Abstract
Background: Brain atrophy, in excess of that seen with normal aging, has been observed early in the clinical course of relapsing-remitting multiple sclerosis (RRMS). Previous work has suggested that at this stage of the disease, gray matter (GM) atrophy progresses more rapidly than the white matter (WM) atrophy.
Objectives: To characterize the evolution of GM and WM volumes over 2 years, and their associations with lesion loads in a cohort of patients with clinically early RRMS.
Methods: Twenty-one patients with RRMS (mean age 37.5 years, mean disease duration from symptom onset 2.1 years) and 10 healthy control subjects (mean age 37.1 years) were studied. Tissue volumes, as fractions of total intracranial volumes, were estimated at baseline and 1- and 2-year follow-up. Brain parenchymal fractions (BPF), GM fractions (GMF), and WM fractions (WMF) were estimated. In subjects with MS, brain lesion loads were determined on conventional T2-weighted along with pre- and post-gadolinium (Gd) enhanced T1-weighted images at each timepoint.
Results: A decrease in GMF was observed in subjects with MS vs normal controls over the 2 years of the study (mean −2.1% vs −1.0%, p = 0.044), while no change was seen in WMF over the same period (mean −0.09% vs +0.09%, p = 0.812). However, when the MS cohort was divided in half, dependent upon change in Gd-enhancing lesion load over 2 years (n = 20), a decrease in WMF was seen in the group (n = 10) with the largest decline in Gd volume, whereas WMF increased in the other half (n = 10) concurrent with a net increase in volume of Gd-enhancing lesions (difference between groups: p = 0.034).
Conclusions: Increasing gray matter but not white matter (WM) atrophy was observed early in the clinical course of relapsing-remitting multiple sclerosis. Fluctuations in inflammatory WM lesions appear to be related to volume changes in WM over this time period.
The concept of multiple sclerosis (MS) as an inflammatory neurodegenerative disease has gained credence in recent years, supported by pathologic1–3 and in vivo MR studies. Axonal damage has been recognized as a key feature of MS and there is increasing evidence that it may be, at least in part, independent of the degree of active inflammation associated demyelination.4–6 Moreover, it has been demonstrated that axonal injury is a major cause of permanent disability in patients with MS7–10 and it is apparent early in the clinical course of the disease.1,5,6,11,12 Both gray matter (GM) and white matter (WM) appear to manifest neurodegeneration, as reflected by tissue atrophy.12
MR measurement of brain atrophy may provide a measure of the neuroaxonal loss in MS pathology, although inflammatory activity with associated tissue edema, demyelination, and glial proliferation may mask or accentuate tissue loss.13 Progressive brain volume loss can be detected over a 1-year period.14–16 A recent preliminary study in 13 patients with clinically early (within 3 years of clinical onset) relapsing-remitting MS (RRMS) followed up for 18 months suggested that GM atrophy evolved progressively after clinical onset and that WM atrophy, while more apparent at baseline, developed less rapidly over the period of observation.17 Work exploring tissue specific atrophy in patients with clinically isolated syndromes (CIS) suggestive of MS has found progressive GM, but not WM, atrophy in those who go on to have a diagnosis of MS established within the subsequent 3 years.18
The present work extends our previous preliminary longitudinal atrophy study,17 following a larger cohort for a longer period of time. The main aims were to explore the relationship between focal lesions and atrophy measures, and evolving disability. In particular we were interested in the relations between focal inflammation, as inferred from gadolinium (Gd) enhancement, neuroaxonal loss, as inferred from atrophy measures, and clinical outcome.
Methods.
Subjects.
Data came from 21 patients with clinically definite RRMS19 (16 women and five men; mean age at the time of first MR imaging 37.5 years, range 26.9 to 56.1 years) and 10 healthy control subjects (4 women and 6 men; mean age at baseline scan 37.1 years, range 31.1 to 52.7 years). They represent a subset of a cohort of 32 MS subjects recruited to a serial MRI and clinical study, for whom 2-year follow-up data were available. Cross-sectional analysis of data derived from a larger subset of the cohort has been presented previously.12,20 At first assessment, the subjects with MS had a mean disease duration of 2.1 years (range 1.2 to 3.7 years) and a median Kurtzke Expanded Disability Status Scale (EDSS)21 score of 1.0 (range 0 to 3.0). At the time of first scanning, none of the patients was treated with disease modifying agents. All subjects were followed up for 2 years with data from annual evaluations included in the present study. At each of the timepoints (baseline, year 1, and year 2), patients with MS underwent a neurologic assessment, which included the rating of disability using EDSS and MS Functional Composite (MSFC)22 scores. Z-scores were calculated for each component of the MSFC (25 foot timed walk test [TWT],23 nine hole peg test [9HPT],24 and paced auditory serial additional test [PASAT]25), with reference to the MS subjects’ baseline data. The mean interval between scanning session and clinical evaluation was 5.8 days. During follow-up, patients received short courses of steroids when clinically indicated for the treatment of acute symptomatic exacerbations, although none received such treatment in the month prior to scanning. During the first year of follow-up, 7 out of the 21 MS subjects started interferon β (IFNβ) treatment, while none did so in the second year. The Joint Medical Ethics Committee of the Institute of Neurology and National Hospital for Neurology and Neurosurgery (Queen Square, London, UK) approved the study. Written informed consent was obtained from all subjects.
MR acquisition protocol.
MRI examinations were performed on a 1.5 Tesla machine (Signa; General Electric Medical Systems, Milwaukee, WI), using a standard quadrature head-coil. A single inversion-prepared three-dimensional fast spoiled gradient recall (three-dimensional FSPGR) sequence (repetition time [TR] = 13.3 msec, echo time [TE] = 4.2 msec, inversion time [TI] = 450 msec, number of excitations [NEX] = 1, matrix 256 × 160 interpolated to 256 × 192, field of view [FOV] = 300 × 225 mm, final in-plane resolution = 1.2 × 1.2 mm, with 124 × 1.5 mm slices covering the whole brain) was acquired in all subjects.
In a separate scanning session, performed a mean of 10 days before or after acquisition of the volumetric data (during which time none of the subjects reported additional symptoms), dual echo fast spin echo (FSE) sequence (TR = 2,000 msec, TE = 19/95 msec, NEX = 2, in-plane resolution = 0.9 × 0.9 mm), yielding proton density (PD) and T2-weighted images with 28 contiguous 5 mm slices, covering the whole brain, and a T1-weighted spin echo sequence (TR = 540 msec, TE = 20 msec, NEX = 1, in-plane resolution = 0.9 × 0.9 mm, with 28 × 5 mm slices covering the whole brain) were obtained in all the subjects. In subjects with MS only, the latter was also acquired 20 minutes post administration of 0.3 mmol/kg Gd-DTPA.
Image analysis.
Tissue volumes were estimated using a technique previously described, which utilizes SPM99 (Statistical Parametric Mapping; Wellcome Department of Cognitive Neurology, Institute of Neurology, Queen Square, London)26,27 to obtain GM, WM, and CSF masks from three-dimensional FSPGR images.28 In patients with MS, a lesion mask was generated using a semiautomatic local thresholding technique, part of the image display package Dispimage (Plummer, Department of Medical Physics and Bioengineering, University College London, UK)29 and then subtracted from GM, WM, and CSF masks using in-house software, yielding four mutually exclusive masks, one for each tissue type (GM, normal appearing WM [NAWM], CSF, and lesions). Tissue specific volumes were expressed as fractions of total intracranial (TI) volume (assumed to be the sum of GM, NAWM, lesion, and CSF volumes), estimating the brain parenchymal fraction (BPF; GM, NAWM plus lesion volume divided by TI volume), white matter fraction (WMF; NAWM plus lesion volume, divided by TI volume), and gray matter fraction (GMF; GM volume divided by the TI volume). T2 hyperintense (from the PD-weighted images), T1 hypointense, and T1 Gd-enhancing lesion volumes were also obtained at each timepoint, by means of the same semiautomatic lesion segmentation technique employed to generate the three-dimensional FSPGR lesion masks.
Statistical methods.
Rates of change over time for atrophy measures were estimated using random subject intercept and fixed slope regression models,30 with age and sex initially included as covariates to allow for possible confounding. Where inclusion of age and sex terms did not materially alter estimates, they were omitted from the model. Atrophy rates in subjects with MS and controls were compared using a patient indicator × time interaction term, and were expressed as percentage difference compared to baseline values. These models were also used to derive estimates of baseline and 2-year differences between patients and controls, potentially yielding slightly different estimates from those derived from purely cross-sectional analyses. Correlations between fractional tissue volumes and both lesion loads and disability measures were explored using Spearman’s rank correlation. Differences in fractional tissue volumes between MS subgroups, and normal controls, were assessed using the Mann-Whitney U test. Changes in lesion volume between baseline and year 2 were assessed using the Wilcoxon sign rank test. Percentage differences in MRI measures over 2 years were estimated relative to baseline values.
Results.
The mean interval between baseline and year 2 MRI scans was 24.8 months (range 22.5 to 26.6 months) in the control group and 24.7 months (range 23.0 to 28.1 months) in the MS cohort.
In one patient with MS and one normal control subject, year 1 follow-up scans were not available; otherwise all subjects completed 2-year follow-up. Administration of Gd was not possible in one patient at year 1 (due to technical difficulties) and in another subject with MS at year 2 (due to a potential allergic reaction to Gd reported during a previous scanning session). In the subjects with MS the EDSS at baseline was 1.0 (range 0 to 3), at 1 year 1.5 (1 to 3), and at 2 years 2.0 (0 to 4). The MSFC scores at each timepoint were 0 (−1.133 to 1.206), −0.006 (−1.648 to 1.466), and −0.124 (−2.508 to 1.407).
Lesion volumes.
Lesion volume estimates in subjects with MS at each timepoint are given in table 1. An increase of T2 hyperintense lesion volume (+23.2%; p = 0.023) was apparent at 2 years vs baseline. No change was observed for the T1 hypointense lesion volume (+2.3%; p = 0.903), while there was a nonsignificant trend to a decrease in Gd-enhanced lesion load (−67.9%; p = 0.079) over the period of the follow-up.
Table 1 Lesion load measurement values in the multiple sclerosis (MS) cohort at each of the three timepoints
Tissue specific atrophy.
Values of GMF, WMF, and BPF at each timepoint are presented in table 2. At baseline, GMF was lower in patients with MS vs the healthy control group (model estimate of difference patients − controls: −0.0160 [95% CI −0.0317 to −0.0003], p = 0.046). Over 2 years GMF decreased both within normal controls (−0.0026 per year [−0.0050 to −0.0002], p = 0.032) and patients (−0.0056 per year [−0.0073 to −0.0040], p < 0.001). The rate of GM atrophy over 2 years was greater in the MS vs the normal control group (mean −2.1% vs −1.0%, p = 0.044), with the result that the difference between the two cohorts was greater at year 2 (model estimate of difference patients − controls: −0.0220 [−0.0377 to −0.0064], p = 0.006). Thirteen out of the 21 patients showed a decrease in GMF over 2 years that exceeded the 95% limit of the change observed in the normal control subjects.
Table 2 Mean (SD) fractional tissue volume at each timepoint in normal control subjects and patients with multiple sclerosis (MS)
At baseline and year 2, WMF was different between patients and normal healthy subjects (model estimate of difference patients − controls: −0.0153 [−0.0269 to −0.0037], p = 0.010 at baseline; and −0.0158 [−0.0273 to −0.0042], p = 0.007 at year 2). WM volumes did not change over the duration of the study within either group (−0.0001 per year [−0.0012 to + 0.0009], p = 0.814 in MS; +0.0001 per year [−0.0014 to + 0.0016], p = 0.899 in normal control) or between the two groups (mean −0.09% vs + 0.09%, p = 0.812).
At entry into the study and after 2 years, BPF was lower in the MS group, vs normal controls (model estimate of difference patients − controls: −0.0313 [−0.0526 to −0.0099], p = 0.004 at baseline; and −0.0378 [−0.0590 to −0.0165], p = 0.001 at year 2). BPF decreased progressively in MS subjects (−0.0058 per year [−0.0079 to −0.0036], p < 0.001), but not in normal controls (−0.0025 per year [−0.0056 to + 0.0006], p = 0.110). The rate of decline in BPF over the 2 years of follow-up showed a trend to be larger in MS vs normal control subjects (mean −1.5% vs −0.6%, p = 0.094).
Changes in any fractional tissue volume from baseline to 2 years were not correlated with changes in disability scores over the same period. No correlation between changes in GMF and changes in any lesion volume over the duration of the follow-up was seen (GMF and Gd-enhancing lesion volume: rs = 0.276, p = 0.238; GMF and T2 hyperintense lesion volume: rs = 0.234, p = 0.308; GMF and T1 hypointense lesion volume: rs = 0.160, p = 0.489). Changes in WMF were correlated with changes in Gd-enhancing lesion load (rs = 0.580, p = 0.007), but not with T1 hypointense (rs = 0.150, p = 0.516) or T2 hyperintense (rs = −0.058, p = 0.801) lesion volumes. Similarly, changes in BPF were correlated with changes in Gd-enhancing lesion volume (rs = 0.525, p = 0.018) but not with T1 hypointense (rs = 0.272, p = 0.233) or T2 hyperintense (rs = 0.127, p = 0.582) lesion volumes. Cross-sectional analysis at baseline found that BPF and GMF were both correlated with T2 hyperintense and T1 hypointense, but not with Gd-enhancing lesion volumes. The correlations with lesion load measures were nonsignificant for WMF (table 3).
Table 3 Spearman correlations between lesion load and fractional tissue volumes at baseline
In order to explore whether change in WMF might be related to concurrent fluctuations in the extent of inflammatory lesions, the relationship between change in Gd volumes and WMF from baseline to year 2 was investigated. As previously reported, there was a correlation between the two measures of change, with a greater decline in Gd volume being associated with a greater decrease in WMF (rs = 0.580; p = 0.007). The concurrent relationship between Gd volume and WMF was further explored by dividing the patient group in half about the median, dependent upon their change in Gd lesion load over the 2 years of follow-up (table 4). In subgroup 1 mean Gd lesion load and WMF concurrently increased, whereas in subgroup 2 there was a net decline in Gd lesion load and decrease in mean WMF. The difference between the groups was significant (respective mean changes in WMF: +1.1% vs −1.1%; p = 0.034), although comparison of either subgroup with normal controls did not reveal any substantial difference (mean changes in WMF: +1.1% in subgroup 1 vs + 0.09% in normal controls, difference p = 0.199; mean changes in WMF: −1.1% in subgroup 2 vs + 0.09% in normal controls, difference p = 0.151). Changes in GMF and BPF were not different between the MS subgroups (p = 0.364 and p = 0.059) over the duration of the study. In subgroup 1, three patients received immunomodulatory treatment (mean 11 days from the baseline scanning); in subgroup 2, four subjects started IFNβ therapy within the first year of observation (mean 239 days from baseline scanning).
Table 4 Mean (SD) Gd-enhancing lesion volume and WMF in subjects with multiple sclerosis (MS) at baseline, year 1, and year 2, divided into two subgroups according to the amount of change in Gd enhancement over 2 years
In order to explore whether the prior extent of inflammatory lesions might be related to change in WMF, the relationship between the volume of Gd enhancing lesions at baseline and WMF change over the subsequent 2 years was investigated. There was a correlation between the two measures with a larger Gd enhancing baseline volume being associated with greater decline in WMF (rs = 0.599; p = 0.004). Similarly, Gd-enhancing lesion load at baseline was correlated with changes in BPF from baseline to 2 years (rs = 0.630; p = 0.002), but not with changes in GMF over the same period (rs = 0.372; p = 0.097).
There were no differences in the rate of changes of GMF (p = 0.149), WMF (p = 0.616), and BPF (p = 0.411) between subjects with MS who started IFNβ treatment within the first year of observation and those who did not.
Discussion.
There is growing evidence to suggest that brain atrophy in MS is an early, if subclinical process, in the pathologic evolution of the disease.12,14–16 Increasing atrophy is apparent over periods as brief as 1 year.14–16 While GM atrophy has been seen in cross-sectional MS studies,12,31–34 it is perhaps surprising that we found it to be progressive so early in the clinical course of the disease, while progressive WM atrophy was not evident. However, this result is consistent with a recent study in subjects with CIS suggestive of MS,18 in which subjects who developed MS within 3 years exhibited progressive GM, but not WM, atrophy.
Several cross-sectional studies have shown moderate correlations between the extent of GM atrophy and T2 lesion load.12,32–34 The degree of these correlations suggests that between a quarter to half of GM atrophy may be accounted for by a relation with T2 lesion loads. Noting that measurement errors may attenuate the apparent strength of associations, this result may be interpreted in several ways, for example: either GM atrophy is only partially related to focal lesion genesis and a more global process needs to be invoked to explain the degree of atrophy seen; alternatively, a temporal delay between the process leading to T2 lesion formation and subsequent atrophy needs to be considered. T2 lesions are not pathologically specific, and while most are thought to have been at one stage inflammatory,35–39 they do not intrinsically tell us when the inflammation occurred; in this regard Gd-enhancing lesions may be a more useful marker. However, in the present cohort, while we do see similar cross-sectional associations, neither a relation between change in T2 nor Gd-enhancing lesion loads and GM atrophy was observed, suggesting that either the delay between the lesion genesis and atrophy exceeds 2 years; that other factors such as ongoing inflammation and gliosis mask atrophy; that both are indirect markers of some other underlying process; or that these processes are truly independent. The concept of a significant delay between focal lesion genesis and atrophy would appear to be supported by findings in a cohort of subjects with MS followed up for 14 years after first onset of symptoms.40 In these subjects, lesion accrual early in the clinical course of their disease appeared to be more closely related to subsequent atrophy than later lesion accumulation, although overall the association was modest, accounting for under a third of estimated tissue volume loss.
The WM results are striking in contrast to those of the GM. While WM atrophy was present at baseline (compared with controls and confirming the results of a previous larger cross-sectional study12), there was no further decrease in WM volume during the 2-year study period in the overall group, but concurrent fluctuations in focal inflammation as marked by Gd-enhancing lesion volume were related to WM volume changes. Volume loss in such circumstances may be seen as the resolution of an inflammatory process, and a decrease in mean WMF was evident in the half of patients exhibiting the largest drop in Gd-enhancing lesion volume between baseline and year 2 (see table 4). On the other hand, in the other half of patients—in whom there was an increase in mean Gd lesion volume—there was an increase in mean WMF. The absence of increasing WM atrophy in the whole group may be due to further inflammation or gliosis compensating for volume loss associated with loss of myelin and axons. Further, some of the WM damage associated with lesions may be reversible; remyelination has been documented in MS lesion and may be extensive during the early stages of the disease, occurring within both lesions and in the NAWM.41–44
In contrast, the absence of a clear association of WM volume change with either T1 hypointense or T2 hyperintense lesion loads is perhaps not surprising, since they reflect more heterogeneous pathologic processes; Gd-enhancing lesions mark focal areas of blood–brain barrier breakdown associated with active inflammation36,38; precontrast T1 hypointense lesions are associated with neurodegeneration45; and T2 hyperintense lesions are pathologically mixed, with elements of edema, demyelination, remyelination, reactive gliosis, and axonal loss.46
Our observation that those patients who had high baseline Gd enhancing volumes exhibited more WMF loss over the subsequent 2 years might indicate that loss of WM tissue occurs as a delayed effect following inflammation, for example due to Wallerian degeneration following axonal transection in acute inflammatory lesions.1,2 Alternatively, a reduction of inflammation per se may have contributed to the decrease in WM volume. This association was observed with enhancing lesion loads determined after administration of triple dose Gd; recent work in subjects with CIS has suggested that this may improve the apparent association between Gd enhancing lesion loads and atrophy,47 and so studies determining enhancing lesion loads after conventional dose Gd may not yield directly comparable results from similarly sized cohorts.
While our study suggests that over a period of up to 2 years, fluctuations in inflammation may limit the ability to detect increasing WM atrophy in early RRMS, it seems likely that with longer term follow-up it will become apparent as myelin and axonal loss increase. That WM tissue loss was evident at entry to the study indicates that it is already present early in the course of MS.
The difference between the serial GM and WM volume changes also suggests that pathologic processes within these tissues are different. This is supported by neuropathologic examination of lesions within both GM and WM, finding that those in WM were more inflammatory than those in GM.48–50 It is possible that differences in the extent of inflammation or glial proliferation may extend beyond focal lesions in WM more than GM and may thereby have contributed to our findings. In support of this concept, both of these pathologic features are observed in the NAWM in MS,51 and MR spectroscopy studies have shown myo-inositol, a putative glial cell marker, to be significantly elevated in MS NAWM, but not in GM.20,52–55
Given the limited clinical progression in the MS cohort (as perceived using the EDSS and MSFC scales), the absence of noticeable relations between atrophy and clinical outcome is not surprising. Other studies addressing the longitudinal relation between atrophy and disability have offered mixed results: some have demonstrated an association between these measures,15,56–58 while others have not.16,59,60 More extensive cognitive test batteries may prove to be a useful addition to EDSS and MSFC estimates and recent work has revealed associations between subtle cognitive impairment and atrophy in subjects with MS.61
The modest size of the MS and the normal control cohorts may have limited the detection of subtle changes in WM tissue volumes and underestimated the extent of GM pathology in patients with MS. Such disease effects may become more apparent with further follow-up or in larger cohorts. The sex characteristics of the MS and control groups were not optimally matched; however, the longitudinal nature of the study renders this relatively less important than may be the case in cross-sectional work; and the statistical models allowed for sex effects, further minimizing its impact upon the results. One third of the patients started treatment with IFNβ within the first year of follow-up; however, this is unlikely to be a confounding factor, given that there was no significant difference between the rate of tissue atrophy in the treated and untreated patients with MS. IFNβ treatment is also unlikely to explain the progressive decline in Gd-enhancing lesion volume observed in subgroup 2 over 2 years (when the MS group was subdivided in half dependent on changes in Gd lesion loads): treatment was started earlier in those subjects with increasing Gd lesion loads (11 vs 239 days). We should also recall the limitations of the measurement techniques used: a range of atrophy measures have been developed and each has its own advantages and disadvantages.13 The reproducibility and sensitivity of the techniques employed in this study has previously been assessed,28 along with the potential for segmentation bias associated with WM lesions. The presence of artificial lesions led to an artifactual increase in GMF and decrease in WMF. A recent study18 has confirmed these findings, further quantifying the effects and finding to be of limited importance. Given this, in the present study it is unlikely that the significant increase in T2 hyperintense lesion load seen may account for the progressive decrease in GMF over 2 years and indeed may, to a degree, have underestimated GMF loss.
Acknowledgment
The authors thank the subjects who participated in this study; C. Middleditch for technical assistance; and R. Gordon, D. MacManus, and C. Benton for scan acquisition.
Footnotes
-
Supported by the MS Society of Great Britain and Northern Ireland (program grant support of the NMR Research Unit). M.T. was supported in part by the Department of Neurologic and Psychiatric Sciences, University of Padua. D.T.C. was supported by a grant from Schering AG, administered by the Institute of Neurology. J.S.-G. was funded through a grant from the Spanish Ministry of Health (BEFI #02/9115).
Received June 9, 2004. Accepted in final form November 29, 2004.
References
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.