Advanced Quantitative MRI Unveils Microstructural Thalamic Changes Reflecting Disease Progression in Multiple Sclerosis

Abstract

Background and Objectives

In patients with multiple sclerosis (PwMS), thalamic atrophy occurs during the disease course. However, there is little understanding of the mechanisms leading to volume loss and of the relationship between microstructural thalamic pathology and disease progression. This cross-sectional and longitudinal study aimed to comprehensively characterize in vivo pathologic changes within thalamic microstructure in PwMS using advanced multiparametric quantitative MRI (qMRI).

Methods

Thalamic microstructural integrity was evaluated using quantitative T1, magnetization transfer saturation, multishell diffusion, and quantitative susceptibility mapping (QSM) in 183 PwMS and 105 healthy controls (HCs). The same qMRI protocol was available for 127 PwMS and 73 HCs after a 2-year follow-up period. Inclusion criteria for PwMS encompassed either an active relapsing-remitting MS (RRMS) or inactive progressive MS (PMS) disease course. Thalamic alterations were compared between PwMS and HCs and among disease phenotypes. In addition, the study investigated the relationship between thalamic damage and clinical and conventional MRI measures of disease severity.

Results

Compared with HCs, PwMS exhibited substantial thalamic alterations, indicative of microstructural and macrostructural damage, demyelination, and disruption in iron homeostasis. These alterations extended beyond focal thalamic lesions, affecting normal-appearing thalamic tissue diffusely. Over the follow-up period, PwMS displayed an accelerated decrease in myelin volume fraction [mean difference in annualized percentage change (MD-ApC) = −1.50; p = 0.041] and increase in quantitative T1 (MD-ApC = 0.92; p < 0.0001) values, indicating heightened demyelinating and neurodegenerative processes. The observed differences between PwMS and HCs were substantially driven by the subgroup with PMS, wherein thalamic degeneration was significantly accelerated, even in comparison with patients with RRMS. Thalamic qMRI alterations showed extensive correlations with conventional MRI, clinical, and cognitive disease burden measures. Disability progression over follow-up was associated with accelerated thalamic degeneration, as reflected by enhanced diffusion (β = −0.067; p = 0.039) and QSM (β = −0.077; p = 0.027) changes. Thalamic qMRI metrics emerged as significant predictors of neurologic and cognitive disability even when accounting for other established markers including white matter lesion load and brain and thalamic atrophy.

Discussion

These findings offer deeper insights into thalamic pathology in PwMS, emphasizing the clinical relevance of thalamic damage and its link to disease progression. Advanced qMRI biomarkers show promising potential in guiding interventions aimed at mitigating thalamic neurodegenerative processes.

Introduction

Gray matter pathology is prominent in patients with multiple sclerosis (PwMS), with degenerative processes initiating early in the disease course and correlating with clinical disability and cognitive performance.^1,2 Among gray matter structures, the thalamus plays a pivotal role, serving as a relay and integration center within the CNS. In PwMS, the thalamus is distinctively affected, because of the complex interplay between heterogeneous pathophysiologic mechanisms.³ Inflammatory processes can directly cause thalamic damage, leading to the development of focal demyelinating lesions.^3,4 Thalamic damage can also occur indirectly due to pathologic processes involving the white matter, where axonal transection in white matter lesions can trigger retrograde, anterograde, and trans-synaptic degeneration, ultimately resulting in thalamic neuronal loss.^3,5 Besides, an increasingly recognized mechanism contributing to thalamic pathology is the damage following a distinctive “surface-in” pattern, hypothesized to be triggered by soluble inflammatory factors contained in the CSF.^6-9 Finally, oxidative stress, mitochondrial dysfunction, and disturbances in iron homeostasis contribute to thalamic degeneration.^3,10,11

Quantification of thalamic damage in multiple sclerosis has been the focus of extensive research, primarily through the measurement of thalamic volume, which exhibits consistent loss across disease phases and closely correlates with neurologic disability and cognitive impairment.^3,12-14 While other neuroimaging techniques such as MR spectroscopy and quantitative MRI (qMRI) approaches have also been used,^3,15-17 our understanding of the microstructural alterations accumulating in the thalamus of PwMS, particularly in relation to disease progression, remains limited. To comprehensively characterize thalamic tissue damage in PwMS and its accumulation across different disease phases, we conducted a multiparametric qMRI study. Our investigation leveraged quantitative T1-relaxometry (qT1) for a comprehensive assessment of tissue microstructural and macrostructural integrity,^18,19 myelin volume fraction (MVF) as a proxy measure of tissue myelin and macromolecular content,^18,20 neurite density index (NDI) as a surrogate measure of axon and dendrite density,²¹ and quantitative susceptibility mapping (QSM) to characterize changes in iron homeostasis and myelin damage.¹⁸ Thalamic atrophy, serving as a marker of the overall accumulation of thalamic neurodegenerative processes, was also measured.

We investigated thalamic microstructural damage and its accumulation over time, by performing a prospective, cross-sectional, and longitudinal study. Pathologic changes were investigated in both focal lesions and the “normal-appearing” tissue, by comparing PwMS with a reference group of healthy controls (HCs). Thalamic alterations associated with different disease phases were examined by comparing 2 subgroups of patients, one with active relapsing-remitting MS (RRMS) and another with inactive progressive MS (PMS), specifically selected to represent the opposing ends of the clinical spectrum. In addition, we investigated the relationship between qMRI measures of thalamic integrity and their rate of change over time with (1) conventional MRI markers of disease burden and (2) clinical measures of neurologic disability and cognitive impairment.

Methods

Participants

Participants were prospectively recruited at the University Hospital of Basel between 2018 and 2022, as part of the INsIDER study (NCT05177523). Inclusion criteria for patients were as follows: (1) age between 18 and 80 years; (2) diagnosis of MS fulfilling the 2017 revised McDonald criteria²²; (3) disease course either active relapsing-remitting (with ≥1 relapse and/or ≥1 gadolinium-enhancing lesion during the year before enrollment) or nonactive primary/secondary progressive (without relapses and MRI activity during the year before enrollment); (4) absence of neurologic or psychiatric comorbidities. HCs were volunteers aged between 18 and 80 years, with a medical history negative for neurologic or psychiatric disorders. Exclusion criteria for patients and controls encompassed pregnancy, contraindication to MRI, and inability to provide consent. This study follows the STROBE guideline for reporting observational studies.²³

A total of 183 PwMS and 105 HCs met the inclusion criteria. At baseline, all underwent a brain MRI scan, with additional neurologic examination for PwMS and cognitive assessment for a subset of 99 PwMS and 100 HCs. 127 PwMS and 73 HCs volunteered to undergo a second brain MRI scan after a period of 2 years (±3 months), with the same acquisition protocol. Concurrent neurologic follow-up was obtained for PwMS. The study design is presented in Figure 1.

Figure 1. Study Design.

*Analyses performed in the group of patients with multiple sclerosis only. ^aAvailable in 99 patients with multiple sclerosis and 100 healthy controls. ^bAvailable in 98 patients with multiple sclerosis and 100 healthy controls. BPF = brain parenchymal fraction; EDSS = Expanded Disability Status Scale; MuSIC = Multiple Sclerosis Inventory Cognition; PMS = progressive multiple sclerosis; qMRI = quantitative MRI; RRMS = relapsing-remitting multiple sclerosis; SDMT = Symbol Digit Modalities Test; T2LV = T2-hyperintense lesion volume; VLMT = Verbal Learning and Memory Test.

Clinical Assessment

The neurologic examination for PwMS included the calculation of the Expanded Disability Status Scale (EDSS) score, by certified raters.^24,25

Cognitive performance was assessed using (1) the oral version of the Symbol Digit Modalities Test (SDMT),²⁶ (2) the Verbal Learning and Memory Test (VLMT),²⁷ and (3) the Multiple Sclerosis Inventory Cognition (MuSIC)²⁸ test. SDMT and VLMT scores were converted to z-scores,^26,27 and the MuSIC score was corrected for age and sex.²⁸

MRI Acquisition

All brain MRI scans were obtained with the same acquisition protocol on a 3T whole-body MR system (Magnetom Prisma, Siemens Healthineers), using a 64-channel phased-array head and neck coil for radio-frequency reception. The MRI protocol included the following: (1) 3D fluid-attenuated inversion recovery (FLAIR) [TR/TE/TI = 5,000/386/1,800 ms; resolution = 1 × 1 × 1 mm³]; (2) 3D magnetization-prepared 2 rapid gradient-echo (MP2RAGE) [TR/TI1/TI2 = 5,000/700/2,500 ms; resolution = 1 × 1 × 1 mm³]; (3) multishell diffusion (TR/TE/δ/Δ = 4,500/75/19/36 ms; resolution = 1.8 × 1.8 × 1.8 mm³; b-values 0/700/1,000/2,000/3,000 s/mm² with 12/6/20/45/66 measurements, respectively, per shell; diffusion acquisition with 12 measurements of b-value 0 s/mm² with reversed phase encoding); (4) 3D segmented echo planar imaging (EPI) [TR/TE = 64/35 ms; resolution = 0.67 × 0.67 × 0.67 mm³]; (5) three 3D radio-frequency spoiled gradient-echo acquisitions with predominantly magnetization transfer–weighted (TR/α = 25 ms/5°), proton density–weighted (TR/α = 25 ms/5°), and T1-weighted (TR/α = 11 ms/15°) contrasts, used to obtain magnetization transfer saturation (MTsat) maps (resolution = 1.33 × 1.33 × 1.33 mm³).²⁹

MRI Processing

qT1 maps were obtained from MP2RAGE images as previously described.^29,30 MTsat maps were computed as previously proposed,^31,32 and MVF was then estimated by multiplying the MTsat signal by a calibration constant (α).^33,34 Diffusion images were denoised and corrected for motion, eddy currents, and susceptibility-induced distortions; microscopic diffusion processes were then modeled using the spherical mean technique (SMT) to obtain NDI, a proxy measure of the signal originating from axons and dendrites.²¹ SMT was preferred over alternative approaches such as neurite orientation dispersion and density imaging (NODDI), given its effectiveness in addressing challenges posed by fiber crossing populations and orientation dispersion, which are particularly prominent within gray matter regions.²¹ QSM was reconstructed from 3D EPI images using the morphology enabled dipole inversion algorithm to compute the susceptibility from the local field, with ventricular CSF used as the zero reference.35

MRI Analysis

In PwMS, T2-hyperintense lesions were detected with a deep learning–based tool³⁶; the resulting masks were manually corrected and used to estimate total T2-hyperintense lesion volume (T2LV).

FIRST³⁷ was applied on MP2RAGE images to obtain an initial reference for thalamic segmentation, and the automatic output was manually refined following the previously proposed protocol for the manual segmentation of the thalamus.³⁸ In PwMS, a mask of T2-hyperintense lesions contained within the thalamus was obtained by intersecting the lesion mask (registered to the MP2RAGE space) with the thalamic mask. Thalamic lesions were then subtracted from thalamic masks to obtain a segmentation of normal-appearing thalamic tissue. For each thalamic lesion, we investigated the morphology and the presence of a central vein or a paramagnetic rim (eMethods).

Thalamus optimized multi atlas segmentation (THOMAS) was used to segment the thalamus into 12 nuclei.³⁹ In addition, the thalamus was partitioned into 3 bands having increasing distance from the CSF, to explore the “ependymal-in” gradient of damage. The 3 bands were identified by performing progressive expansions of a CSF mask (after a 1-mm erosion at the thalamus-CSF and thalamus-internal capsule boundaries, to reduce partial volume effect).

MVF, NDI, and QSM maps were linearly registered to the MP2RAGE space using FLIRT, and the results were visually inspected. The transformation matrices obtained in the process were then used to move the regions of interest (i.e., thalamic lesions and normal-appearing tissue, segmented on the MP2RAGE space) back to the original MVF, NDI, and QSM spaces. For all regions and contrasts of interest, the mean intensity values were extracted in the original space using fslstats. For thalamic lesions, qMRI metrics were extracted exclusively in lesions with volume ≥10 mm³.

Total brain volume and total intracranial volume (TIV) were measured with SAMSEG.⁴⁰ Brain parenchymal fraction (BPF) was calculated as the ratio between total brain volume and TIV; similarly, normalized thalamic volume was estimated as the ratio between thalamic volume and TIV.

Statistical Analysis

All statistical analyses were performed with R (version 4.2.1; R Core Team, 2022).⁴¹ The threshold of statistical significance was set at p < 0.05.

Demographic, clinical, and MRI variables were compared between groups using the Welch t-test, Mann-Whitney U test, Pearson χ² test, and linear regression models, as appropriate.

qMRI measures in the normal-appearing thalamic tissue were compared cross-sectionally, at baseline, between (1) PwMS and HCs and (2) patients with inactive PMS and patients with active RRMS. Comparisons were performed with general linear models, adjusting for age and sex. Differences in qMRI measures were also explored regionally, both in thalamic nuclei and in the 3 bands having increasing distance from the CSF; the results were corrected for multiple comparisons with the false discovery rate approach, using the Benjamini-Hochberg method.

To characterize the microstructural damage within T2-hyperintense thalamic lesions, we compared the qMRI metrics measured within the lesions with those obtained in a 2-voxel area of perilesional normal-appearing tissue. The comparison was performed using the Wilcoxon signed-rank test.

In PwMS, general linear models were used to explore the association of qMRI measures in the normal-appearing thalamic tissue with clinical and conventional MRI features, including (1) neurologic disability, as assessed with the EDSS; (2) cognitive impairment, as assessed with the SDMT, VLMT, and MuSIC test; (3) burden of T2-hyperintense lesion volume (T2LV); and (4) BPF. The relative importance of the different qMRI thalamic measures—together with T2LV and BPF—in (1) explaining neurologic disability (as measured with the EDSS), (2) explaining multidomain cognitive performance (as measured with the MuSIC score), and (3) distinguishing between progressive and relapsing-remitting phenotypes was ranked using multivariable least absolute shrinkage and selection operator (LASSO) regression models. We performed 10-fold cross-validation to select the optimal regularization parameter (lambda). For each patient, we calculated the deviation of the thalamic qMRI metrics (and the BPF) from the HC population, accounting for age and sex. Specifically, we fitted a linear regression in HCs with the MRI measure as the dependent variable and age and sex as independent variables. The deviation from the regression line (residual of the fit) of each MRI measure in PwMS was the variable entered in the LASSO models.⁴²

Longitudinal changes in thalamic qMRI metrics were quantified with mixed-effect models, using the qMRI measurements at each given time point as dependent variables. Models included time (to estimate the rate of change), age at baseline, and sex as covariates and a random intercept for participants. To estimate annual percentage change from the slope over time, the qMRI metrics were log-transformed. Comparisons in the rates of change between (1) PwMS and HCs and (2) patients with inactive PMS and patients with active RRMS were performed by introducing in the mixed-effect models the interaction term between the group and time. Similarly, the association between the rates of qMRI change and (1) baseline EDSS, (2) baseline T2LV, (3) baseline normalized thalamic volume, (4) disease activity over follow-up, (5) EDSS progression over follow-up, (6) change in T2LV over follow-up, and (7) change in BPF over follow-up was explored by introducing in the mixed-effect models the interaction term between the variables of interest and time. EDSS progression was defined as an increase in the EDSS score of ≥1.5 points if the baseline EDSS score was 0, ≥1.0 points if the baseline EDSS score was 1.0–5.5, or ≥0.5 points if the baseline EDSS score was greater than 5.5.⁴³ Mixed-effect models were also used to investigate baseline demographic, clinical, and MRI predictors of subsequent thalamic atrophy.

Sensitivity analyses were conducted to (1) exclude a significant impact of partial volume effect on the estimations of thalamic qMRI measures (by comparing the qMRI values derived with the original thalamic masks with those obtained with thalamic masks that underwent a 1-voxel erosion process); (2) compare the estimations of NDI obtained with SMT with those derived from NODDI; (3) explore between-group differences in thalamic microstructure accounting for treatment effect and presence of disease activity over the follow-up; (4) explore between-group differences in subgroups matched for age and sex; and (5) investigate regional microstructural changes in thalamic nuclei defined with a different probabilistic atlas (as implemented in FreeSurfer; eFigures 1–5).

Additional analyses were conducted to explore (1) between-group differences in thalamic measures of orientation dispersion index (ODI) and isotropic volume fraction (ISOVF) derived with NODDI (eTables 1–4); (2) the extent of microstructural changes as a function of the distance from the CSF (eFigure 6); (3) the correlation between different thalamic qMRI metrics (eFigure 7); (4) the impact of T2-hyperintense lesion burden specifically within thalamocortical bundles as a proxy for thalamic disconnectivity (eTables 5–9); and (5) the relative importance of demographic, clinical, and conventional MRI measures and the global and regional thalamic qMRI metrics explored in the study, in explaining clinical outcomes using random forests (eFigures 8–10).

Further methodological details are available in eMethods.

Standard Protocol Approvals, Registrations, and Patient Consents

Study approval was obtained from the local ethics committee (IRM of Northwest Switzerland); informed consent was obtained from all participants before study entry.

Data Availability

The data that support the findings of this study are available on reasonable request.

Results

The main cohort's demographic, clinical, and MRI characteristics are summarized in Table 1. Baseline characteristics of patients undergoing clinical and MRI follow-up are reported in eTable 10. The qMRI contrasts included in the study are illustrated in Figure 2.

Table 1.

Baseline Demographic, Clinical, and Conventional MRI Characteristics of the Cohort

	Healthy controls	Patients with active RRMS	Patients with inactive PMS	p Values
n	105	101	82	—
Female/male	58/47	66/35	44/38	RRMS vs HCs: 0.18a
				PMS vs HCs: 0.95a
				PMS vs RRMS: 0.15a
Mean (SD) [range] age, y	37.8 (13.0) [18.1–69.0]	37.6 (10.9) [18.3–62.8]	58.1 (9.5) [36.9–77.2]	RRMS vs HCs: 0.92b
				PMS vs HCs: <0.001b
				PMS vs RRMS: <0.001b
Median [IQR] disease duration, y	—	2.5 [0.6–9.3]	15.4 [8.0–25.0]	RRMS vs HCs:/
				PMS vs HCs:/
				PMS vs RRMS: <0.001c
Disease-modifying therapy	—
Platform, n (%)		3 (3)	4 (5)	RRMS vs HCs:/
Oral, n (%)		36 (36)	15 (18)	PMS vs HCs:/
Monoclonal antibodies, n (%)		50 (50)	48 (59)	PMS vs RRMS: <0.001a
Untreated, n (%)		12 (12)	15 (18)
Median [IQR] EDSS score	—	2.0 [1.5–2.5]	5.0 [4.0–6.0]	RRMS vs HCs:/
				PMS vs HCs:/
				PMS vs RRMS: < 0.001c
Mean (SD) SDMT z-score*	0.55 (1.16)	0.29 (1.15)	−0.49 (1.50)	RRMS vs HCs: 0.19b
				PMS vs HCs: <0.001b
				PMS vs RRMS: 0.004b
Mean (SD) VLMT z-score*	1.00 (1.17)	0.89 (1.06)	0.08 (1.08)	RRMS vs HCs: 0.57b
				PMS vs HCs: <0.001b
				PMS vs RRMS: <0.001b
Mean (SD) MuSIC score**	27.0 (3.5)	25.9 (4.6)	19.5 (6.4)	RRMS vs HCs: 0.10b
				PMS vs HCs: <0.001b
				PMS vs RRMS: <0.001b
Mean (SD) BPF	0.74 (0.02)	0.73 (0.02)	0.69 (0.03)	RRMS vs HCs: 0.005d
				PMS vs HCs: <0.001d
				PMS vs RRMS: <0.001d
Median [IQR] T2LV, ml	0.0 [0.0–0.2]	4.0 [1.2–10.3]	12.6 [5.8–26.8]	RRMS vs HCs: <0.001c
				PMS vs HCs: <0.001c
				PMS vs RRMS: <0.001c

Abbreviations: BPF = brain parenchymal fraction; EDSS = Expanded Disability Status Scale; IQR = interquartile range; MuSIC = Multiple Sclerosis Inventory and Cognition; PMS = progressive multiple sclerosis; PPMS = primary progressive multiple sclerosis; RRMS = relapsing-remitting multiple sclerosis; SDMT = Symbol Digit Modalities Test; SPMS = secondary progressive multiple sclerosis; VLMT = Verbal Learning and Memory Test.

^aChi-square test.

^bWelch t-test.

^cMann-Whitney U test.

^dGeneral linear model, age and sex-adjusted. *Available in 100 healthy controls and 99 patients with multiple sclerosis. **Available in 100 healthy controls and 98 patients with multiple sclerosis. Platform disease-modifying therapies included glatiramer acetate and interferon-beta preparations; oral disease-modifying therapies included dimethyl fumarate, fingolimod, siponimod, and teriflunomide; monoclonal antibody disease-modifying therapies included natalizumab, ocrelizumab, and rituximab.

Figure 2. qMRI Contrasts Included in the Study.

The images displayed were obtained by averaging data from all healthy controls included in the study, after nonlinear registration to the MNI152 template. MP2RAGE = magnetization-prepared 2 rapid gradient-echo; MVF = myelin volume fraction; NDI = neurite density index; QSM = quantitative susceptibility mapping; qT1 = quantitative T1-relaxometry.

Owing to insufficient image quality, data from 5 PwMS for MVF analysis and 3 PwMS for NDI analysis were excluded.

Thalamic qMRI Metrics at Baseline

PwMS vs HCs

At baseline, PwMS had reduced thalamic volume (β = −0.232; p < 0.0001) compared with HCs. In addition, in the normal-appearing thalamic tissue, PwMS displayed decreased MVF (β = −0.120; p = 0.032) and QSM (β = −0.138; p = 0.013) values, along with increased qT1 values (β = 0.145; p = 0.016); no difference in NDI values was observed between groups (Figure 3).

Figure 3. Comparisons in Thalamic qMRI Metrics Between Patients With Multiple Sclerosis and Healthy Controls at Baseline (in the Entire Thalamus and Thalamic Nuclei).

The difference between groups is graphically displayed on a segmentation of the thalamus, with colors reflecting the effect size (expressed in terms of standardized regression coefficient, β). Only differences reaching statistical significance are displayed. MVF = myelin volume fraction; NDI = neurite density index; QSM = quantitative susceptibility mapping; qT1 = quantitative T1-relaxometry.

The greatest between-group difference in MVF, QSM, and qT1 values was evident in the thalamic area closest to the CSF (Figure 4). Graphical representations of between-group comparisons at the thalamic nuclei level are displayed in Figure 3 and eFigure 11 while a comprehensive description of the results is reported in eTables 11–12.

Figure 4. Between-Group Comparisons in Thalamic qMRI Metrics in the 3 Bands Presenting Increasing Distance From the CSF.

The 3 bands are associated with different colors, graphically displayed in the panel on top. The height of the bars reflects the magnitude of the effect size (β) of the between-group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001. HCs = healthy controls; MS = multiple sclerosis; MVF = myelin volume fraction; NDI = neurite density index; QSM = quantitative susceptibility mapping; qT1 = quantitative T1-relaxometry.

The differences in thalamic qMRI measures between PwMS and HCs were mainly driven by the subset of patients with inactive PMS (Figure 5). While patients with active RRMS differed from HCs only in thalamic volume (β = −0.145; p = 0.015), patients with inactive PMS differed from HCs in thalamic volume (β = −0.490; p < 0.0001), MVF (β = −0.328; p < 0.0001), qT1 (β = 0.324; p < 0.0001), and QSM (β = −0.258; p = 0.0005) values. Patients with inactive PMS also showed significant differences in thalamic volume (β = −0.356; p < 0.0001), MVF (β = −0.310; p < 0.0001), qT1 (β = 0.270; p = 0.0007), and QSM (β = −0.183; p = 0.013) values compared with patients with active RRMS. The most substantial differences between patients with inactive PMS and patients with active RRMS were observed in the thalamic region close to the CSF for MVF and qT1 and in the central thalamic area for QSM (eFigure 12).

Figure 5. Between-Group Comparisons of Baseline Thalamic qMRI Metrics.

*p < 0.05; **p < 0.01; ***p < 0.001. HCs = healthy controls; MVF = myelin volume fraction; NDI = neurite density index; PMS = progressive multiple sclerosis; QSM = quantitative susceptibility mapping; qT1 = quantitative T1-relaxometry; RRMS = relapsing-remitting multiple sclerosis.

Thalamic Lesions vs Normal-Appearing Thalamic Tissue

T2-hyperintense thalamic lesions were identified in 66 PwMS. The proportion of patients presenting with thalamic lesions was not statistically different between PMS and RRMS groups (43% vs 31%; p = 0.09). Of 112 observed thalamic lesions, 92 were discrete ovoid lesions and 20 were more diffuse periventricular lesional areas. 51.1% of thalamic lesions exhibited the central vein sign, with a higher prevalence in ovoid lesions compared with periventricular lesions (57.4% vs 30.0%; p = 0.031). No lesions displayed a paramagnetic rim (eTable 13). The distribution of lesions across different thalamic nuclei is detailed in eTable 14. Compared with the normal-appearing perilesional tissue, thalamic lesions exhibited reduced MVF and NDI and increased qT1 values (all p < 0.0001). No differences were measured in QSM values between thalamic lesions and the perilesional normal-appearing tissue (eTable 15).

Association Between Thalamic qMRI Measures and Clinical/MRI Features in PwMS

Thalamic volume was associated with the EDSS score (β = −0.471; p < 0.0001), cognitive performance in all tests considered in the study (SDMT: β = 0.345; p = 0.0002; VLMT: β = 0.300; p = 0.0009; MuSIC: β = 0.461; p < 0.0001), T2LV (β = −0.650; p < 0.0001), and BPF (β = 0.824; p < 0.0001).

Various associations with clinical and conventional MRI measures were evident also for qMRI measures of microstructural integrity within the normal-appearing thalamic tissue. Specifically, MVF was associated with the EDSS score (β = −0.246; p = 0.0009), MuSIC score (β = 0.307; p = 0.004), T2LV (β = −0.249; p = 0.0007), and BPF (β = 0.361; p < 0.0001); qT1 was associated with the EDSS score (β = 0.198; p = 0.007), SDMT (β = −0.193; p = 0.036), MuSIC score (β = −0.197; p = 0.034), T2LV (β = 0.229; p = 0.002), and BPF (β = −0.317; p < 0.0001); NDI was associated with the EDSS score (β = −0.198; p = 0.008) and BPF (β = 0.188; p = 0.012); QSM was associated with the EDSS score (β = −0.224; p = 0.002), T2LV (β = −0.248; p = 0.0007), and BPF (β = 0.312; p < 0.0001) (eTable 16).

In the LASSO regression models, the variables selected as predictors of the EDSS score, in order of importance, were BPF (β = −0.382), T2LV (β = 0.155), thalamic MVF (β = −0.051), thalamic QSM (β = −0.022), and thalamic NDI (β = −0.012) [R² of the model: 0.296]; the variables selected as predictors of the MuSIC score, in order of importance, were BPF (β = 0.262), normalized thalamic volume (β = 0.100), and thalamic MVF (β = 0.031) [R² of the model: 0.248]; in the model for discrimination between PMS and RRMS, the selected features, in order of importance, were BPF (odds ratio [OR] = 0.356), thalamic MVF (OR = 0.712), T2LV (OR = 1.231), normalized thalamic volume (OR = 0.783), and thalamic NDI (OR = 0.876) [area under the curve of the model: 0.836].

Longitudinal Thalamic qMRI Changes

PwMS vs HCs

During the 2-year follow-up, PwMS had accelerated thalamic atrophy compared with HCs [mean difference in annualized percentage change (MD-ApC): −0.88 (95% CI −1.56 to −0.20); p = 0.013]. PwMS also exhibited an accelerated rate of i) decrease in thalamic MVF values [MD-ApC: −1.50 (95% CI −2.91 to −0.07); p = 0.041] and ii) increase in qT1 values [MD-ApC: 0.92 (95% CI 0.54–1.30); p < 0.0001].

Compared with HCs, patients with active RRMS had an accelerated rate of increase in qT1 values [MD-ApC: 0.74 (95% CI 0.33–1.15); p = 0.0005]; the difference in thalamic atrophy rates approached but did not reach statistical significance [MD-ApC: −0.71 (95% CI −1.46 to 0.04); p = 0.066]. Compared with HCs, patients with inactive PMS had accelerated thalamic atrophy [MD-ApC: −1.20 (95% CI −2.07 to −0.32); p = 0.008], as well as accelerated rates of decrease in thalamic MVF values [MD-ApC: −2.52 (95% CI −4.28 to −0.72); p = 0.007] and increase in thalamic qT1 values [MD-ApC: 1.24 (95% CI 0.76–1.73); p < 0.0001]. Patients with inactive PMS had accelerated rates of qT1 increase also when compared with patients with active RRMS [MD-ApC: 0.50 (95% CI 0.03–0.97); p = 0.039]. No differences in the rates of change between groups were measured in the other thalamic qMRI metrics (Figure 6).

Figure 6. Between-Group Comparisons of Longitudinal Thalamic qMRI Metric Changes.

*p < 0.05; **p < 0.01; ***p < 0.001. HCs = healthy controls; MD-APC: mean difference in annual percentage change; MS = multiple sclerosis; MVF = myelin volume fraction; NDI = neurite density index; PMS = progressive multiple sclerosis; RRMS = relapsing-remitting multiple sclerosis; QSM = quantitative susceptibility mapping; qT1 = quantitative T1-relaxometry.

Association Between Quantitative Thalamic Changes and Clinical/MRI Features in PwMS

Higher baseline T2LV was associated with accelerated thalamic MVF reduction (β = −0.118; p = 0.039) and qT1 increase (β = 0.088; p = 0.007). Similarly, the increase in T2LV over follow-up was associated with an acceleration in thalamic MVF reduction (β = −0.100; p = 0.038) and qT1 increase (β = 0.075; p = 0.008). Patients with disease activity over follow-up had accelerated qT1 increase compared with those without (β = 0.106; p = 0.034). While the baseline EDSS score was not associated with the rates of thalamic qMRI changes, the EDSS progression over follow-up correlated with accelerated rates of decrease in NDI (β = −0.067; p = 0.039) and QSM (β = −0.077; p = 0.027) values. Baseline normalized thalamic volume was linked to the rates of change in MVF (β = 0.462; p = 0.035), qT1 (β = −0.577; p < 0.0001), and NDI (β = 0.343; p = 0.036) values. The rate of brain volume loss over follow-up was associated with the rate of qT1 change (β = −0.059; p = 0.018) (eTable 17).

Baseline variables associated with the subsequent rate of thalamic volume loss in PwMS included age (β = −0.045; p = 0.034), T2LV (β = −0.084; p = 0.004), and MVF and qT1 in the thalamic region close to the CSF (β = 0.111; p = 0.028 and β = −0.425; p = 0.0001, respectively). In HCs, baseline age was the only significant predictor of the rate of thalamic volume loss (β = −0.103; p = 0.0002) (eTable 18).

Sensitivity Analyses

Excellent agreement was observed between qMRI values obtained with the original thalamic masks and with 1-voxel eroded thalamic masks [qT1: ICC = 0.91 (95% CI 0.90–0.93); MVF: ICC = 0.93 (95% CI 0.91–0.94); NDI: ICC = 0.95 (95% CI 0.94–0.96); QSM: ICC = 0.98 (95% CI 0.98–0.98)].

Good agreement was also measured between the estimations of NDI obtained with SMT and NODDI [ICC = 0.85 (95% CI 0.83–0.88)].

Significant between-group differences in thalamic microstructural changes were confirmed also when accounting for treatment effect (eTables 19–22) and disease activity over the follow-up (eTable 23) and in subgroups matched for age, sex, and disease duration (eTables 24–35).

Discussion

We conducted an in vivo multiparametric qMRI study to investigate the pathophysiologic processes occurring in the thalamus of PwMS. Our selection of qMRI contrasts, which leverage different biophysical tissue properties,¹⁸ aimed to provide a comprehensive characterization of microstructural changes linked to demyelination, neuroaxonal loss, and disturbances in iron homeostasis.

The study revealed substantial thalamic alterations in PwMS compared with HCs at baseline and accelerated rates of thalamic degeneration over a 2-year follow-up. Remarkably, the observed thalamic qMRI changes were more severe in patients with PMS and exhibited extensive correlations with both MRI-derived measures of disease burden and clinical measures of neurologic disability and cognitive impairment.

Thalamic pathology in PwMS manifested at a macroscopic level by significant thalamic atrophy. At a microstructural level, thalamic damage was demonstrated by extensive qMRI alterations, not limited to focal lesions but diffusely extending to the tissue without visible changes on conventional FLAIR images. The reduction in thalamic volume that we observed in PwMS, especially in those with PMS, is a well-recognized hallmark of the disease, reflecting heterogeneous neurodegenerative processes.^3,12 Owing to its dependence on widespread pathophysiologic mechanisms involving the entire CNS, thalamic atrophy has been proposed as a marker of the overall extent of MS-related damage¹² and used as an end point in clinical trials.³ Nevertheless, thalamic volume loss is a relatively unspecific marker, representing the final result of a complex interplay between various pathologic mechanisms occurring at the microstructural level. These microstructural changes in our study were demonstrated by alterations in qT1, MVF, and NDI values within thalamic lesions and qT1, MVF, and QSM values within the normal-appearing thalamic tissue.

qT1, which measures the time of recovery of longitudinal magnetization, was prolonged in both thalamic lesions and normal-appearing tissue, indicating microstructural and macrostructural tissue damage.^18,19 Alterations in qT1 have been previously reported in the thalamus of PwMS.^44,45 Notably, while qT1 exhibits overall low specificity to MS pathology, it is highly sensitive to a large spectrum of pathologic changes, encompassing demyelination, axonal loss, iron loss, and accumulation of free water.¹⁸

As a more specific indicator of macromolecular damage (including demyelination and cellular loss), we observed a diffuse reduction in thalamic MVF values. MVF maps were derived from MTsat images, which quantify the exchange of magnetization between free protons and macromolecular protons.⁴⁶ Previous studies have investigated magnetization transfer changes in the thalamus of PwMS, providing mixed results; in a recent meta-analysis, no difference in thalamic magnetization transfer values was measured between patients with RRMS and controls.⁴⁷ In line with these findings, in our cohort, patients with active RRMS did not differ from HCs in MVF levels within the normal-appearing tissue; conversely, a substantial reduction in MVF values was observed in patients with inactive PMS, reflecting a remarkable degree of demyelination and macromolecular damage in the thalamus of this subgroup of patients compared with both HCs and patients with active RRMS. In addition, MVF also showed high sensitivity to the focal damage in thalamic lesions.

We also observed significant alterations in NDI values in thalamic lesions compared with the surrounding tissue, suggesting increased neuroaxonal loss. By contrast, no differences were found in the normal-appearing thalamic tissue between PwMS and HCs. It is noteworthy that in our study, we chose to model microstructural diffusion processes using SMT because of its effectiveness in addressing challenges posed by fiber crossing populations and orientation dispersion, particularly prominent within gray matter regions.²¹ Nonetheless, it is important to note that, like NODDI, SMT is primarily designed for assessing white matter and limitations may arise when exploring structures such as the thalamus.

Previous research on magnetic susceptibility in the thalamus of PwMS has yielded conflicting results.^48,49 In our cohort, we observed a significant decrease in thalamic QSM values among PwMS, consistent with findings in other QSM studies.^45,49,50 An intriguing interpretation involves the depletion of iron from oligodendrocytes, potentially triggered by chronic microglia activation.⁴⁹ On the contrary, focal thalamic lesions within our cohort exhibited no discernible changes in QSM values when compared with perilesional tissue. Notably, QSM is influenced by not only paramagnetic iron concentration but also diamagnetic myelin content.¹⁸ Therefore, we can speculate that a substantial reduction in myelin content within focal lesions might have mitigated the effect of a potential concomitant decline in iron levels, given the opposing impacts of iron and myelin on the QSM signal.¹⁸

Regionally, the observed qMRI changes in PwMS were predominant in the tissue in proximity to the CSF. Accordingly, a distinctive “ependymal-in” gradient of thalamic damage, characterized by neuroaxonal loss and microglia activation, has been described in MS.⁷ This phenomenon is consistently observed across diverse stages of the disease, being evident from the earliest phases of pediatric MS to the progressive stages in patients with advanced disease.^6-8 In our cohort, qMRI changes were also evident regionally in multiple distinct thalamic nuclei. Of interest, a prevalent involvement of the posterior and medial thalamic compartments was evident. Notably, these compartments, which are closest to the CSF, have previously been shown to be more frequently atrophic in MS.⁵¹ Recent research has also indicated that these compartments are more affected by the disconnection of thalamocortical projections due to MS lesions.⁵²

The clinical relevance of the observed qMRI thalamic changes was demonstrated by their extensive correlations with both conventional MRI and clinical measures of disease burden. All qMRI metrics showed robust associations with both the EDSS score and BPF—a proxy measure of cumulative neurodegenerative processes within the CNS. Furthermore, select qMRI measures exhibited correlations with the T2-hyperintense lesion load and clinical measures of cognitive impairment, including SDMT and MuSIC scores. Notably, qMRI metrics of microstructural integrity also proved to substantially contribute to explaining neurologic and cognitive disability and supporting the differentiation between active RRMS and inactive PMS phenotypes, in multivariable LASSO regression models including normalized thalamic volume, BPF, and T2LV as additional predictors.

In our study cohort, significant differences between PwMS and HCs emerged also in the longitudinal rates of thalamic qMRI changes over a 2-year follow-up period. In comparison with HCs, PwMS exhibited accelerated thalamic atrophy and enhanced rates of MVF decrease and qT1 increase, reflecting accentuated neurodegenerative and demyelinating processes. As for cross-sectional findings, between-group differences in longitudinal rates of change were substantially driven by the subgroup of patients with an inactive PMS phenotype, in whom the neurodegenerative processes were notably accentuated. Nevertheless, a significant difference was measurable between the subgroup of patients with active RRMS and HCs in the pace of thalamic qT1 increase. This observation indicates the accumulation of subtle microstructural changes within the normal-appearing thalamic tissue even in patients with a typical RRMS phenotype. In longitudinal studies, thalamic volume has been shown to decline faster in PwMS compared with HCs, but not differently across disease phenotypes^12,14; such a pattern is in line with our results. In our study, this lack of difference in thalamic atrophy rates between disease phases was confirmed even between phenotypes specifically selected to represent the extremes of the clinical spectrum: active RRMS, dominated by neuroinflammatory processes, and inactive PMS, governed by neurodegenerative mechanisms. Conversely, qT1 emerged as a valuable metric for differentiating between patients with active RRMS and inactive PMS. Specifically, patients with inactive PMS exhibited a notable acceleration in qT1 prolongation, suggesting enhanced accumulation of microstructural alterations.

Similar to the cross-sectional analyses, the longitudinal changes of thalamic qMRI metrics showed substantial clinical relevance. Specifically, longitudinal qMRI changes proved to be associated with MRI markers of disease burden, including the accumulation of T2LV and the rate of brain volume loss over time. Moreover, significant differences in longitudinal qMRI changes were observed between patients with clinical stability and those exhibiting disability progression over the clinical follow-up period. Specifically, patients manifesting disease progression over the 2-year interval exhibited an accelerated reduction in thalamic NDI and QSM values, suggestive of hastened neuroaxonal and iron loss.

Overall, our findings suggest 2 concurrent mechanisms as crucial drivers of thalamic degeneration in PwMS: an indirect damage due to the transection of thalamic projections by white matter lesions and a direct surface-in gradient of damage possibly resulting from toxic factors in the CSF. The impact of focal white matter lesions was evidenced by the association between T2LV and more severe microstructural alterations, as well as accelerated pathologic degeneration over time. Furthermore, the burden of T2LV within thalamocortical projections was significantly more impactful in explaining thalamic microstructural and macrostructural degeneration compared with the general T2LV. The role of surface-in degeneration was suggested by the pronounced pathologic changes in the thalamic regions near the CSF. Notably, the extent of microstructural degeneration in these thalamic areas, but not in other regions, was a significant predictor of the subsequent rate of thalamic volume loss.

Strengths of our study include the large sample size and the inclusion of multiple advanced qMRI contrasts obtained with a standardized acquisition protocol allowing for an extensive concomitant investigation of different pathophysiologic processes. In addition, the availability of clinical and advanced MRI follow-up enabled the investigation of the trajectories of thalamic damage over time, along with their predictors and clinical implications. Furthermore, the study design allowed us to reliably demonstrate the added value of microstructural thalamic changes in explaining disease severity beyond other well-established MRI markers. This study also has some limitations. First, follow-up data were not available for the entire cohort; although the loss to follow-up did not seem to disproportionately affect any specific subgroup of patients, the potential for biases cannot be entirely excluded. Second, given the exploratory nature of this study, we limited the application of multiple comparisons correction to instances where multiple regions of interest were simultaneously investigated. Considering the distinct primary focus of the various qMRI contrasts, each probing different pathophysiologic processes with minimal overlap, we chose not to apply multiple comparisons correction for the remaining hypotheses. This decision was intended to mitigate the risk of type II errors while aligning with the overarching objectives of the investigation. Third, disease progression over the follow-up period was determined solely based on 2 EDSS assessments, conducted at baseline and after 2 years. The absence of an additional follow-up assessment to establish a “confirmed disability progression” criterion potentially limits the accuracy in estimating the actual incidence of disease progression. Fourth, although we ensured the absence of neurologic or psychiatric comorbidities in all participants, we did not systematically collect information on cardiovascular risk factors, which might potentially have had an influence on the extent of the observed neurodegenerative changes.

In conclusion, our extensive qMRI protocol revealed substantial pathologic alterations in the thalamus of PwMS. These changes, extending beyond focal lesions to involve normal-appearing tissue, indicated significant micro/macrostructural alterations, including demyelination and perturbations in iron homeostasis. Thalamic damage accumulated faster in PwMS, particularly in those with inactive PMS, and exhibited extensive clinical correlations. Collectively, these findings contribute to a deeper understanding of the pathologic changes in the thalamus of PwMS, underscoring the clinical importance of thalamic damage and its link to disease progression.

Glossary

BPF

brain parenchymal fraction

EDSS

Expanded Disability Status Scale

EPI

echo planar imaging

FLAIR

fluid-attenuated inversion recovery

HC

healthy control

MTsat

magnetization transfer saturation

MuSIC

Multiple Sclerosis Inventory Cognition

MVF

myelin volume fraction

MS

multiple sclerosis

NDI

neurite density index

NODDI

Neurite Orientation Dispersion and Density Imaging

OR

odds ratio

PMS

progressive MS

PwMS

patients with MS

qMRI

quantitative MRI

QSM

quantitative susceptibility mapping

RRMS

relapsing-remitting MS

SDMT

Symbol Digit Modalities Test

SMT

spherical mean technique

T2LV

T2-hyperintense lesion volume

TIV

total intracranial volume

VLMT

Verbal Learning and Memory Test

Appendix. Authors

Name	Location	Contribution
Alessandro Cagol, MD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland; Dipartimento di Scienze della Salute, Università degli Studi di Genova, Italy	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Mario Ocampo-Pineda, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Po-Jui Lu, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Matthias Weigel, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel; Division of Radiological Physics, Department of Radiology, University Hospital Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Muhamed Barakovic, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Lester Melie-Garcia, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Xinjie Chen, MD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Antoine Lutti, PhD	Laboratory for Research in Neuroimaging, Department of Clinical Neuroscience, Lausanne University Hospital and University of Lausanne, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content
Pasquale Calabrese, PhD	Neuropsychology and Behavioral Neurology Unit, Division of Cognitive and Molecular Neuroscience, University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content
Jens Kuhle, MD, PhD	Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Ludwig Kappos, MD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Department of Neurology, University Hospital Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content
Maria Pia Sormani, PhD	Dipartimento di Scienze della Salute, Università degli Studi di Genova; IRCCS Ospedale Policlinico San Martino, Genova, Italy	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Cristina Granziera, MD, PhD	Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel and University of Basel; Research Center for Clinical Neuroimmunology and Neuroscience Basel (RC2NB), University Hospital Basel and University of Basel, Switzerland	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

The authors report no targeted funding.

Disclosure

A. Cagol is supported by EUROSTAR E!113682 HORIZON2020, and received speaker honoraria from Novartis; M. Weigel has received research funding by Biogen for developing spinal cord MRI; M. Barakovic is an employee of Hays plc and a consultant for F. Hoffmann-La Roche Ltd; A. Lutti was supported by the Swiss National Science Foundation (Grant Number: 320030_184784) and the ROGER DE SPOELBERCH foundation; P. Calabrese has received honoraria for speaking at scientific meetings, serving at scientific advisory boards, steering committees and consulting activities from Abbvie, Actelion, Almirall, Bayer-Schering, Biogen, BMS, EISAI, Genzyme, Lundbeck, Merck Serono, Novartis, Sanofi-Aventis, Schwabe and Teva, he also receives research Grants from the Swiss Insurance Medicine (SIM) and the Swiss National Research Foundation; J. Kuhle received speaker fees, research support, travel support, and/or served on advisory boards by Swiss MS Society, Swiss National Research Foundation (320030_189140/1), University of Basel, Progressive MS Alliance, Bayer, Biogen, Celgene, Merck, Novartis, Octave Bioscience, Roche, Sanofi; L. Kappos has received no personal compensation, his institutions (University Hospital Basel/Foundation Clinical Neuroimmunology and Neuroscience Basel) have received and used exclusively for research support: payments for steering committee and advisory board participation, consultancy services, and participation in educational activities from: Actelion, Bayer, BMS, df-mp Molnia & Pohlmann, Celgene, Eli Lilly, EMD Serono, Genentech, Glaxo Smith Kline, Janssen, Japan Tobacco, Merck, MH Consulting, Minoryx, Novartis, F. Hoffmann-La Roche Ltd, Senda Biosciences Inc., Sanofi, Santhera, Shionogi BV, TG Therapeutics, and Wellmera, and license fees for Neurostatus-UHB products; grants from Novartis, Innosuisse, and Roche; M.P. Sormani received consulting fees from Biogen, Merck, Novartis, Roche, Sanofi, Immunic, Alexion; C. Granziera: The University Hospital Basel (USB), as the employer of C.G., has received the following fees which were used exclusively for research support: (1) advisory board and consultancy fees from Actelion, Genzyme-Sanofi, Novartis, GeNeuro and Roche; (2) speaker fees from Genzyme-Sanofi, Novartis, GeNeuro and Roche; (3) research support from Siemens, GeNeuro, Roche, Cristina Granziera is supported by the Swiss National Science Foundation (SNSF) grant PP00P3_176984, the Stiftung zur Förderung der gastroenterologischen und allgemeinen klinischen Forschung and the EUROSTAR E!113682 HORIZON2020; all other authors report no competing interests. Go to Neurology.org/NN for full disclosures.