Association of DTI-ALPS Glymphatic Index With Differential Phenoconversion in Isolated REM Sleep Behavior Disorder

Abstract

Background and Objectives

Isolated REM sleep behavior disorder (iRBD) is the strongest prodromal marker of synucleinopathies, including Parkinson disease (PD) and dementia with Lewy bodies (DLB). Identifying brain biomarkers that predict progression and distinguish phenoconversion trajectories remains a challenge. The glymphatic system is involved in interstitial waste clearance, and its dysfunction has been associated with pathologic protein accumulation and neurodegeneration. Diffusion tensor imaging along the perivascular space (DTI-ALPS) has been proposed as a noninvasive proxy for glymphatic function. The aim of this study was to determine whether patients with iRBD show a reduced DTI-ALPS index compared with controls and whether a lower DTI-ALPS index predicts future phenoconversion to PD or DLB.

Methods

We conducted a longitudinal, multicenter cohort study using brain MRI scans from patients with polysomnography-confirmed iRBD and healthy controls recruited across 5 international centers. All participants underwent T1-weighted and diffusion-weighted MRI. DTI-ALPS indices were computed from diffusivity along projection and associative fibers adjacent to the lateral ventricles. The primary outcome was time to phenoconversion to synucleinopathy. Linear models assessed baseline group differences and clinical correlates, and Cox proportional hazard models assessed the predictive value of DTI-ALPS for time to phenoconversion.

Results

A total of 250 patients with iRBD (mean age: 66.5 ± 6.8 years; 87% male) and 178 controls (65.7 ± 6.8 years; 81% male) were included. Patients with iRBD showed a lower left DTI-ALPS index compared with controls (mean difference = −0.034, 95% CI −0.067 to −0.001; p = 0.043). Of 224 patients with iRBD followed for a mean of 6.1 ± 3.5 years, 65 phenoconverted to a synucleinopathy. Converters had a lower left DTI-ALPS index than nonconverters (mean difference = −0.050, 95% CI −0.098 to −0.003; p = 0.038). Lower left DTI-ALPS index was associated with an increased risk of conversion to PD over time (hazard ratio = 2.43, 95% CI 1.13–5.25; p = 0.012). Other diffusion metrics inside periventricular masks, namely fractional anisotropy, diffusivity metrics, and free water, did not differ between groups.

Discussion

Patients with iRBD exhibit a reduced DTI-ALPS index, suggesting altered glymphatic function. This reduction was associated with future phenoconversion to PD, supporting the DTI-ALPS index as a potential prognostic MRI biomarker of progression in prodromal synucleinopathies.

Introduction

Isolated/idiopathic REM sleep behavior disorder (iRBD) is a parasomnia characterized by the loss of normal muscle atonia during REM and the onset of abnormal and often violent movements and vocalizations.¹ Approximately 90% of patients with iRBD will develop either dementia with Lewy bodies (DLB), Parkinson disease (PD), or, more rarely, multiple system atrophy (MSA) after 15 years.² Despite this prevalence, predicting whether individuals will develop PD or DLB remains challenging. Several studies using brain imaging demonstrated that patients with iRBD already display neurodegenerative changes reminiscent of overt synucleinopathies.^3-5 However, little is known about the neurodegenerative mechanisms underlying the progression from iRBD to overt synucleinopathies, and phenoconversion-specific imaging markers remain scarce. Such imaging markers are needed for stratifying patients with iRBD based on similar brain disease and benchmarking treatment efficacy against an objective brain-based process.

Synucleinopathies are characterized by abnormal alpha-synuclein accumulation, which aggregates and disrupts cellular function.⁶ This buildup of proteins has been linked to impaired brain clearance mechanisms normally helping remove toxic waste products.⁷ One hypothesized clearance pathway is the glymphatic system, a brain-wide network facilitating waste removal.⁸ In this system, primarily active during non-REM sleep,⁹ CSF enters the brain through periarterial routes and is cleared along perivenous pathways.¹⁰ The glymphatic system allows CSF to flow into perivascular spaces around arteries through diffusion and pulsatile flow, where it mixes with interstitial fluid surrounding neurons, carrying away waste, ions, neurotransmitters, and molecules such as neuropeptides that influence neuronal function. Central to this process is aquaporin-4, a protein located in astrocytes, supporting glial cells in the brain. Astrocytes extend their vascular endfeet to blood vessels, where aquaporin-4 facilitates the movement of CSF between perivascular spaces and interstitial fluid, enabling efficient waste clearance. The waste-laden fluids are ultimately transported out of the brain along perivascular spaces around veins and drained into meningeal and deep cervical lymphatic systems outside the brain, supporting brain health.

Defects in the glymphatic system have been associated with the development and progression of several neurodegenerative diseases,^11,12 including PD, Alzheimer disease (AD), and amyotrophic lateral sclerosis.^9,13,14 In these conditions, impaired glymphatic function may lead to the accumulation of harmful proteins and waste products in the brain,¹⁵ suggesting impaired clearance mechanisms as a contributing factor of pathology and neurodegeneration. Advanced diffusion MRI techniques, specifically diffusion tensor imaging along the perivascular space (DTI-ALPS), provide a noninvasive approach to assess interstitial fluid dynamics by measuring water diffusivity in the periventricular space.¹³ Studies using this proxy identified glymphatic system alterations in PD and DLB, correlating with motor and cognitive impairments in PD.^16-18 Of interest, a unilateral onset of glymphatic index impairment has been reported in PD, being found on the left side in early de novo cases and then later involving the right side as disease progresses.¹⁶ However, it remains unclear whether these changes, such as the hemispheric effect found in PD, are exclusive to the overt phase of synucleinopathies or whether they may emerge in the prodromal stage.

In patients with iRBD, single-site studies have suggested the presence of a reduced DTI-ALPS index compared with controls,^19-22 which is associated with cortical and subcortical atrophy and clinical features.²² These studies were conducted on limited samples of patients with polysomnography-confirmed iRBD (18, 20, and 62 patients),^20,21 or with RBD diagnosed using questionnaires.¹⁹ One study also showed no differences between groups using an automated approach of DTI-ALPS analysis.⁵ Regarding prognosis, glymphatic dysfunction might influence phenoconversion from iRBD to synucleinopathies,²¹ but it remains unknown whether glymphatic index alterations predict differential phenoconversion toward PD or DLB in iRBD. Addressing this gap is important because predictive biomarkers for specific neurodegenerative trajectories in iRBD are needed to monitor disease progression and create homogeneous groups with similar phenotypes for clinical trials.²³

In this study, we examined whether individuals with iRBD exhibit altered DTI-ALPS indices, using a large, multicenter data set of diffusion-weighted MRI scans. We also investigated whether these alterations could predict distinct phenoconversion trajectories over time in iRBD. We hypothesized that the DTI-ALPS index would be significantly lower in patients with iRBD compared with healthy controls and that a lower DTI-ALPS index would be associated with an increased risk of phenoconversion to an overt synucleinopathy.

Methods

Participants

The multicenter iRBD cohort included patients with video-polysomnography–confirmed iRBD and healthy controls recruited from the Centre for Advanced Research on Sleep Medicine at the Centre intégré universitaire de santé et de services sociaux du Nord-de-l'Ile-de-Montreal (CIUSSS-NIM) - Hôpital du Sacré-Coeur de Montréal, Montreal, Canada; the Oxford Discovery Cohort, Oxford, United Kingdom; First Faculty of Medicine at Charles University, Prague, Czechia; the Movement Disorders clinic (ICEBERG and ALICE Cohorts) at the Hôpital de la Pitié-Salpêtrière, Paris, France; and the Parkinson's Progression Markers Initiative study.²⁴ Recruitment was conducted independently at each site through ongoing prospective research cohorts, typically based in clinical sleep centers or movement disorder clinics. Patients were referred for video-polysomnography based on clinical suspicion of iRBD and, on confirmation, were invited to participate in longitudinal cohort studies that included annual clinical evaluations and MRI acquisition. All patients with iRBD had a diagnosis confirmed by overnight video-polysomnography in accordance with the third edition of the International Classification of Sleep Disorders.²⁵ This required repeated episodes of motor behaviors or vocalizations during sleep; video-polysomnography confirmation that these occurred during REM sleep; evidence of REM sleep without atonia; and exclusion of alternative explanations such as epilepsy, medication effects, or other mimics. Patients were invited to take part in site-specific research protocols, which included MRI acquisition and annual follow-up evaluations. For this work, only patients with iRBD and control participants who had both brain MRI (T1-weighted and diffusion-weighted imaging) and a clinical evaluation within a close time frame to the scan were included. All patients with iRBD were in the isolated phase of the disease at the time of the clinical evaluation closest to the MRI scan (average 7.7 ± 25.1 days). Each patient with iRBD was followed longitudinally with annual neurologic and cognitive assessments to determine the point of phenoconversion to a clinically defined overt synucleinopathy. Phenoconversion to PD, DLB, or MSA was determined by movement disorders neurologists using established criteria.^26-28 All participants underwent the Montreal Cognitive Assessment (MoCA) to assess global cognition²⁹ and the motor scale of the Movement Disorder Society–Unified Parkinson's Disease Rating Scale Part III (MDS-UPDRS-III) to evaluate the severity of parkinsonian motor features.³⁰

Data Acquisition

Diffusion-weighted MRI scans in the Montreal cohort were acquired using a 3T Siemens TIM Trio scanner with a 12-channel head coil and echo-planar sequence (b-values of 0 and 700 s/mm²; gradient directions of 63 and 64; isotropic voxel size of 2 mm; repetition time [TR] ranging from 8.6 to 12.7 milliseconds; echo time [TE] ranging from 0.083 to 0.1 milliseconds) or using a 3T Siemens PRISMA scanner with a 32-channel head coil and echo-planar sequence (b-values of 0 and 1,000 s/mm²; gradient directions of 30; isotropic voxel size of 2 mm; TR = 6.9 milliseconds; TE = 64 milliseconds). The Oxford cohort was scanned using a 3T Siemens Trio scanner with a 12-channel head coil and echo-planar sequence (b-values of 0 and 1,000 s/mm²; gradient directions of 60; isotropic voxel size of 2 mm; TR = 9.3 milliseconds; TE = 94 milliseconds). The Prague cohort was scanned using a 3T Siemens Skyra scanner with a 32-channel head coil and echo-planar sequence (b-values of 0 and 1,000 s/mm²; gradient directions of 29 or 30; isotropic voxel size of 2 mm; TR = 10.5 milliseconds; TE = 93 milliseconds). The Paris cohort was scanned using a 3T Siemens TIM Trio scanner with a 12-channel head coil and echo-planar imaging (b-values of 0 and 700 s/mm²; gradient directions of 21 directions; isotropic voxel size of 1.72 mm; TR = 14 milliseconds; TE = 101 milliseconds) or using a 3T PRISMA Fit scanner with a 64-channel head coil and echo-planar imaging (b-values of 0 and 700 s/mm²; gradient directions of 32 directions; isotropic voxel size of 1.7 mm; TR = 10.4 milliseconds; TE = 59 milliseconds).

The T1-weighted MRI scans in the Montreal cohort were acquired using a magnetization-prepared rapid gradient-echo (MPRAGE) sequence with TR = 2,300 milliseconds; TE = 2.91 milliseconds; flip angle = 9°; and isotropic voxel size of 1 mm or with TR = 2,300 milliseconds; TE = 2.98 milliseconds; flip angle = 9°; and isotropic voxel size of 1 mm. The Oxford cohort used MPRAGE sequence with TR = 2,040 milliseconds; TE = 4.7 milliseconds; flip angle = 8°; and isotropic voxel size of 1 mm. The Prague cohort used MPRAGE sequence with TR = 2,200 milliseconds; TE = 2.4 milliseconds; flip angle = 8°; and isotropic voxel size of 1 mm. The Paris cohort used MPRAGE sequence with TR = 2,300 milliseconds; TE = 4.18 milliseconds; flip angle = 9°; and isotropic voxel size of 1 mm or MP2RAGE with TR = 5,000 milliseconds; TE = 2.98 milliseconds; flip angles = 4° and 5°; generalized autocalibrating partial parallel acquisition (GRAPPA) = 3; and isotropic voxel size of 1 mm. The acquisition parameters for the Parkinson's Progression Markers Initiative study have been described elsewhere.²⁸ These T1-weighted MRI scans were part of previous multicentric studies in iRBD.^3,31,32

Diffusion MRI Processing

Diffusion-weighted brain MRI scans were processed using the TractoFlow-Atlas-Based Segmentation (ABS) automated pipeline (Figure 1 shows the imaging protocol).³³ Brain extraction tool (BET) was used to remove nonbrain tissue from the images, and eddy currents and head motion artifacts were corrected using the eddy tool. A brain mask was extracted from the non–diffusion-weighted (b0) image, and the diffusion tensor model was fit to the corrected data with the dtifit tool. This generated fractional anisotropy (FA), mean diffusivity, and directional diffusivity maps along the left-right direction (Dxx), anterior-posterior direction (Dyy), and superior-inferior direction (Dzz). All images were visually inspected in a blinded fashion using FSLeyes, and data with significant artifacts were excluded.

Figure 1. Processing Steps Involved in Deriving the Glymphatic Index.

(A) Diffusion-weighted and T1-weighted MRI scans were processed using the TractoFlow-ABS pipeline. (B) Using the fractional anisotropy map template from the ICBM DTI-81 atlas, 4 masks were positioned on the associative and projection fibers at the level of the lateral ventricle body of each hemisphere and were warped to every participant's native space. Diffusivity measurements were extracted and harmonized for scanner effects using ComBat. The DTI-ALPS index was calculated for each participant as a proxy of glymphatic function. The flowchart is shared courtesy of the TractoFlow development team. ABS = atlas-based segmentation; ALPS = along the perivascular space; BIDS = Brain Imaging Data Structure; DTI = diffusion tensor imaging; DWI = diffusion-weighted imaging; fODF = fiber orientation distribution function; FRF = fiber response function; iRBD = isolated REM sleep behavior disorder; T1w = T1-weighted MRI scan.

For quantitative DTI-ALPS index derivation, measurements were extracted from the diffusivity maps. Using the FA map template from the ICBM DTI-81 atlas,^34,35 we generated four 5-mm cubic masks that were positioned on associative (Montreal Neurological Institute [MNI] coordinates: x = 39, y = −17, and z = 29 on the left and x = −39, y = −17, and z = 29 on the right) and projection (MNI coordinates: x = 26, y = −17, and z = 29 on the left and x = −26, y = −17, and z = 29 on the right) fibers at the level of the lateral ventricle body in each hemisphere, as performed previously.¹³ These masks were registered to each participant's native space using Advanced Normalization Tools (ANTs), and visual quality control was performed on the red-green-blue (RGB) map to ensure accurate coverage of the regions of interest. The masks were then applied to each participant's Dxx, Dyy, and Dzz maps from TractoFlow-ABS to extract diffusivity measurements for left and right associative and projection fiber masks. For each participant, the DTI-ALPS index was calculated as follows, blinded to group assignment, based on Taoka et al.¹³

$$\text{DTI}-\text{ALPS}=\frac{\text{mean}\left({\text{Dxx}}_{\text{proj}},{\text{Dxx}}_{\text{assoc}}\right)}{\text{mean}\left({\text{Dyy}}_{\text{proj}},{\text{Dzz}}_{\text{assoc}}\right)}$$ (1)

To account for the multicentric nature of this data set and the varying acquisition parameters across different scanners, we used the ComBat algorithm,^36,37 a standard tool in neuroimaging studies for harmonizing data across sites, which has been validated in previous studies using T1-weighted and diffusion-weighted MRI scans.³⁶

To determine whether the observed group differences were specific to the DTI-ALPS index, we also extracted other diffusion metrics from the same white matter regions used to calculate the DTI-ALPS index. Standard tensor-based measures, namely FA, mean diffusivity, axial diffusivity, and radial diffusivity, were obtained as part of the TractoFlow-ABS pipeline used in the preprocessing of diffusion-weighted data. In addition, free water content was estimated using the FreeWaterFlow pipeline, which applies a 2-compartment model to separate isotropic (extracellular) from anisotropic (tissue-based) diffusion.³⁸ In this model, free water values approaching 0 indicate restricted water within tissue, whereas values closer to 1 reflect increased extracellular diffusion, potentially indicative of edema or inflammation. Free water–corrected versions of FA, mean diffusivity, axial diffusivity, and radial diffusivity were also computed to account for extracellular contamination in the tensor-derived metrics.

Statistical Analysis

Analyses were performed in IBM SPSS (version 29.0.1.0). Independent samples t tests assessed group differences in age, follow-up time, and DTI-ALPS indices (left/right) between patients with iRBD and controls, and between converters and nonconverters. Chi-squared tests compared gender distribution, and Mann-Whitney U tests were used for non-normally distributed variables (MoCA and MDS-UPDRS-III scores). Paired t tests were used to assess presence of significant asymmetry in DTI-ALPS indices. Spearman correlations tested associations between DTI-ALPS indices and clinical measures (MoCA and MDS-UPDRS-III scores), with false discovery rate (FDR) correction applied for multiple comparisons. To assess phenoconversion risk, binary logistic regression evaluated whether DTI-ALPS predicted conversion to PD or DLB (vs disease-free), adjusting for age, sex, MoCA score, and MDS-UPDRS-III score. Cox proportional hazard models tested the longitudinal impact of DTI-ALPS on conversion risk over time, with age and sex as covariates and z-scored DTI-ALPS indices as predictors. Likelihood ratio tests assessed model fit. Kaplan-Meier analysis evaluated survival proportion in patients divided into high/low left DTI-ALPS index groups (median split), with the log-rank test used to assess significance for overall phenoconversion, as well as PD vs disease-free and DLB vs disease-free trajectories. Finally, t tests compared other diffusion measures (free water, mean diffusivity, axial diffusivity, FA, as well as free water–corrected versions of mean diffusivity, axial diffusivity, and FA) between patients with iRBD and controls, and between converters and nonconverters, to assess specificity of DTI-ALPS findings.

Standard Protocol Approvals, Registrations, and Participant Consents

All study protocols were approved by local ethics committees, and all participants provided written informed consent. This study was approved by the Research Ethics Boards of the McGill University Health Centre (MP-37-2022-7744) and the CIUSSS du Nord-de-l’Île-de-Montréal (MEO-37-2024-2699).

Data Availability

The data used in this study were obtained from multiple collaborating centers, each of which retains ownership of their respective data sets. The principal investigator had authorized access to all data necessary for the analyses performed in this study. However, the accessibility and sharing of data are subject to the local policies and restriction criteria of each center involved. As such, data availability is restricted, and requests for access should be directed to the respective institutions, pending their specific data access and sharing guidelines.

Results

Participants

The multicenter iRBD cohort initially included 592 participants, namely 289 patients with polysomnography-confirmed iRBD and 303 healthy controls. After matching for age and sex, a total of 471 participants were included, namely 276 patients with iRBD and 195 controls, with 129 (75 patients) from the Centre for Advanced Research on Sleep Medicine at the CIUSSS-NÎM – Hôpital du Sacré-Coeur de Montréal, Montreal, Canada; 120 (73 patients) from the Oxford Discovery Cohort, Oxford, United Kingdom; 119 (74 patients) from First Faculty of Medicine at Charles University, Prague, Czechia; 76 (45 patients) from the Movement Disorders clinic (ICEBERG and ALICE Cohorts) at the Hôpital de la Pitié-Salpêtrière, Paris, France; and 27 (9 patients) from the Parkinson's Progression Markers Initiative study.²⁴ After quality control, 43 participants (9.1%, 26 patients with iRBD) were excluded because of diffusion-weighted imaging processing failures, mask misregistration over fibers, and extreme values, yielding a final sample of 250 patients with iRBD and 178 controls for analysis. There were no significant differences between the remaining iRBD and control groups regarding age at MRI acquisition (p = 0.26) or gender distribution (p = 0.10). Patients with iRBD showed significantly lower MoCA scores (iRBD group: 25.4 ± 3.0 vs controls: 26.6 ± 2.3, p < 0.001) and higher MDS-UPDRS-III scores (iRBD group: 6.3 ± 5.6 vs controls: 3.8 ± 6.0, p < 0.001) (Table 1). Excluded participants with iRBD did not differ from those retained in age (p = 0.27), sex (p = 0.81), MDS-UPDRS-III scores (p = 0.17), and MoCA scores (p = 0.49).

Table 1.

Demographic, Clinical, and DTI-ALPS Variables of Patients With iRBD and Controls

Variables	iRBD group (n = 250)	Controls (n = 178)	p Value
Age at MRI, y	66.5 ± 6.8	65.7 ± 6.8	0.26a
Sex, men, n (%)	217 (87)	144 (81)	0.10b
MoCA score	25.4 ± 3.0	26.6 ± 2.3	<0.001 c
MDS-UPDRS-III score	6.3 ± 5.6	3.8 ± 6.0	<0.001 c
DTI-ALPS index, left	1.288 ± 0.198	1.322 ± 0.203	0.043 a
DTI-ALPS index, right	1.264 ± 0.181	1.286 ± 0.185	0.10a

Abbreviations: DTI-ALPS = diffusion tensor imaging along the perivascular space; iRBD = isolated REM sleep behavior disorder; MDS = Movement Disorder Society; MoCA = Montreal Cognitive Assessment; UPDRS-III = motor scale of the Unified Parkinson's Disease Rating Scale.

Values are presented as mean ± SD. Values in bold represent significant differences.

^aStudent's independent 2-sample t test.

^bChi-squared test.

^cMann-Whitney U test.

DTI-ALPS Index Decreased in Patients With iRBD

We first tested group differences in the DTI-ALPS index. Patients with iRBD showed a significantly lower left DTI-ALPS index than controls (1.29 ± 0.20 vs 1.32 ± 0.20, p = 0.043, Cohen d = 0.17) (Table 1, Figure 2). No difference was found for the right DTI-ALPS index (p = 0.10). Paired t tests revealed significant hemispheric differences in both groups (p < 0.001), but this asymmetry did not differ between groups (p = 0.31). These results indicate a statistically lower DTI-ALPS index in patients with iRBD.

Figure 2. Comparison of the DTI-ALPS Index Between Patients With iRBD and Controls.

(A) Violin plots showing the distribution of DTI-ALPS indices (left and right) in healthy controls (blue) and patients with iRBD (orange). The left DTI-ALPS index is significantly lower in patients with iRBD compared with controls. (B) Violin plots showing the distribution of DTI-ALPS indices (left and right) in nonconverted iRBD (blue) and phenoconverted iRBD (orange) groups. The left DTI-ALPS index is significantly lower in the phenoconverted iRBD group compared with the nonconverted iRBD group. The plot includes boxplots within each violin, displaying the median (middle line) and interquartile range of DTI-ALPS indices. ALPS = along the perivascular space; DTI = diffusion tensor imaging; iRBD = isolated REM sleep behavior disorder.

Lower DTI-ALPS Index Is Associated With Lower Cognitive Function

We examined associations between the DTI-ALPS index and clinical measures (MoCA and MDS-UPDRS-III scores). In patients with iRBD and controls, MoCA scores were not associated with left or right DTI-ALPS indices (all p > 0.07). Across all participants, significant but weak correlations emerged between MoCA scores and both left (r = 0.11, p = 0.026) and right (r = 0.11, p = 0.036) DTI-ALPS indices, but neither remained significant after adjusting for age and sex (p = 0.16 and p = 0.39, respectively) or FDR correction (p_FDR > 0.20) (eFigure 1). No significant correlations were found between MDS-UPDRS-III scores and DTI-ALPS indices in the full sample or subgroups (all p > 0.07).

Lower DTI-ALPS Index Is Associated With Phenoconversion in iRBD

Of the 250 patients with iRBD, 224 (90%) were followed longitudinally for an average of 6.1 ± 3.5 years (range: 1–16). At the last follow-up, 65 (29%) developed a synucleinopathy (converters) while 159 (71%) remained disease-free (nonconverters). Converters and nonconverters did not differ in age at MRI (p = 0.81), sex (p = 0.80), or MDS-UPDRS-III scores (p = 0.24), but converters had lower MoCA scores (24.9 ± 2.9 vs 25.9 ± 2.8, p = 0.011) (Table 2). The left DTI-ALPS index was significantly lower in converters (1.25 ± 0.19 vs 1.30 ± 0.20, p = 0.038, Cohen d = 0.26) (Figure 2, Table 2), with no difference on the right (p = 0.24). Paired t tests revealed that nonconverters showed significant hemispheric asymmetry (p < 0.001) while converters did not (p = 0.82), suggesting that a loss of normal left-right asymmetry in the DTI-ALPS index may be associated with disease progression in patients with iRBD.

Table 2.

Demographic, Clinical, and DTI-ALPS Variables of iRBD Converters and Nonconverters

Variables	iRBD nonconverters (n = 159)	iRBD converters (n = 65)	p Value
Age at MRI, y	66.2 ± 7.2	66.4 ± 6.2	0.81a
Sex, men, n (%)	140 (88)	58 (89)	0.80b
MoCA score	25.9 ± 2.8	24.9 ± 2.9	0.011 c
MDS-UPDRS-III score	6.2 ± 5.9	6.8 ± 5.7	0.24c
DTI-ALPS index, left	1.304 ± 0.198	1.254 ± 0.193	0.038 a
DTI-ALPS index, right	1.269 ± 0.181	1.250 ± 0.186	0.24a

Abbreviations: DTI-ALPS = diffusion tensor imaging along the perivascular space; iRBD = isolated REM sleep behavior disorder; MDS = Movement Disorder Society; MoCA = Montreal Cognitive Assessment; UPDRS-III = motor scale of the Unified Parkinson's Disease Rating Scale.

Values are presented as mean ± SD. Values in bold represent significant differences.

^aStudent's independent 2-sample t test.

^bChi-squared test.

^cMann-Whitney U test.

Lower DTI-ALPS Index Is Specifically Associated With PD

Among the 65 iRBD converters, 42 (65%) developed PD, 18 (28%) DLB, 4 (6%) MSA, and 1 (2%) AD. To assess whether DTI-ALPS indices were associated with specific conversion trajectories over time, we performed logistic regressions comparing PD and DLB converters separately with nonconverters. PD and DLB converters did not differ in age, sex, or MDS-UPDRS-III scores (all p > 0.24), but DLB converters had lower MoCA scores (23.1 ± 2.7 vs 25.7 ± 2.7, p < 0.001). In a model including left DTI-ALPS index, age, and sex, the left ALPS index significantly predicted time to conversion to PD vs disease-free (B = −0.42, p = 0.029, OR = 0.65, 95% CI = 0.45–0.96) (Table 3), but not to DLB (p = 0.90). The right DTI-ALPS index was not predictive of either trajectory (p > 0.33). When comparing PD or DLB converters with all other outcomes, the left DTI-ALPS index was still a predictor of PD, but not DLB, conversion (eTable 1). When adding MoCA score to the model, the left DTI-ALPS index remained a significant predictor of PD conversion (B = −0.39, p = 0.045, OR = 0.68, 95% CI = 0.46–0.99), with 32% reduced odds per SD increase (i.e., more preserved glymphatic function). Adding MDS-UPDRS-III score instead of MoCA score yielded similar findings (B = −0.41, p = 0.035, OR = 0.66, 95% CI = 0.45–0.97). The left DTI-ALPS index did not predict DLB conversion in any model (p > 0.89). Kaplan-Meier analysis showed that patients with a low left DTI-ALPS index (median split) had a higher risk of phenoconversion over time compared with those with a high index (p = 0.034) (Figure 3). Cox regression confirmed this association (HR = 1.93, 95% CI = 1.05–3.57, p = 0.035), independent of age and sex (eTable 2). Subtype-specific Cox regressions further demonstrated that a lower left DTI-ALPS index (median split) predicted PD conversion over time (HR = 2.43, 95% CI = 1.13–5.25, p = 0.023), independent of age and sex (Figure 3, eTable 3), but not DLB conversion (p = 0.83) (eFigure 2, eTable 4).

Table 3.

Logistic Models of the Association Between ALPS Index and Trajectories in iRBD

Model term	Conversion to PDa			Conversion to DLBa
	Coefficient (SE)	p Value	OR (95% CI)	Coefficient (SE)	p Value	OR (95% CI)
Model 1: DTI-ALPS
DTI-ALPS index, left	−0.424 (0.20)	0.029	0.654 (0.447–0.958)	0.03 (0.26)	0.901	1.03 (0.62–1.71)
Age	−0.003 (0.03)	0.92	1.00 (0.95–1.05)	0.03 (0.04)	0.42	1.03 (0.96–1.11)
Sex	−0.27 (0.67)	0.68	0.76 (0.21–2.80)	0.37 (0.70)	0.59	1.45 (0.37–5.70)
Constant	−1.19 (1.77)	0.50	0.30	−4.22 (2.51)	0.09	0.02
Model 2: MoCA score
DTI-ALPS index, left	−0.389 (0.19)	0.045	0.678 (0.46–0.99)	−0.004 (0.28)	0.999	1.00 (0.58–1.72)
Age	0.003 (0.03)	0.92	1.00 (0.95–1.06)	0.01 (0.04)	0.80	1.01 (0.93–1.09)
Sex	−0.13 (0.68)	0.84	0.88 (0.23–3.30)	0.45 (0.77)	0.56	1.56 (0.35–7.03)
MoCA score	−0.003 (0.07)	0.968	1.00 (0.88–1.14)	−0.28 (0.09)	0.001	0.75 (0.64–0.89)
Constant	−1.45 (2.71)	0.59	0.24	4.10 (3.62)	0.26	60.26
Model 3: MDS-UPDRS-III score
DTI-ALPS index, left	−0.41 (0.20)	0.035	0.66 (0.45–0.97)	−0.03 (0.27)	0.911	0.97 (0.58–1.63)
Age	−0.01 (0.03)	0.67	0.98 (0.94–1.04)	0.024 (0.04)	0.56	1.02 (0.95–1.11)
Sex	−0.19 (0.68)	0.78	0.83 (0.22–3.13)	0.48 (0.73)	0.51	1.61 (0.39–6.71)
MDS-UPDRS-III score	0.01 (0.03)	0.73	1.01 (0.95–1.08)	0.05 (0.04)	0.19	1.05 (0.98–1.14)
Constant	−0.58 (1.82)	0.75	0.56	−4.16 (2.70)	0.12	0.02

Abbreviations: DLB = dementia with Lewy bodies; DTI-ALPS = diffusion tensor imaging along the perivascular space; iRBD = isolated REM sleep behavior disorder; MDS = Movement Disorder Society; MoCA = Montreal Cognitive Assessment; OR = odds ratio; PD = Parkinson disease; SE = standard error; UPDRS-III = motor scale of the Unified Parkinson's Disease Rating Scale.

Values in bold represent significant predictors.

^aTested against staying disease-free in the longitudinally followed iRBD cohort.

Figure 3. Kaplan-Meier Survival Curves of Conversion Based on the DTI-ALPS Index.

(A) Kaplan-Meier survival curves illustrating the relationship between low and high DTI-ALPS indices (based on the median) and conversion risk in iRBD. A lower left DTI-ALPS z-score is associated with a higher risk of conversion. (B) Kaplan-Meier survival curves illustrating the relationship between low and high DTI-ALPS indices (based on the median) and conversion risk to PD in the iRBD group. A lower left DTI-ALPS z-score is associated with a higher risk of conversion to PD. Shaded areas represent the 95% CIs. ALPS = along the perivascular space; DTI = diffusion tensor imaging; iRBD = isolated REM sleep behavior disorder; PD = Parkinson disease.

Other Diffusivity Measures Did Not Differ in the iRBD Group

To confirm that the observed effects were specific to the DTI-ALPS index, we extracted additional diffusion metrics, including FA, mean diffusivity, axial diffusivity, radial diffusivity, free water, and their free water–corrected counterparts, from the same regions used to compute the DTI-ALPS index. No significant differences emerged between patients with iRBD and healthy controls (uncorrected p > 0.11, FDR-corrected p > 0.50), nor between iRBD converters and nonconverters (uncorrected p > 0.37, FDR-corrected p > 0.98) (eTable 5). This suggests that the DTI-ALPS index captures specific alterations in perivascular diffusion not reflected in conventional or free water–corrected tensor-based diffusion metrics.

Discussion

In this study, we investigated DTI-ALPS as a proxy of glymphatic function in a large, multicentric cohort of patients with polysomnography-confirmed iRBD. Patients with iRBD who phenoconverted to an overt synucleinopathy during follow-up had a lower left DTI-ALPS index at baseline compared with those who remained disease-free. Regression analyses showed that the left DTI-ALPS index predicted phenoconversion to PD versus remaining disease-free, independently of MoCA and MDS-UPDRS-III scores. These findings suggest that lower DTI-ALPS index may serve as a predictive biomarker for progression to PD in iRBD.

Previous studies reported glymphatic dysfunction in neurodegenerative disorders using DTI-ALPS. In PD, studies consistently reported a lower DTI-ALPS index in patients relative to controls, associated with both motor and cognitive impairments.^16,17 Specifically, a lower DTI-ALPS index was associated with lower MMSE scores and higher MDS-UPDRS-III scores in multiple studies.^16,17,39 Similarly, patients with DLB show lower DTI-ALPS values than controls.¹⁸ However, research specifically addressing glymphatic dysfunction in iRBD remains limited.^6,19-22 Existing studies on iRBD were conducted on small, single-center cohorts, some relying on questionnaire-based diagnoses of iRBD, which may be prone to false positives,⁴⁰ and others focusing solely on the bilateral DTI-ALPS index.^6,19-22 A previous study on 18 patients with iRBD reported a reduced DTI-ALPS index compared with controls.²⁰ This finding was confirmed in another small cohort of 20 patients with iRBD²⁶ and in larger cohorts of 119 patients with iRBD with probable RBD based on questionnaires¹⁹ and 62 patients with polysomnography-confirmed iRBD.²² However, 1 study with 68 patients with iRBD found no DTI-ALPS change compared with 50 controls.⁵ In this study, we leveraged the largest data set of diffusion MRI scans from patients with polysomnography-confirmed iRBD. After accounting for harmonization of DTI-ALPS values across multiple centers, our findings confirmed that patients with iRBD exhibit a reduced DTI-ALPS index.

However, in our case, this reduction was specifically observed in the left DTI-ALPS index, aligning with previous PD studies reporting unilateral DTI-ALPS reductions that evolve contralaterally as the disease progresses.¹⁶ Although glymphatic clearance is generally conceptualized as a global brain-wide mechanism, growing evidence suggests that its function may be hemispherically variable. The DTI-ALPS index specifically captures perivascular diffusivity in projection and associative fibers adjacent to the lateral ventricles, where structural asymmetries in fiber orientation and vascular organization may influence local diffusion. Supporting this, controls showed physiologic hemispheric asymmetry in DTI-ALPS values, suggesting that some lateralization of perivascular flow may be a normative feature rather than pathologic. It is important to note that asymmetrical glymphatic disruption has been reported in response to focal cerebral insults. In patients with ischemic stroke, intracerebral hemorrhage, or epilepsy, DTI-ALPS reductions have been shown to occur exclusively in the ipsilateral hemisphere,^41-43 supporting that glymphatic transport may be functionally compartmentalized across hemispheres. By analogy, early unilateral neurodegeneration in iRBD may similarly lead to localized perivascular diffusion disruption, with broader involvement emerging over time. This is in line with PD being generally characterized by asymmetric motor features at onset and unilateral dopaminergic degeneration.⁴⁴ Overall, these findings suggest that although the glymphatic system operates globally, its function may be asymmetrically affected by early, lateralized neurodegeneration. Our findings support the view that in iRBD, a unilateral reduction in the DTI-ALPS index may be an early manifestation of this pathologic process.

Patients with iRBD often convert to PD or DLB 10 years or more after symptom onset.⁴⁵ In this study, we found that the DTI-ALPS index was a significant predictor of conversion trajectories in iRBD, specifically toward PD. This association remained significant even when accounting for MoCA and MDS-UPDRS-III, i.e., indicators of disease progression that are more robust compared to self-reported iRBD symptom duration, which tends to be highly subjective. This highlights that DTI-ALPS index changes predict progression in iRBD, supporting its value as an imaging biomarker in prodromal neurodegenerative conditions. Our findings further align with a previous study that explored the predictive value of the DTI-ALPS index in patients with iRBD and showed that a lower index was associated with disease progression, although specific disease trajectories were not identified.²¹ This may explain why most, but not all, studies assessing DTI-ALPS index changes between patients with iRBD and controls found a significant difference: the clinical progression of the studied sample is a factor that influences the expression of the DTI-ALPS index values. These findings also align with several neuroimaging studies in iRBD showing cortical and subcortical atrophy and abnormal patterns of brain perfusion that can predict specific disease trajectories.^4,31 It also aligns with specific clinical features shown to predict specific phenoconversion trajectories in iRBD.³² Previous studies by our group, using in silico models,^46,47 have shown that neurodegenerative changes in iRBD, namely brain atrophy, can be reproduced based on alpha-synuclein spread in the brain as a function of deafferentation and the accumulation of toxic proteins in specific regions.⁴⁷ While these simulations modeled neurodegeneration using connection strength and gene expression, our findings suggest that altered DTI-ALPS may also play a role in shaping neurodegeneration. In addition, our observation that a lower DTI-ALPS index predicted time to conversion to PD rather than to DLB is intriguing and warrants further investigation. This contrasts with previously identified neuroimaging biomarkers in iRBD, where increased tissue deformation and relative hyperperfusion of the basal forebrain were specific predictors of DLB compared with PD.^4,48 Altogether, these results support the idea that combining MRI-based differential markers, namely DTI-ALPS index and brain atrophy, may allow for more specific prediction of disease trajectories, potentially improving patient stratification in clinical trials and prognosis in neurodegenerative diseases.

Some limitations should be acknowledged. While we used the largest iRBD cohort to date, only a subset of patients had yet phenoconverted. Future studies should continue to monitor these patients longitudinally to capture when they convert and whether they convert to PD or DLB. In addition, the low number of patients phenoconverting to other phenotypes, such as MSA, precluded their inclusion in this analysis of specific trajectories. There are also inherent limitations in the DTI-ALPS methodology itself, which specifically measures periventricular diffusion and may not be sensitive enough to detect glymphatic system dysfunction across the brain.^19,49 Moreover, the DTI-ALPS index reflects water diffusivity, an indirect measure of glymphatic activity, and its interpretation requires caution.⁵⁰ Nevertheless, the DTI-ALPS method has shown potential in associating with clinical features in numerous neurodegenerative diseases. Indeed, even if it primarily reflects water diffusivity, we show that it still provides valuable insights into fluid dynamics that can predict specific trajectories of disease progression in iRBD. In line with this, unlike the DTI-ALPS index, typical diffusion measurements did not differ between iRBD patients and controls nor between conversion subtypes. Furthermore, while all images were visually inspected to ensure that the DTI-ALPS index regions were free of visible abnormalities, white matter hyperintensities were not extracted or quantified. Finally, in this multicentric cohort, the only common clinical assessments across cohorts were MoCA and MDS-UPDRS-III, which are relatively broad measures. While these assessments provide insight into global cognitive function and parkinsonian motor features, they may not capture the full spectrum of disease-related changes. Future studies should incorporate more fine-grained clinical assessments across cohorts to better understand the specific cognitive domains and motor features affected.

In summary, we show a lower DTI-ALPS index in a large multicentric cohort of patients with polysomnography-confirmed iRBD. We found that the DTI-ALPS index was significantly lower in patients with iRBD compared with controls and was a predictor of conversion to PD. These findings suggest changes in the DTI-ALPS index occur early in the PD progression continuum associated with iRBD and may play a role in disease progression.

Glossary

AD

Alzheimer disease

DLB

dementia with Lewy bodies

DTI-ALPS

diffusion tensor imaging along the perivascular space

FA

fractional anisotropy

FDR

false discovery rate

iRBD

isolated REM sleep behavior disorder

MDS-UPDRS-III

Movement Disorder Society–Unified Parkinson's Disease Rating Scale Part III

MNI

Montreal Neurological Institute

MoCA

Montreal Cognitive Assessment

MPRAGE

magnetization-prepared rapid gradient-echo

MSA

multiple system atrophy

PD

Parkinson disease

TE

echo time

TR

repetition time

Appendix. Coinvestigators

Name	Location	Role	Contribution
Violette Ayral	University of Montreal, Canada	Trainee	Data analysis, drafting of the original manuscript
Alexandre Pastor-Bernier	Center for Advanced Research in Sleep Medicine, Montreal, Canada	MRI analyst	Data analysis, review and editing of the manuscript
Véronique Daneault	Center for Advanced Research in Sleep Medicine, Montreal, Canada	MRI analyst	Data analysis, review and editing of the manuscript
Christina Tremblay	Center for Advanced Research in Sleep Medicine, Montreal, Canada	MRI analyst	Data analysis, review and editing of the manuscript
Marie Filiatrault	University of Montreal, Canada	Trainee	Data analysis, review and editing of the manuscript
Celine Haddad	University of Montreal, Canada	Trainee	Data analysis, review and editing of the manuscript
Jean-François Gagnon	Center for Advanced Research in Sleep Medicine, Montreal, Canada	Site investigator	Data acquisition, review and editing of the manuscript
Ronald B. Postuma	Center for Advanced Research in Sleep Medicine, Montreal, Canada	Site investigator	Data acquisition, review and editing of the manuscript
Petr Dušek	Charles University, Prague, Czechia	Site investigator	Data acquisition, review and editing of the manuscript
Stanislav Marecek	Charles University, Prague, Czechia	Site investigator	Data acquisition, review and editing of the manuscript
Zsoka Varga	Charles University, Prague, Czechia	Site investigator	Data acquisition, review and editing of the manuscript
Johannes C. Klein	University of Oxford, United Kingdom	Site investigator	Data acquisition, review and editing of the manuscript
Michele T. Hu	University of Oxford, United Kingdom	Site investigator	Data acquisition, review and editing of the manuscript
Stéphane Lehéricy	Hôpital de la Pitié-Salpêtrière, Paris, France	Site investigator	Data acquisition, review and editing of the manuscript
Isabelle Arnulf	Hôpital de la Pitié-Salpêtrière, Paris, France	Site investigator	Data acquisition, review and editing of the manuscript
Marie Vidaillhet	Hôpital de la Pitié-Salpêtrière, Paris, France	Site investigator	Data acquisition, review and editing of the manuscript
Jean-Christophe Corvol	Hôpital de la Pitié-Salpêtrière, Paris, France	Site investigator	Data acquisition, review and editing of the manuscript

Author Contributions

V. Ayral: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. A. Pastor-Bernier: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. V. Daneault: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. C. Tremblay: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. M. Filiatrault: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. C. Haddad: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. J.-F. Gagnon: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. R.B. Postuma: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. P. Dušek: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. S. Marecek: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. Z. Varga: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. J.C. Klein: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M.T. Hu: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. S. Lehéricy: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. I. Arnulf: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M. Vidailhet: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. J.-C. Corvol: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. S. Rahayel: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

This work was supported by funding grants awarded to S. Rahayel from Parkinson Canada (PPG-2023-0000000122) and Alzheimer Society Canada (0000000082). S. Rahayel holds a research scholar award from the Fonds de recherche du Québec-Santé. The work performed in the Montreal iRBD cohort was supported by the Canadian Institutes of Health Research, the Fonds de recherche du Québec-Santé, and the W. Garfield Weston Foundation. In Montreal, J.-F. Gagnon reports grants from the Fonds de recherche du Québec-Santé, the Canadian Institutes of Health Research, the W. Garfield Weston Foundation, the Michael J. Fox Foundation for Parkinson's Research, and the NIH. R.B. Postuma reports grants and personal fees from the Fonds de recherche du Québec-Santé, the Canadian Institutes of Health Research, Parkinson Canada, the W. Garfield Weston Foundation, the Michael J. Fox Foundation for Parkinson's Research, the R. Howard Webster Foundation, and the NIH. The work performed in the Prague iRBD cohort was funded by the Czech Health Research Council (grant NU21-04-00535) and by The National Institute for Neurological Research (project number LX22NPO5107) financed by the European Union - Next Generation EU. The work performed in Oxford (“Oxford Discovery cohort”) was funded by Parkinson's UK (J-2101) and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). The work performed in the Paris iRBD cohorts was funded by grants from the Programme d'investissements d'avenir (ANR-10-IAIHU-06), the Paris Institute of Neurosciences - IHU (IAIHU-06), the Agence Nationale de la Recherche (ANR-11-INBS-0006), Électricité de France (Fondation d’Entreprise EDF), Biogen Inc., the Fondation Thérèse et René Planiol, the Fonds Saint-Michel; by unrestricted support for research on Parkinson's disease from Energipole (M. Mallart) and Société Française de Médecine Esthétique (M. Legrand); and by a grant from the Institut de France to Isabelle Arnulf (for the Alice Study). The Parkinson's Progression Markers Initiative (PPMI)—a public-private partnership—is funded by the Michael J. Fox Foundation for Parkinson's Research and funding partners, including 4D Pharma, AbbVie Inc., AcureX Therapeutics, Allergan, Amathus Therapeutics, Aligning Science Across Parkinson's (ASAP), Avid Radiopharmaceuticals, Bial Biotech, Biogen, BioLegend, Bristol Myers Squibb, Calico Life Sciences LLC, Celgene Corporation, DaCapo Brainscience, Denali Therapeutics, The Edmond J. Safra Foundation, Eli Lilly and Company, GE Healthcare, GlaxoSmithKline, Golub Capital, Handl Therapeutics, Insitro, Janssen Pharmaceuticals, Lundbeck, Merck & Co., Inc., Meso Scale Diagnostics, LLC, Neurocrine Biosciences, Pfizer Inc., Piramal Imaging, Prevail Therapeutics, F. Hoffman-La Roche Ltd and its affiliated company Genentech Inc., Sanofi Genzyme, Servier, Takeda Pharmaceutical Company, Teva Neuroscience, Inc., UCB, Vanqua Bio, Verily Life Sciences, Voyager Therapeutics, Inc., and Yumanity Therapeutics, Inc. Up-to-date information on the study can be obtained from ppmi-info.org.

Disclosure

R.B. Postuma reports grants from the Canadian Institute of Health Research, the Michael J. Fox Foundation, the Webster Foundation, Roche, and the National Institute of Health; as well as personal fees from Takeda, Biogen, Abbvie, Curasen, Lilly, Novartis, Eisai, Paladin, Merck, Korro, Vaxxinity, Bristol Myers Squibb, and the International Parkinson and Movement Disorders Society, all outside the submitted work. All other authors have no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.