Cortical thinning associated with mild cognitive impairment in Parkinson's disease

The aim of this study was to investigate patterns of cortical atrophy associated with mild cognitive impairment in a large sample of nondemented Parkinson's disease (PD) patients, and its relation with specific neuropsychological deficits. Magnetic resonance imaging (MRI) and neuropsychological assessment were performed in a sample of 90 nondemented PD patients and 32 healthy controls. All underwent a neuropsychological battery including tests that assess different cognitive domains: attention and working memory, executive functions, memory, language, and visuoperceptual‐visuospatial functions. Patients were classified according to their cognitive status as PD patients without mild cognitive impairment (MCI; n = 43) and PD patients with MCI (n = 47). Freesurfer software was used to obtain maps of cortical thickness for group comparisons and correlation with neuropsychological performance. Patients with MCI showed regional cortical thinning in parietotemporal regions, increased global atrophy (global cortical thinning, total gray matter volume reduction, and ventricular enlargement), as well as significant cognitive impairment in memory, executive, and visuospatial and visuoperceptual domains. Correlation analyses showed that all neuropsychological tests were associated with cortical thinning in parietotemporal regions and to a lesser extent in frontal regions. These results provide neuroanatomic support to the concept of MCI classified according to Movement Disorders Society criteria. The posterior pattern of atrophy in temporoparietal regions could be a structural neuroimaging marker of cognitive impairment in nondemented PD patients. All of the neuropsychological tests reflected regional brain atrophy, but no specific patterns were seen corresponding to impairment in distinct cognitive domains. © 2014 International Parkinson and Movement Disorder Society

A B S T R AC T : The aim of this study was to investigate patterns of cortical atrophy associated with mild cognitive impairment in a large sample of nondemented Parkinson's disease (PD) patients, and its relation with specific neuropsychological deficits. Magnetic resonance imaging (MRI) and neuropsychological assessment were performed in a sample of 90 nondemented PD patients and 32 healthy controls. All underwent a neuropsychological battery including tests that assess different cognitive domains: attention and working memory, executive functions, memory, language, and visuoperceptual-visuospatial functions. Patients were classified according to their cognitive status as PD patients without mild cognitive impairment (MCI; n 5 43) and PD patients with MCI (n 5 47). Freesurfer software was used to obtain maps of cortical thickness for group comparisons and correlation with neuropsychological performance. Patients with MCI showed regional cortical thinning in parietotemporal regions, increased global atrophy (global cortical thinning, total gray matter volume reduction, and ventricular enlargement), as well as significant cognitive impairment in memory, executive, and visuospatial and visuoperceptual domains. Correlation analyses showed that all neuropsychological tests were associated with cortical thinning in parietotemporal regions and to a lesser extent in frontal regions. These results provide neuroanatomic support to the concept of MCI classified according to Movement Disorders Society criteria. The posterior pattern of atrophy in temporoparietal regions could be a structural neuroimaging marker of cognitive impairment in nondemented PD patients. All of the neuropsychological tests reflected regional brain atrophy, but no specific patterns were seen corresponding to impairment in distinct cognitive domains. V C 2014 International Parkinson and Movement Disorder Society Key W ords: Parkinson's disease; cortical thickness; cognition Parkinson's disease (PD) is associated with cognitive decline 1-4 that may predict dementia at later stages. [5][6][7] Between 18.9% and 38.2% of patients meet mild cognitive impairment (MCI) criteria. 8 Indeed, the proportion of patients fulfilling MCI criteria increased from one third to approximately 50% of patients without dementia after 5 years from diagnosis. 9 A great variability was seen in the description and proportion of subtypes of MCI in PD, 3,6,10,11 perhaps because of the number and type of tests used and the classification of the tests by domains.
The recognition of PD patients with MCI (PD MCI) has led to studies searching for biological markers associated with this diagnosis. Several magnetic resonance imaging (MRI) studies have investigated the Relevant conflicts of interest/financial disclosures: Nothing to report. Full financial disclosures and author roles may be found in the online version of this article. relationship between brain atrophy and specific cognitive deficits in nondemented PD, such as deficits in memory, [12][13][14][15][16][17][18][19] verbal fluency, 20 visuospatial and visuoperceptual ability, 21 and decision-making and emotional processing. 22,23 Recently, Filoteo et al., 24 using region-of-interest analyses, associated subtle changes in multiple cognitive domains with distinct patterns of regionally specific volume changes in nondemented PD patients. However, to the best of our knowledge, no published MRI studies have focused on whole brain neuroanatomical correlates of the different tests included in cognitive domains assessed by an extensive neuropsychological battery.
Few studies have investigated the neuroanatomical correlates of MCI. Initially, voxel-based morphometry (VBM) analyses showed that PD-MCI had reduced cortical gray matter (GM) density in the left middle frontal gyrus, precentral gyrus, left superior temporal lobe, and right inferior temporal lobe in comparison with cognitively intact PD patients. 25 In contrast, Song et al. 26 reported GM density decreases in frontal regions of PD-MCI patients in comparison with PD without MCI (PD non-MCI). Recently, VBM analyses with a large sample of 148 PD patients did not show any areas of significant GM loss in participants with PD-MCI compared with controls. 27 To clarify these controversial results, certain methodological issues need to be addressed. Both VBM and volumetric analyses may be insufficient to detect early cortical changes in PD MCI patients. Recent studies using cortical thickness measures suggest that this method may be more sensitive than VBM to identify regional GM changes associated with PD. 28 Cortical changes associated specifically with MCI in PD have not been investigated in depth. In a small sample, Biundo et al. 29 showed significant regional thinning in right parietal-frontal areas and in left temporal-occipital areas in PD MCI in comparison with PD non-MCI. Studying a bigger sample, Pagonabarraga et al., 30 using an uncorrected level of significance, showed both increases and decreases in cortical thickness of PD MCI patients in comparison with PD non-MCI, and Hanganu et al. 31 did not find significant cortical thinning in PD MCI subjects compared with PD non-MCI, but detected a small cluster with increased thickness in the left middle temporal gyrus. Recently, Pereira et al 32 studied a large multicentric cohort of drug-na€ ıve PD patients with early PD; they found mainly temporal and parietal cortical thinning in the PD MCI group compared with PD non-MCI patients, using a cognitive-domain approach.
In light of these previous results, the aims of this study were (1) to investigate whether different anatomical patterns of cortical atrophy distinguish PD patients with MCI from patients without cognitive impairment in a large sample of nondemented PD patients and (2) whether different anatomical patterns of cortical atrophy are associated with neuropsychological deficits commonly related to specific cognitive domains. This is of crucial importance for validating MCI criteria, and may help to clarify the neural correlates of cognitive impairment in PD.

Subjects
The study included 121 consecutive PD patients recruited from an outpatient movement disorders clinic (Parkinson's Disease and Movement Disorders Unit, Department of Neurology, Hospital Clinic, Barcelona, Catalonia, Spain) and 49 healthy controls who volunteered to take part in studies addressing agerelated processes at the Institut de l'Envelliment (Aging Institute). The inclusion criteria were: (1) fulfilling the UK PD Society Brain Bank diagnostic criteria for PD 33 ; (2) no surgical treatment with deep brain stimulation. The exclusion criteria were: (1) presence of dementia according to the Movement Disorders Society criteria 33 ; (2) Hoehn and Yahr (H&Y) scale score greater than 3; (3) juvenile-onset PD; (4) presence of psychiatric or neurological comorbidity; (5) low global intelligence quotient estimated by the Vocabulary subtest of the Wechsler Adult Intelligence Scale, 3rd edition (scaled score 7 points); (6) Minimental state examination score below 25; (7) presence of claustrophobia; (8) pathological MRI findings other than mild white matter hyperintensities in long-TR sequences; and (9) MRI artefacts.
Ninety PD patients and 32 healthy volunteers were finally selected. Twelve patients and eight controls were excluded because they fulfilled criteria for dementia or other neurological disease, six PD patients for psychiatric comorbidity, one PD patient who scored higher than 3 on the H&Y scale, one PD patient who presented young-onset PD, three PD patients and one control who presented low global intelligence quotient scores, two PD patients for claustrophobia, three healthy subjects who did not complete the neuropsychological assessment, and two controls and two PD patients because of MRI artefacts. We also excluded four patients and three controls aged younger than 50 years.
Motor symptoms were assessed by means of the UPDRS-III motor section. All PD patients were taking antiparkinsonian drugs, consisting of different combinations of levodopa (L-dopa), Catechol-O-methyltransferase (COMT) inhibitors, monoamine oxidase inhibitors, dopamine agonists, and amantadine.
This study was approved by the ethics committee of the University of Barcelona. Written informed consent was obtained from all study subjects after full explanation of the procedures involved.

Neuropsychological Assessment
We selected a neuropsychological battery to assess cognitive functions usually impaired in PD. 2,3 This battery is recommended by the Movement Disorder Society task force to evaluate cognitive functions in PD 8,34 and is able to detect MCI in PD (level I or level II criteria for PD-MCI, bar language, for which a single measure was used. 10 Attention and working memory were assessed with the Trail Making Test (TMT) (in seconds), part A (TMT A) and part B (TMT B), Digit Span Forward and Backward, and the Stroop Colour-word Test and Symbol Digits Modalities Tests (SDMT); executive functions were evaluated with phonemic (words beginning with the letter "p" in 1 minute) and semantic (animals in 1 minute) fluencies; language was assessed by the total number of correct responses in the short version of the Boston Naming Test, memory through total learning recall (sum of correct responses from trial I to trial V), and delayed recall (total recall after 20 min) through scores on Rey's Auditory Verbal Learning Test (RAVLT). Visuospatial and visuoperceptual functions were assessed with Benton's Judgement of Line Orientation (JLO) and Visual Form Discrimination (VFD) tests.
Initially, z scores for each test and for each subject were calculated based on the control group's means and standard deviations. Expected z scores adjusted for age, sex, and education for each test and each subject were calculated based on a multiple regression analysis performed in the healthy control group. 3 We classified subjects as having MCI if the z score for a given test was at least 1.5 lower than the expected score in at least two tests in one domain, or in at least one test per domain in at least two domains. As expected, 2 most subjects with abnormalities had deficits in more than one function, precluding the creation of patient groups with single-domain impairments. Patients' cognitive complaints were recorded during the clinical interview.
Z composite scores were computed to obtain global cognitive measures (attention and working memory, executive, memory, and visuospatial/visuoperceptual functions).

Neuropsychological and Clinical Statistical Analysis
All statistical analyses were performed using SPSS Statistics 20, release version 20.0.0 (Armonk, NY, http://www-01.ibm.com/software/analytics/spss/). Statistical significance threshold was set at P < 0.05. Pearson's chi-square test was used to compare categorical variables (sex and H&Y stage). Separate-variance test (Welch t) was used to test between-group differences (HC, PD MCI, PD non-MCI) in quantitative clinical and demographic variables. Analyses of covariance including age and education as confounding variables were used to compare the performance on neuropsychological tests and composite scores for each domain. Bonferroni correction was used to control for the number of intergroup comparisons.

Cortical Thickness
Cortical thickness was estimated using the automated FreeSurfer stream (version 5.1; available at: http://surfer.nmr.harvard.edu). The procedures carried out by FreeSurfer software include removal of nonbrain data, intensity normalization, 35 ), tessellation of the GM/white matter boundary, automated topology correction, 36,37 and accurate surface deformation to identify tissue borders). [38][39][40] Cortical thickness is then calculated as the distance between the white matter and GM surfaces at each vertex of the reconstructed cortical mantle. 40 In our study, results for each subject were visually inspected to ensure accuracy of registration, skull stripping, segmentation, and cortical surface reconstruction. Maps were smoothed using a circularly symmetric Gaussian kernel across the surface with a full width at half maximum (FWHM) of 15 mm.
Comparisons between groups were assessed using a vertex-by-vertex general linear model. The model included cortical thickness as a dependent factor and diagnosis (controls, PD non-MCI, PD MCI) as an independent factor, and also included age and education as nuisance variables (https://surfer.nmr.mgh.harvard.edu/fswiki/FsgdFormat).
All results were corrected for multiple comparisons by using a precached cluster-wise Monte-Carlo Simulation. Significance level was set at P < 0.05.
In the PD patient group, the vertex-by-vertex general linear model was used to assess the relationship between cortical thickness and neuropsychological tests. Positive and negative associations between a specific neuropsychological test and cortical thickness were analyzed using Qdec. Initially, a simple model without covariates was tested for each neuropsychological measure. Complementarily, we performed a conservative analysis of covariance including age, education, and sex as confounding variables. An initial vertex-wise threshold was set to P 5 0.005 (2.3) to find clusters. To avoid clusters appearing significant
Total GM volume and total subcortical volumes, as well as mean lateral ventricular volume and estimated total intracranial volume (eTIV) were obtained automatically via whole brain segmentation. 41 An analysis of covariance including eTIV, age, and education was used to compare subcortical volumes between groups. Significant P values were adjusted using post hoc Bonferroni tests considering the number of intergroup comparisons.

Results
Forty-seven patients (52.2%) fulfilled the criteria for MCI. Table 1 shows sociodemographic and clinical data and the corresponding group comparisons. Table 2 shows differences in neuropsychological performance between groups. MCI patient scores were significantly worse than those of non-MCI patients and healthy controls in all tests except Forward and Backward Digits. Forty-two patients (46.6%) showed impairments in attention and working memory, 31 (34.4%) in memory, 26 (28.8%) in executive and also visuoperceptual and visuospatial domains, and only four (0.4%) in language.

Global Atrophy Comparison Between Groups
Significant differences were found in global atrophy between healthy controls and PD patient groups according to MCI status. The PD MCI patients showed decreased total mean thickness and total GM volume, as well as increased mean lateral ventricle volume in comparison with healthy controls and PD non-MCI. No significant differences were found in total subcortical volume (Table 3). Finally, no significant results were seen after the inclusion of disease duration as a nuisance variable in the covaried model.

Cortical Thickness Comparison Between Groups
Surface-based cortical thickness analyses showed a group effect according to cognitive status (Fig. 1). PD MCI (PD MCI < HC) showed cortical thinning in widespread bilateral regions, including parietal (superior and inferior, supramarginal, and also precuneus regions), temporal (posterior middle temporal, inferior temporal, fusiform, and parahippocampal regions), and occipital cortices (bilateral posterior occipital), but also in left frontal superior and rostral middle areas. There were significant differences between controls and PD non-MCI (PD non-MCI < HC) specifically in bilateral superior parietal regions. PD MCI (PD MCI < PD non-MCI) showed significant thinning in right precuneus and supramarginal regions compared to PD with non-MCI. There were no significant results after including disease duration as a nuisance factor in the co-varied model.

Correlations Between Neuropsychological Tests and Cortical Thickness in PD Patients
Vertex-wise regression analyses showed correlations between regional cortical thickness and neuropsychological test performance (Supplemental Data). We found a common posterior atrophy pattern for all the neuropsychological tests evaluated, and there were no specific patterns of atrophy related to neuropsychological domains.
Stroop Test (Words, Colours, and Word-Colours), TMT A, semantic fluency, and RAVLT total learning performance correlated with GM thinning in bilateral . Age showed significant differences between PD MCI and PD non-MCI patients (P 5 0.003, Bonferroni correction). MMSE showed significant differences between PD MCI and both PD non-MCI and HC (P <.001, Bonferroni correction). BDI showed significant differences between PD MCI and both PD non-MCI (P <.010, Bonferroni correction) and HC (P <.001, Bonferroni correction).
medial and lateral areas. Performance on VFD, JLO, and TMT B was associated only with medial temporal-parietal atrophy. Atrophy in anterior regions, mostly left superior frontal gyrus, also correlated with Stroop Test, TMT A, SDMT, phonemic fluency, and RAVLT total learning. In addition, negative correlations with SDMT and JLO were also found in rostral middle frontal gyrus (Supplemental Data Fig. 1 and Table 2). After controlling for the effects of age, education, and sex, only Stroop Words Test and Semantic fluency showed a significant positive correlation with cortical thickness. Stroop Words Test correlated with left medial orbitofrontal, right superior temporal, and right insula. Semantic fluency correlated with right precuneus and lingual gyrus thickness. (Fig.  2 and Supplemental Data Table 3). No significant results were seen after, including disease duration as a nuisance factor in the covaried model. MCI patients' scores were significantly worse than non-MCI patients' and healthy controls' (p<.05, Bonferroni correction).  ANCOVA analyses with age, education, and eTIV as confounding variables. Global cortical thickness showed significant differences between HC and PD MCI patients (P < 0.006, Bonferroni correction). Total GM showed significant differences between HC and PD non-MCI patients (P 5 0.031, Bonferroni correction). Mean lateral ventricular volumes showed significant differences between HC and PD MCI patients (P 5 0.046, Bonferroni correction). MCI, mild cognitive impairment; HC, healthy controls; PD, Parkinson's disease; GM, gray matter; SD, standard deviation; ANCOVA, analysis of covariance; eTIV, estimated total intracranial volume.

Discussion
Patients with MCI showed a posterior pattern of atrophy, characterized by cortical thinning in the bilateral superior parietal and supramarginal regions and in the inferior temporal area, parahippocampal gyrus, fusiform gyrus, and precuneus. Cortical thinning was also observed in the left rostral frontal region. Moreover, PD MCI and PD non-MCI patients differed in right lateral parietal regions and precuneus.
Our results of cortical thickness reductions in the parietal, temporal and frontal regions could be related to previous positron emission tomography data 42 showing that PD-MCI patients exhibited reduced fluorodeoxyglucose (FDG) uptake in the parietal and occipital lobes and in localized areas of the frontal and temporal lobes compared with controls. In addition, longitudinal neuropsychological studies suggest that the dementia process is heralded by posteriorcortically-based cognitive deficits. 43,44 In sum, parieto-occipital changes seem to correspond to the deterioration that may eventually lead to dementia, possibly reflecting gradual loss of synaptic terminals, dendritic arborization, and size of neuronal cell bodies.
In our study, PD groups differed by degree of atrophy in right lateral parietal regions and the precuneus. This finding, together with the right asymmetry of atrophy detected in the PD non-MCI group compared with controls, suggested an asymmetric pattern of deterioration, initially involving parietal and temporal regions, and progressively widespread to bilateral atrophy. Asymmetric atrophy has been found in previous studies when PD patients were compared with controls. 28 Recently, PD MCI patients have been studied using cortical thickness measures. However, no consensus has been reached in relation to cortical thickness differences between PD MCI and PD non-MCI; both increases and decreases have been reported in PD MCI, probably because of small sample sizes. [29][30][31] As expected, we observed thinning in PD MCI involving the supramarginal gyrus and the precuneus.
In addition to differences in regional cortical thickness, PD MCI patients had a reduction of global GM volume together with increased ventricular volume in comparison with PD non-MCI and healthy controls. These results, similar to those of previous studies, [46][47][48] confirm that cognitive deficits seen in PD MCI are related to structural brain changes, probably in combination with neurochemical changes.
The current study also aimed to establish whether specific neuropsychological tests used in clinical practice reflect the degree of regional atrophy in PD patients. Cognitive domains are defined under the assumption that they represent specific functions mediated by specific brain regions. The 'anterior' (frontal) pattern is putatively associated with executive functions, the 'posterior' (temporo-parietal) pattern with visuospatial and visuoperceptual functions, and hippocampal degeneration with the amnestic pattern. 49 Grouping different tests into a single function without knowing the specific correlates of each test may generate confusion. The study of neuroanatomical correlates of specific tests is the first step in the discussion of domains and in determining whether subtypes of mild cognitive impairment exist in PD, and consequently whether they are useful in predicting the evolution to dementia. In line with this statement, our results showed a common posterior atrophy pattern for all of the neuropsychological tests evaluated. Only the Stroop Test, SDMT, and phonemic fluency, classified by recent guidelines 10 as measures of attention and executive function, respectively, had an extended pattern including medial anterior regions. We did not observe any specific dorsolateral prefrontal or limbic pattern of correlations. Analyzing the global patterns of correlations, a group of tests including Stroop Test (Words, Colours, and Word-Colours), TMT A, semantic fluency, and RAVLT total learning correlated with extensive GM loss involving bilateral medial and lateral cortical regions, whereas VFD, JLO, and TMT B only correlated with medial temporal-parietal regions. Analyses of covariance only showed a positive correlation between semantic fluency and temporal-parietal regions and a positive correlation between Stroop words and left medial orbitofrontal, right superior temporal, and right insular regions. In sum, we did not find specific patterns of atrophy related to the neuropsychological domains, either with or without the use of covariates in the analyses.
Interestingly, semantic fluency performance remained significantly correlated with the precuneus after controlling for the effects of age, education, and sex. These results agree in part with current pathophysiological models that dissociate the substrates and prognostic values of different types of cognitive impairment in PD and show that tests with posterior cortical bases (semantic fluency and ability to draw an interlocking pentagon) reflecting probable nondopaminergic cortical Lewy body or Alzheimer's type pathology were associated with dementia, whereas frontostriatal executive deficits were not. 44 Our results indicate that semantic fluency is an easily administered test that should be included in the neuropsychological assessment of PD patients.
The consecutive recruitment of PD patients from an outpatient movement disorders clinic involves certain differences in clinical and demographical variables between PD groups. In this regard, disease duration was longer in PD MCI than in PD non-MCI patients. In our study, additional analyses including disease duration as a nuisance variable in the co-varied models did not show significant results. A previous study has shown that neurodegeneration is likely to occur faster in PD patients with MCI. 31 Cortical degeneration was more advanced in patients who have PD MCI than in those without at the same stage of disease. In accordance with previous longitudinal studies, [43][44][45] disease duration is correlated with PD patients' deterioration, including cognitive impairment and brain atrophy. In this sense, controlling for the effects of disease duration could represent an overcorrection, masking actual intergroup effects.
One possible limitation of our study is that, despite the inclusion of a variety of tests in the main cognitive domains defined in recent guidelines (attention and working memory, executive functions, memory, and visuospatial and visuoperceptive functions), 8 we did not include the same number of tests in each cognitive domain, and language was assessed only with the Boston Naming Test. This limitation may have raised the possibility of false-negative cases in the PD non-MCI group but would not have affected the classification of subjects currently included in PD MCI group.
In sum, degeneration was characterized by a posterior pattern of atrophy mostly involving posterior parietal-temporal areas, which extended to frontal regions in the PD MCI group compared with healthy controls. All of the neuroanatomical correlates associated with neuropsychological tests involve this posterior pattern of atrophy, semantic fluency being the neuropsychological test with the most significant association. Our findings therefore suggest the presence of posterior structural degenerative brain changes in PD MCI patients, evidencing a structural neuroimaging marker of this pathological condition and its possible evolution to dementia.