Article Text

Download PDFPDF

Early central atrophy rate predicts 5 year clinical outcome in multiple sclerosis
  1. Carsten Lukas1,2,
  2. Arjan Minneboo2,
  3. Vincent de Groot3,
  4. Bastiaan Moraal2,
  5. Dirk L Knol4,
  6. Chris H Polman5,
  7. Frederik Barkhof2,
  8. Hugo Vrenken2,6
  1. 1Department of Diagnostic and Interventional Radiology and Nuclear Medicine, St Josef Hospital, Ruhr University Bochum, Bochum, Germany
  2. 2Department of Radiology, MS Centre Amsterdam, VU University Medical Centre, Amsterdam, The Netherlands
  3. 3Department of Rehabilitation Medicine, MS Centre Amsterdam, VU University Medical Centre, Amsterdam, The Netherlands
  4. 4Department of Clinical Epidemiology and Biostatistics, MS Centre Amsterdam, VU University Medical Centre, Amsterdam, The Netherlands
  5. 5Department of Neurology, MS Centre Amsterdam, VU University Medical Centre, Amsterdam, The Netherlands
  6. 6Department of Physics and Medical Technology, MS Centre Amsterdam, VU University Medical Centre, Amsterdam, The Netherlands
  1. Correspondence to Dr C Lukas, Department of Diagnostic and Interventional Radiology and Nuclear Medicine, St Josef Hospital, Ruhr University Bochum, Gudrunstr 56, Bochum 44791, Germany; carsten.lukas{at}


Objective To examine the predictive value of central atrophy in early multiple sclerosis (MS) patients, for medium term clinical outcome.

Methods In 54 patients with recently diagnosed MS, clinical and MRI parameters were obtained at baseline, and after 2 and 5.5 years of follow-up. In addition to conventional MRI parameters and the annualised percentage brain volume change (aPBVC), the annualised percentage ventricular volume change (aPVVC) was quantified. Main outcome measure was disease progression, defined by an increase in Expanded Disability Status Scale of ≥1 after 5.5 years.

Results Disease progression occurred in 29 patients. aPVVC within the first two years was significantly higher in these progressing patients (median 4.76%; IQR 3.05–9.17) compared with stable patients (median 3.23%; IQR −0.1–6.02) (p=0.02). A logistic regression model selected aPVVC within the first 2 years as the only MRI marker predicting progression after 5.5 years (OR 1.17, 95% CI 1.02 to 1.35). When entering all MRI and clinical markers, again aPVVC within the first 2 years was the only MRI marker selected. While aPVVC was correlated between the two consecutive time intervals (ρ=0.41, p<0.01), aPBVC was not. Furthermore, baseline T2 lesion load and gadolinium enhancing lesion load were correlated with aPVVC in the second time interval (2–5.5 years) but not with aPBVC.

Conclusion The rate of ventricular enlargement seems to be even more strongly predictive of disease progression after medium term follow-up than whole brain atrophy rate, and also outperforms lesion measures. Central atrophy rate could therefore be an important prognostic marker, especially in the early stages of MS.

  • MRI
View Full Text

Statistics from


Defining prognostic markers for clinical disease progression in multiple sclerosis (MS) still remains challenging. Besides clinical predictors, such as early relapse rate, several paraclinical markers have been investigated. Widely used are numbers or volumes of T2 lesions or gadolinium enhancing lesions measured using conventional brain MRI. However, these conventional measures have shown only mild correlations with subsequent progression of clinical disability, as expressed by the Expanded Disability Status Scale (EDSS).1–4 Compared with lesion loads, non-conventional MRI measures, such as brain atrophy, are superior in terms of predicting clinical deterioration in MS and therefore are important tools to monitor disease progression.5–8 Atrophic loss of brain parenchyma is accompanied by a corresponding increase in CSF spaces, visible peripherally as widening of sulci and centrally as widening of the ventricular system, hence the terms ‘enlargement of the ventricular system’ and ‘central atrophy’ are used interchangeably.

Ventricular enlargement in MS has been reported early in a CT study,9 was further studied using MRI and was found to be correlated mainly with neuropsychological functioning.10–12 Lateral ventricle volume and third ventricle width were found to be sensitive measures, clearly distinguishing patients with early MS from controls.13 In a cohort of relapsing–remitting MS patients, lateral ventricle volume was identified as an additional MRI parameter predicting EDSS changes at 5 years of follow-up.14 These findings suggest that ventricular enlargement may have additional prognostic value.

The present study investigated early MS patients during 5.5 years of follow-up, aiming to: (1) examine the predictive value of early central atrophy rate (ie, ventricular enlargement rate) for medium term clinical disability in both relapse and progressive onset MS patients; (2) determine whether central atrophy and global atrophy develop simultaneously; and (3) determine whether early lesion load, as measured on T1 and T2 weighted images, is related to the development of central atrophy.



Fifty-four patients with early MS were recruited from an ongoing natural history study of early MS. Patients had to be aged 16–55 years and within 6 months of diagnosis, according to the Poser criteria for clinically definite MS.15 Disease onset type was classified as either relapsing (n=43) or progressive (n=11). Clinical parameters, including EDSS score, and MR images were obtained at baseline and during follow-up visits after 2 years (median 2.2; IQR 2–2.3) and 5.5 years (median 5.4; IQR 5.4–5.6). Patients were classified according to clinical disease progression, defined by an increase in EDSS score of at least 1 point after 5.5 years of follow-up. Clinical characteristics of the patients are given in table 1. In patients who had to be treated during the study, the following disease modifying agents were used: interferon (n=11) and glatiramer acetate (n=3). Furthermore, patients were treated with short courses of steroids when clinically indicated for acute exacerbations. The study was approved by the institutional medical ethics committee and informed consent was obtained from every patient.

Table 1

Demographic and clinical parameters for the entire patient group, as well as separately for patients who showed clinical disease progression and those who did not

MRI and post-processing

Imaging was performed using a 1.0 T scanner (Magnetom Impact; Siemens, Erlangen, Germany). An identical image protocol was used throughout the duration of the study. Furthermore, during this study only regular maintenance was performed, including replacement of defective parts by identical components. No hardware upgrades were performed. The standardised protocol included oblique–axial, dual echo, spin echo PD/T2 weighted images (TR/TE1/TE2: 2700/45/90 ms) and T1 weighted spin echo images (TR/TE: 700/15 ms), all with 5 mm slice thickness, 25 slices, 10% gap and in-plane pixel size of 1.0×1.0 mm. On the baseline visit, the T1 weighted spin echo sequence was repeated after intravenous injection of gadolinium diethylene triamine pentaacetic acid. Measurement of lesion loads from these images at each time point was performed using semi-automated seed growing software developed inhouse (Show_Images). Lesions were identified by an experienced reader (CL, AM), subdivided into supratentorial and infratentorial lesions, and outlined using local thresholding, yielding T2 lesion load (T2LL), T1 hypointense lesion load (T1LL) and gadolinium enhancing lesion load (GdLL).

Brain and ventricular volume changes were calculated from the pre-contrast T1 weighted images. Percentage brain volume change (PBVC) was calculated between each pair of time points (baseline, 2 years, 5.5 years) using an automated technique (SIENA).16 17 Ventricular enlargement was calculated using SIENA, adapted similarly as described by Gunter and colleagues18 and expressed as the percentage ventricular volume change (PVVC). Briefly, operations based on a mask drawn in the MNI-152 standard space were used to select only the brain/non-brain edge points on the ventricular edges; the mean edge displacement of these ventricular edge points was calculated and converted to the PVVC value. In addition, the automated brain extraction performed by SIENA was manually edited to remove remaining non-brain tissue.

To account for small differences in the duration of follow-up between patients, PBVC and PVVC were divided by the time interval in years, yielding the annualised measures aPBVC and aPVVC, respectively.

Statistical analysis

Statistical analyses were performed using SPSS 14 (SPSS, Chicago, Illinois, USA). Because most data were not normally distributed, medians and IQR were used as descriptives. Correlations between clinical and MRI parameters, and among MRI parameters, were assessed using Spearman's rank correlation coefficient, with two tailed p values. The Mann–Whitney U test was used to test differences in aPVVC between disease onset types. Disease progression was defined as an EDSS change ≥1 after 5.5 years. Differences between patients with and without disease progression for both clinical and MRI parameters were assessed using the Mann–Whitney U test or Pearson's χ2 test.

To assess which measures predict clinical progression, logistic regression with forward stepwise selection was used. Three models were investigated: a clinical, MRI and combined clinical–MRI model. The dependent variable was absence or presence of disease progression at 5.5 years. In the clinical model, the variables age, disease duration, onset type, sex, baseline EDSS score and presence of treatment were entered as independent variables. In the MRI model, the independent variables were T1LL, T2LL, and GdLL at baseline, aPBVC and aPVVC within the first 2 years and the change in T1LL and T2LL (∆T1LL and ∆T2LL) during the first 2 years. To corroborate the findings of the prediction model, we also assessed the correlations of these MRI variables with the annual EDSS change, measured over the full 5.5 year period. Finally, the added value of MRI parameters over clinical parameters was tested in a combined model, including all MRI and clinical parameters as independent variables. The final analysis was also performed for the relapse onset-type patients only.


Clinical data

At study entry, disease duration was short (median 1.4 years, IQR 0.5–3.6). Baseline disability was low: median EDSS score was 2.0 (IQR 2.0–3.0), which increased during follow-up to 3.25 (IQR 2.0–4.75). Forty-three patients (79.6%) had relapsing onset disease while 11 patients (20.4%) had progressive onset. During the 5.5 year follow-up period, 29 patients (53.7%) showed clinical progression (∆EDSS ≥1). Median EDSS at baseline did not differ significantly between patients with or without subsequent disease progression. As expected, male and progressive onset patients had a higher likelihood of progressing (table 1).

Conventional MRI parameters

Total lesion loads at baseline were low: median T2LL was 3.07 cm3 (IQR 1.17–7.24), and median T1LL was 0.25 cm3 (IQR 0–0.58). Twenty-two patients (40.7%) had enhancing lesions at baseline. Conventional MRI parameters did not differ between patients with and without progression, either at baseline or after 2 years (table 2).

Table 2

MRI parameters for the whole patient group, as well as separately for patients who showed clinical disease progression and those who did not

Atrophy measures

Annualised whole brain atrophy rates (aPBVC, table 2) were approximately 0.3–0.5%/year and in agreement with previous findings in early MS.19–21 Median rates of ventricular enlargement (aPVVC, table 2) were of the order of a few per cent per year and also in good agreement with previously published results.14 19 20 Median atrophy rates were of smaller magnitude in the second time interval (2–5.5 years) than in the first time interval (0–2 years): aPBVC −0.32%/year versus −0.44%/year; aPVVC 1.89%/year versus 4.28%/year. Patients with relapse onset MS did not differ significantly from patients with progressive onset MS for aPVVC in either time interval (both p>0.20). Patients with disease progression at 5.5 years had a higher aPVVC within the first 2 years (4.76%/y) compared with patients without progression (3.23%/year, p=0.02) whereas this was not observed for aPBVC (p=0.09). For the second time interval, no significant differences between progressing and non-progressing patients were observed for either aPBVC (p=0.49) or aPVVC (p=0.34) (table 2).

Correlations between different MRI parameters

Table 3 lists correlations between conventional MRI parameters and atrophy measures. T2LL and GdLL at baseline correlated negatively with aPBVC in the first time interval (higher lesion loads are related to faster brain atrophy) and positively with aPVVC for both consecutive time intervals (higher lesion loads are associated with faster ventricular widening). Subdividing T2LL by anatomical location, aPBVC and aPVVC were only correlated with supratentorial T2LL.

Table 3

Correlation between conventional MRI parameters at baseline, and aPBVC and aPVVC for both time intervals

Changes in T1LL and T2LL over the first 2 years were negatively correlated with concomitant whole brain atrophy rate (aPBVC) (∆T1LL: ρ=−0.29, p=0.035; ∆T2LL: ρ=−0.27, p=0.047) but not related to concomitant central atrophy rate (aPVVC) (∆T1LL: ρ=0.16, p=0.26; ∆T2LL: ρ=0.22, p=0.12) (table 3). Considering subsequent atrophy rates yielded different results: while ∆T1LL during the first 2 years was positively correlated with aPVVC during the subsequent time interval (ρ=0.34, p=0.013), that correlation failed to reach significance for subsequent whole brain atrophy (p=0.064) (table 3). ∆T2LL was not correlated with subsequent aPBVC (p=0.37) or with subsequent aPVVC although that was borderline significant (p=0.052).

A negative correlation between aPBVC and aPVVC was found during both time intervals (ρ=−0.57 and ρ=−0.50, both p<0.001) (figure 1). Considering the two consecutive time intervals, aPVVC was strongly correlated between the two intervals (ρ=0.41, p<0.01) while aPBVC was not (ρ=0.04, p=0.76) (figure 2).

Figure 1

Scatterplots of annualised percentage brain volume change (aPBVC) versus annualised percentage ventricular volume change (aPVVC) within both time intervals: (A) 0–2 years, (B) 2–5.5 years. Spearman rho and p values are given.

Figure 2

Scatterplots of (A) annualised percentage brain volume change (aPBVC) and (B) annualised percentage ventricular volume change (aPVVC) in the two consecutive time intervals. Spearman rho and p values are given.

Correlation between MRI and clinical parameters

Correlations between annual changes in EDSS and atrophy rates are summarised in table 4. For the first time interval, a positive correlation between EDSS changes and aPVVC was found (ρ=0.28; p=0.04) (more clinical worsening corresponds to faster ventricular enlargement) as well as a negative correlation for aPBVC (ρ=−0.29; p=0.03) (more clinical worsening corresponds to faster whole brain volume loss). Annual EDSS change over the full 5.5 year follow-up period was correlated with aPVVC during the first time interval (ρ=0.32; p=0.02) but not with aPBVC in the same interval or any other MRI measure.

Table 4

Correlation between annual changes in EDSS, and PBVC and PVVC for both time intervals

Logistic regression analysis

The first logistic regression model used only clinical parameters. Type of onset (OR 12.63, 95% CI 1.48 to 107.57) was selected as the only independent parameter, indicating that progressive onset patients had 12.63 times higher odds of clinical progression than patients with relapsing onset. In the second logistic regression analysis, using only MRI parameters, only aPVVC within the first 2 years (OR 1.17, 95% CI 1.02 to 1.35) was selected. The OR indicates that, for a 1%/year increase in aPVVC, the odds of developing clinical progression become 1.17 times higher.

To investigate whether adding MRI information could improve the clinical model, we constructed a combined clinical–MRI model by entering all clinical and all MRI variables. For the prediction of clinical disease progression, the stepwise procedure then selected type of onset (OR 12.22, 95% CI 1.39 to 107.30) and aPVVC (OR 1.17, 95% CI 1.01 to 1.35), indicating that aPVVC added independent information to the clinical model (table 5).

Table 5

Results of logistic regression with all clinical and MRI parameters, with the presence/absence of clinical disease progression after 5.5 years as the dependent variable

Because most of our patients showed a relapsing onset, and type of onset was the only selected clinical predictor, we repeated the combined clinical–MRI analysis on the subset of relapse onset patients. The logistic regression then selected no clinical predictors but only aPVVC (OR 1.18, 95% CI 1.01 to 1.37).


In this study in early MS patients, the rate of early ventricular enlargement was the strongest MRI predictor for clinical disease progression at 5.5 years, outperforming lesion measures and whole brain atrophy rate in progressive onset as well as in relapse onset MS patients. Although whole brain atrophy rate was correlated with the rate of ventricular enlargement, only the latter was correlated between subsequent time intervals and was related to preceding volume changes in T1 hypointense lesions. These results suggest that the early rate of ventricular enlargement is a promising and biologically plausible prognostic marker in MS.

Measurement of brain atrophy using MRI has emerged as an objective and reliable marker of disease progression in MS and several studies have investigated its prognostic value.5 14 22 These studies have used a range of analysis techniques to measure brain atrophy, including whole brain and regional brain measures. The current study used a fully automated measure of ventricular enlargement to investigate the role of both global and central atrophy rates in predicting clinical disability in early MS patients over a follow-up period of 5.5 years. Our results suggest that the rate of ventricular enlargement surpasses even the whole brain atrophy rate in terms of predicting subsequent progression of clinical disability in the medium term in MS. This relationship was further strengthened by the correlation between aPVVC within the first 2 years and the annual EDSS change over 5.5 years, which was not observed for aPBVC or other MRI variables.

Several reasons might explain the closer relation of central atrophy to clinical disability than to global atrophy. Methodological reasons include the fact that a small absolute loss of brain tissue can produce a large relative increase in the initially small ventricular volume.19 As reflected by our results and those of others, ventricular enlargement is therefore a more dynamic measure (rates of up to 10%/year) than global brain atrophy measures (rates of up to 1%/year).19 20 23 24 Furthermore, the SIENA methodology used in the current study and in previous studies calculates the displacement of the brain edge between time points. Compared with common MRI pixel sizes, the brain edge is much smoother at the ventricles than it is at the cortical grey matter, making the calculation of ventricle edge displacement and thereby the derived measure aPVVC potentially more accurate, and therefore more sensitive, than its whole brain counterpart aPBVC.

The rate of ventricular enlargement is also more specific than whole brain atrophy rate. The third ventricle is close to functionally relevant structures such as the thalamus. A close relationship has been demonstrated between cognitive impairment and third ventricular width or volume, possibly reflecting the clinical effect of thalamic atrophy.11 12 25–28 Furthermore, Kezele et al demonstrated a close association between focal tissue loss in periventricular lesions and local ventricular enlargement.29 Considering the central portion of the brain as the most frequent location of MS lesions, especially in the early disease stage, disruption of functionally important white matter tracts may explain the relation between ventricular enlargement and subsequent clinical disability. To a lesser extent, diffuse injury in normal appearing white matter also contributes to ventricular enlargement as these changes are also related to atrophy in MS.24 30 31

In a sample overlapping that of the present study, Jasperse et al demonstrated, using voxel-wise statistical analyses, the relation of clinical and cognitive changes with regional atrophy rates over a 2 year period.32 Specifically, increasing EDSS scores were related to development of widespread atrophy of periventricular regions. The present study corroborates these results using the currently available 5.5 year follow-up data, and using a more general measure of central atrophy (ie, the rate of ventricular enlargement). Our findings are slightly different from those of Horakova et al who observed that both whole brain and central atrophy rates predicted EDSS changes over a 5 year follow-up period and that whole brain and grey matter atrophy rates surpassed central atrophy rate as predictors.14 This may be related to longer disease durations of their patients at study entry (5.5 vs 1.2 years) and to the use of different measures sensitive to slightly different structural changes: volume change of the entire ventricular system versus lateral ventricular volume.

Although ventricular enlargement has been reported in all stages of MS, its behaviour in the different disease types remains unclear. Two studies reported that SPMS patients exhibit faster rates of ventricular enlargement24 33 whereas Pagani et al observed ventricular enlargement almost exclusively in relapsing–remitting MS patients.34 However, ventricular enlargement does seem to be an early phenomenon in relapse onset MS and may help to predict conversion of patients with clinically isolated syndromes suggestive of MS.13 35 36 Future longitudinal studies should investigate its value as such a predictor which may be facilitated by the fully automated adaptation of SIENA used here.

Our findings confirm the predictive value of enhancing lesions regarding ventricular enlargement: the enhancing lesion load at baseline was correlated with aPVVC and to a lesser extent with aPBVC.20 This correlation either could result from the resolution of the initial oedema, sometimes referred to as ‘pseudo-atrophy’ or, alternatively, patients with many enhancing lesions early in the disease could suffer from a more aggressive disease course, resulting in more atrophy.

As expected, the volume of T2 lesions at baseline was correlated with central and global atrophy rates in the first time interval.22 36–38 T1 lesion volumes at baseline were small, presumably because these were very early patients, and this possibly explains the lack of correlations for baseline T1LL which were observed in other studies.33 34 39 However, change in T1LL within the first time interval was correlated with aPVVC in the subsequent second time interval. This ‘delayed’ correlation may reflect a slow process of secondary axonal degeneration due to the axonal damage in these destructive lesions. Interestingly, T2LL and GdLL at baseline were also correlated with ventricular enlargement in the second time interval but not with whole brain atrophy. All such ‘delayed correlations’ with lesions in this study were restricted to central atrophy rates possibly because secondary axonal degeneration especially affects periventricular tracts.

Because central atrophy contributes to global atrophy, it is not surprising that both rates were correlated. However, there were also marked differences: while rates of ventricular enlargement in two consecutive time intervals were strongly correlated, whole brain atrophy rates were not, suggesting that central atrophy may develop more steadily over time than global atrophy. Our findings are in line with a recent study that has demonstrated non-linear temporal dynamics of whole brain atrophy, suggesting that especially in early MS, the prediction of future global atrophy rates is difficult.40 Results from a 1 year follow-up study already suggested a fairly constant rate of ventricular enlargement in MS, and our data now confirm this for a longer observation period.33

Limitations of this study include the fact that we only used longitudinal atrophy measures for lack of an automated accurate cross sectional measure of ventricular volume. Furthermore, MRI data acquisition was performed on a 1.0 T scanner from the beginning of the study. Despite the low given field strength of this scanner which is currently out of date, we agreed to continue studying on this low field strength to gain homogeneous MRI data throughout the whole follow-up period. However, because of the absence of sufficient grey–white matter contrast in these images, the use of separate tissue class analysis, which would have strengthened our results, was not possible. Additionally, in the present patient sample, we were unable to assess relations with cognitive measures as these were not available.

In conclusion, this study demonstrated that in early MS, the rate of early ventricular enlargement is a stronger MRI predictor for medium term clinical disease progression than lesion measures or whole brain atrophy rate. It further evidenced that although global brain atrophy rate and the rate of ventricular enlargement are correlated, the rate of ventricular enlargement is related to preceding changes in T1 hypointense lesion volumes while global brain atrophy rate is not. These results suggest that the early rate of ventricular enlargement is a promising and biologically plausible prognostic marker in MS.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
View Abstract


  • Funding CL was supported by a research grant form Bayer Schering, Germany. The MS Centre Amsterdam, and BM and HV were financially supported by the Dutch MS Research Foundation, The Netherlands (grant Nos 02-358b and 05-358c). The funders of the study had no role in the study design, data collection, data analysis, data interpretation or writing of the manuscript, or in the decision to submit the paper for publication.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the institutional medical ethics committee of the VU, Amsterdam, The Netherlands.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

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.