Cognitive and Neuroanatomical Correlates in Early Versus Late Onset Parkinson’s Disease Dementia

Subjects

The present study enrolled 116 patients with PDD recruited from the Movement Disorders and Dementia outpatient clinic at Yonsei University Severance Hospital from March 2007 to December 2014. PD was diagnosed according to the clinical diagnostic criteria of the UK PD Society Brain Bank [10]. PDD was diagnosed based on the Movement Disorder Society consensus criteria for diagnosis of probable PDD [11]. We enrolled only patients showing decreased dopamine transporter uptake in the posterior putamen on ¹⁸F-FP-CIT PET scans. Patients who experienced cognitive impairment that was sufficient to meet criteria for dementia within 1 year after onset of PD were not included to exclude patients who met research criteria for dementia with Lewy bodies.

Depending on the age at PDD diagnosis, the patients were divided into early-onset PDD (EOPDD <70 years of age, n = 39) and late-onset PDD (LOPDD ≥ 70 years of age, n = 77). Motor symptoms were assessed using the Unified PD Rating Scale part III (UPDRS- III) in the ‘off’ status. Disease duration and memory complaints in each patient were based on interviews with patients and caregivers living with patients. History of vascular risk factors was defined based on the use of agents. All patients participated in the Seoul Neuropsychological Screening Battery (SNSB) [12], consisting of the following cognitive subsets: attention, language and related functions (Korean version of the Boston Naming Test and calculation), visuospatial function (Rey Complex Figure Test; RCFT), verbal memory (three-word registration and recall and the Seoul Verbal Learning Test), visual memory (RCFT, immediate recall, 20-minute delayed recall, and recognition), and frontal/executive function (contrasting program, go-no-go test, phonemic and semantic fluency test, and Stroop test). Among study subjects, MRI scans used for cortical thickness and SI volume analysis with protocols discussed below were available in 65 patients with PDD. Of these 65 patients, 25 were classified as EOPDD and 40 patients as LOPDD. We excluded patients with focal neurological deficits, evidence of focal brain lesions, diffuse white matter intensities, multiple lacunae in the basal ganglia based on conventional MRI, or other past medical comorbidities that could contribute to cognitive decline.

Controls

We recruited 121 healthy age- and sex-matched normal controls who had no history of neurological disease and no abnormalities on neurologic examinations. The normal controls exhibited no objective cognitive dysfunction on the Mini-Mental State Examination (MMSE) and the SNSB. The normal controls were divided into the young control group (<70 years of age) and the old control group (≥70 years of age) and were compared with the respective age-matched PDD group; 49 were allotted to the young controls and 72 to the old controls. Among control subjects, MRI scans for imaging analysis were available in 91 subjects; 39 were allocated to the young control group and 51 to the old control group.

Standard protocol approvals, registrations, and patient consents

We received approval from the Yonsei University Severance Hospital ethical standards committee on human experimentation for experiments using human subjects. Written informed consent was obtained from all subjects participating in this study.

MRI acquisition

All scans were acquired using a Philips 3.0-T scanner (Philips Intera; Philips Medical System, Best, The Netherlands) with a SENSE head coil (SENSE factor = 2). A high-resolution T1-weighted MRI volume data set was obtained from all subjects using a 3D T1-TFE sequence configured with the following acquisition parameters: axial acquisition with a 224×256 matrix; 256×256 reconstructed matrix with 182 slices; 220 mm field of view; 0.98×0.98×1.2 mm³ voxels; TE, 4.6 ms; TR, 9.6 ms; flip angle, 8°; slice gap, 0 mm.

Image processing for cortical thickness

We corrected image intensity inhomogeneity caused by magnetic field inhomogeneity by varying the signal intensity slowly over the image. The N3 algorithm was used to correct images for intensity non-uniformities resulting from inhomogeneities in the magnetic field [13]. Cortical thickness was measured on T1-weighted image by using the fully automated CIVET pipeline [14]. Skull stripping was performed using a Brain Extraction Tool with a deformable model fitted to the brain surface and optimized parameters [15]. Each brain was transformed separately into a standardized stereotaxic space (an ICBM 152 template) and resampled on a 1 mm³ voxel grid to account for inter-individual differences in absolute brain size [16]. An artificial neural network classifier that was validated in human brain images, was applied to identify gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) [17]. Partial volume level and MRI intensity mixing at the tissue interfaces due to the finite resolution of the imaging device were estimated and corrected using a trimmed minimum covariance determinant method [18]. A cortical surface was extracted automatically from each MR volume using the Constrained Laplacian-based Automated Segmentation with Proximities algorithm [19]. Cortical thickness was defined as the Euclidean distance between linked vertices or the inner and outer surfaces. The inner surface was defined by the WM/GM boundary surfaces and the outer surface by GM/CSF boundary surfaces. Then, diffusion smoothing, which generalizes Gaussian kernel smoothing, with 20 mm FWHM (full width half maximum) was used to increase the signal to noise ratio and detect population changes very well. To assess differences in cortical thickness between the PDD group and the control group, a general linear model was constructed with age, sex, and intracranial volume as independent variables and each of the vertices of thickness as a dependent variable. For multiple comparisons, the results were thresholded at a false discovery rate (FDR)-corrected p-value of 0.05 and cluster of 100. In addition, the effect of diagnosis (PDD versus control), age (<70 years versus ≥ 70 years), and their interaction with cortical thinning was assessed using analysis of covariance (ANCOVA) with a threshold of uncorrected p < 0.001 and cluster of 100, controlling for sex, educational level, hypertension, diabetes mellitus, and hyperlipidemia.

Volumetric determination of SI

The analysis was performed using Statistical Parametric Mapping software (SPM8, Wellcome Department of Imaging Neuroscience, London, UK). Each structural MRI scan was bias-corrected, segmented into SPM default tissue probability maps, and then normalized with VBM8 DARTEL templates for registration to MNI space using linear and nonlinear transformations within a unified model. According to a previous study [20], the ROI of SI was defined for the left and right hemispheres based on the location of the anterior commissure, which forms the boundary of the superior part of the end of the anterior third of the SI. The ROI extended 25 mm lateral from the midline, 13 mm ventral from the superior edge of the anterior commissure at the midline, and 3 mm anterior and 9 mm posterior from the middle of the anterior commissure (Supplementary Figure 1). The masks were created using the WFU PickAtlas 2.4 software [21], and volumetry of GM within a selected ROI was performed automatically. To correct for individual brain size, volumes were normalized by dividing by total intracranial volume derived from the masks covering the entire brain. Normalized SI volume was defined by the following formula: total SI volume (mL)/total intracranial volume (mL) X 10,000.

¹⁸F-FP-CIT PET and analysis of data

¹⁸F-FP-CIT PET scans were performed with a GE PET-CT DSTe scanner (GE Discovery STE, GE Healthcare Technologies, Milwaukee, WI), which obtains images with 3D resolution of 2.3 mm full width at half maximum. Quantitative analyses were performed following a modified version of a previously described procedure [22]. Quantitative analyses were based on volumes of interest (VOIs), which were defined based on a template in standard space. All reconstructed PET images were spatially normalized to Talairach space using a standard ¹⁸F-FP-CIT PET template which was made using ¹⁸F-FP-CIT PET and T1 MR images of 13 normal controls to remove inter-subject anatomical variability. The 8 VOIs of bilateral striatal subregions and one occipital VOI were drawn on a coregistered spatially normalized single T1 MR and 18F-FPCIT PET template image. The striatum was divided into the caudate, ventral striatum, anterior putamen, and posterior putamen. The VOI for the ventral striatum was defined according to previously defined criteria [23], and the boundary between the anterior and posterior putamen was the anterior commissure coronal plane. Dopamine transporter (DAT) activity was calculated by the non-displaceable binding potential, which was defined as follows: (mean standardized uptake value of the striatal subregions VOI— mean standardized uptake value of the occipital VOI)/mean standardized uptake value of the occipital VOI [24].

Statistical analysis

The χ² and Mann-Whitney U tests were used for categorical and continuous variables, respectively. The effects of diagnosis (PDD versus control), age (<70 years versus ≥ 70 years), and their interaction on neuropsychological tests and volumes of SI were assessed using ANCOVA, controlling for sex,educational duration, hypertension, diabetes mellitus, and hyperlipidemia. The significance of the interaction was tested in a full factorial model including the two main effects of diagnosis and age and their interaction. Pearson’s correlation analysis was used to evaluate the relationship between normalized SI volume and cognitive performance or cortical thickness. To compare Pearson’s correlation coefficients between young-aged group and old-aged groups, we used Fisher r-to-z transformation through correction of deviations from normal distribution. Statistical analyses were performed using commercially available software (SPSS, version 18.0), and a two-tailed p < 0.05 was considered to indicate statisticalsignificance.

Demographic characteristics of the EOPDD and LOPDD

The demographic characteristics of patients with PDD are shown in Table 1. Mean age at onset was 64.6 years in EOPDD and 75.5 years in LOPDD. There were no significant differences in sex, educational level, duration of Parkinsonism, and duration of cognitive impairment between the EOPDD and LOPDD groups. K-MMSE score tended to be higher in the EOPDD group (19.9) than the LOPDD group (18.5), which may be due to the difference in age. Levodopa equivalent dose and vascular risk factors, such as hypertension, diabetes mellitus, or dyslipidemia, did not significantly differ between the EOPDD and LOPDD groups. The UPDRS motor score was significantly higher in patients with LOPDD (35.5) than EOPDD (28.8, p = 0.005). EOPDD patients and LOPDD patients group did not show significant differences between their control groups in age, sex, K-MMSE score, education duration, and vascular risk factors.

The demographic characteristics of 65 patients with available MRI data were similar to those of patients with PDD (Supplementary Table 1). Mean age at onset in these 65 patients was 64.5 years in the EOPDD group and 74.7 years in the LOPDD group. No significant differences in sex, MMSE, educational level, PD duration, levodopa equivalent dose, UPDRS motor score, or duration of memory complaints were observed between the EOPDD and LOPDD groups. Total intracranial volume was not statistically different between the two patient groups (p = 0.596).

Neuropsychological features of the EOPDD and LOPDD

The neuropsychological features are shown in Table 2. Compared with the young control group, EOPDD patients showed significantly poorer performances in all domains of attention, language, visuospatial, and memory functions. Similarly, the LOPDD group showed a lower score in all cognitive subdomains compared with the old control group. The results of a direct comparison between the EOPDD and LOPDD groups in terms of an interaction between diagnosis and age are illustrated in Table 3. Significant group-by-age interaction effects were observed for digit backward and digit forward span test (p = 0.011 and 0.05, respectively) and visual recognition memory function (p = 0.012) in EOPDD patients compared with LOPDD patients. In addition, EOPDD patients tended to show poorer performance in phonemic generative naming, contrasting program, and verbal memory tests (p = 0.095, 0.078, and 0.118, respectively) than the LOPDD patients, although the difference was not statistically significant. No significant interaction effects between EOPDD and LOPDD groups were observed in language, visuospatial function, or other frontal-executive functions.

The group-by-age interaction effects in the PDD subpopulation with available MRI data are illustrated in Supplementary Table 3. Similar to the overall PDD patients, significant group-by-age interaction effects were observed for backward digit span and visual recognition memory tests (p = 0.009 and 0.006, respectively) in EOPDD patients compared with LOPDD patients.

Cortical thickness analysis between the EOPDD and LOPDD groups

A comparison of cortical thickness between EOPDD patients and age-matched controls is shown in Fig. 1A. As expected, patients with EOPDD exhibited a cortical thinning in extensive cortical areas, involving frontal, temporal, and parietal areas. A comparison of LOPDD and age-matched controls is shown in Fig. 1B. The LOPDD group showedsignificant decrease in cortical thickness in widespread cortical areas of frontal, temporal, and parietal areas, with prominent cortical thinning in bilateral prefrontal areas. A comparison of cortical thickness between the EOPDD and LOPDD groups based on interaction effect demonstrated that, compared with LOPDD patients, EOPDD patients exhibited decrease in cortical thickness in the left anterior cingulate gyrus and a small area of the right inferior temporal gyrus when adjusting for age with a threshold of uncorrected p < 0.001 (Fig. 1C).

Comparison of SI volume between the EOPDD and LOPDD groups

The mean normalized SI volumes among groups are shown in Table 4. Compared with the young control group, the mean normalized SI volume was significantly decreased in EOPDD patients (p < 0.001). In LOPDD patients, the mean normalized SI was decreased compared to the control group, although the difference was not statistically significant. A significant group-by-age interaction effect was observed, suggesting that SI volume was relatively decreased in the EOPDD group comparing to the LOPDD group after adjusting for the aging effect (p = 0.044).

Correlation between the normalized SI volume and performance in cognitive subdomains or cortical thickness in young age and old age groups

The correlation analysis between the normalized SI volume and cognitive performance in each group is shown in Supplementary Table 3. In young age group, the normalized SI volume was significantly positively correlated with all subdomains of cognitive test, whereas it was not correlated with all subdomains except letter cancellation test (p = 0.012) in old age group. A comparison of correlation coefficients between the groups and individual cognitive subdomains revealed that correlation coefficients of EOPDD group were significantly higher in the backward digit spans (p = 0.02), interlocking pentagon (p = 0.05), visual memory (p = 0.01), phonemic generative (p = 0.05), and color Stroop test (p = 0.04). In analysis between the normalized SI volume and cortical thickness, no significant clusters that were significantly correlated were observed in either patients with EOPDD or LOPDD.

DAT activity in ¹⁸F-FP-CIT PET

Comparison of ¹⁸F-FP-CIT uptakes in the striatal subregions between the groups is shown in Supplementary Table 4. DAT activity did not differ between patients with EOPDD and LOPDD in all striatal subregions.