Background Conventional MRI lesion measures modestly predict long term disability in some clinically isolated syndrome (CIS) studies. Brain atrophy suggests neuroaxonal loss in multiple sclerosis (MS) with the potential to reflect disease progression to a greater extent than lesion measures.
Objective To investigate whether brain atrophy and lesion load, during the first year in patients presenting with CIS, independently predict clinical outcome (development of MS and disability at 6 years).
Methods 99 patients presenting with CIS were included in the study. T1 gadolinium enhanced and T2 weighted brain MRI was acquired at baseline and approximately 1 year later. Percentage brain atrophy rate between baseline and follow-up scans was analysed using SIENA.
Results Mean annual brain atrophy rates were −0.38% for all patients, −0.50% in patients who had developed MS at 6 years and −0.26% in those who had not. Brain atrophy rate (p = 0.005) and baseline T2 lesion load (p<0.001) were independent predictors of clinically definite MS. While brain atrophy rate was a predictor of Expanded Disability Status Scale (EDSS) score in a univariate analysis, only 1 year T2 lesion load change (p = 0.007) and baseline gadolinium enhancing lesion number (p = 0.03) were independent predictors of EDSS score at the 6 year follow-up. T1 lesion load was the only MRI parameter which predicted Multiple Sclerosis Functional Composite score at the 6 year follow-up.
Conclusions The findings confirm that brain atrophy occurs during the earliest phases of MS and suggest that 1 year longitudinal measures of MRI change, if considered together with baseline MRI variables, might help to predict clinical status 6 years after the first demyelinating event in CIS patients, better than measurements such as lesion or brain volumes on baseline MRI alone.
Statistics from Altmetric.com
Multiple sclerosis (MS) is a chronic disorder of the CNS, characterised pathologically by multifocal areas of inflammatory demyelination that evolve in the CNS over time (multiphasic), and clinically by a variable course, although most patients develop significant locomotor disability 15–30 years after onset.1–4 In approximately 85% of patients who develop MS, clinical onset is characterised by an acute or subacute episode of neurological disturbance (eg, optic neuritis or an isolated brainstem or partial spinal cord syndrome) due to a single lesion within the CNS, and is known as a clinically isolated syndrome (CIS).2 3
MRI is a well established investigatory tool with a diagnostic and prognostic role in MS, and is increasingly used to provide outcome measures in therapeutic trials.5–13 Follow-up studies ranging from seven to 20 years investigating the prognostic role of MRI in CIS patients2 have shown that clinically definite MS (CDMS) develops in approximately 60–80% of CIS patients with abnormal MRI but in only approximately 10–20% who have normal MRI.14–16 However, conventional MRI measures have correlated less consistently with future disability.14 17 18
Brain tissue loss (atrophy) is thought to reflect the underlying permanent neuroaxonal damage in MS, and hence is seen as a plausible measure of disease progression. Brain atrophy occurs in CIS patients19–21 but few studies have investigated the relative predictive values of early brain atrophy rates and lesion load measures for longer term clinical outcomes—the development of MS and disability.
The utility of longitudinal rather than single time point MRI measures as predictors of future clinical outcome also warrants investigation. A recent 20 year follow-up study from CIS onset demonstrated significant relationships between the rate of lesion load accumulation and clinical evolution17 but sampling of MRI data in this study was at relatively long (approximately 5 yearly) intervals.
The aim of the present study was to investigate, in a cohort of CIS patients who were followed-up at regular intervals with clinical and MRI assessments, whether brain atrophy rate or lesion load measures obtained at baseline and over the first year following presentation were independently predictive of MS diagnosis and disability 6 years after the first demyelinating event suggestive of MS. In particular, we wanted to investigate whether additional predictive information is provided by: (i) measuring brain atrophy in addition to lesion load measures and (ii) including longitudinal MRI measures in addition to single baseline time point measures. This study included a cohort of patients with CIS with longer clinical follow-up than many previous studies (6 years after the first demyelinating event).
Subjects and methods
Ninety-nine patients (36 men, 63 women) were identified from a longitudinal clinical and MRI research study of subjects presenting with a CIS suggestive of MS, being undertaken at the Institute of Neurology, University College London, London, UK, between 1995 and 2007. The original prospective study included 178 patients but only patients who had both a baseline and 1 year scan and had been followed-up clinically at 6 years were included in the present retrospective analysis. The group of 99 patients selected was not significantly different from the whole cohort with regard to baseline characteristics (gender, age, presenting symptoms, lesion loads). A CIS was defined as a single event of acute onset in the central nervous system suggestive of demyelination—for example, unilateral optic neuritis, brainstem and partial spinal cord syndromes. Ethics approval for the longitudinal study was obtained from the Joint Medical Ethics Committee of the Institute of Neurology and National Hospital for Neurology and Neurosurgery, University College London, and the Moorfields Eye Hospital Ethics Committee. In addition, written informed consent was obtained from all study participants and all patients underwent appropriate clinical and laboratory investigations to exclude alternative diagnoses.
At baseline, 76 patients presented with optic neuritis, 13 with a brainstem syndrome, nine with a spinal cord syndrome and one with a visual field defect due to an optic tract lesion. Age at baseline ranged from 18 to 51 years (median 32 years). All patients had a clinical follow-up approximately 6 years from baseline (median 6.5 years; range 3.6–10.2) when they were evaluated for a diagnosis of CDMS (progression from CIS).7 22 If patients were diagnosed with MS, the time that they were diagnosed was also determined (months from baseline). Patients had additionally been assessed at intermediate time points (3 months, 1 year and 3 years from baseline) as part of the prospective study, which aided in this task. Median time to conversion to CDMS was 12 months (range 2–120), with 21 patients converting to MS before the 1 year scan. Only two of these patients started disease modifying treatment before the 1 year MRI scan. The Expanded Disability Status Scale (EDSS)23 and the Multiple Sclerosis Functional Composite (MSFC)24 were also recorded at the 6 year clinical follow-up in order to assess disability. While all patients were scored with the EDSS only 64 patients were scored with the MSFC.
All patients had T1 and T2 weighted brain MRI at baseline and approximately 1 year later (median 12 months; SD 3.06; range 10–25). Baseline MRI was performed within 12 weeks of symptom onset in patients with CIS (median 4.5 weeks; SD 2.9; range 1–12). MRI was acquired on a 1.5 T GE Signa Horizon Echospeed scanner (General Electric Medical Systems, Milwaukee, Wisconsin, USA). Axial T1 weighted images were acquired post-gadolinium-DTPA (Gd) (0.1 mmol/kg) using a conventional spin echo sequence with acquisition parameters TR = 600 ms, TE = 17 ms, matrix 256×256, flip angle 90° and FOV 240×180 mm, resulting in 46 contiguous 3 mm thick slices. In addition, axial dual echo PD and T2 weighted images were acquired using a fast spin echo sequence with acquisition parameters TR = 3200 ms, TE = 19/95 ms, matrix 256×256, flip angle 90° and FOV 240×180 mm, resulting in 46 contiguous 3 mm thick slices.
MR data analysis
Images were transferred to a SUN workstation (Sun Microsystems Inc, Santa Clara, California, USA) for brain atrophy analysis. Percentage brain volume change (PBVC) between baseline and 1 year was estimated from T1 weighted images with SIENA (Structural Image Evaluation using Normalisation of Atrophy25), part of FSL (www.fmrib.ox.ac.uk/fsl). Median control scan–rescan error in SIENA measurements has been reported to be 0.15%.25 SIENA starts by extracting brain and skull images from the two time point whole head input data.26 The two brain images are then aligned to each other27 28 (using the skull images to constrain the registration scaling); both brain images are resampled into the space halfway between the two. Next, tissue-type segmentation is carried out29 in order to find brain/non-brain edge points, and then perpendicular edge displacement (between the two time points) is estimated at these edge points. Finally, the mean edge displacement is converted into a (global) estimate of PBVC between the two time points. T1 weighted images did not always include full coverage of the brain, sometimes excluding the vertex. Thus in order to correct for small differences in the portion of the brain imaged, a standard space based limit of +60 along the z axis was used for each image pair.
In the cohort analysed in this study, the initial automated brain and skull extraction was optimised by using the robust brain centre estimation function. Subsequently, the brain extractions were checked, and manual editing was performed when non-brain regions were still included around the optic nerves, eyes and temporal lobes. Manual editing was performed using the DispImage image analysis package (D Plummer, University College London Hospitals NHS Trust, UK)30 by a single operator (MDF) who was blinded to subject identity. Thirty-five patients required manual editing, of whom 24 required manual editing on both baseline and repeat scans. The final SIENA output was assessed for all subjects by another expert operator (VMA) who was blinded to subject identity, and who rated scan pairs according to the quality of the scans and registration of images.
In addition to brain atrophy rate, T1 and T2 weighted lesion volumes at baseline were calculated, and baseline Gd enhancing lesion number was determined. T1 hypointense lesion count included both enhancing and non-enhancing lesions. Follow-up brain scans were also utilised to determine T2 weighted lesion volume at 1 year and thus to calculate the 1 year change in T2 lesion volume. All lesions were outlined in DispImage using a semiautomated local thresholding technique or, where necessary, by manual outlining using a mouse driven cursor. A computer program summed all of the individual lesion volumes (calculated as the surface area of each lesion multiplied by the slice thickness (3 mm)), and total T2 hyperintense and T1 hypointense lesion volumes were generated.
Data were analysed using Stata V.9.2 (Stata Corp, College Station, Texas, USA). All PBVC measures were annualised prior to statistical analysis. Differences in variables between patients who had developed MS by the 6 year follow-up and those that remained CIS were examined using unpaired t tests and Fisher’s exact test. In addition, Cox regression analysis of time to MS diagnosis was used to determine which MRI variables were independently related to CDMS development.
Ordinal logistic regression analysis was performed to determine which MRI variables were related to disability at the 6 year time point, as assessed by the EDSS. EDSS measurements were categorised into five groups as follows: EDSS = 0, EDSS = 1, EDSS = 1.5–2, EDSS = 2.5–3 and EDSS >3. These groups were chosen as they were considered to represent different levels of disability. Linear regression analysis was used to determine which MRI variables were related to disability at the 6 year follow-up, as assessed by the MSFC.
The following MRI variables were considered for inclusion as potential predictors in all regression analyses: baseline T1 lesion volume, baseline T2 lesion volume, baseline Gd enhancing lesion number and T2 lesion volume change and PBVC between baseline and 1 year. Age and gender were also included as covariates. A manual backwards selection procedure was used to determine which variables were included in the final regression models.
Seven scan pairs were considered to be of unacceptable quality due to motion artefacts on either the baseline or repeat scans, and were thus excluded from the statistical analysis.
Mean baseline to 1 year brain atrophy rate for the remaining group of 92 CIS patients was found to be −0.38%/year (SD 0.55). Mean T1 and T2 lesion volumes at baseline were, respectively, 337.6 mm3 (SD 868) and 2023.5 mm3 (SD 4015). Mean number of Gd enhancing lesions in the baseline scan was 1.54 (SD 3.7). Thirty-one patients had at least one Gd enhancing lesion while 61 patients had no Gd enhancing lesions. The mean increase in T2 lesion load between baseline and 1 year was 1232.3 mm3 (SD 4399).
Relationship between MRI variables and conversion to CDMS
At the 6 year follow-up, 49 patients (53.3%) had developed clinically definite MS while 43 subjects (46.7%) remained CIS. Median EDSS was 1 (range 0–8.5) and only 10/92 (11%) had an EDSS >3, indicating significant disability. Of the 49 patients diagnosed with MS, 45 had an abnormal T2 weighted MRI scan at baseline while of the 43 remaining CIS, only 23 had an abnormal T2 weighted MRI scan at baseline (Fisher’s exact test, p<0.001). Mean brain atrophy rates were significantly different between patients who had developed MS and those who had not (respectively, −0.50%/year (SD 0.63) vs −0.26%/year (SD 0.42), mean difference −0.24%/year (95% CI 0.02 to 0.46); p = 0.035) (figure 1). Considering only patients who had an abnormal MRI scan at baseline, mean atrophy rates were −0.23%/year (SD 0.4) in patients who had not converted to MS and −0.52%/year (SD 0.63) in patients who had converted to MS (p = 0.04, mean difference −0.29%/year (95% CI −0.01 to −0.57)).
Cox regression analysis gave brain atrophy rate and baseline T2 lesion volume as the only independent risk factors for MS diagnosis. The risk (hazard) of diagnosis with MS was reduced by 33.8% (95% CI 12.0% to 50.2%; p = 0.005) for each SD lower (ie, less negative) atrophy rate and increased by 45.6% (95% CI 18.2% to 79.4%; p<0.001) for each SD higher baseline T2 lesion volume.
Relationship between MRI variables and disability measures at 6 years
All patients had EDSS measurement at the 6 year follow-up while only 64 patients had been assessed on all of the individual components of the MSFC. Although univariate ordinal logistic regression gave baseline to 1 year PBVC as a significant predictor of EDSS score at 6 years, with 41% (p = 0.008) reduced odds of higher EDSS per SD lower atrophy rate (figure 2), multiple ordinal logistic regression showed that PBVC did not predict EDSS independently of other MRI measures; baseline to 1 year change in T2 lesion load and baseline Gd enhancing lesion number were the only independent predictors of EDSS score at 6 years. The odds ratio of having a higher EDSS score was 2.21 for each SD higher T2 lesion volume change over the first year (95% CI 1.23 to 3.97; p = 0.007) and was 1.73 for each SD higher Gd enhancing lesion number at baseline (95% CI 1.05 to 2.84; p = 0.030). When patients were divided into only three subgroups according to the degree of disability (“none” (EDSS 0, n = 13), “minimal” (EDSS >0 and <3, n = 64) and “significant” (EDSS ≥3, n = 15)), univariate regression analysis showed that PBVC remained a significant predictor of disability (p = 0.005). In the subset of 64 patients who were scored with the MSFC, univariate regression analysis showed that baseline to 1 year PBVC was not a significant predictor of either MSFC at 6 years (p = 0.309) or its single components (25 foot timed walk, p = 0.340; nine hole peg test, p = 0.567; Paced Auditory Serial Addition Test, p = 0.931). However, in this same subgroup of 64 patients, brain atrophy rate was no longer a significant predictor of EDSS score at the 6 year follow-up (p = 0.08). T1 lesion load was the only MRI parameter which was found to be predictive of MSFC score at the 6 year follow-up (p = 0.005).
Our study confirms that brain atrophy occurs early in patients with MS, and that it is an independent prognostic factor for the development of CDMS in patients with CIS. Although brain atrophy rate was a significant predictor of disability after 6 years in patients presenting with a CIS, this was not independent of other MRI measures taken into account. However, another longitudinal MRI measure, the change over 1 year in T2 lesion volume, was an independent predictor of subsequent disability in this cohort of CIS patients, together with the baseline number of Gd enhancing lesions, suggesting that longitudinal measures of MRI change, if considered together with measurements from baseline MRI, may better predict clinical outcome for patients presenting with a CIS.
Development of clinically definite MS
Our results confirm previous studies showing that brain atrophy rates are higher in those CIS patients who subsequently develop MS compared with those who remain CIS.14 20 21 31 32 In the placebo arm of the ETOMS (Early Treatment of Multiple Sclerosis) trial, mean PBVC in over 100 patients with CIS was found to be −0.83% during the first year.33 In the same study a difference in median annual PBVC was found between patients who developed CDMS versus patients who did not (−0.92% and −0.56%, respectively).33 Although the atrophy rates in that study are slightly higher than those found in our study, one possible explanation is that all subjects in the ETOMS trial had a positive baseline brain MRI scan (at least four T2 lesions or at least three T2 lesions if at least one was infratentorial or enhancing after Gd injection) compared with our unselected cohort of CIS patients. However, analysis of only those patients with four or more lesions in the current study gave atrophy rates that were still noticeably smaller than those in the ETOMS study (data not shown). This may be the result of different image acquisition methods and different standard space limits for the brain in the SIENA analysis.
A new finding in the current study was that brain atrophy emerged as an independent predictor of MS development (in addition to T2 lesions). Thus inclusion of both brain atrophy and lesion load measures in therapeutic trials that aim to prevent the evolution from a CIS to CDMS may provide complementary and relevant information. Future studies of atrophy following CIS might usefully investigate changes in grey matter specifically as there is evidence that most of the brain atrophy seen following a CIS and in early relapsing remitting MS occurs in grey matter.21 34 35
Development of disability
To date, few studies have investigated the relative predictive value of lesion load and brain atrophy measures for disability in CIS patients with a relatively long clinical follow-up. Therefore, a key focus of our study was to investigate the relationship of MRI measures, including atrophy rate, with subsequent disability, measured 6 years after the first clinical episode. In the current study, despite the univariate association of brain atrophy rate with EDSS score, when considering other MRI lesion measures in addition, it did not predict disability independently from these variables. The changes over 1 year in T2 lesion load and baseline Gd enhancing lesion number were found to be the only independent predictors of EDSS score at 6 years in our cohort of CIS patients. This result suggests that longitudinal but still relatively early (1 year) measures of MRI lesion load change, representing a “dynamic” parameter of disease evolution, might predict disease progression and disability better than single baseline lesion measures alone.
A previous study in patients with established relapsing remitting MS reported that brain atrophy rate over 2 years was a predictor of disability after 8 years.36 It may be that over time, as disease duration and disability increase and pathological changes become more widespread and complex, the predictive value of different imaging measures will change. Few patients had acquired significant disability (with an EDSS of more than 3) at the 6 year follow-up in our study. A longer period of observing atrophy following a CIS (eg, 2 or 3 years) may also be informative. Further follow-up of our CIS group should help to elucidate this issue.
We did not find any association between brain atrophy rate and either total MSFC score or its single components measuring leg, arm and cognitive functions. However, only a relatively small number of patients had MSFC measurements at the 6 year time point, and the smaller sample size might have negatively influenced the research of a statistical association between the two parameters. This conclusion is also supported by the fact that in the same subgroup of patients, the association between brain atrophy and EDSS score was not significant. These somewhat inconsistent findings also highlight the complexity of factors to consider when investigating the relationship between a global brain atrophy measure and permanent disability in MS. For example, disability evident on clinical assessment may be limited when neuroaxonal damage causing atrophy occurs in clinically silent areas or does not reach a clinically relevant threshold; and cortical reorganisation may also reduce disability in spite of neuroaxonal loss. It is also worth noting that a crucial role in the determination of disability might be played by the presence of spinal cord lesions37 and spinal cord atrophy.38 Thus the study of the association between brain atrophy and disability might be further complicated by the presence of spinal cord damage.
Several limitations of the study must be noted. Firstly, due to the nature of prospective studies, not all patients were followed-up clinically at exactly 6 years or had their repeat MRI scan at exactly 1 year. However, this affected only a minority of patients and we believe that this should not detract from our finding of an association of brain atrophy rate (all brain atrophy measures were corrected for scan interval) with the development of MS at some point in the future. Secondly, some patients (n = 21) were included in the analysis that converted to MS before the 1 year repeat scan. However, we feel that exclusion of these patients would have biased the results to those patients who had a less severe disease course. Any bias that may be introduced by including these patients in the analysis is extremely unlikely as both brain atrophy and T2 lesion load plausibly increase at a fairly steady rate during the first year, rather than suddenly changing after conversion. Thirdly, the effect of disease modifying treatments in limiting the evolution of brain atrophy has been demonstrated39 40 while steroid treatment has been shown to decrease brain volume after administration.41 Although two patients started treatment with a disease modifying agent before the 1 year MRI scan, this is unlikely to have influenced brain atrophy during this time, or significantly impact on the measurement of brain atrophy rate in the whole cohort of CIS patients. Due to the nature of the study, patients were scanned as close to the initial CIS as possible, and some patients may have been on steroid treatment. This may have resulted in a pseudo increase in brain volume at follow-up. Indeed, 21 patients were found to have an increase in brain volume after 1 year. A pseudo increase in volume may also have been due to the presence of Gd enhancing lesions at the 1 year follow-up. Conversely, pseudoatrophy may also have been detected due to the presence of Gd enhancing lesions at baseline, and these factors should always be taken into account when interpreting the results of brain atrophy studies.
In summary, our study indicates that brain atrophy is a feature of CIS patients who develop MS. Measurement of brain atrophy should be a useful supportive endpoint in trials of disease modifying agents aimed at preventing CDMS development and disability in patients with CIS suggestive of MS. The study also suggests that longitudinal measures of MRI change, such as brain atrophy rate and change in T2 lesion load volume, if considered together with baseline MRI variables, might improve the prediction of future clinical status in CIS patients better than baseline, single time point MRI measurements alone.
Funding VMA is supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland.
Competing interests MDF received a travel grant from Biogen to attend an international congress. VMA, DRA, JKS and GTP have no competing interests. AJT receives honoraria from Novartis (PPMS Advisory Group) and Eisai (Advisory Board, Clinical Neuroscience), and has received honorarium and support for travel from Serono Symposia (invited lecturer). DHM has received grant support from Biogen Idec, Elan, Bayer Schering, Novartis and GlaxoSmithKline for performance of MRI analyses in clinical trials, as well as honoraria for advisory or consultancy work, lectures and related travel expenses from Aventis, Biogen Idec, Bristol Myers Squibb, GlaxoSmithKline, Bayer Schering, Merck Serono, UCB Pharma and Wyeth.
Ethics approval Ethics approval was obtained from the Joint Medical Ethics Committee of the Institute of Neurology and National Hospital for Neurology and Neurosurgery, University College London and Moorfields Eye Hospital Ethics Committee.
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.