Cerebral Glucose Metabolic Features of Parkinson Disease and Incident Dementia: Longitudinal Study

Abstract

Longitudinal studies in non-demented Parkinson disease (PD) subjects offer an opportunity to study the earliest regional cerebral subcortical and cortical metabolic changes underlying incident dementia in this disorder.

Methods:

Twenty-three PD subjects without dementia (Hoehn and Yahr stages I - III, age 61.8 ± 9.7 yr; Mini Mental State Examination 28.0 ± 1.4) and twenty-seven healthy control (HC) subjects (age 59.8 ±11.5 yr) underwent [¹⁸F]-2-fluoro-2-deoxy-D-glucose (FDG) PET imaging at study entry. PD subjects underwent yearly clinical assessment to determine conversion to dementia. The mean duration of follow-up was 3.9 ±1.2 yr (2.0 - 6.8 yr). Follow-up ¹⁸F-FDG-PET was available in a subset of subjects at 2 or more years. Both volume-of-interest and three-dimensional stereotactic surface projections (3D-SSP) analyses were performed.

Results:

Six subjects became demented (PD[+]D) with a mean time to development to dementia of 3.8 ± 1.7 yr (range: 1.9 - 6.0 yr). Mean duration of disease prior to onset of dementia was 9.7 ± 4.2 (range 3.1 - 14) yr. There were significant metabolic reductions in the occipital (−11.8% vs. HCs, F(2,22)= 7.0, P=0.002) and posterior cingulate (−12.1% vs. HCs, F=5.2, P=0.009) cortices in PD[+]D subjects at baseline, before diagnosis of dementia, compared with HCs. Metabolism was most diminished in the visual association cortex (Brodmann Area [BA] 18; −20.0% vs. HCs, F(2,22)=8.45, P=0.0007). There was mild hypometabolism in the caudate nucleus (−8.4% vs. HCs, F(2,22)=3.2, P<0.05). There was no significant hypometabolism in the temporal or frontal lobes. PD subjects who did not become demented (PD[−]D), compared with HCs, had reduced cerebral metabolism in the primary occipital cortex (BA 17) that was revealed only by 3D-SSP analysis. Follow-up scans in 5 PD[+]D subjects at 2-years after study entry demonstrated significant interval within-subject change: thalamus (−11.4%), posterior cingulate (−9%), occipital (−7%), parietal (−7%), and frontal cortices (−7%), and mild reductions in temporal cortex (−5%) and hippocampus (−3%) compared to study entry scans.

Conclusion:

Incident dementia in idiopathic PD is heralded by decreased metabolism in the visual association (BA 18) and posterior cingulate cortices, with mild involvement also of the caudate nucleus. Two-year follow-up data from 5 PD[+]D converters show that progression to dementia is associated with mixed subcortical and cortical changes that involve the mesiofrontal lobes also. These findings provide insights into early metabolic features of parkinsonian dementia.

INTRODUCTION

Impaired cognition is a frequent finding in individuals with Parkinson disease (PD), and ultimately more than 75% of PD patients will become demented (PD[+]D) (1). Reported pathologic changes in PD[+]D include cortical involvement as evidenced by the presence of Lewy body or Alzheimer pathology and subcortical changes, including dopaminergic nigrostriatal neuron degeneration, locus ceruleus neuron loss, and loss of cholinergic neurons in the nucleus basalis of Meynert and pedunculopontine nucleus (2). Cerebral metabolic changes are known to occur in advance of clinically manifest dementia in Alzheimer disease (AD). Using [¹⁸F]fluoro-deoxyglucose-positron emission tomography (¹⁸F-FDG-PET), Minoshima et al. first reported decreased glucose metabolism of the posterior cingulate cortex in Alzheimer disease (AD) (3). Reduced metabolism of the posterior cingulate and cinguloparietal transitional area is observed also in subjects with isolated forgetfulness, before a clinical diagnosis of probable AD can be made and also in clinically normal individuals homozygous for the apolipoprotein E e4 (APOE4) allele (4). Posterior cingulate hypometabolism is present in subjects with PD[+]D (5), but it is unclear whether posterior cingulate or subcortical hypometabolism may herald the onset of dementia in PD.

Longitudinal studies in nondemented PD (PD[−]D) subjects offer an opportunity to study the metabolic pattern of incident dementia. These studies also provide precious opportunities to study the early stage of dementia in Lewy body disease including PD[+]D as well as dementia with Lewy bodies (DLB), the second most common neurodegenerative dementia among elderly people (6). In this prospective cohort study, we evaluated subcortical and cortical metabolic changes in PD subjects who developed incident PD[+]D. ¹⁸F-FDG-PET was obtained for each subject at the entry to the study. We hypothesized that PD[+]D subjects subsequently would have lower regional (particularly, posterior cingulate cortex and parietal association) cortical glucose metabolic activity at baseline compared to PD[−]D subjects and healthy controls (HC). We also analyzed the interval metabolic changes between baseline and follow-up scans in a subset of PD[+]D converters. A single subject with incident dementia had four ¹⁸F-FDG PET scans performed over a period of 5 years.

SUBJECTS AND METHODS

Subjects

This prospective cohort study involved 50 subjects (23 PD subjects and 27 HC). PD subjects had a mean age of 61.8 ± 9.7 yr (range: 43 −79 yr; 17 male, 6 female). At study entry (7), Nine subjects were at Hoehn and Yahr stage I, 12 stage II, and 2 stage III, and all were receiving levodopa treatment. Subjects underwent neuropsychological testing for assessment of dementia at baseline and at subsequent yearly intervals. Subjects were considered demented if they met the research criteria established for the diagnosis in this population assessed by board-certified neurologists in consensus with a neuropsychologist on the basis of criteria from the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R) and the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimers Disease and Related Disorders Association (NINCDS-ADRDA), which reflect a global decline in cognition and impairments in socio-occupational functioning (8, 9). A minimum follow-up period of two years was required for inclusion in the study. PD subjects had a mean Mini-Mental State Examination (MMSE) score of 28.0 ± 1.4 (range 24 - 30) at study entry (10). Demographic information collected included age, duration of education, and age of onset of PD symptoms. Duration of PD was calculated based on the subject’s age at which motor symptoms began.

Twenty-seven age-comparable HCs (age 59.8 ± 11.5 yr; range 41 - 84 yr; 12 male, 15 female) without history of neurologic, psychiatric, or major medical diseases were also included in the study (MMSE, 28.1 ± 1.1; range 26 - 30). Results of the neurologic examinations were confirmed as normal on the day of PET. HCs did not undergo follow-up assessments.

Clinical Assessment

Clinical assessment and a brief neuropsychological screening battery were used at yearly intervals to determine cognitive and dementia status in the PD subjects. The cognitive measures included verbal fluency (VF) (11), Wechsler Adult Intelligence Scale – III Picture Arrangement subtest (PARR) (12), Benton Visual Retention Test Revised (BVRT) (11), and Wechsler Memory Scale III Visual Reproduction subtest (WMS-VI: immediate learning and WMS VI-PREC: percent delayed recall) (13).

PET Cerebral Glucose Metabolic Imaging

PET images were obtained using an ECAT-931 scanner (model 931/08-12; Siemens Medical Systems, Inc.) for 63 scans (37 in PD and 26 in HC subjects) or an ECAT EXACT scanner (CTI) for 16 scans (15 in PD subjects and 1 HC). A transmission scan was obtained using a ⁶⁸Ga/⁶⁸Ge pin source for attenuation correction. Emission scans were acquired in two-dimensional mode, with a total axial field of view of either 10.5 cm (ECAT-931) or 16.2 cm (ECAT EXACT). PET studies were performed with the subjects under resting conditions, with eyes open and ears unplugged, lying comfortably in a dimly-lit and quiet room. All subjects fasted for at least 4 h before PET. Thirty minutes after injection of ¹⁸F-FDG (370 MBq), a sequence of three 10-min frames was acquired and later summated into a single frame. Images were reconstructed by filtered backprojection with a Hanning filter (cutoff frequency at 0.5 cycles per projection element) for both scanners. The dimensions of the reconstructed PET images were 128 x 128 with a pixel size of 1.99 x 1.99 mm and an interslice distance of 3.375 mm.

Data Analysis

The quantitative parametric images were transformed to the bicommissural stereotactic coordinate system using a method described previously (14, 15). Differences in individual brain sizes were removed by linear scaling, and regional anatomic differences were minimized by a non-linear warping technique (15). Subsequently, gray matter activities were extracted to a standard set of pixels covering the entire brain surface using three-dimensional stereotactic surface projections (3D-SSP) (16). This cortical data extraction technique compensates for small anatomical differences in gray matter structures (such as variable depth of gyri) across subjects and minimizes partial volume artifacts and effects of atrophy.

Because ¹⁸F-FDG PET data were obtained on two different scanners, the primary analysis was based on a volume-of-interest (VOI) analysis and complemented by pixel-by-pixel analysis using the 3D-SSP technique (16). Images were reconstructed with slightly different filters for each scanner in order to make the final reconstructed image resolution as similar as possible or the same. Furthermore, these analysis approaches are quite insensitive to differences in resolution, sensitivity and scatter between the 931 and EXACT tomographs. Regional cerebral metabolic rate of glucose consumption was analyzed with stereotactically-defined cortical and subcortical VOIs (16). Predefined cortical VOIs included the following Brodmann areas (BA) predefined in a standard stereotactic atlas: lateral parietal association cortex (BA 5, 7, 39, 40); lateral temporal association cortex (BA 21, 22, 37, 38); lateral frontal association cortex (BA 6, 8, 9, 10, 11, 44, 45, 46, 47); occipital cortex (BA 17, 18, 19); primary sensorimotor cortex (BA 1, 2, 3 4); posterior cingulate cortex (BA 23, 31); and anterior cingulate cortex (BA 24, 32) (14). VOIs were also defined for the striatum, thalamus, amygdala and hippocampus. Data were normalized to the pons as previously described (17). The PD subject data were also compared to the normal data on a pixel-by-pixel basis. Two sample t statistic values were calculated for each pixel, and then converted to corresponding z values. Resultant images were displayed in three dimensions.

Age and education adjusted analysis of covariance was used to compare baseline cognitive performance between the PD[+]D and PD[−]D groups. Age-adjusted ANCOVA with Duncan post-hoc testing was used to determine differences in regional glucose cerebral metabolic activity among PD[+]D, PD[−]D, and HCs. Paired t-testing was used to assess 2-year interval changes in a subset of 5 PD[+]D subjects who completed 2-year follow-up ¹⁸F-FDG PET imaging. Stepwise multiple regression analysis was used for a post-hoc analysis of cognitive-PET relationships for psychometric data obtained at study entry.

RESULTS

Clinical results

The mean duration of follow-up was 3.9 ±1.2 yr (2.0 - 6.8 yr) between study entry and final cognitive evaluation. Six subjects were diagnosed with dementia at follow-up (PD[+]D). The mean baseline MMSE score in the PD[+]D subjects was 27.0 ± 0.9 (range, 26 - 28). The mean time to develop dementia after study entry was 3.8 ± 1.7 yr (range: 1.9 - 6.0 yr). There was no difference in mean symptomatic disease duration at study entry between PD[+]D subjects and those who remained free of dementia at follow-up (PD[−]D) (5.9 ± 2.9 vs 6.2 ± 2.8 yr, respectively). The mean duration of motor disease before onset of dementia was 9.7 ± 4.2 (range: 3.1 - 14.0) yr. As expected, PD[+]D subjects were older than PD[−]D subjects at entry: 72.2 ± 5.6 versus 58.2 ± 8.1 yr, respectively (t = 3.9; P = 0.0009). Cognitive test scores at time of study entry in the PD[+]D and PD[−]D groups are presented in Table 1. Allowing for differences in age and education, there were no significant differences in cognitive performance between the PD[+]D and PD[−]D groups except for lower performance in delayed visual reproduction learning (WMS VI-PREC) in the PD[+]D subjects.

TABLE 1:

Cognitive test scores at time of study entry. ¹

Test	PD[+]D (n=6)	PD[−]D (n=17)	Education Effect2	Age Effect2	Group Effect2	Overall Model2
PARR	9.8 ± 1.5	12.3 ± 2.9	F=1.90; P=0.18	F=7.49; P=0.013	F=0.12; P=0.74	F(3,19)=5.20; P=0.0086
BVRT	5.2 ± 2.1	6.7 ± 1.9	F=0.63; P=0.44	F=15.08; P=0.001	F=1.12; P=0.30	F(3,19)=7.02; P=0.0023
WMS-VI	6.9 ± 2.6	10.0 ± 2.5	F=0.65; P=0.43	F=10.51; P=0.004	F=0.00; P=0.95	F(3,19)=7.11; P=0.0021
WMS VI-PREC	44.6 ± 28.7	82.0 ± 18.5	F=2.02; P=0.17	F=2.46; P=0.12	F=4.95; P=0.039	F(3,19)=6.43; P=0.0034
Verbal Fluency	12.1 ± 4.5	14.8 ± 4.0	F=5.86; P=0.026	F=0.51; P=0.49	F=0.83; P=0.37	F(3,19)=2.73; P=0.075

¹Data are presented as mean ± SD, unless otherwise indicated

²ANCOVA F-values are presented using education and age as covariates.

Abbreviations: Wechsler Adult Intelligence Scale – III Picture Arrangement subtest (PARR); Benton Visual Retention Test Revised (BVRT); Wechsler Memory Scale III Visual Reproduction subtest (WMS-VI: immediate learning and WMS VI-PREC: percent delayed recall).

Table 2 lists the average MMSE scores at years 1-4 in the PD[+]D and PD[−]D groups. MMSE scores become significantly lower in the PD[+]D subjects in years 3 and 4 compared to the PD[+]D group. Scores in the PD[−]D group remained stable over time.

TABLE 2:

Average (SD) follow-up MMSE scores at years 1-4 in the PD[+]D and PD[−]D groups.¹

	PD[+]D	PD[−]D	Statistical significance
Year 1	25.6±2.8	27.7±2.2	tapprox=1.4; P=0.12
Year 2	27.3±2.0	27.8±1.2	tapprox=0.7; P=0.50
Year 3	23.2±3.0	28.1±1.6	tapprox=4.4; P=0.0005
Year 4	22.6±2.5	28.0±1.4	tapprox=4.2; P=0.0004

MMSE scores become significantly lower in the PD[+]D subjects in years 3 and 4 compared to the PD[−]D group. Satterthwaite’s method of approximate t tests (t_approx) were used for group comparisons.

¹Data are presented as mean ± SD, unless otherwise indicated.

Glucose metabolic findings at study entry in prospective PD[+]D converters versus non-converters

Analysis of covariance (using subject age) of the ¹⁸F-FDG PET VOI data from baseline scans demonstrated regional cerebral metabolic reductions in the PD[+]D converters compared with HCs and PD[−]D groups (Table 3). Compared with HCs, PD[+]D subjects had significant metabolic reductions in the occipital (−11.8%, F(2,22)=7.0, P=0.002) and posterior cingulate (−12.1%, F(2,22)=5.2, P=0.009) cortices. Regional hypometabolism was greatest in BA 18 (−20.0%, F(2,22)=8.45, P=0.0007), followed by BA 17 (−13.3%, F(2,22)=4.0, P=0.003). Less prominent reductions were present in BA 19 (−9.3%, F(2,22)=4.0, P=0.02). There was mild hypometabolism of the caudate nucleus (−8.4%, F(2,22)=3.2, P<0.05). There was no significant hypometabolism in the parietal, temporal, or frontal cortices, hippocampus, amygdala, putamen or thalamus of PD[+]D vs. PD[−]D or HC groups.

TABLE 3:

Baseline cerebral glucose metabolism in healthy controls (HC), PD subjects with (PD[+]D) and without (PD[−]D) incident dementia.¹

Region	PD[+]D (n=6)	PD[−]D (n=17)	HC (n=27)	Age Effect	Group Effect	Overall Model
Primary Sensorimotor Cortex	1.40 ± 0.10 n/a	1.40 ± 0.11 n/a	1.42 ± 0.12 n/a	F=9.03; P=0.0043	F=0.23; P=0.79	F=3.74; P=0.017
Parietal Association Cortex	1.35 ± 0.19 n/a	1.42 ± 0.15 n/a	1.45 ± 0.12 n/a	F=10.32; P=0.0024	F=1.46; P=0.24	F=6.64; P=0.0008
Posterior Cingulate Cortex	1.45 ± 0.14 A	1.57 ± 0.18 B	1.65 ± 0.12 B	F=8.43; P=0.0057	F=5.23; P=0.009	F=9.37; P<0.0001
Occipital Cortex	1.35 ± 0.14 A	1.44 ± 0.14 B	1.53 ± 0.12 B	F=11.52; P=0.0014	F=7.02; P=0.0022	F=12.03; P<0.0001
Frontal Association Cortex	1.43 ± 0.08 n/a	1.45 ± 0.12 n/a	1.50 ± 0.13 n/a	F=13.87; P=0.0005	F=1.46; P=0.25	F=7.27; P=0.0004
Anterior Cingulate Cortex	1.34 ± 0.07 n/a	1.35 ± 0.12 n/a	1.39 ± 0.13 n/a	F=27.71; P<0.0001	F=1.49; P=0.24	F=12.97; P<0.0001
Temporal Association cortex	1.24 ± 0.11 n/a	1.28 ± 0.10 n/a	1.30 ± 0.10 n/a	F=6.46; P=0.015	F=1.05; P=0.36	F=4.25; P=0.0099
Hippocampus	1.05 ± 0.09 n/a	1.08 ± 0.06 n/a	1.10 ± 0.08 n/a	F=6.18; P=0.017	F=1.22 P=0.31	F=4.25; P=0.001
Amygdala	1.01 ± 0.09 n/a	1.05 ± 0.09 n/a	1.04 ± 0.08 n/a	F=1.72; P=0.20	F=0.35; P=0.71	F=1.28; P=0.30
Thalamus	1.54 ± 0.06 n/a	1.47 ± 0.10 n/a	1.54 ± 0.12 n/a	F=0.65; P=0.43	F=2.75; P=0.074	F=1.93; P=0.14
Caudate Nucleus	1.42 ± 0.16 A	1.47 ± 0.11 B	1.55 ± 0.20 B	F=11.46; P=0.0015	F=3.24; P=0.048	F=7.80; P=0.0003
Putamen	1.71 ± 0.13 n/a	1.62 ± 0.13 n/a	1.58±0.16 n/a	F=0.65; P=0.43	F=0.27; P=0.77	F=0.28; P=0.83

Analysis of covariance F-values are presented with Duncan post-hoc testing between subgroups: sub-group means with the same letter are not significantly different. Brain regions are averaged for left and right hemispheres.

¹Data are presented as mean ± SD unless otherwise indicated.

The spatial extent and regional magnitudes of cerebral cortical metabolic differences between PD[+]D and PD[−]D subjects at study entry were confirmed by 3D-SSP analysis (Fig 1). Although the PET VOI analysis did not demonstrate significant reductions in the PD[−]D subjects, compared with the HC subjects, 3D-SSP analysis demonstrated isolated metabolic reductions in the more posterior aspect of the primary occipital cortex (BA 17). However, there was metabolic sparing of the occipital association cortex (BA 18 and 19; Fig 1).

FIGURE 1.

Metabolic reduction in PD. 3D-SSP t-statistic maps comparing PD[+]D and PD[−]D subjects with healthy controls (HC) at baseline. ¹⁸F-FDG-PET data are normalized within each subject to the pons. The most prominent metabolic reduction in PD[+]D is evident in the cuneus (especially BA 18; single-lined arrow) and precuneus (double-lined arrow). Mild-to-moderate reductions are also present in the mesiofrontal lobes (stippled arrow). In addition, the PD[+]D subjects demonstrate relative sparing of the primary sensorimotor cortex, a pattern similar to AD. In PD[−]D subjects, metabolic reduction in the posterior calcarine cortex (BA 17) is evident (arrow), while there is relative sparing of BA 18, 19, and the precuneus. RLAT = right lateral, LLAT = Left Lateral, RMED = Right Medial, LMED = Left Medial

Interval changes in glucose metabolic findings at 2 year follow-up, compared with study entry

VOI analysis of the 2-year follow-up scans in 5 PD[+]D subjects demonstrated significant year 2 versus baseline within-subject interval changes in nearly all brain regions, with the exceptions of the primary sensorimotor cortex, anterior cingulate cortex, and amygdala (Table 4). Interval metabolic reductions in the PD[+]D subjects were most prominent in the thalamus (−11.4%, t=5.25, P=0.006) with moderate reductions in the posterior cingulate, occipital, parietal, and frontal cortices and mild reductions in the temporal lobes (Table 4). Occipital cortical sub-region analysis again demonstrated the most significant reductions in BA 18 (−7.7%, t=−8.3, P=0.001) and BA 17 (−7.6%, t=−7.4, P=0.002) and less reductions in BA 19 (−5.7, t=−4.7, P=0.009).

TABLE 4:

Changes in regional cerebral glucose metabolism between baseline and 2-year follow-up imaging in the PD[+]D converters (within-subject analysis).¹

Region	PD[+]D Baseline (n=5)	PD[+]D At 2 years (n=5)	% Decline	t-value; Significance
Primary Sensorimotor Cortex	1.37 ± 0.10	1.33 ± 0.13	−3.0 ± 4.2%	t=−1.61; P=0.18
Parietal Association Cortex	1.31 ± 0.20	1.21 ± 0.17	−7.1 ± 4.6%	t =−3.42; P=0.027
Posterior Cingulate Cortex	1.39 ± 0.16	1.26 ± 0.10	−9.1 ± 5.5%	t =−3.73; P=0.02
Occipital Cortex	1.31 ± 0.15	1.21 ± 0.11	−7.1 ± 2.3%	t =−6.89; P=0.0023
Frontal Association Cortex	1.38 ± 0.12	1.29 ± 0.13	−6.7 ± 3.4%	t =−4.46; P=0.011
Anterior Cingulate Cortex	1.26 ± 0.08	1.20 ± 0.13	−4.6 ± 5.1%	t =−2.04; P=0.11
Temporal Association Cortex	1.20 ± 0.12	1.14 ± 0.10	−4.5 ± 1.8%	t =−5.53; P=0.0052
Hippocampus	1.02 ± 0.09	0.99 ± 0.10	−3.4 ± 2.6%	t =−2.92; P=0.043
Amygdala	1.00 ± 0.10	0.97 ± 0.11	−3.1 ± 3.9%	t =−1.76; P=0.15
Thalamus	1.52 ± 0.07	1.35 ± 0.11	−11.4 ± 4.9%	t =−5.25; P=0.006
Caudate Nucleus	1.35 ± 0.18	1.18 ± 0.14	−11.8 ± 9.6%	t =−2.73; P=0.056
Putamen	1.71 ± 0.12	1.57 ± 0.16	−8.2 ± 7.7%	t =−2.17; P=0.076

Paired Student t values with levels of significance are presented.

¹Data are presented as mean ± SD, unless otherwise indicated.

The spatial extent and regional magnitudes of cerebral cortical metabolic differences between PD[+]D subjects and HCs at baseline and 2-year follow-up imaging were confirmed by pixel-by-pixel 3D-SSP analysis (Fig 2). The most prominent metabolic reduction in PD[+]D was evident in the cuneus and precuneus. Within the frontal lobes, relative regional metabolic reductions were more prominent in the mesiofrontal cortex (Fig 2). In addition, the PD[+]D subjects demonstrated relative sparing of the primary sensorimotor cortex, a pattern similar to other neurodegenerative diseases such as Alzheimer disease.

FIGURE 2.

Progressive metabolic reduction in PD[+]D. 3D-SSP t-statistic maps comparing PD[+]D converters (n=5) to healthy controls (HC) at baseline (t1, upper row) and 2-year follow-up imaging (t2, lower row). FDG-PET data are normalized within each subject to the pons. There is progressive metabolic decline in the cuneus and precuneus and more widespread cortical hypometabolism including the mesiofrontal lobes in the follow-up ¹⁸F-FDG PET.RLAT = Right Lateral, LLAT = Left Lateral, RMED = Right Medial, LMED = Left Medial.

Cognitive-PET analysis

Stepwise multiple regression analysis was used to evaluate the relationship between cerebral metabolic activity in the posterior cingulate, BA 18 and the dorsal caudate nucleus from the baseline ¹⁸F-FDG PET study in the PD subjects (n=23) and a number of cognitive tests obtained at study entry that were available for analysis: VF, PARR, BVRT, WMS-VI and WMS VI-PREC. Stepwise regression analysis for the posterior cingulate cortex demonstrated a significant model effect for the WMS VI-PREC variable (R²=0.31, F(1,21)=9.2, P=0.007) with a non-significant regressor effect for the BVRT variable (R²=0.10, F(1,21)=3.2, P=0.08). Stepwise regression analysis for BA 18 demonstrated a significant model effect only for the BVRT variable (R²=0.43, F(1,21)=14.9, P=0.001) with no other significant regressors in the model. Stepwise regression analysis for the dorsal caudate nucleus demonstrated a significant model effect only for the WMS-VI variable (R²=0.31, F(1,21)=9.0, P=0.007) with no other significant regressors in the model.

Case studies of subjects with 3 and 4 longitudinal ¹⁸F-FDG PET studies

Interval changes between scans of years 2 and 4 in 2 PD[+]D subjects and years 2, 4 and 5 in one PD[+]D subject with multiple ¹⁸F-FDG-PET scans demonstrated the most prominent changes in the subcortical structures and in the occipital, cuneus, parietal and frontal cortices. Progressive metabolic changes from a cognitively grossly normal state to mild cognitive impairment and progressive dementia in a single subject with 4 FDG PET scans over a 5-year interval are illustrated in figure 3. This subject became clinically demented at 5 years after study entry (i.e., at the time of the fourth PET scan). Evolving progressive reductions in the cuneus and precuneus occur before less severe, but more widespread, neocortical reductions. Sparing of the primary sensorimotor cortical strip is present (Fig 3).

FIGURE 3.

Progressive metabolic reduction from within reference range at baseline (Year 0), through stages of mild cognitive impairment (Year 2, and Year 4), and clinical diagnosis of dementia (Year 5) in a PD[+]D converter. 3D-SSP t-statistic maps comparing a single PD[+]D converter to healthy controls (HC) with ¹⁸F-FDG PET data normalized to the pons. Progressive metabolic reductions in the cuneus and precuneus occur before less severe, but more widespread, cortical reductions. Sparing of the primary sensorimotor cortical strip is present. RLAT = Right Lateral, LLAT = Left Lateral, RMED = Right Medial, LMED = Left Medial, Yr = Year

DISCUSSION

Our findings indicate that incident dementia in idiopathic PD is heralded by metabolic changes within visual association (BA 18) and posterior cingulate cortices. Although hypometabolism of the cuneus and precuneus remained prominent, progression of dementia was associated with mixed subcortical (especially thalamic) and widespread cortical changes that also involve the mesiofrontal lobes. On follow-up imaging, PD[+]D subjects demonstrated relative sparing of the primary sensorimotor cortex, a pattern similar to other neurodegenerative dementias such as AD. Anterior cingulate cortex, which has distinct connections and behavioral attributes distinct from those of the posterior cingulate cortex, was relatively spared. Cuneus hypometabolism was observed first in parkinsonian dementia by Kuhl et al. in 1985 (18). In subsequent studies of severely affected PD[+]D subjects, glucose metabolic and regional cerebral blood flow studies demonstrated reduced activity in the parietal and temporal neocortices, similar to that seen in AD (19, 20). Vander Borght et al., for example, compared metabolic differences between AD and PD[+]D subjects matched for severity of dementia (mean MMSE scores of 18 in both groups) and found similar glucose metabolic reductions globally, and regionally involving the posterior cingulate, lateral parietal, lateral frontal, and lateral temporal association cortices when compared to HCs (5). PD[+]D subjects, however, exhibited greater metabolic reductions in the occipital cortex and relatively preserved metabolism in the medial temporal lobe. In the present study, we followed subjects converting from mild cognitive impairment to dementia and found less prominent reductions in the lateral parietotemporal cortices, compared with metabolic changes seen typically in subjects with mild cognitive impairment at the time of conversion to clinical dementia of the AD type (21). Cuneus involvement as a herald of dementia may, thus, be specific for PD[+]D. Our present findings indicate that incident dementia preferentially involves BA18 (visual association cortex), whereas, reductions in the primary visual cortex (BA 17) can be seen in PD without dementia (22). Extension of occipital hypometabolism from the primary to the visual association cortices, together with precuneus hypometabolism, may be the early cortical metabolic “signature” of incident dementia in PD. The involvement of the cuneus is distinct from findings in longitudinal ¹⁸F-FDG-PET in AD. For example, Mosconi et al. reported the first clinicopathological series of longitudinal ¹⁸F-FDG-PET scans in post-mortem elderly subjects who were verified to be cognitively normal and were followed to the onset of AD (23). The ¹⁸F-FDG PET scans indicated a progression of glucose metabolic deficits from the hippocampus to the parietotemporal and posterior cingulate cortices, but sparing the cuneus. One AD subject had combined AD and diffuse Lewy body disease at post-mortem examination, and her last ¹⁸F-FDG PET showed occipital as well as parietotemporal hypometabolism.

Visual association cortex hypometabolism as a herald of dementia is interesting in view of the prominence of visual hallucinations in more advanced PD and PD[+]D. Prior metabolic imaging studies suggest that visual hallucinations are associated with hypometabolism of visual association cortices (24). Visual hallucinations are by far the most common form of hallucination in PD and incidence of visual hallucinations in PD may herald progression of dementia (25).

PD[+]D may be a diverse entity, although pathologically, dementia with Lewy bodies (DLB) likely accounts for many cases of PD[+]D (6). A generally accepted distinction in current International Consensus diagnostic criteria is between subjects presenting with parkinsonism for at least 1 year prior to the onset of dementia (PD[+]D), versus those developing parkinsonism and dementia concurrently or with dementia in advance of parkinsonism (DLB) (6). We do not have autopsy diagnosis available for our PD[+]D subjects, however, following the above 1-year rule, our PD[+]D subjects would not be categorized as DLB. However, recent β-amyloid and post-mortem literature indicates significant overlap between subjects clinically as either PD[+]D or DLB (26). Cuneus hypometabolism, is also a common metabolic feature of both DLB and PD[+]D (27), and it is plausible that some of our findings may reflect a pre-dementia stage of DLB in clinically diagnosed PD[+]D subjects.

Prior resting cerebral glucose metabolic and blood flow studies in mild, early PD showed increased striatal activity contralateral to the clinically most affected body side, which may represent a compensatory mechanism within the striatum (28). Striatal glucose metabolism may then decrease with advancing disease, especially in subjects with PD[+]D (5). Interval changes in striatal metabolism or perfusion are more difficult to interpret because of early compensatory change related to motor impairments in PD and possible effects of dopaminergic medications. Our data suggest that the most prominent early metabolic changes in incident dementia in PD occur at the posterior and medial neocortical levels and that mild caudate hypometabolism is also present. Progression of dementia is associated with mixed subcortical, and diffuse neocortical changes that involve the mesiofrontal cortices. The involvement of mesiofrontal regions in our PD[+]D subjects may potentially reflect effects of dopaminergic nigrostriatal denervation, including the caudate nucleus, and dopaminergic mesocortical pathways (29). Posterior cortical activity changes may reflect cholinergic denervation secondary to loss of nucleus basalis of Meynert afferents. The cuneus is thought to be the most vulnerable region to cholinergic deafferentation as there is more severe loss of cholinergic projections to the cuneus than to any other cortical region (30, 31). Recently, Shimada et al. also made an important observation of significant reductions in cortical acetylcholinesterase activity in early drug naïve PD subjects. These authors found most prominent cholinergic reductions in the medial occipital association visual cortex (BA 18) (32). The observed anterior and posterior cerebral metabolic changes may in part correspond to mixed effects of dopaminergic and cholinergic denervation in PD[+]D (31).

Analysis of cognitive test results obtained at the time of study entry demonstrate regionally specific correlations with ¹⁸F-FDG-PET changes. The Benton Visual Retention Test, which assesses visuospatial perception, visuomotor and visuoconstructive abilities, and visual memory, best correlates with BA 18 metabolism. Immediate recall on the Wechsler Memory Scale III Visual Reproduction Learning best correlated with glucose metabolism in the dorsal caudate nucleus. In contrast, delayed recall on this test had the most robust correlation with glucose metabolic activity in the posterior cingulate cortex. This finding may reflect the cognitive difference in visual delayed recall between the PD[+]D and PD[−]D subjects at the time of study entry. In this respect, functions of working memory may be more related to the caudate nucleus and its connections, and delayed memory recall to the posterior cingulate and its mesial temporal lobe connections. This may correspond to data from animal studies showing that posterior cingulate cortical lesions lead to specific learning impairment in the later stages of data acquisition (33). Our preliminary data indicate that specific brain regions that are affected in incident dementia in PD may have differential contributions to the cognitive profile in these subjects.

CONCLUSION

The results of this prospective cohort study demonstrate that incident dementia in idiopathic PD is heralded by hypometabolism in visual association (BA 18) and posterior cingulate cortices, with additional mild reduction in the caudate nucleus. Although hypometabolism in the cuneus and precuneus remain prominent, two-year follow-up data from 5 PD[+]D converters show that progression of PD dementia is associated with mixed subcortical, especially thalamic, and cortical changes that also involve the mesiofrontal lobes. These results suggest that visual association and posterior cingulate cortex metabolic changes may be a useful biomarker for enriching studies aimed at studying progression to dementia in PD.