Introduction
Isolated REM sleep behaviour disorder (iRBD) patients are at high risk of developing clinical syndromes of the α-synuclein spectrum [
1]. Most iRBD patients (> 90%) will develop Parkinson’s disease (PD), dementia with Lewy bodies (DLB), or (rarely) multiple system atrophy (MSA) within 5–20 years after diagnosis [
2]. iRBD patients are ideal candidates for potential disease-modifying trials [
3]. However, the progression and clinical trajectory rates are heterogeneous [
4], making it difficult to predict which iRBD patient will develop a particular condition and when. Longitudinal neuroimaging studies may be of use in making such predictions.
18F-2-Fluoro-2-deoxy-
d-glucose positron emission tomography (
18F-FDG PET) allows assessment of cerebral glucose metabolism and is used in the workup of parkinsonism [
5] and dementia [
6]. Using single-subject statistical parametric mapping (SPM) t-maps, expert readers can identify typical patterns of hypometabolism in individual cases [
7]. This semi-quantitative approach can be used to distinguish between PD and MSA [
8] and to predict the development of dementia in PD patients [
9]. Quantitative methods include univariate voxel-wise or region of interest (ROI)–based analyses. Glucose metabolism changes have also been investigated in several neurodegenerative disorders with spatial covariance analysis, which provides disease-related patterns that disclose the network-level brain changes in specific conditions [
10]. Such patterns are applied case-by-case to calculate to which degree an individual’s scan is compatible with the disease pattern.
Several
18F-FDG PET studies have been performed in iRBD ([
11] for review). However, most of these studies are cross-sectional and based on group analysis, and the results are conflicting. Longitudinal neuroimaging studies are scarce. Kim et al. [
12] applied spatial covariance and regional analyses to scans of 20 iRBD patients at baseline and after 2 and 4 years. They found that metabolism in the putamen increased over time, and that metabolism in the motor cortices and frontal cortex decreased. These changes were related to conversion to PD/DLB. A study of brain perfusion in 37 iRBD patients scanned at baseline and after 1 year found that regional hypoperfusion in the frontal cortex normalised over time, perhaps suggesting early compensatory mechanisms [
13]. Contradictory to cross-sectional studies [
11] and two older longitudinal perfusion studies [
14,
15], Baril et al. [
13] and Kim et al. [
12] did not find the expected increases in the hippocampus or decreases in the occipital cortex over time.
Spatial covariance analysis has also been applied to
18F-FDG PET in iRBD by us and other authors [
12,
16‐
21]. We previously investigated the expression of a PD-related network (PDRP) in a cohort of 20 iRBD patients scanned twice, approximately 4 years apart [
21]. This was characterised by a relatively increased activity in the cerebellum, pons, thalamus, putamen/globus pallidus, and motor cortex and moderately decreased metabolism in the lateral premotor and parieto-occipital cortex. We showed that PDRP expression increased from baseline to follow-up and that a greater rate-of-change of PDRP expression is associated with conversion to PD. However, in that study, the contribution of single brain regions to PDRP expression was not addressed. The abovementioned inconsistencies warrant further evaluation of regional longitudinal brain metabolic changes. So far, no study has evaluated the topography of hypometabolism and hypermetabolism patterns over time in iRBD patients at the individual level.
This study evaluates the longitudinal regional brain changes in iRBD with
18F-FDG PET. Although most iRBD patients share a final common clinical syndrome, their trajectories towards this common phenotype may differ. Therefore, it is useful to investigate regional changes on a case-by-case basis instead of at the group level. We hypothesised that this heterogeneity would be reflected in their individual brain hypometabolism and hypermetabolism maps. Specifically, we expected different topographical changes over time in iRBD patients who converted to PD at follow-up compared to those who did not. To this end, we applied an optimised statistical parametric mapping (SPM) procedure [
7] to obtain single-subject hypo/hypermetabolism maps at baseline and follow-up (approximately 4 years later), which were visually rated by experts. In addition, we investigated the progression of regional hypometabolism without an a priori hypothesis using two data-driven univariate approaches (voxel-based and ROI analyses). We also evaluated the contribution of regional changes to PDRP expression over time and their relationship to clinical follow-up measures. Furthermore, cardiac noradrenergic sympathetic and presynaptic dopaminergic nigrostriatal innervations were assessed in 17 out of 20 subjects, using
123I-MIBG and
123I-FP-CIT SPECT, respectively.
Discussion
This study describes longitudinal brain metabolic changes in a cohort of 20 iRBD patients scanned at baseline and after 4 years. Four patients converted to PD during follow-up. Although most iRBD patients share a final common clinical syndrome, their trajectories towards this common phenotype may differ. We hypothesised that this heterogeneity might be reflected in the changes in brain hypometabolism and hypermetabolism individual maps over time. To understand the individual trajectories of our iRBD patients, we analysed SPM t-maps at the single-subject level, which revealed three scenarios: (1) a normal
18F-FDG PET scan at both baseline and follow-up; (2) a normal scan at baseline, with occipital/occipito-parietal hypometabolism at follow-up; and (3) occipital hypometabolism at both baseline and follow-up. Subjects who converted to PD during the study pertained to this third scenario. The presence of occipital hypometabolism on the SPM t-map at baseline gave a 93.6% accuracy (sensitivity: 1.000 and specificity: 0.875) for the prediction of conversion to PD during follow-up. It must be mentioned that these values (accuracy, sensitivity, and specificity) should be interpreted with caution. The high accuracy of occipital hypometabolism in identifying converters might be overestimated because of the sample size; it needs validation in a larger cohort of patients to estimate its effective performance as a predictive biomarker. However, ROC curves at least suggest—in accordance with the single-subject t-maps’ visual rating—an early occipital vulnerability in patients with a short-term conversion (3–4 years) to PD. Occipital hypometabolism is a hallmark feature of DLB [
29] and is associated with a faster progression towards dementia in PD [
9]. This might suggest that our converters may develop a phenotype with more pronounced cognitive deterioration (PD dementia) as the disease progresses. It is still largely unknown what drives occipital hypofunction [
30]. It is probably not related to α-synuclein aggregation in the occipital brain regions [
31], but perhaps to an epiphenomenon related to the degeneration of cholinergic projections to the occipital cortex [
32‐
34]. Importantly, our results show that initially negative
18F-FDG PET scans can become abnormal in just 4 years. Four of the 20 iRBD subjects originally had a normal scan but developed occipito/occipito-parietal hypometabolism at follow-up. Amongst the iRBD converters, we observed the presence of occipital hypometabolism 3–4 years before conversion to PD (years between baseline and conversion: mean ± SD: 3.63 ± 0.49, range: 3.14–4.16). This suggests that patients who show occipital hypometabolism (at follow-up) are likely to phenoconvert in the upcoming years.
Ten patients (50%) had a normal
18F-FDG PET scan at both baseline and follow-up. We speculate that a negative
18F-FDG PET scan in iRBD could indicate three possible scenarios: (1) the patient does not have an α-synucleinopathy, (2) the patient has an α-synucleinopathy but are early in the disease course, or (3) the patient will develop PD within a few years but will only have limited cognitive symptoms. The
123I-FP-CIT SPECT and
123I-MIBG SPECT results support the third scenario: two iRBD patients with a normal
18F-FDG PET scan at baseline and follow-up had abnormal
123I-FP-CIT and
123I-MIBG SPECT at baseline. Pathological values of
123I-MIBG SPECT highly indicate the presence of prodromal PD or DLB [
35,
36]. Moreover,
123I-FP-CIT SPECT abnormalities suggest that a short-term phenoconversion (3–5 years) can be expected [
37]. At follow-up, our two patients had a dopaminergic deficit for at least 3–4 years (pathological
123I-FP-CIT SPECT already at baseline), thus having a high risk of conversion in the upcoming years. However, contrary to the other patients with abnormal
123I-FP-CIT SPECT at baseline (Figures
S3 and
S4), they did not develop occipital hypometabolism at follow-up (Figure
S2). On this basis, we speculate that these iRBD subjects may develop a PD subtype without dementia
. A previous longitudinal
18F-FDG PET study in PD has shown that PD patients who do not deteriorate cognitively during long-term clinical follow-up have either a completely normal
18F-FDG PET scan or very limited cortical hypometabolism, for instance, in the sensory-motor cortex [
9]. It could be argued that having a normal
18F-FDG PET scan has a limited negative predictive power for phenoconversion in the longer term (approximately > 4 years), but this conclusion requires further long-term follow-up. In addition, these findings could also indicate that abnormal
123I-FP-CIT and
123I-MIBG SPECT reflect earlier events compared to
18F-FDG PET (all patients with an abnormal
18F-FDG PET scan also had an abnormal
123I-FP-CIT and
123I-MIBG SPECT), representing sensitive early biomarkers for the underlying pathological process.
Our longitudinal results explain why findings in cross-sectional
18F-FDG PET studies are heterogeneous and sometimes conflicting [
38]. iRBD is intrinsically heterogeneous: in any cohort, a variable proportion of patients will develop DLB, PD, or MSA at an unknown time interval. In a cross-sectional setting, each individual is scanned at a different time point on his/her disease progression towards manifest PD/DLB (or MSA). This aspect can explain why different cohorts of iRBD patients present variable prevalence of abnormal and normal brain
18F-FDG PET scans. In our study, 50% of iRBD showed normal brain
18F-FDG PET scans. This differs from what was described in a previous cross-sectional study (only 5 iRBD patients showed normal brain scans) [
39]. The two cohorts might have patients in different disease stages and/or clinical trajectories. The prevalence of abnormal scans in our iRBD population might increase over time with the disease progression. On the other hand, a higher percentage of negative scans can also represent a higher percentage of patients converting to a more benign form of PD without cognitive deterioration [
9]. Another source of variability may be the presence of compensatory mechanisms. Baril and colleagues [
13] found significant regional hypoperfusion in the anterior frontal and lateral parietal-temporal cortex, which disappeared at follow-up. We found a similar trajectory in one of our cases (Figure
S2, case 17). This patient showed hypometabolism in the frontal cortex at baseline, which was normalised completely on follow-up imaging. The cause of this phenomenon is unclear. All the above underscores the necessity of longitudinal studies with a large sample and multiple brain scan acquisitions to understand the neurodegenerative process underlying this proteinopathy spectrum.
The SPM t-maps of hypermetabolism showed a high degree of topographical variability. Specifically, hypermetabolism variably involves the cerebellum, sensory-motor cortex, thalamus, putamen, and limbic regions (hippocampus and parahippocampus). Such relative increases have been regarded as artefacts of global mean normalisation [
40], but there is also evidence that hypermetabolic areas are important reflections of the pathophysiological process [
41,
42]. Findings of hypermetabolism in iRBD have been inconsistent, possibly due to the aforementioned heterogeneity of iRBD cohorts. In one study, hyperperfusion of the hippocampus was also found at baseline and was associated with conversion to PD or DLB within 3 years of follow-up [
14]. Kim et al. studied
18F-FDG PET scans of 20 iRBD patients at baseline and after 2 and 4 years and found progressive hypermetabolism in the bilateral putamen, cuneus, and lingual gyrus. Hypermetabolism in the bilateral putamen was associated with phenoconversion (3 to PD, 4 to DLB). In our cohort, only 2 out of 6 patients with occipital hypometabolism at baseline did not convert during follow-up. These two patients were also the ones without any significant hypermetabolism at both time points (Figure
S4, case 2 and case 5), suggesting that hypermetabolism may also have a role in predicting short-term phenoconversion. Indeed, the four patients who converted at follow-up all showed hypermetabolism in the cerebellum, sensory-motor cortex, and limbic structures and either putamen, thalamus or globus pallidus (Figures
S4 and
S5).
The results of the voxel-wise and ROI-based analyses mirrored those of the single-subject SPM t-maps. These data-driven analyses also showed a significant progression in hypometabolism at follow-up in frontal, occipital, and parietal regions and hypermetabolism in cerebellar and limbic regions (bilateral hippocampus and parahippocampus). Again, hypometabolism in occipito-parietal regions was significantly more severe in converters compared to non-converters at both baseline and follow-up. The presence of baseline hypometabolism in the right occipital middle gyrus, left inferior occipital gyrus, or right angular gyrus gave the highest accuracy (AUC: 0.922) compared to the other tested ROIs (lingual gyrus, superior occipital gyrus, and parahippocampus) for identifying the converters at follow-up (Table
S3). In addition, iRBD converters showed a faster progression of hypermetabolism in the cerebellum and cerebellar vermis.
The brain regions with a significant progression of hypo- or hypermetabolism resemble the topography of the PDRP (Fig.
3). The PDRP is consistently characterised by relatively increased activity in the cerebellum, pons, thalamus, putamen/globus pallidus, and motor cortex and relatively decreased metabolism in the lateral premotor and parieto-occipital cortex [
10]. We previously described PDRP expression
z-scores in this cohort [
21]. We showed that PDRP expression increased from baseline to follow-up and that rates-of-change (Δ/years) in UPDRS-III scores were significantly correlated with rates-of-change (Δ/years) in PDRP
z-scores (Fig.
4). However, in that study, the contribution of single brain regions to PDRP expression in individuals with iRBD was not addressed. In the current study, we show that PDRP
z-score Δ/years is correlated with the progression of hypometabolism in the occipital cortex and hypermetabolism in the cerebellar vermis (Fig.
4). These results suggest that the increased expression of PDRP in iRBD patients over time is mainly driven by progressive hypometabolism in occipital regions and progressive hypermetabolism in the cerebellum.
This study has some limitations, mainly related to the small sample size and relatively short follow-up. Furthermore, the workup of cognitive changes was limited as we did not perform an extensive neuropsychological evaluation. This limits the assessment of hypometabolic changes in relation to cognitive deterioration and the risk of DLB. Moreover, we do not have repeated measures for age-matched healthy controls for comparison. That said, our results remained significant even after correction for an age effect.
In conclusion, our results suggest that occipital hypometabolism at baseline is a risk factor for short-term conversion from iRBD to manifest PD. Increases in PDRP expression are mainly driven by progressive hypometabolism in this region and progressive cerebellar hypermetabolism. We have shown that repeat scanning is pertinent in this condition and that single-subject SPM t-maps can help map the individual trajectories of patients. Also, our findings support using multiple biomarkers for the stratification of iRBD patients. To further explore the predictive value of 18F-FDG PET semi-quantifications and other molecular biomarkers (e.g. 123FP-CIT SPECT), a larger, multicentre longitudinal 18F-FDG PET study in iRBD is ongoing.
Acknowledgements
REMPET Working Group list of authors: Jan Booij, MD, PhD, Department of Radiology and Nuclear Medicine, Amsterdam University Medical Centers, location Academic Medical Center, Amsterdam, the Netherlands; Kathrin Reetz, MD, Department of Neurology and JARA-BRAIN Institute Molecular Neuroscience and Neuroimaging, RWTH Aachen University, Aachen, Germany; Sebastiaan Overeem, MD, PhD, Kempenhaeghe Sleep Medicine Center, Heeze, the Netherlands; Angelique Pijpers, MD, PhD, Kempenhaeghe Sleep Medicine Center, Heeze, the Netherlands; Fanni Geibl, PhD, MD, Department of Neurology, Philipps-Universität Marburg, Marburg, Germany; Martin Henrich, MD, Department of Neurology, Philipps-Universität Marburg, Marburg, Germany
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.