Article Text

Download PDFPDF

A diffusion tensor MRI study of cervical cord damage in benign and secondary progressive multiple sclerosis patients
  1. B Benedetti1,2,
  2. M A Rocca1,2,
  3. M Rovaris1,3,
  4. D Caputo3,
  5. M Zaffaroni4,
  6. R Capra5,
  7. A Bertolotto6,
  8. V Martinelli2,
  9. G Comi2,
  10. M Filippi1,2
  1. 1
    Neuroimaging Research Unit, Scientific Institute and University Ospedale San Raffaele, Milan, Italy
  2. 2
    Department of Neurology, Scientific Institute and University Ospedale San Raffaele, Milan, Italy
  3. 3
    Multiple Sclerosis Centre, Scientific Institute Don Gnocchi, Milan, Italy
  4. 4
    Multiple Sclerosis Centre, Ospedale di Gallarate, Gallarate, Italy
  5. 5
    Multiple Sclerosis Centre, Spedali Civili, Brescia, Italy
  6. 6
    Multiple Sclerosis Centre, Ospedale San Luigi, Orbassano, Italy
  1. Correspondence to Dr M Filippi, Neuroimaging Research Unit, Institute of Experimental Neurology, Scientific Institute and University Ospedale San Raffaele, via Olgettina, 60, 20132 Milan, Italy; m.filippi{at}


Background: Diffusion tensor (DT) MRI enables quantification of the severity of brain and cervical cord pathology in multiple sclerosis (MS).

Objective: To investigate DT MRI patterns of cervical cord damage in patients with benign MS (BMS) and secondary progressive MS (SPMS), in order to achieve a better understanding of the mechanisms underlying the development of irreversible disability in MS.

Methods: Conventional and DT MRI scans of the cervical cord and brain were acquired from 40 BMS patients, 28 SPMS patients and 18 healthy individuals. Cervical cord and brain mean diffusivity (MD) and fractional anisotropy (FA) maps were created and average MD and FA were calculated. Cross sectional cord area (CSA) was also computed.

Results: 37 (92%) BMS patients and all (100%) SPMS patients had macroscopic cervical cord lesions. Compared with healthy individuals, BMS patients had higher average cord MD while SPMS patients had higher average cord MD, lower average cord FA and lower average CSA. Compared with BMS patients, SPMS patients had lower cord average FA and lower average CSA. In MS patients, Expanded Disability Status Scale (EDSS) was correlated with CSA (r = −0.47, p<0.0001), average cord FA (r = −0.37, p = 0.002) and brain T2 lesion volume (LV) (r = 0.34, p = 0.005). A multivariate regression model identified CSA, average cord FA and brain T2 LV as variables independently influencing the EDSS score (r = 0.58, p<0.0001).

Conclusions: Cervical cord damage outside focal macroscopic lesions is limited in patients with BMS. The assessment of cord and brain pathology provides complementary information to improve the understanding of disability accumulation in MS.

Statistics from

Spinal cord damage is likely to be one of the major determinants of disability in patients with multiple sclerosis (MS).1 2 3 4 However, despite the presence of macroscopic lesions in the spinal cord, some patients show a particularly favourable course of the disease, with mild or no accumulation of disability several years after the clinical onset. This condition, termed benign MS (BMS), has an estimated prevalence in cross sectional surveys that ranges from 6% to 64%, depending on the criteria used to define it.5 6 7 At present, the most accepted consensus definition of BMS is that of “a disease in which the patient remains fully functional in all neurological systems 15 years after disease onset”.8

During the past decade, due to the paucity of correlation between conventional MRI measures and disability,9 quantitative MR based techniques have been increasingly applied for the assessment of CNS involvement in patients with MS. Among these techniques, diffusion tensor (DT) MRI has been demonstrated to be able to grade the severity of brain and cervical cord pathology in MS and other neurological conditions.10 11 12 13 14 DT MRI enables the random diffusional motion of water molecules to be measured, and provides quantitative indices of the structural and orientational features of tissues.15 Such indices include mean diffusivity (MD), which is a directionally averaged measure of the apparent diffusion coefficient, and fractional anisotropy (FA), which summarises the orientational dependence of diffusivity. The pathological elements of MS have the potential to alter the permeability and geometry of structural barriers to water molecular diffusion in the CNS. While a few studies have applied DT MRI for the quantification of the extent and distribution of microscopic disease related damage in the brain in patients with BMS,16 17 18 19 no study has been performed to assess the extent of cervical cord involvement in these patients. Based on the criteria used for the definition of such a disease phenotype, it is tempting to speculate that the relative structural integrity of the cervical spinal cord might be among the factors contributing to limit accumulation of disability over long periods of time. To test this hypothesis, we applied DT MRI to investigate the pattern of cervical cord damage in a relatively large sample of patients with BMS. As a control group, we selected patients with secondary progressive (SP) MS with, by definition, a higher Expanded Disability Status Scale (EDSS) score, in whom the extent of spinal cord involvement is likely to have an important role in the accumulation of irreversible disability.

Patients and methods


Forty consecutive patients with BMS and 28 patients with SPMS were studied. Criteria for a diagnosis of BMS were an EDSS score ⩽3.0 and a disease duration ⩾15 years.8 Patients with MS, an SP course8 and persistent walking limitations (ie, a confirmed EDSS score of 4.0 or higher), were enrolled to serve as controls. To be included, patients had to be relapse and steroid free for at least 3 months before MRI acquisition. All patients underwent a complete neurological examination, with rating of the EDSS score20 within 24 h of the MRI scanning session. The main clinical and demographic characteristics of the study groups are reported in table 1. Age did not differ between the three groups of subjects. Disease duration was longer in BMS compared with SPMS patients (p = 0.002) and the median EDSS score was lower in BMS compared with SPMS patients (p<0.001).

Table 1

Main clinical and demographic characteristic of the study groups

Twenty-five BMS patients were not receiving any disease modifying treatment, eight were being treated with β interferon, six with glatiramer acetate and one with azathioprine. Eight SPMS patients were not receiving any disease modifying treatment, 14 were being treated with β interferon and six with azathioprine. Eighteen healthy individuals with no previous history of neurological, major medical or psychiatric disorders, and no alcohol or drug abuse and a normal neurological examination were also studied (table 1). All subjects signed a written informed consent prior to study entry and the study design was approved by the local ethics committee.

Image acquisition

Using a 1.5 T system, the following pulse sequences were acquired from the cervical cord of all subjects, using a tailored cervical spine phased array coil for signal reception: (1) fast short tau inversion recovery (fast-STIR) (TR = 2288, TE = 60, TI = 110, echo train length = 11, field of view (FOV) = 280×280 mm, matrix size = 264×512, number of signal averages = 4; number of slices = 8; sagittal plane; slice thickness = 3 mm; inter-slice gap = 0.3 mm); (2) pulsed gradient, DT sensitivity encoded (SENSE) single shot echo planar imaging (reduction factor, R = 2; TR = 7000, TE = 100; FOV = 240×90 mm; matrix = 128×48; echo train = 40 ms; number of slices = 5; sagittal plane; slice thickness = 4 mm). This sequence collects 16 images per section, including two images with no diffusion weighting (b = 0 s/mm2) and 14 images with the same b factor of 900 s/mm2, but with gradients applied in different directions.10 11 12 13 14 The non-diffusion weighted images are needed to compute the DT, and the gradient orientations were chosen according to the algorithm proposed by Jones and colleagues,21 which was designed to optimise DT MRI acquisition. The measurement was repeated four times to improve the signal to noise ratio. Three saturation bands were used, positioned in the anterior part of the neck and transversely at the edges of the FOV in the superior–inferior direction. A detailed description of this sequence is given elsewhere10 11 12 13 14; (3) sagittal T1 weighted three-dimensional magnetisation prepared rapid acquisition gradient echo (MP-RAGE) (TR = 9.7, TE = 4, flip angle = 12°, slab thickness = 160 mm, FOV = 280×280 mm, number of partitions = 128).

In the same scanning session, the following pulse sequences were used to image the brain: (1) dual echo turbo spin echo (TR = 3300, TE = 16–98, echo train length = 5) and (2) pulsed gradient DT echo planar imaging (inter-echo spacing = 0.8, TE = 123), with DT gradients applied in eight different directions.14 As for cervical cord imaging, only two b factors were used: b1 = 0 and b2 = 1044 s/mm2. For the turbo spin echo, 24 contiguous interleaved axial slices were acquired with 5 mm slice thickness, 256×256 matrix and 250×250 mm FOV. The slices were positioned to run parallel to a line that joins the most infero-anterior and infero-posterior parts of the corpus callosum. For the DT scans, 10 axial slices with 5 mm slice thickness, 128×128 matrix and 250×250 mm FOV were acquired, with the same orientation as the dual echo scans, and with the second last caudal slice positioned to match exactly the central slices of the dual echo set. This brain portion was chosen as these central slices are less affected by the distortions due to B0 field inhomogeneity which can affect image co-registration and because they cover the brain regions most frequently involved in MS pathology.

Image analysis

Cervical cord hyperintense lesions were identified on the fast STIR scans. After offline reconstruction of cervical cord DT MRI data, cervical cord MD and FA maps were derived and MD and FA histograms produced, as previously described.10 11 12 13 14 For each histogram, only the average MD and FA were a priori chosen to enter the analysis, to minimise the number of comparisons and hence reduce the risk of type I errors.

Reformatting of the original MP-RAGE data was performed using the standard, vendor supplied multiplanar reformatting software available on the operator’s console of the scanner. For each subject, a set of five contiguous, 3 mm thick axial slices (perpendicular to the spinal cord) was reconstructed using the centre of the C2–C3 disc as the caudal landmark. Then, a semiautomated technique developed by Losseff and colleagues22 was used to segment the cord tissue and to measure the cross sectional cord area (CSA) at the level of each slice. Values from the five slices were averaged to obtain a single value for each subject, as previously described.10 14

Brain T2 lesion volume (LV) was measured using a computer assisted technique, as previously described.10 11 12 13 14 Brain MD and FA maps were derived for every pixel, as for the cord, and MD and FA histograms of the normal appearing brain tissue (NABT) were produced, after masking of T2 visible lesions, as previously reported.10 14 As for the cord, only average MD and FA entered the analysis.

Statistical analysis

An analysis of variance model was used to compare conventional and DT MRI derived measures from healthy individuals and from BMS and SPMS patients. Univariate correlations were assessed using the Spearman Rank Correlation Coefficient. To address group membership (BMS vs SPMS), logistic regression analysis was run. A multivariate linear regression model was used to identify MRI variables independently correlated with EDSS. To this end, all of the MRI variables significantly different between groups were included in the model. The variables retained in the final model were correlated with the EDSS using the Spearman Rank Correlation Coefficient. Considering that the definition of BMS is also based on disease duration, a multivariate regression model was also used to identify MRI variables independently correlated with the rate of disability accrual defined as EDSS/disease duration. All statistical analyses were performed using SPSS software.


In table 2, cervical cord MRI derived metrics from patients and healthy volunteers are reported. No macroscopic abnormalities were seen on the cervical cord MR images from healthy individuals. Cord lesions were identified on fast STIR images in 37 (92%) BMS patients (median number of lesions = 2; range 1–6) and in all (100%) SPMS patients (median number of lesions = 2; range 1–5). Compared with healthy individuals, BMS patients had higher average cord MD (p = 0.009), and SPMS patients higher average cord MD, lower average cord FA and lower average CSA (p<0.0001 for all comparisons). Finally, compared with BMS patients, SPMS patients had lower cord average FA (p = 0.01) and lower average CSA (p<0.0001).

Table 2

Cervical cord and brain conventional and DT MRI derived metrics from the study groups

No macroscopic T2 visible lesions were identified on brain T2 weighted scans from healthy individuals. Brain T2 LV was significantly higher in SPMS than in BMS patients. MD and FA metrics of the NABT were significantly heterogeneous between the three study groups (table 2). At post hoc analysis, NABT average MD was significantly higher in SPMS patients in comparison with both healthy individuals (p = 0.002) and BMS patients (p = 0.04). Similarly, NABT average FA was lower in SPMS patients in comparison with healthy individuals (p<0.0001) and BMS patients (p = 0.02). BMS patients had lower NABT average FA in comparison with healthy individuals (p<0.0001).

Logistic regression analysis identified CSA (p = 0.007, odds ratio = 0.91) and T2 LV (p = 0.01, odds ratio = 1.05) as variables associated with the risk of having SPMS compared with BMS.

In MS patients, EDSS was correlated with cord CSA (r = −0.47, p<0.0001), average cord FA (r = −0.37, p = 0.002) and brain T2 LV (r = 0.34, p = 0.005) (fig 1). No correlation was found between the number of cervical cord lesions and EDSS. Cord CSA was correlated with brain T2 LV (r = −0.32, p = 0.007), NABT average FA (r = 0.27, p = 0.03) and NABT average MD (r = −0.29, p = 0.02).

Figure 1

Scatter plots showing the correlation between Expanded Disability Status Scale (EDSS) scores and cervical cord cross sectional area (CSA) (A), cord average fractional anisotropy (FA) (B) and brain T2 lesion volume (LV) (C) in patients with benign (triangles) and secondary progressive (circles) multiple sclerosis. See text for further details and statistical analysis.

A multivariate regression model with EDSS as the dependent variable was run, including cord CSA, cord average MD, cord average FA, brain T2 LV, NABT average MD and NABT average FA as independent variables. The final model retained CSA (p = 0.001, beta = −0.38, standard error (SE) = 0.02, partial correlation = −0.39), average cord FA (p = 0.05, beta = −0.21, SE = 0.004, partial correlation = −0.23) and brain T2 LV (p = 0.04, beta = 0.21, SE = 0.01, partial correlation = 0.24) as variables independently influencing the EDSS score. The r value of the correlation between the EDSS score and the composite score obtained combining these three variables was 0.58 (p<0.0001). When the multivariate regression model was run with EDSS/disease duration as the dependent variable, only CSA (p = 0.03, beta = −0.25, SE = 0.003) was retained in the final model.


MS is a highly variable disease in its clinical presentation, course and prognosis. While some patients remain clinically stable for relatively long periods of time, others tend to accumulate irreversible disability. Although clinical and epidemiological studies have defined several factors which might contribute towards identifying patients with a more favourable disease evolution, including gender, age at onset, time to second relapse, type and duration of initial symptom,6 23 the definition of BMS patients still remains a challenge. Similarly, MRI studies tailored towards the assessment of the extent of focal16 17 18 19 24 25 and diffuse2 16 17 24 26 brain damage in these patients have yielded conflicting results. Considering that EDSS, the most common scale used for grading disability in MS, is heavily weighted towards locomotor disability, we argued that the relative sparing of the cervical cord in BMS is likely to be a factor explaining the favourable clinical status of these patients.

In line with previous studies, which showed only a mild correlation between the extent of cord MRI abnormalities and clinical disability in MS patients,26 27 quantification of cervical cord macroscopic lesions did not demonstrate a significant difference between SPMS and BMS patients. Such a finding might be related to the fact that this measure is only semiquantitative. However, a quantitative assessment of lesion load in the cervical cord is usually not performed because of methodological limitations which include contaminations from partial volume effects and movement artefacts. To achieve a comprehensive estimate of cervical cord damage in our patients, we did not limit the analysis to the assessment of focal lesions, but used additional MRI techniques, which have been proven to provide a measure of irreversible tissue loss (ie, atrophy) and “diffuse” tissue abnormalities (ie, damage possibly associated with axon and myelin loss).28 29 30 Post mortem studies have demonstrated that demyelination and axonal injury occur not only inside T2 visible cord lesions but also in the normal appearing cord tissues in patients with MS.28 29 30 Demyelination and axonal loss might lead to increased extracellular spaces between surviving fibres which can harbour cell debris, inflammatory infiltrates, fibrillary gliosis and astrocytic proliferation.31 Moreover, similarly to what occurs in the brain, the loss of neurons secondary to lesions in the grey matter or to axonal transection in the white matter can also occur in the cord.32 All of these pathological changes have the potential to alter the diffusivity and volumetric characteristics of the cervical cord.

In line with the results of two seminal studies,26 33 the analysis of cervical cord atrophy demonstrated significant loss of cervical cord area in SPMS patients, but not in BMS patients. For DT MRI, we used a sequence tailored for imaging the cervical cord which has already been applied in patients with different neurological conditions, including MS,10 11 14 neuromyelitis optica12 and neuropsychiatric systemic lupus erythematosus.13 Analysis of DT MRI characteristics of the individuals enrolled in this study revealed some interesting differences between BMS and SPMS patients. The two diseased groups showed a significant increase in cord MD values in comparison with healthy individuals, without significant differences between them. Remarkably, SPMS patients also had a significant decrease in cord average FA values in comparison with both healthy individuals and BMS patients. Admittedly, we studied a cohort of SPMS patients with a disease duration shorter than that of BMS cases. However, such a difference is expected to reduce the magnitude of MRI abnormalities we found in the SPMS group. We can also reasonably exclude the fact that the differences in diffusivity of the cord are secondary to partial volume effects from the surrounding CSF. Nevertheless, contamination from the CSF was minimised by considering in the analysis only pixels located in the central slice of the sagittal slab and by including only those from C1 to C5. This was at the price of restricting the portion of the cervical cord studied but we believe this made our observations more robust and reliable. As a consequence, these results indicate that the cervical cord in BMS patients without locomotor disability is relatively spared by the most destructive components of MS pathology outside focal macroscopic lesions and, hence, suggest that this might contribute towards explaining the favourable clinical status of these patients. Such a finding is not unexpected, considering that the clinical definition of BMS, which is based on the EDSS, assumes implicitly that these patients would have relative spinal cord sparing. The discrepancy between FA and MD measures in BMS patients also suggests a relative preservation of the orientation of the major white matter fibre bundles in the cervical cord, despite the reduction of structural barriers to water molecular motions. Admittedly, we limited out analysis to the cervical spinal cord, which is the most commonly affected by the disease and, due to technical challenges, we did not investigate if similar differences are also found in the thoracic and lumbar segments. Furthermore, we used as a control group patients with SPMS, who had a similar disease duration and higher disability compared with those with BMS. An additional helpful control group would have been RRMS patients. Future studies are now warranted to assess whether spinal cord diffusivity is different between patients with different disease durations but similar levels of disability.

Another important issue to be addressed is how much of the cord pathology of patients with MS is due to local damage and how much is due to retrograde and anterograde degeneration of neurons secondary to injury of fibres passing through T2 visible lesions of the brain. In line with previous pathological29 and MRI10 11 14 27 studies, we found only a relatively modest correlation between cord CSA and brain T2 lesion load and NABT damage, suggesting that distant pathology in the brain results in only a marginal contribution to the observed cord MRI changes.

As our main objective was to define the factors associated with the accumulation of irreversible disability in long lasting MS, we analysed the contribution of cord and brain measures of macroscopic and microscopic tissue abnormalities separately, in determining EDSS scores in the entire sample of patients. The univariate analysis of correlation showed that EDSS was significantly correlated with cord CSA, cord average FA and brain T2 lesion volume. Several studies have convincingly demonstrated that cord area has a role in predicting the accumulation of disability in patients with MS.22 27 33 In this regard, our study provides an additional piece of information by showing that the assessment of intrinsic cord pathology is a useful tool to further improve the understanding of the mechanisms responsible for the development of “fixed” disability in this disease. It is worth noting that, contrary to previous studies,11 14 brain T2 lesion volume was also significantly correlated with EDSS in our sample. Such a finding is in agreement with the results of a seminal study34 in patients with progressive MS, which demonstrated that brain T2 lesion burden can contribute to differentiate patients with and without locomotor disability (defined, as in the present study, by an EDSS ⩾4.0).

All of the previously reported MRI variables were retained in the final multivariate model as independent factors influencing the disability status. This suggests that brain and cord pathology accumulates, at least partially, independently of each other, and both may contribute to the development of disability. In addition, it seems that different factors in these two CNS compartments are likely to play a role in disability accumulation over relatively long periods of time since we found that in the brain, T2 lesion burden rather than microscopic NABT changes may be relevant, whereas in the cord, “diffuse” abnormalities and tissue loss more than focal lesions are likely to be the central aspects of the disease. Assessing a potential treatment effect on the observed findings was beyond the scope of this study. In addition, the cohort of subjects studied was too small and the treatment regimen too heterogeneous to address this issue.

The finding of an independent contribution of brain and cord damage to disability status in MS confirms previous reports,15 34 and points to the use of multicomposite MRI scores as potential useful measures of outcome in studies aimed at monitoring disease evolution. Admittedly, since this is a cross sectional study, caution must be exercised before drawing firm conclusions on MS dynamics. Longitudinal studies are indeed warranted to confirm or reject our speculation.



  • See Editorial Commentary, p 4

  • Funding This study was supported by a grant from FISM (2005/R/18).

  • Competing interests None.

  • Provenance and Peer review Not commissioned; externally peer reviewed.

  • Ethics approval The ethics committees of each of the centres involved approved the study.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles