Abstract
The authors sought to identify clinical and MRI predictors of outcome in primary progressive multiple sclerosis (PPMS). Clinical and MRI assessments were performed at baseline and 2 and 5 years (clinical only). At baseline, disease duration, expanded disability status scale (EDSS) and brain volume predicted outcome. Adding short-term change variables, baseline EDSS, changes in T2* lesion load and cord area, and number of new lesions were predictive. Clinical and MRI variables predict long-term outcome in PPMS.
MRI is a potential surrogate marker for relapse activity in MS; however, few studies are available for disability prediction.1–4 Long-term, prospective, natural history studies are required to elucidate the relationship between MRI changes and subsequent disability.2
Primary progressive multiple sclerosis (PPMS) is characterized by disease progression from onset with a relentless accumulation of clinical disability.5 Clinicoradiologic correlations in PPMS have been limited as cranial MRI displays few abnormalities.6 In this prospective, long-term clinical follow-up study, we aimed to identify MRI and clinical predictors of long-term disability in PPMS. We evaluated baseline clinical and MRI variables to predict disability at 5 years. We also investigated the value of short-term changes (baseline to year 2) in these variables to predict 1) disability changes over the length of the study (baseline to year 5) and 2) disability changes over the ensuing part of the study (year 2 to year 5).
Methods.
This PPMS cohort has been followed prospectively since 1997 within the MAGNIMS initiative. At 2 years, MRI and clinical data from 145 patients were available from five centers7 (table 1).
Table 1 Availability data at 5 years for EDSS and MSFC assessments
Previous MRI acquisition and analysis has shown 2-year changes in T1- and T2*-weighted lesion loads, partial brain volume, and cervical cord area.7 In the present study, T1-weighted spin-echo brain images from baseline and 2-year time points were reanalyzed using SIENA to obtain a percentage of brain volume change quantification over time and only baseline T1-weighted images with SIENAx to obtain a normalized total volume for brain tissue, gray matter, and white matter.8
In comparing patients completing vs not completing follow-up, we used two sample t tests except for expanded disability status scale (EDSS) (Mann-Whitney U test). Ordinal logistic regression was carried out with EDSS step changes (over 5 years, categorized ≤0, 1, 2, 3, ≥4 and from years 2 to 5, categorized ≤0, 1, 2, ≥3) and on an ordered raw change variable (seven ordinal categories) as outcome variables, with baseline and early change (up to year 2) potential predictors. Step changes implicitly but only partially allow for baseline EDSS level, and EDSS was included as a baseline covariate. To elucidate the relationship between disease duration and baseline EDSS as prognostic factors, logistic regression analyses were also carried out on dichotomized EDSS raw changes (≥1 vs <1). All multivariate models are adjusted for duration of follow-up.
Results.
Missing data at 5 years.
We compared baseline demographic, clinical, and radiographic parameters between patients with (n = 122) and without (n = 23) an EDSS available at 5 years. The only significant difference was higher Z scores for the timed walk test at baseline in patients with complete EDSS follow-up.
Value of baseline variables to predict EDSS change over 5 years.
In the multivariate analysis using the ordered step variable with demographic and baseline clinical and radiologic variables, the only independent predictors were center and, with borderline significance, disease duration. When center was included, baseline, clinical, and radiologic variables, other than disease duration, were not predictive of changes using the step variables. However, in models with baseline demographic, clinical, and radiologic variables, but excluding center, the best model for the ordered step changes variable included disease duration, baseline partial brain volume, and (with borderline significance) gender.
To further assess the predictive value of baseline EDSS, dichotomized and ordered raw variables were also analyzed, without center in the models. In models with baseline demographic and clinical variables, but without baseline radiologic variables, the association between disease duration and dichotomized raw change was modified by the baseline EDSS level (with a significant interaction term, p = 0.005), although the effect of disease duration seems to be more important (table 2).
Table 2 Interaction between disease duration and baseline EDSS on worsening EDSS over 5 years
Value of short-term changes (baseline to year 2) to predict long-term outcome.
Predicting EDSS change between baseline and year 5.
In models for the ordered step variable, change in T2* lesion load (odds ratio [OR] of worsening per 1 mL additional increase in load over first 2 years, 1.10; 95% CI: 1.02 to 1.18; p = 0.019), and the number of new lesions (OR of worsening per additional new lesion, 1.10; 95% CI: 0.99 to 1.23; p = 0.063) independently predicted outcome at 5 years. Higher baseline EDSS still predicted lower odds of worsening (OR 0.80, p = 0.054 and for the ordered raw variable, p < 0.001). Center no longer contributed to these models. For the ordered raw change outcome, T2* lesion load change could be substituted by change in cord area to give a borderline significant OR for worsening of 0.86 (95% CI: 0.75 to 1.00; p = 0.049) per 1 mm2 decrease in cord area.
Predicting EDSS change between years 2 and 5.
In models for the ordered step variable between years 2 and 5, higher baseline EDSS was associated with reduced odds of worsening (OR per extra EDSS point, 0.64; 95% CI: 0.49 to 0.83; p = 0.001). Larger partial brain volume at baseline was the only atrophy measure associated with reduced odds of worsening (OR per 1 mL, 0.97; 95% CI: 0.95 to 0.99; p = 0.003). Greater increase in T1 lesion load was associated with higher odds of worsening (OR per 1 mL, 1.17; 95% CI: 1.00 to 1.37; p = 0.043). T1 lesion load change could be substituted by T2* in the model with broadly similar prediction (OR per 1 mL, 1.08; p = 0.064).
Discussion.
In this study, patients with longer disease duration displayed less progression over the study duration. The nature of the interaction between baseline EDSS and disease duration shows that the disease duration may be more important. This is in accordance with previous natural history studies showing steeper disease progression curves earlier in the condition.9 Although this is unlikely to be of use in clinical practice, it might have implications for the design of clinical trials and research projects. Choosing patients close to clinical onset, for a given EDSS level, might improve the correlations between clinical variables (EDSS change) and other surrogate markers.
An increase in T2* lesion load and the development of new lesions and cord atrophy over 2 years predicted clinical outcome on long-term follow-up. The predictive role of new or enhancing lesions has already been investigated,3,4 and it has been suggested that the formation of new T2* lesions might be more relevant than the increase in T2* lesion load.4 In a recent study in patients with early PPMS, the presence of enhancing lesions has been associated with worse clinical outcome.10 Our results suggest that both markers might be independently helpful. Cross-sectional studies of cervical cord atrophy have shown it to be related to clinical outcome, especially in patients with PPMS. The findings of the present study strengthen the value of combined cervical cord atrophy and brain T2* lesion detection in predicting outcome in patients with PPMS.
When the time frame used for prediction did not overlap with the period of outcome change, greater increases in T2* and T1 lesion load over the first 2 years predicted increasing EDSS scores in the ensuing period. Interestingly, the development of new lesions or cervical cord atrophy was not of predictive value. This might imply that the effect of cord atrophy and new lesions reflects more immediately on the clinical status of patients with PPMS. Interestingly, smaller baseline brain volumes were predictive of more EDSS change between year 2 and long-term follow-up, suggesting that the impact of brain atrophy on clinical outcome is delayed.
Acknowledgment
The authors thank the subjects for their participation.
Footnotes
-
Supported by the Spanish Ministry of Health (J.S.G., BEFI #02/9115) and the Wellcome Trust (G.T.I.). The IoN NMR Research Unit is funded by the MS Society of Great Britain and Northern Ireland. This project is an initiative of the E.U. funded MAGNIMS consortium.
Disclosure: The authors report no conflicts of interest.
Received December 3, 2004. Accepted in final form May 5, 2005.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. 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.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.