Spinal cord atrophy and disability in MS

Involvement of the spinal cord in MS has long been known to be extremely common and of particular importance in the development of disability.¹ The advent of MRI demonstrated lesions within the spinal cord in 75% of patients with MS, more commonly in the cervical than the thoracic cord.² However, the number and extent of T2-weighted lesions have not correlated with disability in cross-sectional or longitudinal studies.^2,3 The presence of atrophy, which may be focal at the site of lesions or can become generalized after subsequent degeneration of axons, has also been noted.^4,5 Several groups have looked at the potential of measurement of cord atrophy as a marker of axonal loss and have shown promising correlations with disability.^2,3,6,7

Kidd et al.² in 1993 evaluated 5-mm axial sections taken at four vertebral levels; C5, T2, T7, and T11. The cords were manually outlined and atrophy was considered to be present when the measured area was 2 SD below that of the mean for healthy control subjects. The mean cord areas of the patients were significantly smaller than that of the control subjects at each of the four levels. Those patients with atrophy were found to have significantly higher levels of disability as measured by Kurtzke's expanded disability status scale (EDSS)⁸ than those without atrophy. Another study measuring cross-sectional area at C5 showed a significant difference between patients with benign and secondary progressive MS.⁶

Subsequent studies concentrated on measuring serial cord cross-sectional area at C5. In a study comparing primary and secondary progressive MS, both groups showed a decrease in mean cord area during 1 year, but there was no difference between the two groups and no significant correlation with disease progression.³ The intrarater reliability of the measurement technique was 2%, but the scan-rescan variability was in the order of 6% and changes detected were within the 95% confidence limits for measurement variation.

A similar study of relapsing-remitting patients over 1 year using the same technique also failed to demonstrate a significant change in cord area, but the mean intrarater variability was high at 4.8% (intrarater limits of agreement - 11.6 to 12.9%).⁷

All of the aforementioned studies depended on two-dimensional imaging with a gradient echo sequence and a manual outlining technique for cross-sectional area measurement. The poor reproducibility of this technique made the detection of small change, an essential prerequisite for serial studies, impractical.

Subsequently, we have developed a technique to address these difficulties.⁹ By using a volume-acquired acquired inversion-prepared fast spoiled gradient echo (FSPGR) acquisition, the contrast between the gray cord and nulled black CSF is markedly improved. The cord area is assessed at the level of the C2-C3 intervertebral disc using a semiautomated technique. This level was chosen because there is little variability in cross-sectional area over this segment, and the CSF pool is capacious, thus optimizing cord/CSF contrast. Using this methodology, reproducibility was greatly improved with a scan-rescan coefficient of variation of 0.79%. In a cross-sectional study, Losseff et al.⁹ studied 30 control subjects and 60 patients (15 in each subgroup of relapsing-remitting, primary progressive, secondary progressive, and benign MS patients). The cord cross-sectional areas of the benign, primary, and secondary progressive groups were significantly smaller than that of the control subjects, whereas the relapsing-remitting group had no significant atrophy. Cord cross-sectional area correlated strongly with disability (r = -0.7, p < 0.001) and with disease duration (r = -0.52, p < 0.001).

We have now applied this technique in a serial study. Because of the extremely small changes expected in cord cross-sectional area during 12 months, all parameters during data acquisition and analysis must remain absolutely constant. During our first serial study, there was a major hardware upgrade. This resulted in minor changes in the pulse sequence and spatial signal intensity uniformity. As a consequence, all of the control subjects exhibited an artefactual increase in their measured cord areas.¹⁰

In the current study, we have continued the annual assessment of available patients first assessed in the aforementioned articles by Losseff et al. to give two sets of comparable data, 12 months apart, allowing us to assess whether changes in cord cross-sectional area can be measured during such a period in MS patients.

Methods. Patients. Patients were recruited from the cohort previously studied.⁹ Thirteen of the control subjects and 28 patients attended at baseline and 1 year follow-up. Of the 28 patients, 12 had primary progressive, 6 secondary progressive, 6 relapsing-remitting, and 4 benign disease (relapsing-remitting disease of at least 10 years' duration and a disability of EDSS 3 or less). All subjects gave informed written consent to continue the study, which was approved by the ethical committee of the National Hospital for Neurology and Neurosurgery.

All patients underwent a neurologic examination and evaluation of the Kurtzke EDSS before undergoing MRI of the cervical spine. None of the patients was experiencing an acute relapse at the time of assessment. A definite change in EDSS during the 12-month period was defined as an increase of 1.0 if the EDSS was ≤5 or an increase of 0.5 if it was >5.0.

Imaging. All MR imaging was carried out on a Signa 1.5-T system(General Electric, Milwaukee, WI). An FSPGR acquisition was performed (60 1-mm slices, inversion time = 450 ms, echo time = 4.2 ms, repetition time = 15.6 ms, flip angle = 20°, matrix 256 × 256, acquisition time = 7 minutes), and from the data set a series of 5 contiguous 3-mm axial slices(perpendicular to the spinal cord) were reformatted using the center of the C2-C3 disc as the caudal landmark.

Assessment of degenerative vertebral disease. All of the sagittal data sets were examined for the presence of spondylosis or disc protrusion by two experienced neuroradiologists who were blinded to the subjects' details. Each data set was scored on a four-point scale: 0(normal), 1 (thecal indentation only), 2 (thecal indentation touching the cord), and 3 (cord compression).

Cross-sectional area measurement and reproducibility. This was achieved using the semiautomated methodology as previously described⁹ after image uniformity correction.¹¹ The analysis was carried out by a single observer (V.L.S.) blinded to the clinical status of the subjects.

Intrarater and interrater reproducibility was assessed in 10 random subjects and scan-rescan in 5 subjects. To ensure reproducibility over time, a quality assurance protocol was designed that involved weekly scanning of control subjects.¹²

Statistics. Vertebral degenerative scores of the patients and control subjects were compared using the independent t-test. The paired t-test was used to assess significant changes in the cord cross-sectional area over time, and the independent t-test assessed differences between the groups. Correlations between cord area, disease duration, and the EDSS were calculated using Spearman rank correlation coefficient.

Results. Intrarater reproducibility of the cross-sectional area measurement technique was found to be 0.51%, which is comparable to the original assessment of reproducibility by Losseff et al.⁹ of 0.73%. Interrater reproducibility was 1.75% and scan-rescan variation was 0.87%.

The characteristics of the 13 control subjects and 28 patients are detailed in tables 1 and 2. There was no significant difference between the ages of patients and control subjects, although as expected the relapsing-remitting patients were the youngest. The benign and secondary progressive patients had significantly longer disease durations than the other patient groups, and with regard to the EDSS, both the primary and secondary progressive patients were significantly more disabled than the other groups. There was no significant difference in the scores for vertebral degenerative disease between the patients and control subjects (patient mean score = 1.23, SD 0.86; control mean score = 1.38, SD 1.10; p = 0.66).

At baseline, there was a significant difference in cord cross-sectional area between control subjects and both primary and secondary progressive groups, but not between control subjects and the relapsing-remitting and benign groups.

There was a significant difference in both the baseline cord area(p = 0.001) and the change in cord cross-sectional area during the 12 months (p = 0.05) between the whole patient group and control subjects. In the whole patient group and in the primary progressive and relapsing-remitting subgroups, the follow-up measures were significantly smaller (p ≤ 0.001). There was no significant difference in cord area at baseline and 1 year in the control subjects, secondary progressive, or benign patient subgroups (figure).

There was a strong correlation between baseline cord area and EDSS(r = -0.52, p = 0.005) and to a greater extent disease duration (r = -0.75, p < 0.001). The EDSS did not correlate significantly with disease duration (r = -0.34, p = 0.07).

Eight of the 28 patients had a definite increase in their EDSS during the 12-month period, but they exhibited no significant differences in cord area at baseline (p = 0.69) or change in cord area during the year(p = 0.51) compared with the 20 patients without a definite increase in EDSS.

Discussion. This study shows that it is possible to reproducibly measure and detect change in cord cross-sectional areas during a 12-month period in patients with MS. It also confirms the previous findings of a strong correlation between a clinical measure of disability (the EDSS) and spinal cord atrophy. This has been possible because of improvements in the measurement technique. With the application of an inversion-prepared FSPGR sequence and assessment at the C2-C3 level, excellent cord/CSF contrast is obtained. This increase in contrast combined with a semiautomated contouring technique eliminates much of the measurement variability that occurred with the previous method of manual outlining of the cord at the C5 level using a gradient echo sequence with a poorer cord/CSF contrast. Ongoing quality assurance measures ensure that there is no drift in control measurements over time and in particular no effects secondary to hardware services or upgrades.

These findings are independent of vertebral degenerative disease, which was mild and comparable in both the control and patient groups; the most common site of spondylosis was the C5-C6 disc space, and no subjects had abnormality at or above the C2-C3 disc space where atrophy measures were obtained. The scoring system was applied with a very low threshold for abnormality, scores of 1 or 2 were not thought to be of any clinical significance. Six subjects had evidence of mild cord compression (not warranting clinical intervention) at one or both of their examinations; two of these were control subjects and four were patients (i.e., 15% of control subjects and 14% of patients).

There was no significant change in cord area during 1 year in the control group. Cord atrophy was present at baseline in the benign and primary progressive groups, but most marked in the secondary progressive group. Increasing atrophy during the 12-month period was only detectable in the relapsing-remitting and primary progressive patient groups.

To try to understand the mechanisms and clinical importance of atrophy, it is relevant to consider the pathophysiology of the development of disability, which results from two causes: first, from incomplete recovery after relapse; and second, and most importantly, as the result of insidious disease progression.

The relapsing-remitting patient group who are very early in their disease process have normal-size cords, but have a high rate of cord cross-sectional area loss. These patients are experiencing relapses that are predominantly associated with conduction block secondary to inflammatory demyelination, although some element of axonal loss can occur acutely.^13,14 Demyelination per se has been shown to result in a reduction of axonal diameter¹⁵ and, combined with an element of acute axonal loss, could well result in atrophy of a structure such as the spinal cord after a relapse.

The benign patients begin with a similar relapsing course resulting in associated cord loss; however, minimal change in function occurs. This may be caused by predominant demyelination rather than axonal loss. Subsequently in their disease course, they experience very few relapses and little further change in cord area.

The largest degree of cord atrophy over time was seen in the primary progressive group, which is not surprising if we consider their slowly progressing disability as a consequence not of inflammatory demyelinating lesions but of progressive axonal loss. Certainly less inflammation is seen in this group as shown by pathologic¹⁶ and MRI studies.¹⁷

The secondary progressive patient group who had the smallest cords may have a combination of both demyelination and acute axonal loss secondary to relapses and subsequently progressive axonal loss underlying the later progressive phase. The small sample in this study is not entirely representative of the secondary progressive disease group, because all four patients were very disabled and their disease may have "burned out"; consequently, no detectable change in cord cross-sectional area was seen.

The suggestion that atrophy may have two mechanisms, one nondisabling(demyelination), the other disabling (axonal loss), is supported by the almost identical atrophic cord areas seen in the disabled primary progressive and nondisabled benign cohorts. However, the development of axonal loss is linked to demyelination within lesions as shown by pathologic studies. Trapp et al.¹⁸ have recently demonstrated extensive axonal transection within both active and chronic active MS lesions, which was related to the degree of demyelination within the lesions. Examination of areas of normal-appearing white matter (NAWM) also revealed many more transected axons than seen in control brains. Another factor that should be considered is the role of reactive gliosis; this in theory may partly compensate for the loss of myelin by filling space but could also lead to contraction of the underlying tissue. Evidence of gliosis comes from histologic¹⁹ and MR spectroscopy studies.²⁰ Rooney et al.²⁰ demonstrated a reduced N-acetyl aspartate (NAA)/creatine ratio is areas of NAWM. This appeared to be produced by a rise in the level of creatine and could be accounted for by an increase in gliosis. This study did not find evidence of diffuse axonal loss (decrease in NAA levels), although the patient group was not defined and may have comprised only early relapsing-remitting MS cases.

These arguments are hypothetical and could be investigated further by combining the assessment of atrophy with markers of both demyelination and axonal loss. The former includes magnetization transfer ratio (MTR), a technique that indirectly measures the degree of macromolecular structure and consequently may reflect demyelination. Measurement of MTR is now feasible within the cervical cord and has recently been shown to be lower in patients with MS than in normal control subjects.²¹ Potential markers of axonal loss include quantification of NAA levels by MR spectroscopy and diffusion tensor imaging.²² The reported reduction of NAA in both lesions and NAWM in primary progressive MS²³ suggests that diffuse axonal loss is present in this subgroup. Further evidence of pathologic change distinct to new lesion formation in this patient group comes from the presence of diffuse signal change on proton density-weighted conventional spin echo images of the spinal cord; the degree of this signal change was shown to correlate with cord atrophy.²⁴ The combination of MR studies with postmortem data will further enable us to correlate changes in the size of the spinal cord with pathologic changes.^25,26

The findings in the primary progressive group are of particular importance in relation to monitoring disease activity, because this patient group does not show the rate of new lesion development and enhancement seen in the relapsing-remitting and secondary progressive groups. Serial evaluation of cord and possibly brain atrophy²⁷ may be extremely valuable in the assessment of these patients. A large longitudinal study of 158 such patients is already under way to investigate this further,²⁸ and data are being collected in the recently established phase II trial of beta-interferon 1a in primary progressive MS.²⁹

In conclusion, the fact that significant changes in cord cross-sectional area can be detected within 12 months is important both in increasing our understanding of the underlying mechanisms of disability and in the monitoring of therapeutic trials. This study does have limitations: the sample size is small, in particular there are few patients in the secondary progressive or benign patient groups, and follow-up is for only 1 year. These factors may have contributed to our failure to detect a correlation between change in cord size and change in EDSS.

Larger serial studies are now needed to confirm the detected differences between disease subgroups and ideally should commence in the early stages of MS, allowing evaluation of cord atrophy as a potential prognostic marker.

Acknowledgment

We thank Dr. I. F. Moseley for his help and useful discussion.