Prognostic value of magnetic resonance imaging in patients with clinically isolated syndrome conversion to multiple sclerosis: A meta-analysis
Background: Numerous studies have investigated the associations of brain or spinal cord MRI with the risk of developing Multiple Sclerosis (MS) in people with Clinically Isolated Syndrome (CIS), however, the findings are uncertain. Therefore, we performed a meta-analysis based on 24 publications to comprehensively evaluate such associations. Materials and Methods: The databases of EMBASE and MEDLINE (January 1980-August 2011) were searched electronically for all relevant studies. Data were extracted from each study independently by both reviewers using a predefined structured spreadsheet. The quality of each study was assessed independently by two reviewers according to Newcastle-Ottawa Scale for reading cohort study proposed by Deeks et al. The meta-analysis including 24 qualified studies was performed by using the Cochrane Collaborations RevMan5.0 software. Results: Twenty-four identified studies met the inclusion criteria and minimum quality threshold. A meta-analysis of cohort studies indicated that the CISs having MRI lesions did have significantly increased risk for MS (risk ratio [RR] = 3.71, 95% confidence interval [CI], 3.27-4.21, P < 0.00001). In the subgroup analysis (according to the number of T2 lesions at baseline), the risk of developing MS in CIS patients with the medium MRI burden (4-9 lesions) was higher than with the low MRI burden (1-3 lesions) (RR = 0.66,95% CI, 0.45-0.95, P < 0.00001). While, no correlation was found in group between the medium MRI burden and the high MRI burden(>9 lesions) (RR = 0.97, 95% CI, 0.82-1.15, P = 0.72). Meanwhile, the CIS patients with abnormal baseline MRI, especially with infratentorial lesions, had a high risk of conversion to MS compared to patients without the such infratentorial lesions (RR = 1.37, % CI, 1.09-1.73, P = 0.0008). Conclusions: Despite some limitations, this meta-analysis established solid statistical evidence for an association between the presence or absence of MRI lesions within the brain or spinal cord MRI and the risk of developing MS, particularly for studies with large sample size. The CIS patients with abnormal baseline MRI, especially with infratentorial lesions, had a high risk of conversion to MS. However, this association warrants additional validation in larger and well designed studies.
Keywords: Clinically isolated syndrome, cohort study, magnetic resonance imaging meta-analysis, multiple sclerosis
Approximately 90% of patients with multiple sclerosis (MS) initially present with a clinically isolated syndrome (CIS), but convert to clinically definite multiple sclerosis (CDMS) when they develop a second attack.  It is important to identify patients at high risk of conversion to CDMS as early immunomodulatory treatment has the potential to delay conversion. Studies have shown that the age of first attack, season, clinical manifestations, cerebrospinal fluid, magnetic resonance imaging (MRI), gender, ethnicity and other factors can affect the prognosis of CIS. Advancements in imaging technologies, such as MRI offer the promise of more effective management strategies for MS. Until recently, confirmation of the diagnosis of MS has generally required the demonstration of clinical activity that is disseminated in both time and space. Nevertheless, with the advent of MRI techniques, occult disease activity can be demonstrated in 50-80% of patients at the time of the first clinical presentation.
So MRI is a useful tool in CIS patients, both to exclude other diseases and to establish the risk and probability of developing MS.  CIS patients with an abnormal cerebral MRI at presentation have a substantially higher risk of conversion to CDMS than those with a normal cerebral MRI. Prospective studies have shown that the presence of such lesions predicts future conversion to CDMS. Indeed in a young to middle-aged adult with a CIS, once alternative diagnoses are excluded at baseline, the finding of white matter lesions on a T2-weighted MRI scan (especially if one of these lesions is located in the periventricular region) is a very sensitive predictor of the subsequent development of CDMS within the next 7-10 years. By contrast, normal results on MRI at the time of clinical presentation make the future development of CDMS considerably less likely.
The aim of the present study was to evaluate whether the presence of demyelinating lesions of T 2 WI increases the risk of developing MS after a first attack and if MRI adds information in predicting CIS conversion to CDMS, so as to provide objective proof for early screening and treating CIS patients with the high risk of suggestive of MS.
We searched the electronic literature MEDLINE and EMBASE databases for all relevant articles using the search terms: "CIS," "optic neuritis," "acute partial transverse myelitis," "MRI," "Magnetic resonance image," "MS" and "cohort study." The search terms were combined with the Boolean operators OR and AND. All titles and abstracts were screened to determine their eligibility. If the title or abstract indicated possible relevance, their full texts were more fully examined for inclusion according to the inclusion or exclusion criteria. Any discrepancies were resolved by consensus.
Inclusion and exclusion criteria
The primary studies included in this analysis were to meet the following inclusion criteria: (1) the study must evaluate the associations between brain MRI and risk of MS in CIS patients; (2) the study must be a cohort study; (3) exposure factor must be having demyelinating lesions in first scanning; (4) the diagnosis of MS was based on Poser, McDonald or Schumacher criteria; and (5) the study must provide the total number of the exposed group and non-exposed group.
The major reasons for exclusion were: (1) being a review, editorial, or comment; (2) duplicated studies; (3) laboratory molecular or animal studies; and (4) the rate of dropout was greater than 20%. If more than one study was published using the same data series, all results on the different time points were retained.
Data extraction and quality assessment
Two authors extracted data independently by using a standard form and disagreements were resolved by discussion with co-authors. The following data were extracted: The first author's surname, year of publication, study design, number of subjects, characteristics and number of expose group and non-expose group, diagnostic methods for MS and follow-up time.
At first, the statistical heterogeneity assumption among studies was verified by Chi-square-based Q test and the Higgins I2 statistic. If the studies were found to be homogeneous (with P > 0.10 for the Q test or I2 ≤ 50%), the pooled risk ratio (RR) estimate of all studies were calculated by the fixed effects model (the Mantel-Haenszel method). If homogeneity could not be assumed, a random-effects model (the DerSimonian and Laird method) would be used. However, if heterogeneity was shown, the sources of heterogeneity were explored and subgroup or sensitivity analyses were performed. Subgroup analyses were performed by the number of T2WI lesions and the site of the lesion. Then, the strength of association between the brain or spinal cord MRI and the risk of developing MS in patients with CIS was assessed by calculating RRs with the corresponding 95% CIS. All statistical analyses were performed with RevMan5.0 software. Funnel plots were constructed to evaluate potential publication bias.
As shown in [Figure 1], a total of 1751 titles and abstracts were identified. Of these, after the primary screening of abstracts, 1636 were rejected, most of which were duplicates (n = 203), review articles (n = 473), or irrelevant (n = 960). Of the remaining 139 articles, 125 were excluded because they did not meet the inclusion criteria or published information incomplete. Ultimately, a total of 24 articles were included in the meta-analysis.
The basic information of all eligible studies included in our meta-analysis is given in [Table 1]. ,,,,,,,,,,,,,,,,,,,,,,, In the collective of 24 studies, there were 3,522 participants aged 10-50 year. The ratio of men to women was about 1:2, with one study  that did not declare non-expose group. All studies were cohort studies. Seven studies ,,,,,, were conducted optic neuritis (ON) participants and acute partial transverse muelitis (APTM) was the object of one study.  Six studies, ,,,,, which described CIS patients were treated with drugs during the study period. MS was diagnosed in 15 studies ,,,,,,,,,,,,,, by Poser criteria; CIS patients developed MS according to McDonald criteria in eight studies, ,,,,,,, and one study according to Schumacher criteria.  Eight studies ,,,,,,, had performed subgroup analyses based on the number of different lesions. Eleven studies ,,,,,,,,,, did not report the number of exclusions or lost to follow-up during the study period. Two studies , described the incidence of MS in CIS patients for the first time MRI in different parts of demyelinating lesions. All relevant demographic and clinical data of eligible studies were provided in [Table 2]. ,,,,,,,,,,,,,,,,,,,,,,,,
Comparing the risk of conversion to MS in CIS patients with abnormal and normal baseline MRI
Twenty-three studies were available for the meta-analysis; however, only five studies described CIS patients were treated with drugs during the study period. ,,,, The heterogeneity among studies was shown to be high (I 2 = 59%). However, the heterogeneity was not significant when we examined subgroups stratified by treatment situation during the study period. Hence, the fixed effects model was used. Separate analyses were performed to determine whether an abnormal scan had any effect on the risk of conversion to MS in CIS patients. And the result from treatment (RR = 2.51 95% confidence interval [CI], 2.12-2.97) and non-treatment (RR = 4.89, 95% CI, 4.07-5.88) were almost identical and indicated that having T2WI white matter lesions on their baseline scans was associated with risk of MS [Figure 2].
Association between different parts of lesions of baseline MRI and MS risk
Two studies described the incidence of MS in CIS patients for the first time MRI in different parts of demyelinating lesions. , The results indicated that the presence of infratentorial lesions was associated with risk of developing MS in patients with CIS (RR = 1.37, 95% CI, 1.09-1.73, [Figure 3] for with infratentorial lesion vs. without infratentorial lesion). Due to limited data of occurrence rate of MS between in the CIS patients with and without infratentorial lesion, further subgroup analysis was not possible.
Association between different number of lesions at the baseline scans and MS risk
Eight studies described the incidence of MS in the CIS patients with different lesions. ,,,,,,, In the subgroup analysis grouped by lesion numbers 1-3, 4-9, >9, a statistically significant association was found for studies between 1and 3 lesions and 4-9 lesions [RR = 0.66, 95% CI, 0.45-0.95, [Figure 4]. However, the results suggested the number of MRI lesions (between 4 and 9 lesions and >9) was not associated with the risk of developing MS in CIS patients [RR = 0.97, 95% CI, 0.82-1.15, [Figure 5].
It was reported that MRI findings of CIS patients at baseline were related to the risk of conversion. When clinically silent brain lesions are seen on MRI, the likelihood of developing MS is also high. In patients with an abnormal scan who were followed-up for a long period, the conversion rate was more than 50% in patients with optic neuritis and about 90% in patients with another CIS, although with a little indication that the number of lesions increases the already high risk  . This is not surprising. Patients with abnormal brain MRI already have morphologic evidence of disseminated disease and as such it is expected that most of these patients will eventually develop additional neurologic events sufficient for a diagnosis of CDMS. Evidence is accumulating that treatment may delay the onset of CDMS. Most clinicians who see patients with a CIS or suspected early MS would now agree that it is important to diagnose early and having done so, to weigh-up the potential benefits, risks and uncertainties of disease modifying treatment while ensuring the patient is fully informed and participates in the decision-making process. 
The number of lesions at baseline had an influence on course of the disease and was related to the risk of MS in patients with CIS.  These data indicate that in a mixed CIS population, the number of baseline lesions predicts not only conversion to CDMS, but also the likelihood of developing mild and even moderate disability; note that all three deaths due to MS occurred in patients with four or more lesions at baseline. 
In a CIS of the brainstem, the dissemination in space criteria are not as specific in predicting conversion to CDMS as when they are applied to other CISs; this is mostly because the presence of an infratentorial lesion is not used to show dissemination in space in this subgroup of patients. In all CISs, abnormal brain MRI is the strongest predictor of conversion. However, spinal-cord imaging did not affect subsequent diagnosis when the McDonald criteria, in which one spinal-cord lesion can be substituted for a brain lesion, were applied.  Relevant reports showed that the presence of an infratentorial lesion could be closely related to the expanded disability status scale (EDSS).
Sailer et al. found that the CIS patients with infratentorial lesions had a higher EDSS scores at 10 year of follow-up.  The patients with CIS whose lesions located in brainstem had a poor prognosis.  Our study disclosed the presence of MRI lesions at the beginning, especially the number of lesions, and parts of lesions were interrelated with the risk of development of MS in CISs.
However, there were more inherent limitations of the literature in this meta-analysis, including the quality of included studies and study design. And the type of CIS, time of follow-up and the clinical criteria for diagnosis of MS were not uniform, especially when the follow-up is either too short or too variable. Most studies of the predictive validity of MRI in MS are confounded by this limitation. Moreover, if the period of follow-up is variable, each patient will not have had an equal chance to develop CDMS. Then, the number of cases were small; thus, it could affect the demonstrate strength and comprehensiveness of this systematic review. We look forward to a rigorous design, detailed and high - quality prospective cohort.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2]