Cognitive reserve hypothesis: Pittsburgh Compound B and fluorodeoxyglucose positron emission tomography in relation to education in mild Alzheimer's disease




The reduced risk for Alzheimer's disease (AD) in high-educated individuals has been proposed to reflect brain cognitive reserve, which would provide more efficient compensatory mechanisms against the underlying pathology, and thus delayed clinical expression. Our aim was to find possible differences in brain amyloid ligand 11C-labeled Pittsburgh Compound B ([11C]PIB) uptake and glucose metabolism in high- and low-educated patients with mild AD.


Twelve high-educated and 13 low-educated patients with the same degree of cognitive deterioration were studied with PET using [11C]PIB and 18F-fluorodeoxyglucose as ligands. The between-group differences were analyzed with voxel-based statistical method, and quantitative data were obtained with automated region-of-interest analysis.


High-educated patients showed increased [11C]PIB uptake in the lateral frontal cortex compared with low-educated patients. Moreover, high-educated patients had significantly lower glucose metabolic rate in the temporoparietal cortical regions compared with low-educated patients.


Our results suggesting more advanced pathological and functional brain changes in high-educated patients with mild AD are in accordance with the brain cognitive reserve hypothesis and point out the importance of development of reliable markers of underlying AD pathology for early AD diagnostics. Ann Neurol 2007

Several epidemiological studies have found a lower incidence of Alzheimer's disease (AD) in high-educated populations, suggesting that education provides protection against the disease.1 This reduced risk for AD in high-educated individuals is proposed to reflect brain cognitive reserve that provides greater brain capacity to compensate for disruption caused by disease pathology, and thus delays the clinical expression of AD.2 At a particular level of AD pathology, highly educated individuals are less likely to manifest clinical symptoms of dementia as compared with less educated individuals.3 After the diagnosis of AD, high-educated individuals show more rapid progression of dementia and lower survival compared with low-educated AD patients.4–6 The different course of the disease between these groups has been hypothesized to reflect a more advanced AD pathology at the time of diagnosis in high-educated patients, and a rapid cognitive deterioration after the compensatory capacity becomes insufficient.

The association of educational level with the severity of brain damage in AD has been evaluated in vivo with positron emission tomography (PET) and single-photon emission computed tomography by using the brain glucose metabolism or cerebral blood flow as measures of brain functional changes.7–12 These studies have shown an inverse relation between the level of education and brain glucose metabolism or blood flow after adjusted for the dementia severity. This relation was evident especially in brain regions typically affected in AD, such as temporal and parietal cortices. The effect of education on the relation between brain AD pathology (ie, amyloid plaques and neurofibrillary tangles) and cognitive level has been evaluated in two postmortem studies.13, 14 In these studies, education was found to modify the relation of amyloid, but not tangle load, to cognitive performance, suggesting that brain cognitive reserve can partly compensate the effects of amyloid accumulation on cognition but fail to modify the effect of tangle pathology.

In vivo PET studies have shown an increased uptake of amyloid ligand 11C-labeled Pittsburgh Compound B ([11C]PIB) in AD and mild cognitive impairment patients, especially in the frontal, parietal, and temporal cortices and in the posterior cingulate, which indicates an increased amyloid accumulation in these areas.15–17 There are no previous studies on the association between education and [11C]PIB uptake in AD. Our objective was to study [11C]PIB uptake in high- and low-educated patients with mild AD using a voxel-based statistical analysis and to evaluate possible differences in amyloid pathology between these groups. Furthermore, as an indicator of brain functional changes, possible differences in 18F-fluorodeoxyglucose ([18F]FDG) uptake between the groups were also evaluated. Our hypothesis was that high-educated patients with mild AD have greater [11C]PIB uptake and lower [18F]FDG uptake, indicating more marked pathological and functional brain changes compared with low-educated patients with the same clinical disease stage.

Patients and Methods


All patients were diagnosed by an experienced neurologist (J.O.R.) according to the National Institute of Neurological and Communication Disorders-Alzheimer's Disease and Related Disorders Association criteria. All patients had progressive disease with impairment of memory and at least one additional field of cognitive function. The patients were cognitively assessed using a comprehensive neuropsychological test battery that included Consortium to Establish a Registry for Alzheimer's Disease, Wechsler Memory Scale–Revised, parts of Wechsler Adult Intelligence Scale–Revised, Trail Making Test and Stroop or Alzheimer's Disease Assessment Scale Cognitive subscale, Trail Making Test, clock drawing, copy of modified complex figure, digit span forward, Boston Naming Test, and fluency. The tests showed typical impairment in memory functions and in at least one separate cognitive function compared with age-appropriate norms and estimated premorbid levels. Results from central neuropsychological tests are presented in Table 1. We studied 13 low- and 12 high-educated AD patients. The 13 (4 men, 9 women) low-educated (elementary school, 6 years of formal education) patients with mild AD had a mean age of 71.0 (standard deviation, [SD] 7.0; range, 55–80) years. The mean Mini-Mental State Examination (MMSE) score in these patients was 25.0 (SD, 2.1; range, 22–29). Five patients were receiving cholinesterase inhibitor treatment (one rivastigmine, two galantamine, and two donepezil) at the time of the study, and one patient also received memantine. Eight patients in the low-educated group had never received AD medication, and cholinesterase inhibitor treatment was initiated after the PET studies. Apolipoprotein E genotype was analyzed in 12 of 13 low-educated patients with 10 apolipoprotein E ε4 carriers (83%). Furthermore, 12 (9 men, 3 women) high-educated (academic degree, at least 15 years of formal education) patients with mild AD with a mean age of 73.0 (SD, 3.6; range, 64–77) years were studied. In the high-educated group, five patients were receiving cholinesterase inhibitor treatment (three rivastigmine, two donepezil). The remaining patients had never received AD medication, and cholinesterase inhibitor treatment was initiated after the study. The mean MMSE score of the high-educated group was 26.0 (SD, 1.3; range, 24–28). There was no difference in the mean MMSE score (p = 0.23) or in the mean age between the low- and high-educated groups (p = 0.32).

Table 1. Neuropsychological Test Performances in the High- and Low-Educated Groups
inline image

[11C]PIB PET scanning was performed for all patients and FDG for all in the low-educated group. In the high-educated group, FDG-PET was performed in 11 in 12 patients with mean age of 73.2 years and mean MMSE score of 26.1. Patients underwent a magnetic resonance imaging, which was interpreted by an experienced neuroradiologist (R.P.). Magnetic resonance imaging showed central or cortical atrophy, or both, and hippocampal atrophy. No findings incompatible with the diagnosis of AD were found.

The study was approved by the Ethical Committee of Southwest Finland Health Care District. All patients gave their written informed consent. In cases where there was doubt that a patient was competent to give consent, it was obtained from the patient's next of kin.

Positron Emission Tomography Imaging

[11C]PIB was produced by the reaction of 6-OH-BTA-0 and [11C]methyl triflate, as reported earlier.16 The specific radioactivity of [11C]PIB at the time of administration was 32.5 (SD, 7.8) MBq/nmol. The radiochemical purity of the tracer was more than 98 % in all [11C]PIB studies. [18F]FDG was synthesized with an automatic apparatus by a modified method of Hamacher and colleagues.18 The radiochemical purity of the tracer was more than 95% in all [18F]FDG studies. In [11C]PIB studies, all patients underwent a 90-minute dynamic PET scan with a GE Advance PET scanner (General Electric Medical Systems, Milwaukee, WI) in the three-dimensional scanning mode as described earlier.16 [11C]PIB was injected into an antecubital vein as a bolus with a mean dose of 449MBq (SD, 92MBq) and flushed with saline. No blood sampling was performed during the scan. Imaging data were reconstructed into a 128 × 128 matrix using a transaxial Hann filter with a 4.6mm cutoff and an axial Ramp filter with an 8.5mm cutoff. In [18F]FDG imaging, the same scanner, scanning mode, patient positioning (orbitomeatal and sagittal lines), and reconstruction matrix were used as in [11C]PIB studies. Before scanning, two cannulas were placed in both antecubital veins, one for injection of [18F]FDG (3.7MBq/kg) and the other for blood sampling. Twenty-one arterialized blood samples were drawn from the preheated arm during the 55-minute study to measure plasma activity. Additional blood samples were obtained for plasma glucose concentration measurement before the tracer injection and 25 and 55 minutes after the beginning of the scan. The mean of these three measurements was used in data analysis.

Data Analysis

Before the voxel-based statistical analysis and automated region-of-interest (ROI) analysis, dynamic images were first computed into quantitative parametric images. Parametric images representing [11C]PIB uptake in each pixel were calculated as a region-to-cerebellum ratio of the radioactivity concentration over 60 to 90 minutes, as described earlier.16 Cerebral glucose metabolic rate (GMR) in each pixel in [18F]FDG images were calculated using tracer kinetic modeling and plasma activity data.19

Statistical Parametric Mapping Analysis

Voxel-based statistical analyses of [11C]PIB and [18F]FDG data were performed using Statistical Parametric Mapping version 99 (SPM99) and Matlab 6.5 for Windows (Math Works, Natick, MA), using procedures described in detail elsewhere.16, 20 In brief, spatial normalization of parametric images was performed using ligand-specific [11C]PIB and [18F]FDG templates.16, 20 The between-group comparison equaling two-sample t tests and testing the difference in ratio values was performed as an explorative analysis covering the whole brain. Multiple-comparison corrected p values less than 0.05 were considered significant.

Automated Region-of-Interest Analysis

To obtain quantitative regional values of [11C]PIB uptake and GMR, we performed automated ROI analysis as described earlier.16, 20 In brief, the ROIs were defined using Imadeus software (version 1.50; Forima, Turku, Finland) bilaterally on the frontal cortex, lateral temporal cortex, inferior parietal lobe, occipital cortex, hippocampus, parahippocampal gyrus, cerebellar cortex, and subcortical white matter.17 The average regional ratio values of [11C]PIB uptake and GMR were calculated using these ROIs from spatially normalized parametric ratio images (see earlier SPM analyses) and subjected to statistical analysis conducted using SPSS for Windows (1989-2001, release 12.0.1; SPSS, Chicago, IL).


Statistical Parametric Mapping Analysis

The between-group SPM analysis showed that high-educated patients with mild AD had significantly greater [11C]PIB uptake in the ventrolateral frontal cortex compared with low-educated patients (Fig, A). Furthermore, SPM analysis showed that high-educated patients had significantly lower GMR in the temporal and parietal cortices compared with low-educated patients (see Fig, B). There was no difference in GMR in the frontal cortex between the groups.

Automated Region-of-Interest Analysis

The ROI analyses supported the results of the SPM analyses and showed significant increases in [11C]PIB uptake in the ventrolateral frontal cortex in high-educated compared with low-educated patients. In this region, high-educated patients showed 14% greater [11C]PIB uptake values than the low-educated patients. High-educated patients had significantly lower GMR in the parietal cortex and a trend toward lower GMR in the lateral temporal cortex. High-educated patients showed less than 85% of the GMR in these areas compared with low-educated patients. The mean [11C]PIB uptake values, mean GMRs, and percentage differences in the studied brain regions are shown in Table 2.

Table 2. Results from Automated Region-of-Interest Analysis of 11C-Labeled Pittsburgh Compound B Uptake (Mean Region-to-Cerebellum Ratio at 60–90 Minutes [SD]) and Glucose Metabolic Rate (Mean [SD]) in High- and Low-Educated Patients with Mild Alzheimer's Disease and the Percentage of the High-Educated Mean of the Low-Educated Mean
inline image


This study demonstrates significantly increased brain uptake of the amyloid PET ligand [11C]PIB in the frontal cortex in high-educated patients with mild AD compared with low-educated AD patients at the same stage of the disease. Furthermore, high-educated patients showed significantly lower GMR in the temporal and parietal cortices compared with low-educated patients. Both groups were clinically diagnosed as having mild AD according to National Institute of Neurological and Communication Disorders-Alzheimer's Disease and Related Disorders Association criteria, and there was no between-group difference in the MMSE scores and Consortium to Establish a Registry for Alzheimer's Disease total scores corrected for age, education, and sex.21 Despite the lack of differences in the more global measures of general cognitive functioning, the high-educated group showed significantly better performances in verbal neuropsychological tests compared with the low-educated group. Tests of naming, verbal fluency, and verbal memory have been shown to be affected by educational attainment in older age,22 and thus may also be particularly resistant to decline in high-educated AD patients. Patients with high education may also gain an advantage by being more familiar with the kinds of tasks used in neuropsychological assessments. However, after AD incident faster decline in memory and executive speed is associated with higher education.23

The lower incidence of AD in high-educated populations has been shown in several epidemiological studies, but the protective mechanism of education against the disease is still unclear.1 Brain reserve hypothesis relies on the theory that more efficient neural networks may compensate the disruption caused by pathology, and thus delay the expression of the disease.2 Moreover, one explanation for the lower risk for AD in high-educated populations might be the different pattern of metabolic activity, that is, lower basal or default activity and greater specific activity in high-educated compared with low-educated individuals.24, 25 This pattern of mental activity has been hypothesized to decrease neuronal Aβ release and protect against AD because the presence of brain regions with high basal activity has been suggested to lead to greater Aβ release.24–26 These results indicate a greater amyloid pathology and more diminished brain metabolism in high-educated compared with low-educated AD patients and led to the proposal that education would provide cognitive reserve and later expression of the disease rather than protection against disease pathology. Accordingly, a functional magnetic resonance imaging study showed that in older adults, higher education is associated with greater and more widespread activation of the frontal cortex than low education, suggesting compensatory network engaged to aid cognitive function.27 The cross-sectional study design used in this study does not shed light on the development of brain pathological and functional changes during the presymptomatic phase of the disease. It is thus possible that the pathological changes in the high-educated group have started to develop earlier compared with the low-educated group, or that the brain changes in the high-educated group have developed more rapidly. The difference in cognitive reserve between individuals could result from genetic factors or from a lifelong mental stimulation due to education, or both. In addition to cognitive reserve hypothesis, there are also alternative explanations for the lower incidence of AD among highly educated people. Higher education is often associated with healthier lifestyle, fewer diseases, and lower exposure to toxic factors, and this could contribute to the differences between educational groups, as suggested for heart disease mortality.28 The brain reserve hypothesis implies that people with higher education have greater reserve capacity, and that greater pathological changes are needed for dementia to manifest itself. This reserve could be innate and people with greater reserve would be prone to have more education. The cognitive reserve hypothesis, however, emphasizes functional aspects such as more efficient and flexible usage of the existing neural network.2

Previous in vivo PET studies have shown increased [11C]PIB uptake in anterior and posterior cingulate; frontal, parietal, temporal and occipital cortices; and in the striatum and thalamus in AD and mild cognitive impairment patients compared with healthy control subjects.15–17, 29 Our results showed a regionally different pattern of [11C]PIB uptake increase and GMR decrease in high-educated compared with low-educated patients with increased [11C]PIB uptake in the frontal cortex and decreased GMR in the temporoparietal cortical regions in the high-educated group. These brain regions show the most prominent changes in [11C]PIB and [18F]FDG PET studies in AD as the greatest [11C]PIB uptake values and the most significant between-group differences have been reported in the frontal cortex and posterior cingulate.15, 16 On the other hand, the strongest decrease in GMR takes place in the temporoparietal cortical regions in AD patients.30 It has been suggested that regional [11C]PIB uptake indicating amyloid accumulation is not directly associated with glucose metabolism and neuronal function in the same region. Increased [11C]PIB uptake has been demonstrated especially in the frontal cortex, which shows preserved glucose metabolism but clearly increased [11C]PIB uptake in AD.29 Thus, decreased metabolism and neuronal dysfunction in certain brain regions might be mediated by other factors such as formation of neurofibrillary tangles, which spreads to frontal cortex relatively late during the disease.26

The study has certain limitations. The number of low- and high-educated groups is relatively small, reducing the power to detect differences between the groups. The patient groups were not perfectly matched regarding cognitive functions because the low-educated group did not perform as well as the high-educated group in some of the cognitive tests. This, however, would, if anything, underestimate the difference in [11C]PIB and [18F]FDG uptake between low- and high-educated groups. As an extension to this study one could in the future perform a study to explore the relation between education and [11C]PIB and [18F]FDG uptake in a population that would be more homogenous according to distribution of level of education.

Our results indicate more pronounced amyloid accumulation and neuronal dysfunction in high-educated compared with low-educated patients with mild AD. These results are in accordance with cognitive reserve hypothesis and delayed clinical expression of the disease in high-educated patients with marked AD pathology. The late appearance of clinical symptoms in these patients makes it challenging to detect the disease process in its early phase and points out that this group of patients would especially benefit from reliable markers of AD pathology. Possible candidates for such a marker are, for example, in vivo PET imaging of amyloid accumulation or glucose metabolism, hippocampal magnetic resonance imaging, biomarkers from cerebrospinal fluid, and detailed neuropsychological assessment. This becomes increasingly important when treatments affecting the disease progression become available.1

Figure 1.

Visualization of the results of Statistical Parametric Mapping (SPM) analysis. Regions with statistically significant increases (corrected p value at cluster level < 0.01) in [11C]PIB uptake (A) and decreases (corrected p value at cluster level < 0.01) in glucose metabolic rate (GMR) (B) in high-educated compared with low-educated patients are indicated with colors. The red-yellow scale indicates the level of statistical significance of the differences in [11C]PIB uptake (yellow indicates most significant difference).


This study was supported by Turku University Hospital (J.O.R.), the Maud Kuistila Foundation (N.M.K.), the Sigrid Juselius Foundation (J.O.R.), and the Research Council for Health of the Academy of Finland (J.O.R.; project #205954).

We thank the staff of the Turku PET Centre for technical assistance and GE Healthcare Medical Diagnostics for providing [11C]PIB for this study.