For a long time, multiple sclerosis (MS) has been considered to be a disease with predominantly inflammatory features, followed by secondary neurodegeneration in advanced disease stages, with both processes occurring predominantly in the white matter (WM) of the central nervous system (CNS). It now appears that degeneration is not an end-stage disease phenomenon, but rather an early feature of MS,1 2 probably starting simultaneously with focal inflammation both in the WM and the grey matter (GM).3 4 Evidence is mounting that other factors, especially age,5 6 may play an important role in the evolution of individual clinical course, development of disability, and the inter-relationship between inflammatory and neurodegenerative changes in the WM and GM.

Several studies have shown more robust correlations between clinical disability and measures of tissue atrophy than had been shown previously by examining correlations between disability and lesion volumes (LVs).7^–13 However, it is still not clear whether the rate and location of tissue loss is constant or varies over time and between clinical subgroups.14 Because of the presence of GM lesions in cortical and deep GM structures,15 and the increasingly important evidence for neurodegeneration in MS, measurement of GM atrophy is becoming an important endpoint in clinical studies. Preliminary data from several studies suggest that GM atrophy develops at a much faster rate than WM or whole brain atrophy.16^–18 However, it has not been fully elucidated whether GM atrophy is a result of the primary disease process in MS or is secondary to the damage in the WM. Establishing the rate of GM and WM atrophy progression over the mid- to long-term in patients with different disease courses, and exploring the relationship between the development of focal inflammatory disease and progression of GM atrophy, is an important step in validating these newer tissue compartment MRI measures in clinical trials. Another question that still has not been completely answered is whether the pathological processes in the periventricular regions and deep GM structures might be better reflected by measurement of central atrophy rather than the global atrophy measures. Some studies showed that these measurements14 19^–21 may be sensitive markers of the tissue destruction process.

With regards to the differences in WM and GM evolution that have been recently reported in patients with MS,16^–18 22 an extremely interesting fact is that different ageing patterns of GM and WM tissue compartments are also seen physiologically in normal populations.23^–26 The results from these normal control (NC) studies have shown that GM volume reaches its maximum at age 6–9 years, thereafter starts to decline, and that this decline appears to be a constant, linear function of age throughout adult life. However, WM volume loss seems to be delayed until middle adult life.25 26

Therefore, in the present study, we investigated the evolution of different inflammatory and neurodegenerative MRI indices in a group of early relapsing-remitting patients with MS with high clinical activity and in NCs over 2 years. We also evaluated the progression of these MRI measures in a subset of patients who were followed-up for 5 years. Finally, we aimed to investigate the relationship between accumulation of T2-LV over time and development of brain atrophy in different tissue compartments.

METHODS

The current analysis is based on MRI data from the patients enrolled in a randomised, double-blind, placebo-controlled ASA (Avonex, Steroid, Azathioprine) study.27 28 In the original trial, 181 patients were randomised equally into one of three treatment groups: Group 1, IFNβ-1α 30 µg IM once weekly plus two placebos; Group 2, IFNβ-1α 30 µg IM once weekly plus Azathioprine 50 mg PO once a day plus placebo; or Group 3, IFNβ-1α 30 µg IM once weekly plus Azathioprine 50 mg PO once a day plus prednisone 10 mg PO every other day.

In the original study, there were no significant differences regarding clinical or MRI outcomes during the 2- and 5-year studies, except at 2 years when the change from baseline in T2-LV favoured the triple-agent combination therapy versus IFNβ-1α monotherapy.27 28 Therefore, in this study, we present the combined clinical and MRI findings of the 147 (81.2%) patients who had completed treatment, clinical and MRI assessment at all three time points (months 0, 12, 24) and were on blinded originally assigned therapy. Clinical and MRI baseline characteristics of this patient subgroup did not differ significantly from the subgroup not included in the study (data not shown). In addition to the 147 patients completing 24-month follow-up, 66 patients at 36 months (49 still on blinded treatment), 51 at 48 months (28 on originally assigned therapy) and 43 at 60 months (10 on original therapy) obtained full MRI and clinical assessment. A subgroup of 36 patients, who were followed for 5 years, underwent complete serial clinical and MRI assessment at each time point and were therefore separately analysed. In the extended phase of the study (3–5 years), the rate of drop-out from the original medication increased due to patient decision or clinical worsening. In cases of worsening, the blinded study medication was stopped and treatment was escalated in different ways according to general recommendation guidelines (usually switched to a higher dose of interferon-β treatments or a combination with chemotherapeutic drugs).

The inclusion criteria were clinically definite MS according to the Poser criteria29 confirmed by MRI and the presence of at least two oligoclonal bands in cerebrospinal fluid (CSF), age 18–55 years, Expanded Disability Status Scale (EDSS)30 ⩽3.5 on the day of screening, and active disease defined by two relapses in the previous 12 months or 3 relapses in the previous 24 months. Exclusion criteria were IFNβ therapy in the previous 12 months or immunosuppressive treatment with either pulse cyclophosphamide or mitoxantrone in the previous 6 months, ineffective contraception and active major organ disease.

Patients were clinically assessed using the EDSS score every 2 months over the first 12 months and then every 3 months until month 60. As well as patients with MS, a group of 27 NCs was recruited prospectively during the study and scanned with the same MRI protocol. Control subjects were recruited from among healthy hospital volunteers. The study was approved by the local Medical Ethical Committee.

MRI methods

Image acquisition

Brain MRI was performed on a Philips Gyroscan 1.5 Tesla unit (Best, The Netherlands). Axial images of the brain were obtained with 1.5 mm slice thickness using fast attenuated inversion recovery (FLAIR) (TR/TE/TI 11000/140/2600 ms, matrix size 256 × 181, flip angle 90) and axial T1-weighted 3-dimensional (3D) SPGR images (TR/TE 25/5 ms, matrix size 256 × 204, flip angle 30) obtained with 1 mm slice thickness. All images were non-gapped. The scans were acquired on a bi-monthly basis in patients with MS for the first 2 years and then yearly thereafter up to 5 years. The NCs underwent MRI scans at baseline, 12 and 24 months. MRI scans of patients with MS acquired at baseline, 12, 24, 36, 48 and 60 months were evaluated.

Image analysis

Image analysis was performed at the Buffalo Neuroimaging Analysis Center, The Jacobs Neurological Institute, Department of Neurology, State University at Buffalo, NY, USA. Buffalo investigators were blinded to the patients’ clinical characteristics and clinical status.

Semi-automated edge detection methods with manual correction for region of interest definition was used for measurement of T2-lesion volume (T2-LV),31^–33 lateral ventricle volume (LVV) and the third ventricle width (3VW).13 21

For brain extraction and tissue segmentation, we used SIENA/X cross-sectional and longitudinal brain atrophy analyses methods.34 35 Normalised volumes of whole brain (NBV), GM (NGMV), WM (NWMV) and CSF were obtained via this process.21 36 To measure normalised peripheral grey volume (NPGV), a standard space mask was used to separate neocortical GM from non-neocortical volumes.21 To compare tissue compartments longitudinally, we performed cross-sectional SIENAX analysis and then calculated absolute compartment-wise differences. Using these, we then calculated median percentage compartment volume changes. The SIENA method utilises algorithms similar to those of automated SIENAX; however, it is used specifically between longitudinal scans.21 34 35 37 The baseline scan was co-registered with the follow-up scan and an analysis of relative edge motions was used to calculate the percentage brain volume change (PBVC) between all available time points.37

Statistical analysis

Statistical analyses were performed using SPSS 13 (SPSS, Chicago, IL, USA). Data were first examined statistically and visually to ensure that they met the assumption of normality. Multivariate repeated measure analysis of variance (MANOVA) was performed using the SPSS General Linear Model (GLM) procedure to analyse the effect of time, group (NCs vs patients with MS) and time by group interactions for all MRI measures (NBV, NGMV, NWMV, NPGV, 3VW and LVV). This process was then repeated using multivariate analysis of covariance (MANCOVA) to quantify effects after adjusting for group differences in age and sex at baseline. A final set of models were constructed with group as the between-subject factor. Post hoc Wilcoxon tests were used to compare medians when a significant F test ratio was found for main effects or interaction terms. Median percentage absolute changes—from baseline to 12 months, 12 to 24 months, and baseline to 24 months—were also determined for the six atrophy measures in patients with MS vs NCs using each patient’s baseline values as the denominator. Percentage change during the period 0–12, 12–24 and 0–24 months was compared separately for patients vs controls for each of the MRI measures using the Wilcoxon test. Following this, differences in median absolute percentage changes for each time point were assessed with analysis of covariance (ANCOVA), adjusting for baseline age group differences. In addition, forward conditional logistic regression analysis adjusted for age and sex differences was performed to determine the set of MRI measures that best differentiated patients with MS from normal controls. Finally, a subset of patients with MS had additional MRIs performed annually for up to 5 years following baseline scan (5-year serial completers). Therefore, separate analyses were conducted at each time point to compare the relative change in MRI measures in the whole group of patients (147) as well as the 5-year completers (36). Friedman’s test for multiple related samples was performed, followed by Wilcoxon rank sum post hoc test, separately for any possible time period combination to show differences between particular MRI measures. Spearman rank correlation coefficients were used to assess the univariate relationship between different MRI variables. Also conducted were partial correlation analyses adjusted for age at baseline MRI scan, followed by various multivariate stepwise linear regression analyses. Post hoc Bonferroni correction was performed for all MANOVA, MANCOVA and Wilcoxon analyses in order to correct for multiple comparisons directly using the SPSS-related procedure, and two-sided p-values were used to perform all significance tests.

RESULTS

Baseline demographic, clinical and MRI data

Table 1 summarises the baseline demographic and clinical data of patients with MS and NCs. There was a significant difference in sex (p = 0.001) and age (p = 0.003) when NCs were compared with patients with MS.

Table 2 includes baseline MRI metrics for both study groups. After adjusting for baseline age and sex, there were significant differences between patients with MS and NCs in baseline NBV (p = 0.002), NPGV (p = 0.002) and NVMV (p = 0.014) (table 2).

Evolution of different MRI measures over 2 years: Comparison of patients and normal controls

Figure 1 shows evolution of atrophy MRI measures in NCs and patients with MS over the 2-year period. Using MANCOVA, we found a statistically significant difference between the two groups at all time points for all MRI measures except the NWMV. In the patients with MS, a post hoc analysis (Wilcoxon) corrected for multiple comparison (threshold of p<0.003) showed significant within-group median changes in NBV, NGMV, NPGV, LVV and 3VW at all three time points (0–12, 12–24 and 0–24) (p<0.001), except for the NBV at time period 12–24 months (p = 0.01) and the NWMV at any of these three time points. No significant within-group changes of the above MRI parameters were found in the NC group except for the NPGV, with borderline significant change (p = 0.032) at time period 0–24 months (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Evolution of normalised MRI measures in patients with MS and normal controls over 2 years. 3VW, third ventricle width; GMV, grey matter volume; LVV, lateral ventricle volume; n, number of patients; PBVC, percentage brain volume change; PGV, peripheral grey volume; T2-LV, T2 lesion volume; WMV, white matter volume.

Median absolute and percentage changes over 5 years for different MRI measures in patients with MS are given in table 3 and in figure 2. Corresponding data for the NC group at time period 0–24 months were: −0.22% for PBVC, −0.72% for GMV, −1.03% for PGV, −0.1% for WMV, +0.55% for LVV and 0% for 3VW.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Median percentage MRI changes over 5-year period. Patients with all available MRI scans (upper raw) and 5-year serial completers (lower raw) are presented. 3VW, third ventricle width; GMV, grey matter volume; LVV, lateral ventricle volume; n, number of patients; PBVC, percentage brain volume change; PGV, peripheral grey volume; T2-LV, T2 lesion volume; WMV, white matter volume.

The differences between SIENA/X-derived variables (PBVC, GMV, WMV and PGV) were compared at each time point using the Friedman test. Significant differences (p<0.001) were found at each time point. PBVC, percentage brain volume change; GMV, grey matter volume; WMV, white matter volume; PGV, peripheral grey volume; T2-LV, T2 lesion volume; 3VW, third ventricle width; LVV, lateral ventricle volume; n, number of patients.

There was a significant difference between patients with MS and NCs for several MRI-related median percentage changes when analyses were age- and sex-adjusted (fig 1). In particular, patients with MS lost significantly more PBVC over 0–24 months (p<0.001), GMV over 12–24 months (p<0.003) and 0–24 months (p<0.001), PGV over 0–24 months (p<0.001), 3VW over 0–24 (p = 0.003) and LVV at all three time points (p<0.003).

Forward conditional logistic regression analysis including age and sex as covariates, and NBV, NGMV, NWMV, NPGV, 3WV and LVV or their median percentage changes (0–24 months) as independent variables (including PBVC), investigated differences between patients with MS and NCs at baseline and over the 24-month period. At baseline, the NPGV (p = 0.001), NGMV (p = 0.021) and 3VW (p = 0.016) were retained in this model. Over 2 years, median percentage changes in GMV and LVV were retained as the best predictors of MS versus NC atrophy accumulation (all p<0.001).

Evolution of different MRI measures over 5 years in patients with MS

The evolution of median percentage changes of various MRI measures over a 5-year period (at time points 0–12, 0–24, 0–36, 0–48 and 0–60 months) is shown in tables 3 and 4, and figure 2. The data are provided separately for patients with all available MRI scans (n = 147) (table 3, fig 2 upper raw) and for 5-year serial completers (n = 36) (table 4, fig 2 lower raw). In patients with all available MRI scans (n = 147), the non-parametric statistic (Friedman test) showed that the median percentage changes during that time period were statistically different when PBVC, GMV, WMV and PGV were compared (p<0.001 at all study time points) (table 3). A post-hoc Wilcoxon analysis corrected for multiple comparisons (threshold of p = 0.006) revealed significant differences between GM atrophy measures (GMV and PGV), and PBVC and WMV, favouring a faster decrease of GM volumes. A similar finding was observed for the subgroup of 36 serial 5-year completers who showed significant differences at the time points of 48 and 60 months (p<0.001) (table 4).

Annual median percentage changes are summarised in figure 3. No significant differences were found comparing year-to-year changes except for the PBVC over months 24–36 versus 36–48 (p<0.03) in the group of 5-year completers.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Annual median percentage MRI changes over 5 years: 2-year treatment (left) and 5-year completers (right) are shown. Key shows time in months. GMV, grey matter volume; PBVC, percentage brain volume change; PGV, peripheral grey volume; WMV, white matter volume.

Correlations between different MRI measures

Spearman cross-sectional analysis at each study time point (0, 12, 24, 36, 48 and 60 months) showed modest-to-strong correlations between T2-LV and atrophy measures, especially with normalised GM and central atrophy measures (data not shown). When partial correlation, adjusted for age at baseline MRI scan, was performed, almost all correlations increased in their magnitude.

Second, we explored a correlation between the absolute changes of T2-LV for 0–24 and 0–60 months and atrophy markers at months 24 and 60 (table 5). No significant correlation was found at 24 months, except for the 3VW (r = 0.152, p = 0.048). Nevertheless, T2-LV change at months 0–24 correlated with GMV (r = 0.32, p = 0.034), PGV (r = 0.34, p = 0.027) and 3VW (r = 0.33, p = 0.029) at month 60. Additionally, the correlation between absolute change of T2-LV (0–24 and 0–60 months) and median percentage changes of atrophy markers (0–24 and 0–60 months) was performed. The change of T2-LV at 0–24 months correlated with PBVC for months 0–60 (r = 0.36, p = 0.018).

Finally, various stepwise linear regression analyses models were performed, including age as covariate, and NBV, NGMV, NWMV, NPGV, 3WV and LVV or their median percentage changes, including PBVC (0–24 and 0–60 months), as independent variables and T2-LV as the dependent variable. T2-LV data were normalised using a cube root transformation of values prior to performing stepwise linear regression analyses. At baseline and 24 months, 3VW was the best predictor of T2-LV (r² = 0.24 and 0.25, p<0.001), whereas at 5 years it was the NBV (r² = 0.24, p = 0.001). At 5 years, the median percentage change in WMV was the only variable that predicted the accumulation of T2-LV (r² = 0.18, p = 0.019). Using the same regression model, we explored which MRI measure predicts the evolution of GM atrophy over 2 and 5 years. At 2 years, the progressions of 3VW (r² = 0.07, p = 0.014) and WMV (r² = 0.05, p = 0.04) correlated with evolution of GM atrophy, whereas at 5 years, progression of LVV (r² = 0.31, p = 0.001) was the only variable retained in the model.

DISCUSSION

To the best of our knowledge, this study represents the largest and longest follow-up study reported to date on the evolution of different tissue compartment atrophy measures in a homogenous group of highly active, early relapsing-remitting patients with MS treated with IFNβ-1α 30 µg IM once weekly alone or in combination with Azathioprine and/or prednisone. In addition, a group of 27 NCs was prospectively collected in order to compare, over a 2-year period, the evolution of different MRI indices between the two study groups. The study confirmed different evolution patterns of tissue compartment atrophy measures in patients with MS, with a faster decline in cortical and deep GM, and periventricular WM regions over periods of 2 and 5 years. In addition, central and GM atrophy measures evolved more rapidly in patients with MS over 2 years, when compared with NCs.

Evolution of different MRI measures over 2 years in patients and normal controls

The MRI comparison between patients with MS and NCs, despite an older NC group (30.7 vs 36.6 years), showed a significantly lower amount of brain tissue (NBV and PGV) in patients with MS at baseline, which is indicative of a destructive process occurring in the whole brain and in the GM compartment. These results agree with recent longitudinal studies on clinically isolated syndromes16 and relapsing-remitting MS,17 18 which showed, during an observation period of 9 months to 3 years, that increasing GM, and not WM, atrophy is most responsible for the progression of whole brain atrophy. Conversely, no significant changes (except for a borderline significant change in PGV) were found in the NC group over 2 years (fig 1). When comparing within-percentage MRI change differences between patients with MS and NCs over a 2-year period, the largest differences in effect were observed in central, WM and cortical regions. These results are supported by logistic regression analysis. An important finding in this study is the effect of ageing on GM and WM changes in patients with MS and NCs. The results from NC studies showed that GM volume increases linearly from early (mean age of 26 months) to later childhood (6–9 years)23 24 and decreases linearly thereafter throughout adult life.25 38 There is a large increase in volume of WM compartment from infancy to adolescence; thereafter, it increases in a slower manner and reaches a plateau by the fourth decade, and then decreases until later life.38 In our NC group, we did not discover a significant change in any of the atrophy parameters followed over 2 years, but a trend for higher loss in PGV (p = 0.032) does support the concept of an ageing effect on GM atrophy evolution. This suggests that the loss of GM tissue observed in patients with MS is not only caused by the disease-related MS process, but is also due to the effect of ageing on the GM compartment. In fact, the evolution of GM atrophy was higher in patients with MS compared with NCs. Although, over a 2-year period, the patients with MS increased more in WMV than NCs, this difference was not significant. This suggests that WMV loss in patients with MS might be masked by inflammatory processes that simultaneously occur in the CNS and increased variability in the measurement of true WMV changes.

In the present study, we also examined the evolution of central atrophy measures (LVV and 3VW) between patients with MS and NCs. Over a 2-year period, the LVV progression was 30.2 times higher (+16.6% vs +0.55%) in patients with MS compared with NCs. In addition, enlargement of the third ventricle was significantly higher than that in the NC group (+9.3 vs 0%). This supports the idea that extensive periventricular WM and deep GM pathology contribute to selective central atrophy development in patients with MS from the earliest phases of the disease. Therefore, LVV and 3VW should be considered as sensitive outcomes in early RR MS clinical trials.14 21

Evolution of different MRI measures over 5 years in patients with MS

At 5 years, the percentage change of SIENAX-derived measures in all study participants was −4.2% for PBVC, −6.2% for GMV, −5.8% for PGV and −0.5% for WMV (table 3, fig 2 upper raw). The median percentage change over 5 years was +68.7% for LVV and +4% for 3VW. These data confirmed a much faster decline in GM atrophy measures, when compared with whole brain or WM atrophy (table 3). This information may have an important implication on the design of future MS studies in which atrophy will be the primary endpoint. Probably, lower sample size and shorter follow-up may be needed to show significant changes over time in patients with early relapsing-remitting MS when GMV and PGV changes are primary endpoints of the study. All patients in this 2-year analysis were on active disease-modifying therapy (mono- or combination therapy). Therefore, it can be hypothesised that GM atrophy evolves at an even faster rate in the non-treated MS population39 than shown in this study. At this time, natural history studies on GM and WM atrophy in the early relapsing-remitting MS population are rare or impossible to conduct. Useful information on the evolution of GM and WM atrophy over a short time period was achieved in a recent 9-month follow-up of the placebo arm participating in the European-Canadian glatiramer acetate study.18 In that study, the time–trend analysis revealed a progressive decrease in GM volume with an estimated mean percentage change per month of 0.3%, whereas WM volume did not change significantly over the study period. In the present study, the GMV decreased by 1.4% over the first year and by 2.3% over the 24-month period, indicating that treatment slowed down the rate of GM atrophy progression.39

Several possible pathogenic mechanisms could explain the different dynamics of GM and WM atrophy evolution in patients with MS over 2–5 years. The ongoing inflammation process may mask real changes in the WM. In our study, we showed modest-to-strong correlations at all study time points between T2-LV and GMV, but not T2-LV and WMV. These data support the hypothesis that the accumulating volume of T2 lesions in the early relapsing-remitting phase of the disease replaces real WM tissue loss.

Second, advanced GM pathology may be indirectly caused by a combination of progressive accumulation of cortical and deep GM lesions15 40 41 and cumulative neuronal damage due to retrograde Wallerian degeneration, with probable delay between focal lesion formation and GM atrophy development. Our data support the delay between pathological changes occurring in WM and GM. The regression analysis showed that the only variable that predicted the accumulation of T2-LV over 5 years was the median percentage change in WMV, whereas at baseline and 2 years, 3VW and NBV were the best predictors. The delay between focal lesion genesis and GM atrophy development in this study may be explained by the lower rate of secondary Wallerian degeneration in the early phases of the disease, or by the fact that these processes are truly independent.42 On the other hand, GM atrophy was predicted at 2 years by the development of deep GM atrophy (3VW)21 43 and progression of WM atrophy, whereas at 5 years the LVV was the best predictor of GM atrophy development. These findings may indicate that separate neurodegeneration and inflammatory processes are ongoing simultaneously in WM and GM.

Advantages and limits of the study

It is worth mentioning that, in the present study, advanced, fully automated MRI image analysis software (SIENA/X) was applied to a quality MRI dataset acquired on a 1.5T scanner with 3D T1-weighted SPGR images of 1 mm slice thickness. It is important to emphasise that all of the patients were on the same medication for at least 2 years. The results of the original trial showed no difference between the three treatment groups, except at 2 years, when the change from baseline in T2-LV favoured the triple-agent combination therapy versus IFNβ-1α monotherapy.27 28 This allowed us to combine the three treatment groups in one cohort and analyse them as a whole group in this study. The decreasing numbers of patients at months 36, 48 and 60, and switching to different treatments could have influenced the evolution of MRI measures. Nevertheless, the subanalysis with only 5-year serial clinical and MRI completers confirmed similar trends in the evolution of atrophy and lesion measures. For purposes of statistical analysis, it was impossible to account for each treatment change, as the number of comparisons would be too high and the sample size too small to yield any meaningful results. This is a common problem in long-term follow-up studies of original pivotal trials.44 45 Nevertheless, most of the relapsing-remitting patients with MS receive or switch disease-modifying therapy in the first couple of years of their disease. Additionally, any direct comparison between SIENA (PBVC) and SIENAX (GMV, WMV, PGV and their percentage changes) measures has to be made carefully because they are methodologically different from each other. Whereas SIENA is a voxel-wise technique tailored for longitudinal studies with the possibility for direct comparison of the raw scans, SIENAX is a cross-sectional measure.46 In order to minimise this effect, we used individual absolute changes rather than the normalised ones. Finally, we did not have a Gd sequence in this study, which did not allow us to compare active lesions with the development of atrophy in different tissue compartments. No clinical results are provided in this paper as it focused only on the evolution of different MRI measures, especially atrophy indices. Detailed analysis of the evolution of clinical and MRI parameters is currently under way.

Conclusions

In conclusion, our study confirmed different patterns of GM, WM and central atrophy evolution over 2 and 5 years in patients with MS and NCs. Measurement of GM and central atrophy provides important additional information about disease progression in early relapsing-remitting patients with MS.