interval of expected variance in the analysis of PIB retention (Price et al., 2005).


Production of FDG and PIB was carried out according to the standard good manufacturing process at Uppsala Imanet. Synthesis of N-methyl[11C]2-(4′-methylaminophenyl)-6-hydroxy-benzothiazole (PIB) was performed by means of the method described previously (Mathis et al., 2003; Klunk et al., 2004).

PET scanning

The PET scans were performed using Siemens ECAT EXACT HR+ scanners (CTI PET-systems Inc.), with an axial field of view of 155 mm, providing 63 contiguous 2.46-mm slices with 5.6-mm transaxial and 5.4-mm axial resolution. The patients were scanned after fasting for 4 h under resting conditions in a dimmed room. The orbitomeatal line was used to centre the heads of the subjects. The data were acquired in three-dimensional mode. The doses of PIB and FDG, and the scanner protocol for transmissions, emissions and reconstructions were the same as used in the previous study (Klunk et al., 2004). The subjects were given 287 ± 65 (mean ± standard deviation) MBq of 11C-PIB in the baseline study and 254 ± 86 MBq in the follow-up study. They received 231 ± 40 MBq of 18F-FDG in the baseline study and 226 ± 33 MBq in the follow-up study.

Regions of interest (ROIs)

The set of ROIs applied for statistical analyses in the follow-up and the test–retest study was the same as in the previous study, described in detail earlier (Engler et al., 2003; Klunk et al., 2004). The following areas were included in the analyses: the frontal, parietal, temporal, occipital and cerebellar cortices, pons, white matter and striatum and posterior cingulum.

A computerized reorientation procedure applied in the first study was used to align consecutive PET images for accurate intra- and inter-individual comparisons (Andersson and Thurfjell, 1997). The FDG images obtained in the follow-up study were realigned to the FDG images from the previous study, and the PIB images from both studies were co-realigned using the respective FDG images as templates. In addition, FDG images were analysed according to a clinical routine set of ROIs (Engler et al., 2003). In three of the Alzheimer patients the calculation of FDG uptake could not be performed according to the protocol in the follow-up study. In one patient this was a result of technical problems during acquisition and in the other two patients, a result of the fact that blood samples could not be obtained. For one of these patients, regional mean uptake values could be obtained, whereas no data concerning glucose uptake were available for the two other patients. MR images were not used to delineate ROIs, nor were such data used to do any partial volume correction.

Data management

For the FDG examinations, venous arterialized blood samples were obtained and parametric maps of the regional cerebral metabolic rate for glucose (rCMRGlc) were generated by means of the Patlak method, using the time course of the tracer in the arterialized venous plasma as an input function (Patlak et al., 1983). The rCMRGlc values were normalized to the pons value (ROI/ref) to allow inter- and intra-individual comparisons (Minoshima et al., 1995). In the earlier baseline PIB study (Klunk et al., 2004), PIB retention data were given as standard uptake values (SUVs). In the present follow-up study the mean uptake values of the ROI obtained in a late time interval (40–60 min) were normalized to the corresponding uptake in a reference region (ROI/ref) (Lopresti et al., 2005). The cerebellar cortex was chosen as reference because of its previously reported lack of Congo red- and thioflavin-S-positive plaques (Yamaguchi et al., 1989; Mirra et al., 1994). Among several different PIB evaluation methods the late scan reference method has shown a large size effect (Lopresti et al., 2005).

Statistical analysis

The statistical method used to compare the Alzheimer patients and the HC was a two-sample, unequal variance, two-tailed Student's t-test. Paired t-tests were used to study changes in PIB retention and rCMRGlc in the Alzheimer patients over the time period. Analysis of correlation between PIB retention and rCMRGlc parameters was conducted, yielding Pearson's product moment correlation coefficient r.

Role of the funding source

No funding source had a role in the preparation of this article or the decision to submit it for publication. The authors had full access to all data in the study and final responsibility for the decision to submit for publication.


Test–retest examinations

Four Alzheimer patients were recruited for a test–retest PIB study within 20 days. The individual results are shown in Table 2. Low variability in PIB retention was observed in the frontal cortex (7.3%), parietal cortex (4.1%), temporal cortex (3.2%) and occipital cortex (3.7%), while the highest variability was observed in the striatum (12.7%). Variability was also low in the pons (3.4%), white matter (6.4%) and the cerebellum (3.3%).

View this table:
Table 2

Changes in PIB retention (ROI/ref) in four Alzheimer patients undergoing repeated PET scans

% Diff−1.8−3.70.1−1.5−4.7−6.76.7−2.3
% Diff−6.6−1.5−3.7−2.2−−4.9
% Diff13.
% Diff−7.9−4.3−4.8−5.2−−0.3
Mean abs (% Diff)7.324.143.223.6512.73.356.393.29
SD %4.612.
  • Three of the patients underwent two PET investigations within 12 h (T1–3) while one Alzheimer patient (T4) underwent a second PET study after 20 days; AD 1 to 3 were retested within 12 h; AD 4 was retested after 20 days; percentage difference calculated: [(R−T)/(R+T)] × 200; ROI = region of interest; ref = reference; SD = standard deviation.

Clinical follow-up of Alzheimer patients

The 16 Alzheimer patients who had undergone baseline studies with PIB (Klunk et al., 2004) were clinically followed for 1.5–2.5 years and then they underwent the follow-up PET studies with PIB and FDG (Table 1). Eleven Alzheimer patients were being treated with cholinesterase inhibitors at the baseline PET scans, while all 16 patients were on cholinesterase inhibitor treatment at the time of the follow-up studies. In addition, 5 of the 16 patients were also being treated with the N-methyl-d-aspartate (NMDA) antagonist memantine at the time of follow-up (Table 1).

Cognitive testing showed a non-significant decrease in MMSE score of 1.6 points at follow-up compared with baseline. Five of the 16 patients showed a very clear clinical deterioration at follow-up, with a decrease of 3 points or more (3–9) in the MMSE test. In these five Alzheimer patients the mean MMSE score was significantly lower (P < 0.01) at follow-up [15.6 ± 3.9 (SD)] compared with baseline (21.4 ± 3.5). This group of patients was considered to comprise those who had clinically deteriorated (AD-P; progression) (Table 1). The change in MMSE score among the other 11 patients was <3 (−2 to +3) at follow-up versus baseline. This group of Alzheimer patients was considered to be clinically relatively stabile at follow-up (AD-S) (Table 1). The AD-S group (n = 10) differed significantly in RAVL test at follow-up (Z-score) compared with the AD-P group (n = 4), P < 0.01. The AD-P group was significantly younger than the AD-S group (P < 0.05). They showed significantly higher CSF tau values (P < 0.05) while there was no difference in CSF Aβ 1-42 values between the two groups (Table 1). Three of the 5 patients in the AD-P group carried two APO e4 alleles, while 5 out of 10 carried e4 alleles in the AP-S group (Table 1).

Regional PIB retention at follow-up

Significantly greater retention of PIB (expressed as ROI/ref) (P < 0.006) was observed at follow-up in the frontal, parietal, temporal and occipital cortices as well as in the striata of the Alzheimer patients versus age-matched HC (Fig. 1). No significant changes in PIB retention were observed between baseline and follow-up in any brain region of the Alzheimer patients as an overall group (Fig. 1). Retention of PIB was somewhat higher at both baseline and follow-up in the cortical brain regions and striata of the AD-P group versus the AD-S group. This difference was of statistical significance in the posterior cortex cinguli at baseline (P < 0.05) (Table 3). Owing to error in the scanning procedure at baseline, data for one Alzheimer patient (02) are not included in the statistical analysis.

Fig. 1

Comparison of PIB retention in the frontal cortex (Fr), parietal cortex (Par), temporal cortex (Temp), occipital cortex (Occ), striatum, pons, subcortical white matter (SWM) and cerebellum (Cb) between healthy controls (HC) and Alzheimer patients at baseline (AD 1) and follow-up (AD 2). Mean ± standard deviation. Number of Alzheimer patients = 15 and healthy age-matched controls = 6. Significant difference between groups (HC versus AD 1 and AD 2, respectively) indicated with *P < 0.05; **P < 0.01; ***P < 0.001; #significant difference between AD 1 and AD 2 (paired t-test; P < 0.05). Missing data for one patient with Alzheimer's disease owing to errors in the scanning procedure at baseline.

View this table:
Table 3

Changes in PIB retention (ROI/ref) in the frontal cortex (Fr ctx), parietal cortex (Par ctx), temporal cortex (Temp ctx), cingulum posterior (Cing post), occipital cortex (Occ ctx), striatum, pons, subcortical white matter (SWM) and cerebellum cortex (Cb ctx) in 10 Alzheimer's patients cognitively relatively stable at follow-up changes of less than 3 MMSE scores (AD-S) and 5 Alzheimer patients with progression in cognitive decline (change of 3 MMSE scores or more at follow-up) (AD-P)

AD-S (n= 10)AD-P (n= 5)
BaselineFollow-up (P-value)BaselineFollow-up (P-value)
Fr ctx1.86 ± 0.741.90 ± 0.72 (0.346)2.38 ± 0.352.32 ± 0.25 (0.684)
Par ctx1.81 ± 0.551.85 ± 0.55 (0.344)2.21 ± 0.272.18 ± 0.26 (0.790)
Temp ctx1.51 ± 0.551.52 ± 0.37 (0.740)1.82 ± 0.221.78 ± 0.19 (0.737)
Cing post1.84 ± 0.65*1.88 ± 0.63 (0.480)2.43 ± 0.33*2.29 ± 0.32 (0.441)
Occ ctx1.49 ± 0.411.54 ± 0.45 (0.291)1.89 ± 0.332.02 ± 0.39 (0.057)
Striatum1.97 ± 0.691.92 ± 0.69 (0.292)2.23 ± 0.302.35 ± 0.34 (0.271)
Pons1.81 ± 0.241.83 ± 0.32 (0.828)1.64 ± 0.221.65 ± 0.22 (0.940)
SWM1.73 ± 0.311.69 ± 0.26 (0.642)1.72 ± 0.291.71 ± 0.21 (0.923)
Cb ctx1.09 ± 0.081.07 ± 0.12 (0.668)1.06 ± 0.081.08 ± 0.08 (0.051)
  • Mean values ± standard deviation; probabilities were calculated by using Student's paired t-test; S = stable/normally progressing Alzheimer patients; P = faster progressing Alzheimer patients; *significant difference between AD-S and AD-P group (P < 0.05); missing data for one patient owing to errors in the scanning procedure at baseline.

Regional FDG uptake at follow-up

Analysis of rCMRGlc in the entire cohort according to routine clinical standard assessment revealed significant decreases in rCMRGlc (ROI/ref) in several cortical brain regions (Table 4). Both the AD-P group and the AD-S group showed significant decreases in rCMRGlc in various cortical brain regions at 2.0 ± 0.5 years of follow-up versus baseline (Table 4). Cortical rCMRGlc values were in general lower in the AD-P group both at baseline and follow-up versus the AD-S group (Table 4). Owing to technical problems at follow-up data are missing for two of the patients (08, 20). For one Alzheimer patient (09) SUV values were generated. When these three patients (8,20,9) were excluded the statistical analysis were based upon 13 Alzheimer patients.

View this table:
Table 4

Changes in regional cerebral glucose metabolism (rCMRGlc; μmol/min/100 ml) in different cortical brain regions, caudate nucleus, whole brain and pons in nine Alzheimer patients cognitively relatively stable at follow-up and in four patients showing cognitive decline

BaselineFollow-up (P-value)AD-stable (AD-S) n = 9AD-progressive (AD-P) n = 4
BaselineFollow-up (P-value)BaselineFollow-up (P-value)
Cing post dx33.98 ± 10.5127.65 ± 10.28 (0.0019)38.29 ± 9.5031.30 ± 10.27 (0.0145)24.29 ± 4.56**19.44 ± 3.42** (0.0448)
Cing post sin35.67 ± 11.1329.45 ± 9.38 (0.0019)40.05 ± 10.6333.25 ± 9.27 (0.0556)25.81 ± 2.70**20.69 ± 3.10** (0.0564)
Fr ctx dx35.48 ± 10.7628.71 ± 7.74 (0.0081)37.75 ± 12.0930.85 ± 7.32 (0.0500)30.36 ± 4.7823.91 ± 7.23 (0.0700)
Fr ctx sin36.16 ± 8.4430.36 ± 7.56 (0.0069)38.42 ± 9.3033.18 ± 4.25 (0.061431.67 ± 2.11*24.02 ± 6.87 (0.0592)
Par ctx dx31.49 ± 8.7625.67 ± 8.87 (0.0091)35.31 ± 6.4329.26 ± 8.22 (0.0486)22.90 ± 7.38*17.61 ± 3.09** (0.0964)
Par ctx sin33.67 ± 7.3827.41 ± 8.10 (0.0041)36.65 ± 6.4830.65 ± 7.66 (0.0430)26.96 ± 4.36**20.11 ± 1.83** (0.0964)
Par temp dx30.44 ± 6.4724.43 ± 8.96 (0.0058)32.81 ± 6.1127.77 ± 8.75 (0.0720)25.09 ± 3.54*16.90 ± 2.81** (0.0273)
Par temp sin32.84 ± 7.1826.01 ± 9.01 (0.0037)35.08 ± 7.5629.29 ± 8.71 (0.0503)27.78 ± 2.06*18.62 ± 4.12** (0.0356)
Caudate nucleus dx41.55 ± 8.4936.50 ± 6.18 (0.0941)43.47 ± 9.5937.36 ± 5.87 (0.1417)37.25 ± 2.5034.58 ± 7.34 (0.5180)
Caudate nucleus sin42.39 ± 6.8933.93 ± 6.59 (0.0042)43.92 ± 7.7535.05 ± 7.69 (0.0320)38.95 ± 2.6631.41 ± 1.90 (0.0342)
Whole brain32.47 ± 4.5727.25 ± 4.65 (0.0051)35.42 ± 4.9628.31 ± 5.01 (0.0453)30.34 ± 3.0424.85 ± 2.93 (0.0400)
Pons24.20 ± 4.0322.54 ± 3.32 (0.2336)24.63 ± 4.6622.16 ± 3.90 (0.2138)23.26 ± 2.3023.41 ± 1.85 (0.9030)
  • There were technical problems with the 18F-FDG investigations in two subjects in the AD-S group and one subject in the AD-P group (data not included); mean values ± standard deviation; P-values (paired t-test) indicate difference between baseline and follow-up; *significant difference between AD-S and AD-P groups (Student's t-test); P < 0.05; **P< 0.01. Bold indicates significant P values of 0.05 and less.

Comparison of changes in regional PIB retention and FDG uptake at follow-up

In order to compare the changes over time in PIB retention and rCMRGlc, we focused our analysis on the parietal region, where a significant correlation between rCMRGlc and PIB was present both at baseline (Klunk et al., 2004) (Fig. 2A) (P = 0.002) and at follow-up (Fig. 2B) (P = 0.009). Low rCMRGlc values were defined as those >1 SD below the mean value for HC and high PIB values as those >1 SD above the mean value for the HC subjects.

Fig. 2

Correlation between rCMRGlc (ROI/ref) and PIB retention (ROI/ref) in the parietal cortex of individual Alzheimer patients at baseline (A) and at follow-up (B). Dotted lines indicate mean values for the healthy controls plus 1 SD for PIB (mean = 1.35) and minus 1 SD for rCMRGlc (mean = 1.47). Open circles represent Alzheimer patients cognitively relatively stable, with changes in MMSE score of <3.0 at follow-up (AD-S). Filled circles represent Alzheimer patients who deteriorated cognitively with a decrease in MMSE score of ≥3.0 at follow-up (AD-P). Only SUV rCMRGlc values were available for one patient (09) indicated with filled squares. Activity in the parietal cortex was normalized to that in the pons as regards glucose uptake.

At follow-up (Fig. 2B), most of the Alzheimer patients showing changes of <3 in the MMSE score, and high PIB retention at baseline, showed only slight changes in PIB ROI/ref values at follow-up, and a slight decrease in rCMRGlc. The five patients with low MMSE scores at baseline and a decrease in MMSE score of 3 or more at follow-up showed clearly lower rCMRGlc values (Fig. 2B, filled circles). Parametric images of one of these five Alzheimer patients (02), based on PIB (ROI/ref) and rCMRGlc (ROI/ref) ratios at baseline and follow-up, are presented in Fig. 3.

Fig. 3

Parametric images concerning PIB (ROI/ref) and rCMRGlc (ROI/ref). The Alzheimer patient showed deterioration in cognition at follow-up (7 points in MMSE). The images show only slightly increased PIB retention (upper) but a pronounced decrease in glucose uptake (lower).

Three Alzheimer patients (04, 12, 14; Fig. 2B) with high MMSE scores and low PIB retention at baseline (Klunk et al., 2004) showed unchanged MMSE scores and low PIB retention at follow-up, although some increase in cortical PIB retention was observed in all three patients. Patient 14 showed a slight increase in PIB retention in the frontal and parietal cortices and Patient 04 showed a slight increase in PIB retention in the frontal and temporal areas with somewhat decrease in rCMRGlc but within the range for HC.

A fourth Alzheimer patient (05, an 82-year-old man, APOE ɛ2/4 carrier) showed relatively unchanged PIB retention, at the lower level for these patients (Fig. 2B). His MMSE score (20 out of 30) at follow-up was similar to that at baseline although his relatives were complaining of increasing problems in coping with daily activities of living.

The OHC, who showed high PIB retention at baseline (Klunk et al., 2004), showed unchanged PIB retention and rCMRGlc in the frontal cortex and striatum at follow-up. A slight increase in PIB retention was observed in the temporal and parietal cortices at follow-up. The glucose uptake was in these brain regions unchanged or slightly increased, respectively. The observed changes in PIB retention and rCMRGlc were <1 SD compared with test–retest data. The OHC showed quite normal cognitive performance in neuropsychological test at follow-up.

Correlation between cognition, PIB retention and rCMRGlc

Correlation analysis was performed to investigate the relationship between cognitive status, measured by MMSE score, RAVL test, PIB retention and rCMRGlc. The MMSE score showed significant negative correlation with PIB retention at baseline in three areas: the frontal cortex (r = −0.64; P = 0.010), parietal cortex (r = −0.60; P = p.018) and occipital cortex (r = −0.56; P = 0.028). There were, however, no significant correlations between MMSE score and PIB retention at follow-up. Changes in MMSE score and PIB retention expressed as percentages of baseline values did not show significant correlation.

Significant correlation was observed between MMSE score and rCMRGlc both at baseline and at follow-up. Two brain regions showed significant positive correlations at baseline, the parietal cortex (r = 0.56; P = 0.005) and temporal cortex (r = 0.51; P = 0.044). Four areas showed significant positive correlation at follow-up, namely the frontal cortex (r = 0.74; P = 0.002), parietal cortex (r = 0.79; P = 0.001), temporal cortex (r = 0.56; P = 0.036) and cerebellar cortex (r = 0.64; P = 0.013). A significant positive correlation was also observed between percentage change in MMSE score and percentage change in rCMRGlc in the parietal cortex (r = 0.59; P = 0.027).

The RAVL test score (expressed as Z-score) showed significant negative correlation with PIB retention at follow-up in four areas: the frontal cortex (r = 0.67; P = 0.009), parietal cortex (r = 062; P = 0.018), cingulum posterior (r = 0.64; P = 0.014) and striatum (r = 0.62; P = 0.018) (Fig. 4A).

Fig. 4

Correlation between RAVL and PIB retention (ROI/ref) (A) and RAVL and rCMRGlc (ROI/ref), respectively (B), in the parietal cortex of individual Alzheimer patients at follow-up. Open circles represent Alzheimer patients cognitively relatively stable, with changes in MMSE score of <3.0 at follow-up (AD-S). Filled circles represent Alzheimer patients who deteriorated cognitively with a decrease in MMSE score of ≥3.0 at follow-up (AD-P) only SUV rCMTglc values were obtained for patient 9. Patient 9 is indicated as filled squares in both A and B. Activity in the parietal cortex was normalized to that in the pons as regards glucose uptake. Z-scores of RAVL were generated from raw scores by comparing data with the mean value and standard deviation of a large control group.

Significant positive correlation was observed between RAVL test score (expressed as Z-score) and rCMRGlc at follow-up in seven areas, the frontal cortex (r = 0.77; P = 0.004), parietal cortex (r = 0.85; P = 0.001), temporal cortex (r = 0.76; P = 0.005), cingulum posterior (r = 0.90; P = 0.001), occipital cortex (r = 0.68; P = 0.015), striatum (r = 0.69; P = 0.012) and cerebellum (r = 0.62; P = 0.031) (Fig. 4B).


This study represents a 1.5–2.5-year follow-up with the amyloid PET ligand PIB, and FDG, in a group of 16 Alzheimer patients. When re-examined, the Alzheimer patients, as earlier, showed significantly higher retention of PIB in several cortical brain regions and the striatum compared with age-matched HC. No significant change in regional PIB retention was observed at follow-up versus baseline. The group, however, was somewhat heterogeneous, showing small variations (both increase/decrease) in PIB retention. The number of patients examined in the test–retest study is too small to give a definitive answer as to the accuracy of the method, and studies with larger numbers are needed. Similar variations in test–retest PIB retention, using the reference Logan model, have recently been reported (Price et al., 2005).

Interestingly, PIB retention remained unchanged despite the fact that the Alzheimer patients showed a decrease in rCMRGlc and a subgroup showed significant clinical deterioration (AD-P) at follow-up. This PIB retention was generally higher in the AD-P group than in the Alzheimer patients with less progression of disease at follow-up (AD-S group). The cortex cinguli posterior, which showed the highest PIB retention in the AD-P group, significantly differed from the corresponding brain area in the AD-S group. It seems also that the patients with the highest parietal PIB retention (both at baseline and follow-up) showed the largest decrements in rCMRGlc. A larger number of Alzheimer patients are needed to estimate the significance of these observations.

The results of the present study reveal relatively stable PIB retention in patients with mild Alzheimer's disease over a mean time period of 2 years (1.5–2.5 years) despite progressive deterioration in rCMRGlc as well as a clear decrease in cognitive function in some cases. This relatively stable PIB retention, with only minor variations (increase/decrease), might reflect a dynamic process in amyloid deposition reaching an equilibrium. The results are in agreement with those in earlier human post-mortem studies showing a dynamic balance between amyloid deposition and resolution in senile plaques, or amyloid burden (Hyman et al., 1993). An in vivo multiphoton microscopy study of thioflavin-S-positive senile plaques in the Tg2576 transgenic mouse model of Alzheimer's disease did not reveal detectable changes in plaque size over extended periods (Christie et al., 2001). Nonetheless, the authors described rare examples of growth or shrinkage of individual plaques, and the appearance of new plaques between imaging sessions, and they suggested that glial interaction with amyloid stabilizes the size of plaques and prevents continued enlargement. The dynamic changes over time of the amyloid depositions in the brain may produce variation in the binding properties of amyloid to the tracer. Structural changes in the evolution of the plaques, with an increment in the number of neuritic plaques, could, for example, result in a slight decrease in the binding of PIB. Furthermore, concomitant inflammatory processes also influence the binding properties (Su et al., 1996). Our results suggest that the accumulation of fibrillar amyloid may increase to a certain level and then become relatively stable, while degeneration of the neurons continues. It is quite possible that the amyloid plaques might be growing but that only a relatively stable part is accessible to exogenous tracer binding. It was recently suggested that there are three distinct binding sites for thioflavin compounds on β-amyloid peptide fibrils (Lockhart et al., 2005). Further PET multi-tracer binding studies involving visualization of not only amyloid but also inflammatory processes, microglial activation and perhaps also neurotransmitter activity might provide further insight into these ongoing processes in the brains of patients with Alzheimer's disease. The fact that high cortical PIB retention can be observed in patients with only mild cognitive impairment (Lopresti et al., 2005) indicates that further studies with PIB should now be performed among subjects at a genetically high risk of developing Alzheimer's disease.

Since all patients with Alzheimer's disease in this study were on cholinesterase inhibitor treatment (and some also on memantine treatment) at the follow-up PET studies we cannot exclude the possibility of interactions between cholinesterase inhibitor treatment and amyloid deposition (Ballard et al., 2005; Inestrosa et al., 2005). The results of several PET studies have revealed stabilizing effects on cognition, rCMRGlc and cerebral blood flow (CBF) after long-term treatment with cholinesterase inhibitors (Nobili et al., 2002; Stefanova et al., 2003; Tune et al., 2003; Stefanova et al., 2005). Experimental data also suggest that cholinesterase inhibitors might influence amyloid deposition (Francis et al., 2005). Only a placebo-controlled study analysing the effect of cholinesterase inhibitor treatment on PIB retention can rule out this possibility. The relatively low decrease in MMSE score measured at follow-up might be expected to be larger if the Alzheimer patients had not been on cholinesterase inhibitor treatment (Birks, 2005). The patients in this study represented those with mild Alzheimer's disease, for whom it is known that the MMSE is a relatively insensitive measure of cognitive status. An extensive evaluation of the outcome of consecutive neuropsychological tests in this patient group is ongoing and will be presented later. In the present study we presented the outcome of the episodic memory test, RAVL test, at follow-up.

The three patients with low PIB retention and normal rCMRGlc presented in the baseline study (Klunk et al., 2004) still showed low PIB retention and normal rCMRGlc at follow-up. Their MMSE scores at follow-up were still high and the RAVL test Z-scores were low. These patients may have very slow development of the disease with a slight or no increment in amyloid deposition over time. In addition to showing impairment in neuropsychological testing, these subjects also showed deficits in cortical rCMRGlc before initiation of cholinesterase inhibitor treatment. One of the subjects showed a pathologically low CSF Aβ 1-42 level (365 pmol/ml) but a low CSF tau value (75 pmol/ml) This illustrates the complexity in findings, as also recently described in CSF data by Iqbal et al. (2005). These three patients should today be considered as mild cognitive impairment patients. We can conclude that in two consecutive PIB PET studies we have measured consistently low cortical PIB retention and unchanged MMSE values in these subjects. Their clinical symptoms may have a different pathological basis than Alzheimer's disease.

The OHC who showed high PIB retention and relatively well-preserved rCMRGlc at baseline (Klunk et al., 2004) still showed high PIB retention at follow-up but normal cognitive performance in testing. This finding illustrates and supports the assumption that amyloid deposition can be detected in the brains of elderly subjects without cognitive deficits (Schmitt et al., 2000). Such amyloid deposition might solely be related to ageing. We cannot exclude the possibility that it may lead to cognitive symptoms in future years.

The PIB retention data in this study are presented as ROI/ref values for 60 min of scanning time. Sixty minutes were chosen in order to be able to compare follow-up data with baseline data. The cerebellum was chosen as reference region. The influence of the time interval chosen for late scan ratio has recently been investigated by Lopresti et al. (2005). PIB data from Alzheimer patients as well as control subjects were used to compare the intervals 40–60 and 60–90 min. The 60–90 min interval gave slightly better values than the 40–60 min interval in the frontal cortex, slightly worse values in the mesial-temporal cortex and the same values in the posterior cingulate gyrus (Lopresti et al., 2005). In the same study, several methods of analysis of PIB data were compared, and no method was better than the late scan ratio. An important question is whether changes in CBF in patients with Alzheimer's disease might influence the measured PIB retention in cortical brain regions of Alzheimer patients. We tested this crucial question recently in a monkey model (Blomquist et al., 2005). The effect of changes in CBF on the late target to cerebellum ratio of PIB was tested. An increase in CBF of 50–80% caused an increase in target to reference of ∼10% in the monkey. Since PIB redistribution has already taken place at 60 min, when PIB retention is measured, CBF changes in Alzheimer patients will probably have only a minor influence on the late scan ratio of PIB.

The results of this study show, for the first time, that amyloid deposition in vivo as measured by using the PET ligand PIB remains high but stable despite further decreases in cortical rCMRGlc and cognitive function. It is possible that the increase in amyloid deposition in the brains of Alzheimer patients represents a dynamic pathological brain process reaching equilibrium or plateau very early in the course of Alzheimer's disease. The time course for increasing PIB retention might be different from ongoing processes regarding cortical hypometabolism, which might be more related to cognitive function. This was also reflected in the observation of a significant correlation between MMSE score and rCMRGlc both at baseline and follow-up. This indicates a link between these two parameters that seems to be stronger than with PIB. This assumption was also strengthened by the stronger positive correlation between RAVL Z-score and rCMRGlc in the parietal cortex compared with the negative correlation between RAVL and PIB retention in the parietal cortex at follow-up. The combination of PIB and rCMRGlc in a dual tracer application might increase understanding and interpretation of the underlying pathological mechanisms in Alzheimer's disease. The small changes in PIB retention observed in patients with mild Alzheimer's disease over 2 years strongly suggest that PIB might be a suitable tracer in studies involving evaluation of the outcome of drug treatments with a focus on amyloid clearance. Further studies are necessary to enlarge the patient material and to study the time course of Alzheimer's disease, including pre-symptomatic carriers. In addition, other suitable PET tracer combinations must be explored in order to understand better the in vivo pathological changes in Alzheimer's disease.


Financial support from the Swedish Medical Research Council (project 05817, A.N.), the Foundation for Old Servants (A.N.), the Stohnes foundation (A.N.), the KI foundation (A.N.), the Swedish Brain Power (A.N., B.L.) and the EC-FP5-project NCI-MCI, QLK6-CT-2000-00502 (A.N.) is gratefully acknowledged. We thank all the patients who have participated in this study as well as their relatives. We thank the staff at the Department of Geriatric Medicine, Karolinska University Hospital Huddinge and Uppsala Imanet for their dedication and high level of professionalism in performing these studies and Mrs Marianne Grip for professional help with preparation of the manuscript.


View Abstract