Abstract
INTRODUCTION
Dementia with Lewy Bodies (DLB) is the second most common neurodegenerative dementia with an incidence of 3.5/100.000 persons each year [1]. It is characterized by parkinsonian syndrome, cognitive impairments and sleep disorders, variably associated at onset, frequently accompanied by hallucinations, fluctuations in alertness, and autonomic failures [2]. Functional loss is also typical of the syndrome, differentially related to motor and cognitive dysfunctions [3]. Neuroimaging with Positron Emission Tomography (PET) and Single-Photon Emission Tomography (SPECT) is considered supportive for diagnosis [2]. 18F-FDG-PET can detect specific patterns of brain hypometabolism involving particularly the occipital cortex, whereas DAT-SCAN can prove a loss of Dopaminergic Transporters Activity (DAT) in the basal ganglia.
REM sleep behavior disorder (RBD) is a parasomnia characterized by a loss of normal muscle atonia and by complex motor activity during REM sleep, usually associated with dream mentation [4]. RBD can occur in the absence of any other obvious associated neurologic disorder or in association with a neurodegenerative disease, in which case it is considered as symptomatic RBD. RBD is frequently associated with Parkinson’s disease (PD) [5–7], DLB [8], and Multiple System Atrophy (MSA) [9, 10]. In several cases it may even antedate the occurrence of motor symptoms by decades [11].
Polysomnography (PSG) confirmation of RBD is essential to make the diagnosis of definite RBD, but when PSG confirmation of RBD is not feasible in all cases and by investigators who are not able to perform sleep studies, it is possible to make the diagnosis of clinically probable RBD or probable RBD (pRBD) [12, 13].
Aims of this study were: (i) to evaluate the presence of pRBD in a sample of DLB patients with an ad hoc RBD questionnaire (RBD1Q) [14] testing also for its sensitivity in distinguishing DLB from AD cases in early phase, and (ii) to assess the patterns of brain glucose hypometabolism using 18F-FDG-PET in DLB patients, with and without sleep disturbances.
METHODS
Subjects
With regards to the RBD1Q evaluation, we included 38 patients of which 27 DLB patients (mean age ± std: 72.40 ± 7.87) and 11 AD patients (66.54 ± 8.02) fulfilled the respective criteria for clinical diagnosis [2, 15].
To limit possible artifacts due to different sample sizes in the 18F-FDG-PET metabolic comparison between DLB patients with and without RBD, we considered an additional set of patients (N = 13) with a clinical diagnosis of DLB [2], and a negative history for sleep disturbances as ascertained by close relative (when possible with partners). A total of 20 DLB RBD+ patients were thus compared to 20 DLB RBD–patients (of which 7 RBD- were part of the questionnaire assessment). All the patients were admitted at the Department of Neurology, San Raffaele Hospital and Vita-Salute San Raffaele University, Milan, Italy. Informed written consent was obtained for all the patients. The protocol was approved by the San Raffaele Hospital Local Ethical Committee. Descriptive statistics and offline analysis were performed with SPSS software (http://www.ibm.com/spss, v.20). Demographic summary is available in Table 1.
REM sleep behavior disorder assessment
The presence of pRBD was ascertained by a self-administered RBD Single-Question Screen (RBD1Q) [14], followed by a sleep-structured interview by experts in sleep disorders (LFS, SM), in the presence of bed partner.
The RBD1Q consists of a single question, answered “yes” or “no,” as follows: “Have you ever been told, or suspected yourself, that you seem to ‘act out your dreams’ while asleep (for example, punching, flailing your arms in the air, making running movements, etc.)?”. This question has been validated in several languages (French, German, Japanese, Italian, Spanish, Czech, and Danish) by native-speaking medical translators or by RBD expert investigators fluent in English and the local language [14].
Structured interview and RBD1Q data were available for all the patients. Interviewers were blinded to clinical information at the sleep interview and RBD1Q administration time.
18F-FDG-PET functional imaging
Acquisition
All subjects underwent an 18F-FDG-PET imaging scan, using a multi-ring General Electric Discovery STE PET/CT at the Nuclear Medicine Unit, San Raffaele Hospital, Milan, Italy, following a standard clinical acquisition protocol described elsewhere in detail [16].
Preprocessing
Image preprocessing was carried out with SPM5 software (Wellcome Department of Imaging Neuroscience, London, UK; http://www.fil.ion.ucl.ac.uk/spm) on MATLAB 8 (MathWorksInc, Sherborn, Mass). Images were first spatially normalized using a custom optimized 18F-FDG “dementia-specific” template [17] and then underwent a Gaussian Kernel smoothing (FWHM: 8 mm) prior statistical comparisons.
Single-subject and group analysis
Normalized and smoothed 18F-FDG-PET images in each patient were voxel-wisely compared to a large validated normal controls database as previously reported [16]. Briefly, a two-sample t-test design, implemented in SPM, was adopted to evaluate whole-brain hypometabolism at individual level, entering age as a nuisance variable. Cerebral global mean computation (CGM) and proportional scaling (PS) were used to remove global variation in PET intensities. Family-Wise Error (FWE) correction for multiple comparisons at pFWE < 0.05 threshold was adopted, limiting the minimum significant cluster size to k = 100. This study design and significance was used also to verify the DLB pattern of hypometabolism at the group-level.
Two experts in neuroimaging and two experts in sleep disorders checked the SPM-t maps at the individual level and the clinical records, respectively. A cross-check comparison was performed afterwards to identify the DLB cases with SPM typical hypometabolism pattern positive or negative for history of sleep disturbances or to the RBD1Q questionnaire.
18F-FDG-PET comparison between DLB RBD+ and DLB RBD– groups
We sought to investigate the brain metabolic differences in DLB patients with or without sleep disturbances by mean of group comparisons. These were first run with a 1st level two-sample t-test implemented in SPM5 software, comparing directly DLB RBD+ and DLB RBD– patients 18F-FDG-PET images and taking into account age and disease duration (as reported duration of cognitive deficits prior hospital admittance, CDO), with CGM computation and proportional scaling. An additional 2nd level analysis with a two-sample t-test was also run, comparing the DLB RBD+ and DLB RBD–patients individual contrast images delivered by the SPM single-subject analysis. Age and disease duration were taken into account. For both the above analyses, statistical significance was set to p < 0.025 with a minimum cluster size of k = 100, not corrected for multiple comparisons. To further examine the difference between presence of sleep disturbances and metabolic changes, we also assessed separately DLBRBD+ and DLB RBD– group-level hypome-tabolism in comparison to the normal database, taking into account age differences, with CGM computation and proportional scaling. Statistical significance was set to p < 0.05 FWE-corrected for multiple comparisons with a minimum cluster size of k = 100.
CSF assessment
Cerebrospinal fluid (CSF) measurements for amyloid-β (Aβ), total-tau, and phospho-tau were also evaluated, considering published pathological cut-off values to assess biomarker positivity [18], namely (i) Aβ42 <500 ng/L for amyloid, (ii) t-tau >350 ng/L for total-tau, and (iii) p-tau >61 ng/L for phospho-tau.
RESULTS
REM sleep behavior disorder assessment
The RBD1Q questionnaire identified 20/27 DLB RBD+ and 7/27 DLB RBD–. None of the AD patients was positive to the RBD1Q test, therefore leading to high discrimination abilities of the questionnaire for diagnosis of pRBD (1.00 specificity, 0.74 sensitivity, and 0.82 accuracy).
Within-group demographic comparisons
With regards to DLB patients, RBD1Q positivity was not associated with statistically significant demographic differences except for the CDO, which was significantly higher in DLB RBD+ patients (Mann-Whitney non-parametric independent samples test p = 0.002) (see Table 1 for details). As a further evaluation, we compared in the DLB RBD+ patients the CDO and the time from onset of sleep disorders (SDO). The medians of the two distributions were similar (CDO 3 yrs, SDO 2.87 yrs) and the non-parametric Wilcoxon signed rank test for matched pairs was not statistically significant (p = 0.507). More in detail, the slight majority (11/20, 55%) of the DLB RBD+ patients showed cognitive symptoms before RBD onset (see Fig. 1).
18F-FDG-PET functional imaging
Brain hypometabolism in single-subject and group analysis
Each DLB patient showed at 18F-FDG-PET SPM map brain posterior hypometabolism, involving temporo-parietal and occipital associative cortices, reported in literature as the metabolic hallmark of the disease, supporting diagnosis [2]. In addition, some patients also showed clusters of reduced metabolism in dorsolateral prefrontal cortex. The same hypometabolism pattern was confirmed at the group level (see Fig. 2).
18F-FDG-PET comparison between DLB RBD+ and DLB RBD– groups
The first and second level analyses consistently revealed a more severe metabolic decrease in the DLB RBD+ with respect to DLB RBD–patients, in the dorsolateral and medial frontal regions, left precuneus, bilateral superior parietal lobule and rolandic operculum, and amygdala (pUNCORRECTED < 0.025). These differences were not related to age or disease duration (see Fig. 3).
In addition, the separate comparisons between the DLB RBD+ and DLB RBD–group-level hypometabolism and normal database revealed a more extensive involvement in posterior brain regions in the DLB RBD+ group (see Supplementary Figure 1).
CSF assessment
CSF measurements were available for 34 patients: 14 DLB RBD+, 10 DLB RBD–, and 10 AD.
We found significant differences among Aβ, total-tau, and phospho-tau CSF levels in AD compared to DLB patients (Mann-Whitney non-parametric independent samples test: p = 0.010, p = 0.002, and p = 0.013, respectively; see Table 1). In the whole DLB group, ~71% of the patients (n = 17/24) were amyloid positive. Of these 17, 4 patients were also positive for pathological total-tau and phospho-tau levels. As for comparisons between DLB RBD+ and DLB RBD–patients, there were no significant differences (Aβ p = 0.437, total-tau p = 0.709, and phospho-tau p = 0.796, Mann-Whitney U test). Eight DLB RBD–patients (80%) were amyloid positive, and three of them were also positive for total-Tau and phophos-Tau. Within the DLB RBD+ group, 64% (n = 9/14) of the patients were amyloid positive and only one patient was also positive for total-Tau and phospho-Tau.
DISCUSSION
This 18F-FDG-PET imaging study in DLB, using a voxel-based SPM approach at single-subject level, showed the specific and typical DLB brain metabolic pattern in each individual DLB patient. A number of studies in literature have reported brain glucose hypometabolism in DLB, highlighting the specific occipital pattern that is considered supportive for clinical diagnosis in the currently adopted criteria [2]. In the present study, by using an optimized 18F-FDG-PET SPM analysis, we confirmed the pattern of predominant occipital hypometabolism in each DLB case, which extended to parietal, temporal and dorsolateral prefrontal cortex, possibly following long-distance occipito-frontal deafferentations and alpha-synuclein spreading.
With regards to the evaluation of RBD1Q and the comparisons between the AD and DLB dementia groups, 74% of the evaluated DLB patients but none of the AD patients were affected by pRBD. This is coherent with the rarity of RBD in AD-pathology and its frequent occurrence in synucleinopathy conditions [19], witnessed by its inclusion among the current DLB diagnostic criteria [2]. It has been suggested that RBD1Q could be used for the assessment of RBD prevalence in broad-scale epidemiologic surveys of disease [19]. Our study seems to indicate that RBD1Q may also be useful in the clinical practice in the evaluation of demented subjects. It is of note that some patients (n = 5) had a very long history of RBD symptoms (up to 10 years) with only a more recent manifestation of cognitive impairments (2–3 years history) (see Fig. 1). Therefore, these subjects first had an idiopathic RBD (iRBD) presentation and then after several years developed cognitive deficits, progressing to a DLB diagnosis. A recent observational cohort study with postmortem assessment reported in 44 iRBD patients the development of DLB (n = 14), PD (n = 16), MSA (n = 1), and mild cognitive impairment (MCI) (n = 5) within a median of 12 years (range 3–23) between estimated RBD onset and final diagnosis [20]. This predementia phase in alpha-synuclein pathology is characterized by iRBD that has been shown to begin even 50 years before the appearance of the other motor and/or cognitive symptoms [21]. Notably, in our sample, the majority of the DLB RBD+ patients (11/20, 55%) showed cognitive symptoms before RBD onset. It is known that DLB patients might show no RBD before development of dementia [22]. Moreover, a diagnostic delay of RBD (8.7 ± 11 years) has been reported in relation to mild or infrequent occurrence of sleep behavior [23]. DLB RBD+ patients presented a CDO (reported duration of cognitive deficits prior hospital admittance) significantly longer than DLB RBD- patients, despite a comparable Mini-Mental State Examination score (see Table 1). This result seems to suggest that, for DLB RBD- patients, it takes a shorter time to reach a significant degree of cognitive decline, while it takes longer for DLB patients with RBD. This is consistent with the hypothesis by [24], suggesting that these two clinical phenotypes might reflect two different underlying pathologies. In particular, DLB RBD- patients are more likely to bear a mixed AD-DLB pathology, thus possibly leading to a faster cognitive decline.
To the best of our knowledge, there is no evidence in literature on metabolic differences in DLB patients with and without sleep disturbances. Here, we showed that significant differences in the amount of brain glucose metabolism may occur. The DLB RBD+ with respect to DLB RBD–patients showed a more severe metabolic decrease in the posterior typically affected cerebral regions and particularly in the bilateral superior parietal lobule and rolandic operculum, and in the left precuneus, and in addition in the dorsolateral and medial frontal regions and amygdala. These results survived the statistical corrections for age and disease duration. The amygdala is known for its involvement in synucleinopathies [25, 26]. A previous study found a significant negative association between Lewy bodies burden and amygdala volume [27], while this relationship was only trending toward significance in another study [28]. The present findings consistently show a more severe hypometabolism in DLB RBD positive patients, who are expected to carry a more “pure” synuclein pathology [24].
With regards to the metabolic involvement of posterior parietal and frontal regions, given their role in dorsal and ventral visual streams and their relation with presence of visual hallucinations in DLB [29, 30], we also assessed whether visual hallucinations were differentially present within the DLB RBD+ and DLB RBD– groups. The DLB RBD+ patients had a slightly higher frequency of concomitant visual hallucinations (50%, 10/20) with respect to the DLB RBD–group (35%, 7/20). This difference, however, was not significant (Fisher’s Exact test, p = 0.523). Therefore, in the present findings, visual hallucinations represent a limited factor. Further studies on a larger dataset will be necessary to shed more light on this crucial aspect. The brain metabolic differences we found between DLB RBD+ and RBD– groups might also reflect different patterns of both underlying pathologies and of the associated neurotransmission alterations. Consistently, in a previous study, Kotagal and co-workers reported PD RBD+ patients with a more severe neocortical, limbic, and thalamic cholinergic denervation with respect to the PD RBD– as measured in vivo with 11C-DTBZ [31]. The authors concluded that the presence of RBD symptoms might index a degeneration of the cholinergic system [31]. Given the common neuropathology (alpha-synuclein accumulation) in DLB and PD [32, 33], we speculate that the present 18F-FDG-PET findings might indicate a more severe alteration of the cholinergic system in DLB with RBD. Previous studies investigated the differences in white- and gray-matter abnormalities between early PD patients with or without RBD [34, 35]. With regards to brain atrophy, there were some structural differences, namely greater degeneration in early PD with RBD in several cerebral regions. A postmortem study in AD and DLB cases showed that in autopsy-confirmed cases with and without RBD, there were variable atrophy patterns [28]. The RBD negative cases presented more severe atrophy in AD-signature regions, and a higher Braak neurofibrillary tangle stage. The pathologic probability of clinical DLB was, however, higher in the RBD positive group, as well as the higher frequency of parkinsonism and visual hallucinations. Therefore, different studied cohorts reported variable findings at difference to the present study where all the RBD positive and negative patients had a clinical diagnosis of probable DLB and comparable frequency of parkinsonism and visual hallucinations (see Table 1).
The present findings show higher CSF amyloid in DLB RBD+ patients when compared to DLB RBD–patients, even if this difference was not statistically significant (p = 0.437 Mann-Whitney U test). These findings may still indicate that DLB RBD+ patients may present with a more “pure” synuclein-driven pathology, with respect to a greater concomitant Alzheimer pathology in DLB RBD–patients [24]. On the basis of the neuropathological differences found between DLB RBD+ and DLB RBD– patients, some authors have suggested different DLB subtypes [24]. Future neuroimaging studies in DLB larger samples, with or without RBD, measuring either amyloid- or tau-burden and neurodegeneration (18F-FDG-PET) might be of utmost interest for the identification of possible DLB subtypes.
A limitation of our study is the lack of PSG evaluation. However, other recent studies enrolled patients with the diagnosis of “pRBD” in absence of PSG study [28, 37].
Some studies reported that clinical characteristics differ between PD with and without RBD [38, 39]. Our study also shows that 18F-FDG-PET findings differ between DLB with and without RBD. As recently suggested for PD patients [40], the identification of different subtypes of neurodegenerative disorders also in relation to the presence of RBD, could be of fundamental importance for potential neuroprotective intervention with disease-modifying therapy.
DISCLOSURE STATEMENT
Authors’ disclosures available online (http://j-alz.com/manuscript-disclosures/15-1000r1).