Objective To examine the temporal evolution of spinal cord (SC) atrophy in multiple sclerosis (MS), and its association with clinical progression in a large MS cohort.
Methods A total of 352 patients from two centres with MS (relapsing remitting MS (RRMS): 256, secondary progressive MS (SPMS): 73, primary progressive MS (PPMS): 23) were included. Clinical and MRI parameters were obtained at baseline, after 12 months and 24 months of follow-up. In addition to conventional brain and SC MRI parameters, the annualised percentage brain volume change and the annualised percentage upper cervical cord cross-sectional area change (aUCCA) were quantified. Main outcome measure was disease progression, defined by expanded disability status scale increase after 24 months.
Results UCCA was lower in SPMS and PPMS compared with RRMS for all time points. aUCCA over 24 months was highest in patients with SPMS (−2.2% per year) and was significantly higher in patients with disease progression (−2.3% per year) than in stable patients (−1.2% per year; p=0.003), while annualised percentage brain volume change did not differ between subtypes (RRMS: −0.42% per year; SPMS −0.6% per year; PPMS: −0.46% per year) nor between progressive and stable patients (p=0.055). Baseline UCCA and aUCCA over 24 months were found to be relevant contributors of expanded disability status scale at month-24, while baseline UCCA as well as number of SC segments involved by lesions at baseline but not aUCCA were relevant contributors of disease progression.
Conclusions SC MRI parameters including baseline UCCA and SC lesions were significant MRI predictors of disease progression. Progressive 24-month upper SC atrophy occurred in all MS subtypes, and was faster in patients exhibiting disease progression at month-24.
- Multiple Sclerosis
Statistics from Altmetric.com
MRI derived brain atrophy quantification has become an established predictor of neurological impairment in multiple sclerosis (MS).1 Central nervous system atrophy in patients with MS, assumed to be driven by neuraxonal loss,2 is however not limited to the brain but also involves the spinal cord (SC). Several MRI studies have investigated SC atrophy in MS.3–16
While correlations between SC atrophy and concurrent disability status have been observed cross-sectionally,9 ,10 ,13 ,16 ,17 it is currently unclear whether SC atrophy evolution is related to ongoing clinical worsening. Conflicting results on associations between SC area changes and worsening disability have been described in former studies,6 ,7 ,9 ,15 ,18 probably due to the relatively small disease subtype groups and/or short follow-up periods.6 ,14 ,18 ,19
Two recent cross-sectional studies have demonstrated that SC atrophy as well as focal and diffuse SC lesions are independent predictors of clinical outcomes,13 ,16 but large longitudinal studies focusing on the clinical relevance of changes in SC atrophy in MS are still missing.
In this 2-year follow-up study we determined SC atrophy rates in a large MS cohort and assessed the independent clinical impact of SC atrophy in relation to other brain and SC measures of the disease.
In total 352 patients were included from two MS centres participating in the GeneMSA consortium: the VU University Medical Center in Amsterdam and the University Hospital in Basel. Patients were diagnosed with MS according to established criteria,20 ,21 and classified either as relapsing remitting (RR), secondary progressive (SP), or primary progressive (PP) MS.22 Disability was scored using the expanded disability status scale (EDSS) score.23 Patients were classified according to clinical disease progression, without further dividing according to whether progression was sustained or not. Progression was defined by an increase of 1 point in EDSS score after 2 years of follow-up if the entry EDSS score was ≤5.5. In case of an entry EDSS score >5.5, an increase of 0.5 point in EDSS score after 2 years was scored as progression. For all patients, use of disease modifying therapy (DMT) before start of the study and during follow-up was recorded. At MRI, all patients had been relapse-free and steroid-free for at least 1 month. The local ethics committee approved the study. All participants gave written informed consent.
Imaging was performed on two 1.5T MRI scanners (Amsterdam: Siemens Vision using the standard circularly polarised head-coil; Basel: Siemens Avanto with a 12-element head matrix coil) using a standardised protocol at baseline and during follow-up visits after 12 months and 24 months. Neither scanner was changed during the follow-up period, nor were any major upgrades performed. For brain volume measurement, three-dimensional (3D) T1 weighted images were acquired (repetition time [TR]: 9.7–20.8 ms; echo time [TE]: 2–4 ms; inversion time [TI]: 300–400 ms), consisting of 1.0 mm-thick slices and a 1.0×1.0 mm2 inplane resolution. Additionally, dual echo proton density (PD)-T2-weighted images (TR: 2000–4000 ms; TE: 14–20/80–108 ms), with interleaved axial 3.0 mm-thick slices and an inplane resolution of 1.0×1.0 mm2 and postcontrast T1-weighted spin-echo images (TR: 467–650 ms; TE: 8–17 ms; axial 3.0 mm-thick slices with an inplane resolution of 1.0×1.0 mm2) were obtained. Upper SC atrophy was measured on each available 3D-T1 data set of the brain on which the upper cervical cord region was sufficiently visible.
SC scanning included a cardiac-triggered sagittal PD and T2-weighted dual-echo spin echo sequence with a slice-thickness of 3.0 mm covering the whole SC (TR: 2500–3000 ms; TE1: 20–30; TE2: 80–100 ms), with a gap between slices of 0.3 mm and an inplane resolution of 1×1 mm. The total scanning time was 30 min.
Brain volume analyses were performed at the Image Analysis Center in Amsterdam, using SIENAX (part of FSL) to obtain head-size normalised total brain volume (NBV), normalised grey matter volume (NGMV) and normalised white matter volume (NWMV) at baseline.24 Percentage brain volume change (PBVC) over each interval was calculated using the fully automated method SIENA (part of FSL).25 ,26 Segmentation and registration results of all subjects were visually inspected for errors, and only those who passed quality control were used for further analysis.
Marking and measurement of focal brain lesions was performed in the University Hospital in Basel, using commercial semiautomatic software (AMIRA V.3.1.1; Mercury Computer Systems). For brain images, T2-hyperintense lesions were outlined on the PD images, T1-hypointense lesions were outlined on T1 weighted spin echo images, and lesion volumes for T2 lesions (T2LV) and T1-hypointense lesions (T1LV) were calculated. The number of enhancing brain lesions on postcontrast T1-weighted spin-echo images was scored.
For SC images, the number of focal lesions as well as the number of involved segments and the presence of diffuse abnormalities were scored. Diffuse abnormalities were defined as poorly delineated areas with increased signal intensity compared with spinal cerebrospinal fluid (CSF) on PD-weighted images (figure 1).27
For each time point, upper cervical cord cross-sectional area (UCCA) was measured between the upper border of C2 and the intervertebral disc between C2 and C3 as previously described,16 using a semiautomated volumetry method (NeuroQLab, Fraunhofer-Mevis, Germany), from the high-resolution 3D-T1 brain images, if the upper cervical cord region was sufficiently covered and of sufficient image quality in that region. Measurements were performed in a randomised order by two authors (CL and BB), both blinded to the date of the MRI and to consecutive results of previous measurements.
The percentage change of UCCA between time points was calculated. To account for small variation of follow-up duration, PBVC and UCCA per cent change were divided by the time interval in years, yielding the annualised measures denoted as aPBVC and aUCCA.
All analyses were performed using SPSS V.18 (SPSS, Chicago, USA).
Comparisons of the demographical, clinical and MRI data between the disease types, progressive and non-progressive patients and among centres were made using the Mann-Whitney U test or Pearson χ2 test. We performed univariate and multivariate linear regression analyses with annualised SC atrophy rate over 24 months (24-month aUCCA) as dependent variable, statistically adjusted for centre through a two-level fixed factor. Independent variables were gender, MS subtype, age, disease duration, use of DMT, baseline EDSS score; baseline NGMV, NWMV, NBV; T1LV, T2LV and presence of enhancing lesions at baseline; number of SC lesions and affected segments; presence of diffuse abnormalities; baseline UCCA; 24-month change of T1LV, T2LV and brain volume.
To improve normal distribution, a natural log transformation was applied to lesion metrics, after adding the number 1 to the number of SC segments and number of SC lesions. In the multivariate linear regression of 24-month aUCCA, all independent variables identified as relevant in the univariate analysis were entered and then removed one by one using backward stepwise selection until all remaining variables had p<0.1. We also conducted univariate and multivariate analysis with month-24 EDSS score as the dependent variable. Finally, binary logistic regression was performed with the absence or presence of disease progression at 24 months as dependent variable. In the multivariate binary logistic regression a combined clinical-MRI model was analysed including relevant clinical and MRI variables using stepwise backward selection. All regression analyses were statistically adjusted for centre.
Differences between participating centres
Patients from Amsterdam had a slightly lower fraction of RRMS in conjunction with higher baseline EDSS scores and progression, although the median disease duration was shorter. Mean UCCA did not differ between centres, but Amsterdam patients exhibited higher annualised SC atrophy rates. Additionally, more diffuse SC abnormalities were found in Amsterdam patients. Detailed information about centre differences is provided in table 1.
Differences between disease types
Table 2 summarises clinical and MRI parameters by disease subtype. At each time point, UCCA was lower in the progressive subtypes (SPMS and PPMS) than in RRMS, but SPMS and PPMS never differed from each other. All disease types showed a reduction of UCCA over 24 months (figure 2). The annualised 24 month SC atrophy rate was significantly higher in SPMS than in RRMS (p=0.019), but did not distinguish PPMS from either RRMS or SPMS. Among brain atrophy measures, NGMV differed most between subtypes, being lowest for patients with SPMS. However, 24-month brain atrophy rates did not differ between subtypes.
Correlations between brain and SC MRI measures
Correlations between brain and SC MRI measures are summarised in table e-1 (web only). For the whole group, correlations between baseline brain and SC measures are in agreement with our previous published cross-sectional study. On follow-up, brain T1LV change was weakly associated with the SC atrophy rate over 24 months, mainly in patients with SPMS, while change in brain T2LV was not associated with any SC measure. Brain atrophy rate was weakly associated with the SC atrophy rate which was dominated by patients with RRMS.
Relationship between SC atrophy rate and other MRI measures
In the univariate linear regression analyses, 24-month aUCCA was associated positively with disease subtype, 24-month aPBVC and NGMV; negative associations were detected between 24-month aUCCA and disease duration, number of SC segments involved by SC lesions, T1LV change and UCCA. No other MRI parameters were significantly associated (table 3). Backward stepwise multivariate regression analysis yielded a combined model of clinical variables: disease type and disease duration, and following MRI variables (coefficient b / p value): baseline UCCA (−0.06/0.001), number of affected SC segments (−0.49/0.01), presence of diffuse SC abnormalities (−0.96/0.02) and change in brain T1LV (−1.4/0.04) (table 3). Among these, baseline UCCA was found to be most significantly associated with aUCCA. The model had low R2:0.14.
MR predictors of EDSS after 24 months and disease progression
Univariate and multivariate results for EDSS at month-24, are summarised in table e-2 (web only). In the multivariate model statistically significant MRI parameters were (coefficient b/p value): NGMV (−0.003/0.002), brain T2LV change (0.37/0.06), number of affected SC segments (0.24/0.001), presence of diffuse SC abnormalities (0.32/0.036), baseline UCCA (−0.01/0.047) and 24-month SC atrophy rate (−0.06/0.02). Among these, NGMV and number of affected SC segments were the most significant, with a R2-value of 0.83 for the final model.
When comparing patients according to disease progression after 24 months, SC parameters, including baseline UCCA, SC atrophy rate and SC lesions differed significantly between both groups (table 4). Patients with disease progression had lower UCCA at all time points (all: p≤0.005) and higher SC atrophy rates (all: p<0.01). Especially SC atrophy rate (p=0.003) and not brain atrophy rate over 24 months (p=0.055) was significantly higher in patients with disease progression.
In the binary logistic regression with disease progression over 24 months as dependent variable, significant univariate associations were found for disease type, baseline EDSS, number of SC lesions, number of affected SC segments, baseline UCCA, 24-month aUCCA, baseline NGMV and baseline NBV. In the multivariate binary logistic regression, significant associations were found for baseline UCCA, number of affected SC segments, presence of enhancing brain lesions at baseline, T1LV change and 24-month aPBVC (table 5).
Among these MRI variables the number of segments involved by SC lesions was found to be most significantly associated with disease progression. Nagelkerke's R2 was 0.297.
The present study investigated cervical SC atrophy rates in a large cohort of patients with MS. Faster atrophy rates were observed for patients with SPMS compared with patients with RRMS. SC atrophy rates were independently related to baseline cervical cord area, the number of SC segments affected by lesions and presence of diffuse SC abnormalities. Finally, although SC atrophy rate was significantly faster in patients exhibiting disease progression and univariately associated with clinical disease progression over 2 years, it was not independently predictive over and above the effects of baseline cord area and number of affected SC segments.
Like brain atrophy, cervical cord atrophy has been suggested as a potentially valuable surrogate marker of irreversible tissue damage in MS. However, while brain atrophy measures have emerged as a secondary end point in many MS trials this has not yet been established so far for SC atrophy.28 An important reason is that the development and relevance of SC atrophy in relation to clinical disability in MS is still unclear, due to the lack of sufficiently large longitudinal studies.
Our present longitudinal study involving a large number of patients confirms previous findings that a reduction of UCCA over time occurs in all MS subtypes.6 ,12 ,18 ,19 The median cervical atrophy rate in our total MS cohort was 1.5% per year, in line with previous findings.6 ,8 In our study patients with RRMS exhibited median annual atrophy rates of up to 1.7%, whereas another study found smaller SC atrophy rates around 1% per year.12 Fastest atrophy rates were found in SPMS and PPMS, with median rates up to 2.2% per year. Other authors found slightly slower rates for patients with SPMS (1.6% per year),6 and faster rates (about 3.75% per year) for patients with PPMS.18 Such differences in atrophy rates between studies may be related to imaging and analysis differences, or patient selection differences, and the relatively small number of patients with PPMS in our current study warrants some caution.
In line with several previous studies,11 ,29 ,30 our results indicate that SC atrophy develops mostly independently from brain pathology. Rather than brain pathology, the most significant independent predictors of cervical cord atrophy rates were SC parameters such as the number of SC segments involved by lesions and the presence of diffuse SC abnormalities.
We previously reported results of cross-sectional analyses on the same cohort of patients, in which we found that these same SC parameters also best explained the cervical cord area measured at a single time point.16 The importance of these measures of SC tissue damage is confirmed here by our results that underline their close relation to the rate of change of UCCA.
Surprisingly, higher UCCA at baseline in our study were associated with faster SC atrophy rates, whereas one would expect the opposite association. It could be speculated that the rate of SC tissue loss could slow down over time, with already highly atrophied structures exhibiting slower atrophy rates. Future, more detailed studies may allow a better interpretation of these findings.
The best distinction between patients who suffered disease progression and patients who did not, was provided by cervical cord results including the annualised SC atrophy rate over 24 months and the number of spinal T2 lesions, respectively, the number of cord segments with T2 lesions, rather than brain atrophy or other brain MRI parameters. Previous longitudinal studies of cord atrophy have shown conflicting results regarding the relation of SC atrophy to clinical progression. Furby et al6 followed patients with SPMS over a time period of 2 years. They found that grey matter and SC atrophy explained multiple sclerosis functional composite (MSFC) worsening over time, but no MRI measures differed between stable and progressing patients defined according to EDSS. By contrast those with worsening disease according to the MSFC criteria had a significantly greater rate of whole brain and SC atrophy (−2.77 vs −1.30% per year, p=0.02) compared with those with stable disease. Given our results of the total patient group, similar differences with almost doubled annual atrophy rates in patients with worsening disease were found when using EDSS progression as an end point. In patients with PPMS no correlation between SC atrophy and clinical disability could be established in a cross-sectional,11 as well as in longitudinal studies over 1-year follow-up and 5 year follow-up,18 ,19 although another 5 year study on PPMS showed a weak correlation between increase in EDSS and decrease in cord area.31
Finally, conflicting reports also exist for patients with RRMS. Over a follow-up of 48 months a correlation between EDSS change and UCCA change was found in a small group of patients with RRMS and SPMS.9 However a contribution of baseline SC atrophy was noted only for SPMS and no statistically significant difference between UCCA change rates between SPMS and RRMS was found. By contrast, other studies including patients with RRMS did not observe a correlation between UCCA change and clinical progression,12 ,15 ,18 although a reduction of UCCA in early RRMS over time was noted.12 Although we were not able to demonstrate that the SC atrophy rate was a relevant independent contributor to disease progression, baseline UCCA and the number of SC segments affected by T2-lesions could be identified among the most relevant MRI parameters for disability in the short term. The number of cord segments affected by SC lesions represents a simplified measure of the cervical cord lesion volume. Thus, the number of affected cord segments probably has a stronger association with clinical disability than the plain number of SC lesions, which was reflected by the results of our binary regression of disability progression over 24 months where the number of SC segments was predictive while the number of cord lesions was not. As the EDSS used as a measure for clinical disability is strongly weighted towards locomotion it is plausible, that clinical progression is more related to SC lesions and/or SC atrophy rather than brain pathology. In line with previous reports on the relevance of SC pathology for disability in MS the results of our longitudinal study therefore highlight the importance of SC imaging in MS.13 ,16 ,32
Treatment with DMT in our study did not differ between patients who progressed and those who did not and treatment was not a relevant independent predictor of SC atrophy rates. Similarly a previous study did not find significantly different SC atrophy rates in the placebo arm compared with patients treated with interferon beta-1a over a 12 month follow-up.9 However, our results have to be treated with caution. Although treatment did not differ between centres, treatment allocation was not random and we did not further differentiate groups according to the different treatments or treatment duration. Longer prospective studies are warranted to provide more definitive answers regarding possible treatment impact on SC pathology.
Some limitations of our study have to be mentioned. The current study aimed to analyse follow-up data of a group of patients with MS that was previously studied.16 Due to loss of patients during follow-up and incomplete data sets the number of patients differs between the current and previous patient cohort, so we cannot exclude a potential bias related to the incomplete follow-up. Furthermore the lack of a representative group of healthy controls prohibited a distinction between MS-related SC atrophy and any UCCA volume decrease due to healthy aging. However the observed decrease of UCCA over time in RRMS can be interpreted mainly as a disease related effect, as in the normal population significant SC atrophy occurs late, normally after the fifth decade and the observed atrophy rates are in line with previous reports.33 Oedematous cord swelling which might alter SC atrophy quantification was not considered in our analyses since contrast enhanced images of the SC were not acquired. Methodological issues such as potential role of different scanners and the herein used method for UCCA quantification could have influenced the results as well.34 However, none of the scanners were changed during the follow-up period and no major upgrades have been performed at both sites. Statistical analyses were controlled for the centre to exclude a possible scanner-related bias.
Methodologically, SC atrophy rates were derived from two separated segmentations. Registration-based methods such as those used here to measure brain atrophy might lead to a more precise estimation of longitudinal SC atrophy, however such a technique was not available at present and in general no standardised technique is yet available to estimate SC atrophy. Several approaches for SC atrophy quantification in MS have been published making a direct comparison between other results difficult.35–39 Most of the above-mentioned studies normalise SC volumes or mean cross-sectional areas, although normalisation showed only limited gain in reliability.38 Furthermore raw data seems to have the best reliability and the most sensitivity in discriminating between patients with MS and controls.37
In conclusion, the results of our longitudinal study including a high number of patients with MS highlight the importance of MRI detected SC pathology for disease progression in MS. Especially our proposed approach to derive SC atrophy from 3D brain images has been proven feasible in a longitudinal setting, thus eliminating the need for an additional volumetric SC scan which has important implications for clinical studies including treatment trials. Nevertheless, our results highlight that assessment of SC lesion is of important relevance.
Further longitudinal studies aiming at the assessment of SC atrophy rates over a longer period seem to be helpful to establish SC atrophy as a relevant end point in MS trials.
Contributors CL: Study concept and design, drafting and revising of manuscript, acquisition of data, statistical analysis, analysis and interpretation of data, final approval of the version to be published. DLK: Analysis and interpretation of data, statistical analysis, critical revision of the manuscript for important intellectual content, final approval of the version to be published. MHS: Study concept and design, drafting of manuscript, statistical analysis, analysis and interpretation of data, final approval of the version to be published. BB: acquisition of data, statistical analysis, analysis and interpretation of data, drafting/revising the manuscript for content, final approval of the version to be published. HKH: interpretation of data for the work, revising the manuscript for content, contribution of vital tools, final approval of the version to be published. VP: Acquisition of data, drafting/revising the manuscript for content, final approval of the version to be published. KW: Acquisition of data, drafting/revising the manuscript for content, final approval of the version to be published. EWR: Substantial contributions to the conception or design of the work, revising the manuscript for content, study supervision or coordination, final approval of the version to be published. AG: acquisition of data for the work, revising the manuscript for content, final approval of the version to be published. LK: Substantial contributions to the conception or design of the work, revising the manuscript for content, study supervision or coordination, final approval of the version to be published. YN: interpretation of data for the work, revising the manuscript for content, acquisition of data, final approval of the version to be published. BMJU: interpretation of data for the work, revising the manuscript for content, study supervision or coordination, final approval of the version to be published. JJGG: interpretation of data for the work, revising the manuscript for content, final approval of the version to be published. FB: interpretation of data for the work, drafting/revising the manuscript for content, study supervision or coordination, final approval of the version to be published. HV: Study concept and design, drafting and revising of manuscript, study supervision, analysis and interpretation of data, final approval of the version to be published. All authors declare that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding The MS Center Amsterdam is funded by the Dutch MS Research Foundation programme grant 98-358, 02-358b, 05-358c, 09-358d. The MS Center in Basel received grants by the Swiss National Research Foundation, the Swiss MS Society, the European Union, Gianni Rubatto Foundation, and the Novartis and Roche Research Foundations. Part of the study was sponsored by GlaxoSmithKline.
Competing interests None.
Ethics approval Local ethics committee VU University Medical Center (Amsterdam) and University Hospital (Basel).
Provenance and peer review Not commissioned; externally peer reviewed.
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.