Abstract
Abstract
ABBREVIATIONS
Brodmann area Parkinson’s disease fluorodeoxyglucose positron emission tomography Mattis dementia rating scale puncorrected
INTRODUCTION
In Parkinson’s disease (PD), cognitive impairment can occur early in the course and is heterogeneous [1]. Even though the incidences of mild cognitive impairment (MCI) and dementia have been well characterized [2–5], little is known about the worsening of cognitive impairment for a given patient after the “honeymoon period", i.e. once the motor signs are bilateral and motor complications are present.
It would be useful if data on brain metabolism were predictive of the risk of greater cognitive impairment - especially since deep brain stimulation is now a treatment option for this type of patient [6]. At present, most of the prognostic information on the possible conversion of MCI to dementia comes from clinical observations, including neuropsychological parameters (e.g. semantic fluency, and the ability to copy intersecting pentagons) and the motor phenotype (e.g. the non-tremor dominant subtype) [7].
Brain metabolic profiles of PD patients with MCI or dementia are now available [8]. Patients with MCI show prefrontal and parietal hypometabolism [9, 10], which extends to the parietal, occipital and temporal areas in patients with dementia [8]. However, only one study has explored prognostic metabolic factors for future cognitive alterations in PD. Bohnen et al. [11] observed significant hypometabolism in the occipital cortex and the posterior cingulate cortex in six patients with PD and dementia 1.9 to 6.0 years after diagnosis [10]. However, disease progression and initial cognitive performance varied significantly among the 23 PD patients studied [11] so these data provide little prognostic information on the future cognitive status of advanced PD patients.
The primary objective of the present longitudinal study was thus to determine whether a specific brain metabolic pattern was associated with the occurrence of cognitive decline in a very homogeneous group of patients with advanced PD.
METHODS
Study participants underwent [18F]-fluorodeoxyg-lucose positron emission tomography ([18F]-FDG PET) of the brain. To maintain homogeneity of our population, the inclusion criteria were as follows (i) a Mattis Dementia Rating Scale (MDRS) score of 130 or more out of 144 recorded in the 6 months preceding the PET acquisition [12], (ii) levodopa response for 5 years or more to exclude atypical forms of parkinsonism [13], (iii) a Hoehn and Yahr stage ≥2 [14] and (iv) the presence of complications of disease and therapies [15].
The participants had been on their usual, stable medication regimen for at least 3 months prior to the PET acquisition. PET procedures were all performed under off-drug conditions, after the withdrawal of dopaminergic therapy for at least the preceding 12 hours [16]. As part of routine clinical follow-up, 16 patients were scored for MDRS about two years after the PET acquisition. Patients were then defined as “cognitive decline group” (defined as an MDRS score of below 130 or a loss of more than five points [12], and an overall assessment by a the neuropsychologist) or “cognitive stable group” free of overall cognitive decline. All cognitive evaluations were performed in on-state.
At baseline, we evaluated each patient’s brain metabolism of [18F]-FDG at rest. The participant lay quietly in the supine position during the tracer accumulation period. An emission scan was initiated 30 minutes after injection of the radiotracer.
The study’s objectives and procedures were approved by the local independent ethics committee (CPP Nord-Ouest IV, Lille, France; reference 2010-A01329-30) and each participant gave his/her prior, written, informed consent to participation in the study. The study sponsor was Lille University Hospital (Lille, France).
DATA ACQUISITION AND ANALYSIS
PET data
All participants underwent PET scans at Lille University Hospital’s Nuclear Medicine Department. Data were acquired on an Advance SL PET/CT device (GE Medical Systems, Chalfont St. Giles, UK) with a 4 to 5 mm full-width at half-maximum and a 30 cm transaxial field of view. Participants were instructed to fast before the scans, and the blood glucose level was always checked prior to intravenous injection of 185 to 198 MBq [18F]-FDG. Thirty minutes later, a low-dose CT scan of the brain was acquired for attenuation correction of the PET data. Emission images were subsequently acquired in three-dimensional mode. Images were reconstructed iteratively using an ordered-subset expectation-maximization algorithm (with two iterations and 21 subsets) in a 256 × 256 matrix. The same acquisition and reconstruction procedures were used for all patients.
Data processing and statistical analysis were performed using the SPM8 package (Wellcome Department of Cognitive Neurology, London, UK), implemented in Matlab 7.9 (MathWorks Inc., Sherborn, MA, USA). All the reconstructed [18F]-FDG brain PET images were spatially normalized into the Montreal Neurological Institute standard template (McGill University, Montreal, Canada) using an affine transformation (with 12 parameters for rigid transformations) [17]. In order to increase the signal-to-noise ratio, the images were then smoothed by convolution with an isotropic Gaussian kernel and a 12-mm full-width at half-maximum. Overall normalization was applied by including each subject’s mean overall activity as a covariate of no interest.
Whole-brain analyses were performed in all cases. Clusters of at least 50 contiguous voxels with a threshold two-tailed p value of 0.001 were considered to be statistically significant.
In a second step, to test the predictive value at the individual level, we performed a 3D-SSP t statistic maps analysis for each patient (CortexID®), with normalization to the pons [11]. The PD subject data were compared with age stratified normals on a pixel-by-pixel basis. Two sample t-statistic values were calculated for each pixel and then converted to corresponding z values, presented in a color scale (Fig. 1).
Statistical analysis
For clinical data, intergroup differences were evaluated with an unpaired Student’s t test for continuous variables with a normal distribution and the Wilcoxon test for variables with a non-normal distribution. A chi-squared test was used for categorical variables. The threshold for statistically significance was set to p < 0.05. All analyses of clinical data were performed with IBM SPSS for Windows software (version 16.0, IBM, Armonk, NY, USA).
Functional imaging data were evaluated in a voxel-wise manner using the flexible factorial design in SPM8. We included each of the following variables in turn: subject and group (decline, stable). We tested the group effect by using the “compare-populations one scan/subject” routine: for each voxel, a simple fixed-effect T test was used to compare two groups.
Secondly, three different raters (CT, FD, and AD) classified patients as having or not a significantposterior hypometabolism at baseline, blind to their cognitive evolution. The intraclass correlation coefficient Fleiss [18] was calculated in order to evaluate inter-rater reliability. We then calculated sensitivity, specificity, positive and negative predictive values of the posterior hypometabolism to predict cognitive decline (based on clinical and neuropsychological evaluation at the second evaluation).
RESULTS
Population (Supplemental Table 1)
The two groups did not differ in terms of baseline age, disease duration, levodopa equivalent daily dose, score on the Unified Parkinson’s Disease Rating Scale (I, II, III, including the axial subscore, and IV), the MDRS, or the duration of follow-up (p > 0.05). At baseline, the mean MDRS was 140 for the decliners and 138 for the stables. At the second assessment (after about 30 months), the mean MDRS score was 128 for the decliners and 139 for the stables. Four of the 6 patients in the decliner group and none of the patients in the stable group met the criteria for dementia at the second visit [19]. One patient in each group had initiated deep brain stimulation between the two MDRS assessments.
PET
Comparison of the two populations according to the clinical follow-up (Fig. 2)
Compared with the stable group, the decliner group showed hypometabolism in several clusters: the left precuneus (BA31, 177 voxels, (–10, –64, 18), puncorrected (punc) <0.001), the right precuneus (including BA23, cingulate gyrus, 74 voxels, (12, –46, 32), punc < 0.001), the left middle temporal gyrus (BA21; 83 voxels, (–70, –32, –8), punc < 0.001) and the left fusiform gyrus (BA37; 87 voxels, (–44, –64, –12), punc < 0.001) (presented in the MNI space).
Due to the small sample size, only the left precuneus tended to withstand correction for multiple testing (family-wise-error 0.094).
Individual FDG PET data at baseline compared with age-stratified normals on a pixel-by-pixel basis
The sensitivity of a posterior hypometabolism to predict a cognitive decline according to our definition was 77.2% , with a specificity of 46.7% . The positive predictive value was 44.8% and the negative predictive value was 73.7% . The intraclass correlation coefficient was 0.706 (p = 0.002).
DISCUSSION
Our results suggest that cortical hypometabolism in the two precunei, the left middle temporal and fusiform gyri is associated with cognitive decline in hitherto cognitively intact patients with advanced PD. Pathological changes in Parkinson’s disease occur in a predictable sequence, with Lewy body deposition first in the frontal regions of the cortex and then in occipital and temporal regions [20]. The present neural correlates of cognitive decline argue in favour of a posterior impairment. As previously described [7], the impairment in anterior executive function usually seen in PD [1] does not seem be predictive of cognitive decline, whereas more posterior cortical impairment (involving the precuneus) is predictive.
The precunei and the left middle temporal and fusiform gyri are involved in visuoconstructive functioning [21]. Indeed, the precunei play a central role for the integration of visuospatial sensory information and behaviour [22] whereas the left middle temporal and fusiform gyri participate in the ventral visual stream involved in object recognition [23]. This finding fits with the fact that the previously described difficulty in figure-copying is a predictor of conversion to dementia [2]. It’s also necessary to highlight that only 6 of 16 patients were decliners, this low power may explain the failure to demonstrate relative bilateral baseline FDG reductions in the decliner subgroup rather than the left sided emphasis seen here.
A posterior hypometabolism in advanced PD patients has a good sensitivity and a good negative predictive value to predict a further cognitive decline. At the opposite, the specificity and the positive predictive value were low. In case of neuropsychological data at the lower band (close to 130 for MDRS, [6], the normality of the resting [18F]-FDG PET could be an argument to further discuss the eligibility of the patient to deep brain stimulation. At the opposite, an abnormal [18F]-FDG PET can’t be considered as contraindication for deep brain stimulation. False positive patients could also be explained by a too short follow-up period.
The present study had some obvious limitations: the sample size was small and the follow-up time was short (2.5 years, on average). However, the decliners and stables did not differ in terms of baseline clinical characteristics. Hence, assessment of hypometabolism in the areas identified here may be of valuable for routinely detecting patients at risk of cognitive decline.
Our findings need to be confirmed in prospective studies of consecutive patients. Indeed, it was an exploratory study (proof of concept) but a longer follow-up is necessary to perform an appropriate survival analysis. Ultimately, [18F]-FDG PET brain imaging at rest may have a role in the treatment strategy as a biomarker of cognitive decline, in adjunction to an extensive neuropsychological analysis for example before deep brain stimulation discussion [6, 24].
CONCLUSION
We demonstrated that the occurrence of a significant cognitive decline in advanced PD patients was associated with a hypometabolism of the two precunei, the left middle temporal and fusiform gyri a few yearspreviously (i.e. at a time when their cognitive status was maintained). [18F]-FDG PET brain imaging may be a useful tool in the multidisciplinary management of advanced PD patients.
CONFLICTS OF INTEREST
None.