Original ResearchFree Access

Clinical Relevance of Multiparametric MRI Assessment of Cervical Cord Damage in Multiple Sclerosis

Published Online:https://doi.org/10.1148/radiol.2020200430

Abstract

Background

In multiple sclerosis (MS), knowledge about how spinal cord abnormalities translate into clinical manifestations is incomplete. Comprehensive, multiparametric MRI studies are useful in this perspective, but studies for the spinal cord are lacking.

Purpose

To identify MRI features of cervical spinal cord damage that could help predict disability and disease course in MS by using a comprehensive, multiparametric MRI approach.

Materials and Methods

In this retrospective hypothesis-driven analysis of longitudinally acquired data between June 2017 and April 2019, 120 patients with MS (58 with relapsing-remitting MS [RRMS] and 62 with progressive MS [PMS]) and 30 age- and sex-matched healthy control participants underwent 3.0-T MRI of the brain and cervical spinal cord. Cervical spinal cord MRI was performed with three-dimensional (3D) T1-weighted, T2-weighted, and diffusion-weighted imaging; sagittal two-dimensional (2D) short inversion time inversion-recovery imaging; and axial 2D phase-sensitive inversion-recovery imaging at the C2-C3 level. Brain MRI was performed with 3D T1-weighted, fluid-attenuated inversion-recovery and T2-weighted sequences. Associations between MRI variables and disability were explored with age-, sex- and phenotype-adjusted linear models.

Results

In patients with MS, multivariable analysis identified phenotype, cervical spinal cord gray matter (GM) cross-sectional area (CSA), lateral funiculi fractional anisotropy (FA), and brain GM volume as independent predictors of Expanded Disability Status Scale (EDSS) score (R2 = 0.86). The independent predictors of EDSS score in RRMS were lateral funiculi FA, normalized brain volume, and cervical spinal cord GM T2 lesion volume (R2 = 0.51). The independent predictors of EDSS score in PMS were cervical spinal cord GM CSA and brain GM volume (R2 = 0.44). Logistic regression analysis identified cervical spinal cord GM CSA and T2 lesion volume as independent predictors of phenotype (area under the receiver operating characteristic curve = 0.95). An optimal cervical spinal cord GM CSA cut-off value of 11.1 mm2 was found to enable accurate differentiation of patients with PMS, having values below the threshold, from those with RRMS (sensitivity = 90% [56 of 62], specificity = 91% [53 of 58]).

Conclusion

Cervical spinal cord MRI involvement has a central role in explaining disability in multiple sclerosis (MS): Lesion-induced damage in the lateral funiculi and gray matter (GM) in relapsing-remitting MS and GM atrophy in patients with progressive MS are the most relevant variables. Cervical spinal cord GM atrophy is an accurate predictor of progressive phenotype. Cervical spinal cord GM lesions may subsequently cause GM atrophy, which may contribute to evolution to PMS.

© RSNA, 2020

Online supplemental material is available for this article.

See also the editorial by Zivadinov and Bergsland in this issue.

Summary

Cervical cord damage assessed with multiparametric MRI explains a substantial proportion of disease course in multiple sclerosis.

Key Results

  • ■ In patients with multiple sclerosis (MS), multivariable analysis showed that phenotype, cervical spinal cord gray matter (GM) cross-sectional area, lateral funiculi fractional anisotropy, and brain GM volume were independent predictors of Expanded Disability Status Scale score (R2 = 0.86).

  • ■ A GM cross-sectional cut-off value of 11.1 mm2 was found to enable accurate differentiation of patients with progressive MS from those with relapsing-remitting MS (sensitivity, 90%; specificity, 91%).

Introduction

Spinal cord damage is one of the most important causes of long-term disability in patients with multiple sclerosis (MS) (1,2). MRI can help evaluate the different damaging processes that occur within this small structure (35), including lesions, atrophy, and microstructural damage—both direct and indirect secondary to dying-back and other neurodegenerative phenomena (6,7). Contrary to the thoracic cord, evaluation of the cervical spinal cord is technically feasible with advanced MRI techniques (2,8). Given its precise anatomic organization, the cervical spinal cord is attractive for determining the respective contributions of these different pathologic processes to clinical findings in MS (9).

With regard to lesions, a previous study showed that cervical spinal cord lesion load was correlated with Expanded Disability Status Scale (EDSS) score in a mixed cohort of patients with MS who were studied with use of a single axial slice (10). A recent multicenter study performed with a mixture of two-dimensional (2D) sagittal and three-dimensional (3D) axial images found higher cervical spinal cord lesion frequency in patients with MS who had more severe disease (11). However, the accuracy with axial 3D T2-weighted images is superior to that with sagittal images when assessing cervical spinal cord MS lesions (11,12). In addition, global cervical spinal cord atrophy has been extensively studied (5,1317) in MS by using 3D T1-weighted images; however, cervical spinal cord gray matter (GM) atrophy was demonstrated to be a stronger predictor of disability when compared with brain measures (18). Finally, cervical spinal cord microstructural damage assessed with diffusion-tensor (DT) imaging (9,19) was correlated with severity of clinical disability in patients with MS. However, none of these studies provided a comprehensive picture of cervical spinal cord damage using lesion volume, diffusion indexes, and atrophy at the same time.

Against this background, we aimed to (a) assess cervical spinal cord damage in patients with MS with the main disease clinical phenotypes by applying a multiparametric MRI approach and (b) identify the most accurate predictors of disability and disease course in these patients.

Materials and Methods

The local ethical standards committee on human experimentation approved the study, which was a retrospective hypothesis-driven analysis of longitudinally acquired data. Written informed consent was obtained from all participants.

Participants

Patients with MS were consecutively screened for enrollment between June 2017 and April 2019. Inclusion criteria were diagnosis of MS according to current operational criteria (20) and age at onset greater than 18 years. Clinical exclusion criteria were relapses or use of corticosteroids within 4 weeks before MRI; presence of major medical conditions other than MS; cervical cord damage not due to MS abnormality (eg, cervical trauma); and evidence of cervical spinal cord compression and/or deformity on previous MRI scans. Five patients were excluded owing to movement artifacts and eight were excluded because it was not possible to perform GM segmentation on phase-sensitive inversion-recovery (PSIR) images. Our final cohort included 120 patients with MS (68 women aged 23–70 years [mean age, 46 years] and 52 men aged 24–71 years [mean age, 45 years]). Of these 120 patients, 58 had relapsing-remitting MS (RRMS) and 62 had progressive MS (PMS) (47 with secondary PMS and 15 with primary PMS) (21). See Figure 1 for the complete study flowchart. Thirty age- and sex-matched healthy volunteers (19 women aged 22–69 years [mean age, 42 years] and 11 men aged 25–70 years [mean age, 44 years]) were also included in the study.

Study flowchart for patients with multiple sclerosis (MS). cSC =                        cervical spinal cord, GM = gray matter, PSIR = phase-sensitive                        inversion recovery.

Figure 1: Study flowchart for patients with multiple sclerosis (MS). cSC = cervical spinal cord, GM = gray matter, PSIR = phase-sensitive inversion recovery.

At the time of MRI, patients with MS underwent neurologic examination, which helped determine the EDSS score. The nine-hole peg test and timed 25-foot walk were administered to assess limb function.

MRI Scan Acquisition

With use of a 3.0-T scanner, images of the cervical spinal cord were obtained with the following sequences: 3D T1-weighted turbo field echo; 3D T2-weighted fast field echo; cardiac-gated, reduced field-of-view 2D diffusion-weighted pulsed gradient spin-echo single-shot echo-planar with diffusion weighting (b factor, 700 sec/mm2) applied along six noncollinear directions (22); 2D short inversion time inversion-recovery; and 2D PSIR at the C2-C3 intervertebral disk level. During the same scanning session, images of the brain were acquired with the following sequences: 3D T1-weighted turbo field echo; 3D T2-weighted fluid-attenuated inversion-recovery turbo spin echo; and 3D T2-weighted turbo spin echo. Whenever possible, in case of movement artifacts (particularly with the diffusion-weighted and PSIR sequences), the acquisition was repeated. Details of the MRI protocol are available in Appendix E1 (online).

MRI Analysis of the Cervical Spinal Cord

T1-weighted and short inversion time inversion-recovery images were registered to the axial 3D T2-weighted image by using the Spinal Cord Toolbox (23). Focal T2 hyperintense lesions were identified on axial 3D T2-weighted images by consensus of two experienced observers (P.P. and E.D.M., with 10 and 5 years of experience, respectively) who were blinded to demographic and clinical information. Lesions were then segmented by using a semiautomated local thresholding segmentation technique (Jim 8; Xinapse Systems, Colchester, England). T1-weighted and short inversion time inversion-recovery images aided lesion identification and separation from artifacts. Normalized T2 hyperintense lesion volume between C1 and C5 vertebral levels was calculated for the whole cervical spinal cord, GM, white matter (WM), dorsal columns, and lateral funiculi by using the T2 lesion masks and a probabilistic regional atlas available in the Spinal Cord Toolbox (24). Normalized lesion volume was calculated by dividing the total lesion volume within a region by the volume of this region, thus indicating the percentage of tissue affected by lesions in the respective region (11).

Total cervical spinal cord cross-sectional area (CSA) was calculated on PSIR images by using the active surface method (25). The GM was manually segmented three times by the two experienced readers, who were unaware of the participant’s identity and clinical information, and the mean GM CSA was calculated (18). When the presence of lesions at the C2-C3 vertebral level precluded accurate GM border definition, the participant was excluded. Intra- and interrater reliability for GM area assessments showed intraclass correlation coefficients of 0.98 and 0.90, respectively. WM CSA was calculated as the difference between total and GM CSAs.

After correction for motion and distortions, diffusion-weighted images were registered to the T2-weighted image. The DT was estimated with linear regression (26) by using FSL software (version 5.0.5; Analysis Group, FMRIB, Oxford, England), and fractional anisotropy (FA) and mean diffusivity (MD) were derived. Average FA and MD were obtained within lesions and in the lesion-free cervical spinal cord, which was defined as those voxels free of T2 hyperintense lesions within the WM, dorsal columns, and lateral funiculi.

Figure 2 illustrates the cervical spinal cord analysis pipeline. Detailed cervical spinal cord MRI analysis is available in Appendix E1 (online).

Spinal cord MRI analysis. Central slice from a sagittal short                        inversion time inversion-recovery image in a 32-year-old man with secondary                        progressive multiple sclerosis. A, Total cord (green) and gray matter (blue)                        segmentation on axial phase-sensitive inversion-recovery (PSIR) image at                        C2-C3 vertebral level. B, Lesion segmentation (red) on axial T2-weighted                        images; T1-weighted images were used to improve lesion identification. C,                        Axial fractional anisotropy (FA) and mean diffusivity (MD) maps. D, Spinal                        cord (SC) regional atlas registered to the T2-weighted image (gray matter in                        purple, white matter in yellow, dorsal columns in light blue, lateral                        funiculi in red).

Figure 2: Spinal cord MRI analysis. Central slice from a sagittal short inversion time inversion-recovery image in a 32-year-old man with secondary progressive multiple sclerosis. A, Total cord (green) and gray matter (blue) segmentation on axial phase-sensitive inversion-recovery (PSIR) image at C2-C3 vertebral level. B, Lesion segmentation (red) on axial T2-weighted images; T1-weighted images were used to improve lesion identification. C, Axial fractional anisotropy (FA) and mean diffusivity (MD) maps. D, Spinal cord (SC) regional atlas registered to the T2-weighted image (gray matter in purple, white matter in yellow, dorsal columns in light blue, lateral funiculi in red).

Brain MRI Analysis

Brain T2 hyperintense lesion volume was measured by using Jim software (version 8.0; Xinapse Systems, Colchester, England). Volumes normalized for head size of the whole brain, GM, and WM were assessed by using SIENAX on lesion-filled 3D T1-weighted images.

Statistical Analysis

Continuous demographic and clinical variables were compared between groups by using the two-sample t test or Mann-Whitney U test, according to variable distribution. The χ2 test was applied for categorical data. Inverse scores were calculated for the nine-hole peg test (right and left hand) and timed 25-foot walk measures due to their skewed distribution. Inverse nine-hole peg test (average of the left and right hands) and timed 25-foot walk measures were converted to z scores by using means and standard deviations of the healthy control group (27). A composite limb function score was calculated by averaging nine-hole peg test and timed 25-foot walk z scores.

Brain and cervical spinal cord T2 hyperintense lesion volume measures underwent square-root transformation. MRI variables were compared by using age- and sex-adjusted linear models, with false-discovery rate (Benjamini-Hochberg procedure) correction (28) to account for the overall number of pairwise contrasts.

In patients with MS, the associations of MRI variables with disability scores were explored with age- and sex-adjusted linear models. The analysis was performed with or without phenotype as a covariate. Differences between phenotypes were tested by examining specific interaction terms. Standardized beta coefficients were reported to compare the relative strength of association of each variable. False discovery rate (Benjamini-Hochberg procedure) correction (28) was applied for each outcome separately to adjust for the overall number of associations tested.

Age- and sex-adjusted multivariable linear models with step-wise variable selection were used to identify independent predictors of disability scores. The analyses were repeated after removing phenotype from the set of explanatory variables. Interaction term selection was hierarchically contingent upon significance of the corresponding principal effects. Similar models were built for patients with RRMS and PMS separately. Standardized beta coefficients and the proportion of variance explained (R2 index) were reported for each model. Age- and sex-adjusted step-wise multivariable logistic regression was performed to identify independent predictors of MS phenotype. Model discrimination was assessed with the area under the curve.

For all analyses, a statistically significant difference was set at P < .05 (SAS Software, version 9.4 [SAS Institute, Cary, NC]; R Software, version 3.6.0 [R Core Team, Vienna, Austria).

Results

Demographic and Clinical Characteristics

Demographic and clinical characteristics of patients with MS and the healthy control group are summarized in Table 1.

Table 1: Summary of Demographic and Clinical Characteristics of Healthy Control Participants and Patients with MS

Table 1:

MRI Findings

Table 2 summarizes the main MRI measures from the healthy control group and patients with RRMS and PMS. No cervical spinal cord lesions were found in the healthy control group. Compared with the healthy control group, both patient groups had higher brain T2 hyperintense lesion volume (P < .001) and atrophy of the brain (ie, reduced volumes) and cervical spinal cord (ie, reduced CSA) in toto and in WM compartments (P < .01 for RRMS and P < .001 for PMS). Compared with the healthy control group, patients with PMS also had GM atrophy of the brain and cervical spinal cord (P < .001). DT indexes did not differ between patients with RRMS and healthy control participants. Decreased FA and increased MD were found in patients with PMS compared with control participants for all cervical spinal cord WM compartments (P < .05), except the lateral funiculi MD (P = .12). Compared with patients with RRMS, patients with PMS had higher T2 hyperintense lesion volume in the brain (P < .05) and cervical spinal cord (P < .001), GM and global atrophy of the brain and cervical spinal cord (P ≤ .001), more pronounced cervical spinal cord WM atrophy (P < .05), and more severe DT abnormalities in lesion-free cervical spinal cord WM compartments (decreased FA [P < .05] and a trend toward increased MD [uncorrected P < .05]). DT indexes did not differ between patient groups for lesions and among all groups for GM compartment.

Table 2: Between-Group Comparison of MRI Variables

Table 2:

Associations with EDSS and Composite Limb Function Scores

Table 3 and Figure 3 summarize the associations between MRI variables and EDSS score. In patients with MS, EDSS score was associated with cervical spinal cord lateral funiculi and GM normalized T2 hyperintense lesion volume (P < .05); brain T2 hyperintense lesion volume (P = .004); cervical spinal cord GM, brain GM, and global atrophy (P < .01); and lesion-free cervical spinal cord WM diffusivity indexes (P < .001 for FA; P < .05 for MD). Without phenotype adjustment, significant associations were found for cervical spinal cord normalized T2 hyperintense lesion volume (P < .001), brain T2 hyperintense lesion volume (P = .005), brain GM and global atrophy (P < .001), and cervical spinal cord compartment FA (P < .001) and MD (P < .05). No significant associations were found for brain WM atrophy (P = .06) and diffusivity indexes within cervical spinal cord GM and lesions (P > .05, see Table E1 [online]).

Table 3: Associations of MRI Variables with EDSS Score within Groups

Table 3:
Graphs show associations of MRI variables with Expanded Disability                        Status Scale (EDSS) at univariable analysis. Associations (age- and                        sex-adjusted linear models) of the main MRI variables were selected by                        multivariable models with EDSS score in patients with relapsing-remitting                        multiple sclerosis (RRMS) (blue), progressive multiple sclerosis (PMS)                        (red), and all multiple sclerosis (MS) (phenotype unadjusted, dashed gray                        line). cSC = cervical spinal cord, FA = fractional anisotropy, GM                        = gray matter, GMV = gray matter volume, lf = lesion free,                        NBV = normalized brain volume, nT2-LV = normalized T2 hyperintense                        lesion volume.

Figure 3: Graphs show associations of MRI variables with Expanded Disability Status Scale (EDSS) at univariable analysis. Associations (age- and sex-adjusted linear models) of the main MRI variables were selected by multivariable models with EDSS score in patients with relapsing-remitting multiple sclerosis (RRMS) (blue), progressive multiple sclerosis (PMS) (red), and all multiple sclerosis (MS) (phenotype unadjusted, dashed gray line). cSC = cervical spinal cord, FA = fractional anisotropy, GM = gray matter, GMV = gray matter volume, lf = lesion free, NBV = normalized brain volume, nT2-LV = normalized T2 hyperintense lesion volume.

In patients with RRMS, the EDSS score was associated with cervical spinal cord global and regional normalized T2 hyperintense lesion volume (P values ranged from .02 to .002, except for the dorsal column, with P > .05) and brain T2 hyperintense lesion volume (P = .009); brain GM and global atrophy (P = .02 for both); and lesion-free cervical spinal cord diffusivity indexes (P < .01 for FA, P <.05 for MD values). In patients with PMS, EDSS score was significantly associated with cervical spinal cord and brain GM (P = .003 and P = .03, respectively) and cervical spinal cord WM and global brain atrophy (P = .02 and P = .04, respectively). In addition, there was a trend of association for cervical spinal cord diffusivity indexes in WM compartments (uncorrected P < .05). Significant heterogeneities of association with EDSS score between patients with RRMS and those with PMS were found for cervical spinal cord lateral funiculi normalized T2 hyperintense lesion volume and FA in lesion-free lateral funiculi, WM, and GM.

Similar findings were shown for the associations between MRI variables and composite limb function score, as summarized in Table E2 (online).

From an anatomic point of view, the lateral funiculi and the GM were the regions in which cervical spinal cord normalized T2 hyperintense lesion volume had the strongest correlation with disability scores in patients with RRMS (with EDSS: standardized beta coefficient = 0.24 and 0.21, respectively; with composite limb function score: standardized beta coefficient = −0.17 and −0.23, respectively; P < .05 for all) and all patients with MS (with EDSS: standardized beta coefficient = 0.14 and 0.13, respectively; with composite limb function score: standardized beta coefficient = −0.18 and −0.21, respectively; P < .05 for all). The GM was the cervical spinal cord compartment where CSA reduction showed the strongest correlation with disability in patients with PMS (with EDSS: standardized beta coefficient = −0.29; with composite limb function score: standardized beta coefficient = 0.41; P < .01 for both) and all patients with MS (with EDSS: standardized beta coefficient = −0.24; with composite limb function score: standardized beta coefficient = 0.35; P = .001 for both). Furthermore, the lateral funiculi were the compartment in which FA had the strongest correlation with EDSS (standardized beta coefficient = −0.23 for all patients with MS [P < .001], −0.34 for patients with RRMS [P < .001], and −0.12 for patients with PMS [P = .06]) and composite limb function score (standardized beta coefficient = 0.25 for all patients with MS [P < .001], 0.17 for patients with RRMS [P = .06], and 0.35 for patients with PMS [P < .001]) in all groups of patients.

Table 4 summarizes the results of multivariable analyses. In patients with RRMS, cervical spinal cord lesion-free lateral funiculi FA, normalized brain volume, and cervical spinal cord GM normalized T2 hyperintense lesion volume were independent predictors of disability. In patients with PMS, cervical spinal cord GM atrophy, brain GM volume, and lesion-free lateral funiculi FA were independent predictors of disability.

Table 4: Independent Predictors of EDSS and Composite Limb Function Scores within Groups

Table 4:

Logistic Multiple Regression Analysis

In patients with MS, logistic multivariable regression analysis identified cervical spinal cord GM CSA (odds ratio associated with a 1 standard deviation increase = 0.08, P < .001) and cervical spinal cord GM normalized T2 hyperintense lesion volume (odds ratio associated with a 1 standard deviation increase = 2.50, P = .03) as independent predictors of phenotype (area under the curve = 0.95). Given the prominent role of cervical spinal cord GM CSA in predicting MS phenotype, the optimal (sensitivity = 90% [56 of 62], specificity = 91% [53 of 58]) cut-off point of 11.1 mm2 for differentiating patients with PMS from those with RRMS was identified (Fig 4).

Cervical spinal cord (SC) gray matter (GM) atrophy and disease                        phenotype. Graph shows distribution of cervical spinal cord GM                        cross-sectional area according to multiple sclerosis (MS) phenotype, with an                        optimal cut-off point (dashed green line) of 11.1 mm2 (sensitivity =                        90% [56 of 62], specificity = 91% [53 of 58]) enabling differentiation                        of patients with progressive MS (red) from those with relapsing-remitting MS                        (blue) (logistic regression analysis). Distribution in healthy control                        participants (green) is shown for comparison.

Figure 4: Cervical spinal cord (SC) gray matter (GM) atrophy and disease phenotype. Graph shows distribution of cervical spinal cord GM cross-sectional area according to multiple sclerosis (MS) phenotype, with an optimal cut-off point (dashed green line) of 11.1 mm2 (sensitivity = 90% [56 of 62], specificity = 91% [53 of 58]) enabling differentiation of patients with progressive MS (red) from those with relapsing-remitting MS (blue) (logistic regression analysis). Distribution in healthy control participants (green) is shown for comparison.

Discussion

We aimed to investigate the complex pathophysiologic processes occurring in the cervical spinal cord of patients with multiple sclerosis (MS) and their role in explaining clinical disability. For this purpose, we adopted a comprehensive multiparametric MRI approach based on the combination of lesion volume, atrophy measures, and diffusivity metrics, which reflect underlying pathologic substrates of the disease (focal inflammation, tissue loss, and microstructural disruption, respectively). Brain MRI measures were also included, thus providing a global assessment of MS-related damage with the aim of identifying the main determinants of clinical disability.

To study focal inflammation, cervical spinal cord lesion volume was assessed by using an axial 3D T2-weighted sequence, with wide coverage of the cervical spinal cord (from C1 to C5). This is a substantial improvement over previous studies, which assessed correlations with EDSS score by using less accurate methods: lesion counts instead of lesion volume (29) and lesion volume on a single axial slice (10) or on a mixture of axial and sagittal 2D and 3D images (11). Indeed, lesion volume assessment was shown to be superior to lesion count in reflecting damage in MS (30), and axial slices showed superior sensitivity for the detection of MS lesions in the cervical spinal cord compared with sagittal ones (12,31). In addition, we normalized cervical spinal cord T2 hyperintense lesion volume in each compartment, thus obtaining measures easier to interpret and compare between individuals. In line with previous data (11), our results show that cervical spinal cord lesions are not generally seen in healthy control participants, whereas patients with PMS have higher normalized T2 hyperintense lesion volume in all cervical spinal cord compartments compared with patients with RRMS.

To assess neurodegeneration, global cervical spinal cord atrophy has been extensively studied (1417,32) in MS by using 3D T1-weighted images, which unfortunately do not have sufficient contrast or spatial resolution to allow separation of GM and WM. High-spatial-resolution axial PSIR has recently emerged as a suitable pulse sequence for accurately measuring GM, WM, and global cervical spinal cord CSAs as a surrogate for atrophy. A previous study demonstrated GM CSA reduction in both patients with RRMS and those with PMS and WM and global CSA reduction only in those with PMS (18). Conversely, we found WM and global cervical spinal cord atrophy in both patients with RRMS and those with PMS and significant GM atrophy only in those with PMS (compared with both the healthy control group and patients with RRMS). These results are in line with those of a meta-analysis of 94 studies reporting global cervical spinal cord atrophy in patients with PMS versus those with RRMS and in patients with RRMS versus healthy control participants (15).

Microstructural tissue abnormalities, which are quantifiable with DT MRI, are known to occur before the onset of atrophy (9). With use of manual segmentation of GM and WM on T2-weighted images, a previous study found altered DT indexes in all cervical spinal cord compartments in patients with relapse-onset MS, with more severe abnormalities in those with secondary PMS (19). In our study, we derived cervical spinal cord DT measures with use of a probabilistic atlas. This method may be prone to partial volume effects. However, this was minimized by the use of cervical spinal cord images (including diffusion-weighted images) with high spatial resolution in the axial plane (voxel area equal to or less than 1 mm2) and by eroding one voxel at the cervical spinal cord–cerebrospinal fluid interface. This strategy avoided human bias and allowed assessment of the WM compartment over its full extent (not just a central manually selected circle as in the previous work) (19). Our results confirmed a trend toward greater damage (decreased FA and increased MD) transitioning from the healthy control group to patients with RRMS and to patients with PMS in all cervical spinal cord WM compartments. DT metrics within GM and lesions did not differ among groups, likely due to inadequate sensitivity of the DT model in these compartments.

Results of group comparisons for brain MRI variables are in line with those in the previous literature (33). GM atrophy is an orderly process in MS, occurring in selected brain regions (eg, deep GM nuclei) from the earliest stages of the disease and then extending to global atrophy, correlating with clinical disability (33,34). Accordingly, given their mild disability, our patients with RRMS did not have significant global GM atrophy of the brain compared with the control group, a result also obtained in a previous cross-sectional study (35). Instead, they had significant WM atrophy, which has been demonstrated to occur as a global measure from the earliest stages of MS—even in patients with clinically isolated syndrome (36).

Results of univariable analyses indicate a prominent role of lateral funiculi damage, in terms of both focal inflammation and microstructural tissue disruption, in explaining disability across all MS phenotypes. In addition to lateral funiculi, cervical spinal cord GM is another important target of clinically meaningful damage in MS, manifesting on MRI scans as focal lesions and atrophy. Our results underscore the importance of identifying the location of damage in addition to its severity in this condition.

Results of multivariable analysis identified lesion-induced damage in cervical spinal cord (particularly if it affects the lateral funiculi and GM) and brain atrophy as central in determining disability in patients with RRMS. As a matter of fact, previous pathologic studies have demonstrated axonal loss and damage (reflected by decreased FA in our study) as a consequence of focal lesions along the corticospinal tracts (ie, in lateral funiculi), strongly impacting clinical disability (7). Lateral funiculi damage (lesions and microscopic disruption) is disproportionately more relevant in explaining disability in RRMS compared with PMS, as demonstrated by multivariable models and the significant heterogeneities of associations of lateral funiculi normalized T2 hyperintense lesion volume and FA in univariable models.

Results of multivariable analysis identified GM damage as central in explaining disability in patients with PMS, with a more prominent role of cervical spinal cord compared with the brain, in line with previous results (18). Lateral funiculi microstructural damage (reflected by decreased FA) also seems to play a role.

At logistic multivariable regression analysis, cervical spinal cord GM normalized T2 hyperintense lesion volume was retained as an independent predictor of phenotype in addition to atrophy. This probably either reflects the presence of a subset of patients with PMS without substantial cervical spinal cord GM atrophy, in whom, however, the cervical spinal cord is heavily affected by lesions and/or points at the existence of a subset of patients in whom intrinsic cervical spinal cord GM lesions—not identified with T2-weighted and PSIR pulse sequences—may lead to falsely elevated CSA values. The role of cervical spinal cord normalized T2 hyperintense lesion volume in predicting a progressive phenotype is in line with findings in a previous study that highlighted a preferential distribution in the GM of cervical spinal cord lesions in patients with PMS (11).

Collectively, our findings demonstrate a prominent and central role of cervical spinal cord GM damage in determining clinical disability and a progressive phenotype in MS. Lesions in this compartment were an independent predictor of disability in patients with RRMS and in all patients with MS. Patients with PMS showed cervical spinal cord GM atrophy (probably a result of neuronal loss and decreased synaptic complexity) (6,7,37) and higher lesion volumes in this compartment compared with patients with RRMS. Finally, cervical spinal cord GM atrophy was the main independent predictor of a progressive phenotype and of clinical disability in patients with PMS and all patients with MS. According to these results, it is tempting to speculate that GM atrophy may be the end-stage result of a damaging mechanism initiated by lesions. GM atrophy may ensue after a certain threshold of damage is reached, then progressively develop in patients with RRMS, eventually being crucial in the transition to a progressive disease stage.

If confirmed in a longitudinal setting, these findings may have important clinical applications. Indeed, the introduction of axial T2-weighted and PSIR sequences for lesion and GM atrophy assessment could provide important pieces of information for prognosis and follow-up and for the identification of a PMS phenotype. In addition, our findings might also influence treatment decision making by offering highly active disease-modifying therapy to patients with cervical spinal cord lesions, particularly in the case of GM involvement.

Our study has a number of limitations. First, a few patients were excluded due to movement artifacts or the difficulty in accurately delineating the GM border on PSIR images owing to lesions. Second, cervical spinal cord GM segmentation was performed manually, and this could have resulted in operator-dependent bias. However, a fully automated segmentation method is not currently available, and the interrater reliability of GM area assessments was excellent (intraclass correlation coefficients, 0.90). Third, the small number of patients with PMS did not permit separate analysis of this group. Moreover, the cervical spinal cord could only be studied from the C1 to C5 vertebral levels because of the prevalence of artifacts at lower levels, impacting the accuracy of lesion identification and segmentation and DT metrics. Furthermore, our study included a retrospectively selected healthy control group. Finally, this was a cross-sectional study, and longitudinal associations between cervical spinal cord MRI variables and EDSS score progression remain to be determined.

In conclusion, cervical spinal cord MRI involvement has a central role in explaining disability in multiple sclerosis (MS): Lesion-induced damage in the lateral funiculi and gray matter (GM) in relapsing-remitting MS and GM atrophy in patients with progressive MS (PMS) are the most relevant variables. A cervical spinal cord GM cross-sectional area less than 11.1 mm2 was an accurate predictor of a progressive phenotype. Longitudinal studies are needed to validate our hypothesis that cervical spinal cord GM lesions may subsequently cause GM atrophy, which may lead to evolution to PMS.

Disclosures of Conflicts of Interest: R.B. disclosed no relevant relationships. E.P. disclosed no relevant relationships. A.M. disclosed no relevant relationships. L.C. disclosed no relevant relationships. P.P. disclosed no relevant relationships. E.D.M. disclosed no relevant relationships. M.F. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: is a paid consultant for Bayer, Biogen Idec, Merck-Serono, Novartis, Roche, Sanofi Genzyme, Takeda, and Teva Pharmaceutical Industries; institution has grants/grants pending from Biogen Idec, Merck-Serono, Novartis, Roche, and Teva Pharmaceutical Industries; receives payment for lectures including service on speakers bureaus from Bayer, Biogen Idec, Merck-Serono, Novartis, Roche, Sanofi Genzyme, Takeda, and Teva Pharmaceutical Industries. Other relationships: disclosed no relevant relationships. M.A.R. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: receives payment for lectures including service on speakers bureaus from Bayer, Biogen Idec, Celgene, Merck-Serono, Novartis, Roche, and Teva Pharmaceutical Industries disclosed no relevant relationships. Other relationships: disclosed no relevant relationships.

Author Contributions

Author contributions: Guarantors of integrity of entire study, R.B., M.A.R.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, R.B., E.P., L.C., P.P., E.D.M., M.A.R.; clinical studies, R.B., L.C., E.D.M., M.A.R.; experimental studies, R.B., E.P., P.P., E.D.M.; statistical analysis, R.B., A.M., E.D.M.; and manuscript editing, R.B., A.M., L.C., E.D.M., M.F., M.A.R.

References

  • 1. Gass A, Rocca MA, Agosta F, et al. MRI monitoring of pathological changes in the spinal cord in patients with multiple sclerosis. Lancet Neurol 2015;14(4):443–454. Crossref, MedlineGoogle Scholar
  • 2. Kearney H, Miller DH, Ciccarelli O. Spinal cord MRI in multiple sclerosis--diagnostic, prognostic and clinical value. Nat Rev Neurol 2015;11(6):327–338. Crossref, MedlineGoogle Scholar
  • 3. Filippi M, Brück W, Chard D, et al. Association between pathological and MRI findings in multiple sclerosis. Lancet Neurol 2019;18(2):198–210. Crossref, MedlineGoogle Scholar
  • 4. Filippi M, Bar-Or A, Piehl F, et al. Multiple sclerosis. Nat Rev Dis Primers 2018;4(1):43 [Published correction appears in Nat Rev Dis Primers 2018;4(1):49.]. Crossref, MedlineGoogle Scholar
  • 5. Filippi M, Preziosa P, Rocca MA. Microstructural MR Imaging Techniques in Multiple Sclerosis. Neuroimaging Clin N Am 2017;27(2):313–333. Crossref, MedlineGoogle Scholar
  • 6. Gilmore CP, DeLuca GC, Bö L, et al. Spinal cord neuronal pathology in multiple sclerosis. Brain Pathol 2009;19(4):642–649. Crossref, MedlineGoogle Scholar
  • 7. Petrova N, Carassiti D, Altmann DR, Baker D, Schmierer K. Axonal loss in the multiple sclerosis spinal cord revisited. Brain Pathol 2018;28(3):334–348. Crossref, MedlineGoogle Scholar
  • 8. Moccia M, Ruggieri S, Ianniello A, Toosy A, Pozzilli C, Ciccarelli O. Advances in spinal cord imaging in multiple sclerosis. Ther Adv Neurol Disord 2019;12:1756286419840593. CrossrefGoogle Scholar
  • 9. Oh J, Saidha S, Chen M, et al. Spinal cord quantitative MRI discriminates between disability levels in multiple sclerosis. Neurology 2013;80(6):540–547. Crossref, MedlineGoogle Scholar
  • 10. Kearney H, Altmann DR, Samson RS, et al. Cervical cord lesion load is associated with disability independently from atrophy in MS. Neurology 2015;84(4):367–373. Crossref, MedlineGoogle Scholar
  • 11. Eden D, Gros C, Badji A, et al. Spatial distribution of multiple sclerosis lesions in the cervical spinal cord. Brain 2019;142(3):633–646. Crossref, MedlineGoogle Scholar
  • 12. Galler S, Stellmann JP, Young KL, et al. Improved Lesion Detection by Using Axial T2-Weighted MRI with Full Spinal Cord Coverage in Multiple Sclerosis. AJNR Am J Neuroradiol 2016;37(5):963–969. Crossref, MedlineGoogle Scholar
  • 13. Rocca MA, Horsfield MA, Sala S, et al. A multicenter assessment of cervical cord atrophy among MS clinical phenotypes. Neurology 2011;76(24):2096–2102. Crossref, MedlineGoogle Scholar
  • 14. Rocca MA, Valsasina P, Damjanovic D, et al. Voxel-wise mapping of cervical cord damage in multiple sclerosis patients with different clinical phenotypes. J Neurol Neurosurg Psychiatry 2013;84(1):35–41. Crossref, MedlineGoogle Scholar
  • 15. Casserly C, Seyman EE, Alcaide-Leon P, et al. Spinal Cord Atrophy in Multiple Sclerosis: A Systematic Review and Meta-Analysis. J Neuroimaging 2018;28(6):556–586. Crossref, MedlineGoogle Scholar
  • 16. Tsagkas C, Magon S, Gaetano L, et al. Spinal cord volume loss: A marker of disease progression in multiple sclerosis. Neurology 2018;91(4):e349–e358. Crossref, MedlineGoogle Scholar
  • 17. Tsagkas C, Magon S, Gaetano L, et al. Preferential spinal cord volume loss in primary progressive multiple sclerosis. Mult Scler 2019;25(7):947–957. Crossref, MedlineGoogle Scholar
  • 18. Schlaeger R, Papinutto N, Panara V, et al. Spinal cord gray matter atrophy correlates with multiple sclerosis disability. Ann Neurol 2014;76(4):568–580. Crossref, MedlineGoogle Scholar
  • 19. Kearney H, Schneider T, Yiannakas MC, et al. Spinal cord grey matter abnormalities are associated with secondary progression and physical disability in multiple sclerosis. J Neurol Neurosurg Psychiatry 2015;86(6):608–614. Crossref, MedlineGoogle Scholar
  • 20. Thompson AJ, Banwell BL, Barkhof F, et al. Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria. Lancet Neurol 2018;17(2):162–173. Crossref, MedlineGoogle Scholar
  • 21. Lublin FD, Reingold SC, Cohen JA, et al. Defining the clinical course of multiple sclerosis: the 2013 revisions. Neurology 2014;83(3):278–286. Crossref, MedlineGoogle Scholar
  • 22. Wheeler-Kingshott CA, Hickman SJ, Parker GJ, et al. Investigating cervical spinal cord structure using axial diffusion tensor imaging. Neuroimage 2002;16(1):93–102. https://doi.org/10.1006/nimg.2001.1022. Crossref, MedlineGoogle Scholar
  • 23. De Leener B, Kadoury S, Cohen-Adad J. Robust, accurate and fast automatic segmentation of the spinal cord. Neuroimage 2014;98:528–536. Crossref, MedlineGoogle Scholar
  • 24. Lévy S, Benhamou M, Naaman C, Rainville P, Callot V, Cohen-Adad J. White matter atlas of the human spinal cord with estimation of partial volume effect. Neuroimage 2015;119:262–271. Crossref, MedlineGoogle Scholar
  • 25. Horsfield MA, Sala S, Neema M, et al. Rapid semi-automatic segmentation of the spinal cord from magnetic resonance images: application in multiple sclerosis. Neuroimage 2010;50(2):446–455. Crossref, MedlineGoogle Scholar
  • 26. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66(1):259–267. Crossref, MedlineGoogle Scholar
  • 27. Cutter GR, Baier ML, Rudick RA, et al. Development of a multiple sclerosis functional composite as a clinical trial outcome measure. Brain 1999;122(Pt 5):871–882. Crossref, MedlineGoogle Scholar
  • 28. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Series B Stat Methodol 1995;57(1):289–300. Google Scholar
  • 29. Dula AN, Pawate S, Dortch RD, et al. Magnetic resonance imaging of the cervical spinal cord in multiple sclerosis at 7T. Mult Scler 2016;22(3):320–328. Crossref, MedlineGoogle Scholar
  • 30. Filippi M, Horsfield MA, Tofts PS, Barkhof F, Thompson AJ, Miller DH. Quantitative assessment of MRI lesion load in monitoring the evolution of multiple sclerosis. Brain 1995;118(Pt 6):1601–1612. Crossref, MedlineGoogle Scholar
  • 31. Breckwoldt MO, Gradl J, Hähnel S, et al. Increasing the sensitivity of MRI for the detection of multiple sclerosis lesions by long axial coverage of the spinal cord: a prospective study in 119 patients. J Neurol 2017;264(2):341–349. Crossref, MedlineGoogle Scholar
  • 32. Rocca MA, Sormani MP, Rovaris M, et al. Long-term disability progression in primary progressive multiple sclerosis: a 15-year study. Brain 2017;140(11):2814–2819. Crossref, MedlineGoogle Scholar
  • 33. Rocca MA, Battaglini M, Benedict RH, et al. Brain MRI atrophy quantification in MS: From methods to clinical application. Neurology 2017;88(4):403–413. Crossref, MedlineGoogle Scholar
  • 34. Eshaghi A, Marinescu RV, Young AL, et al. Progression of regional grey matter atrophy in multiple sclerosis. Brain 2018;141(6):1665–1677. Crossref, MedlineGoogle Scholar
  • 35. Bergsland N, Horakova D, Dwyer MG, et al. Subcortical and cortical gray matter atrophy in a large sample of patients with clinically isolated syndrome and early relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 2012;33(8):1573–1578. Crossref, MedlineGoogle Scholar
  • 36. Rocca MA, Preziosa P, Mesaros S, et al. Clinically Isolated Syndrome Suggestive of Multiple Sclerosis: Dynamic Patterns of Gray and White Matter Changes-A 2-year MR Imaging Study. Radiology 2016;278(3):841–853. LinkGoogle Scholar
  • 37. Wegner C, Esiri MM, Chance SA, Palace J, Matthews PM. Neocortical neuronal, synaptic, and glial loss in multiple sclerosis. Neurology 2006;67(6):960–967. Crossref, MedlineGoogle Scholar

Article History

Received: Feb 12 2020
Revision requested: Mar 9 2020
Revision received: Apr 27 2020
Accepted: May 1 2020
Published online: June 23 2020
Published in print: Sept 2020