CIS-PREDICT
Predictors of Multiple Sclerosis After Clinically Isolated Syndrome: A Systematic Review and Meta-Analysis
Clinical Question
Which demographic, clinical, MRI, and CSF factors predict which patients with a clinically isolated syndrome will go on to be diagnosed with multiple sclerosis?
Bottom Line
In adults presenting with CIS, younger age, multifocal presentation, higher T2 lesion burden, lesions in periventricular, corpus callosum, infratentorial, and spinal cord locations, gadolinium-enhancing lesions, CSF oligoclonal bands, and CSF pleocytosis all significantly increase the likelihood of a subsequent MS diagnosis — these factors can be used for early risk stratification and to guide personalized treatment decisions after CIS.
Major Points
- 72 observational studies (9915 adults, 69 unique cohorts) were pooled with random-effects meta-analysis; median follow-up 37.5 months (IQR 24–60).
- Younger age at CIS onset predicted MS diagnosis (OR 1.60, 95% CI 1.30–2.00, p=0.0001; 34 studies, 4598 patients; I²=76.4%), robust on leave-one-out analysis.
- Multifocal CIS presentation predicted MS (OR 1.55, 95% CI 1.10–2.19, p=0.022; 6 studies, 939 patients; I²=0%).
- MRI predictors: higher number of T2 lesions (OR 7.46, 95% CI 1.53–36.43, p=0.019), periventricular lesions (OR 4.08, 95% CI 2.99–5.57, p<0.0001), corpus callosum lesions (OR 14.88, 95% CI 3.42–64.75, p=0.0156), infratentorial lesions (OR 2.16, 95% CI 1.10–4.26, p=0.0317), spinal cord lesions (OR 1.37, 95% CI 1.05–1.79, p=0.0275), gadolinium-enhancing lesions (OR 1.91, 95% CI 1.26–2.90, p=0.007).
- CSF predictors: oligoclonal bands (OR 3.57, 95% CI 2.72–4.67, p<0.0001; 38 studies, 6066 patients) and pleocytosis (OR 3.39, 95% CI 1.38–8.34, p=0.0176; 6 studies, 747 patients).
- Not significantly associated: female sex, baseline EDSS, optic neuritis presentation, spinal cord syndrome presentation, smoking, vitamin D, and CSF neurofilament light chain.
- 61.1% of included studies were rated good quality (Newcastle-Ottawa); main methodological concerns were insufficient follow-up (76.4%) and lack of confounder adjustment (51.4%).
Design
Study Type: Systematic review and random-effects meta-analysis of observational studies (cohort, cross-sectional, case–control)
Randomization:
Blinding: Not applicable (independent duplicate screening, extraction, and risk-of-bias assessment)
Allocation: Not applicable
Enrollment Period: Literature search of EMBASE, PubMed, and Scopus up to December 31, 2025
Follow-up Duration: Median 37.5 months (IQR 24–60) across included studies
Centers: 0
Countries: Spain, Italy, Germany, Czech Republic, China, United Kingdom
Sample Size: 9915
Analyzed: 9915
Analysis: Random-effects meta-analysis of ORs with Knapp–Hartung adjustment; heterogeneity by Cochran's Q and I²; leave-one-out sensitivity analyses; publication bias by funnel plots and Egger's test (≥10 studies); meta-regression (study design, diagnostic criteria, follow-up duration, sample size) for high-heterogeneity outcomes; R 4.4.2 (meta, metafor, dmetar)
Power Calculation: Not applicable; meta-analysis performed for factors with data from at least 3 cohort studies
Registration: PROSPERO CRD42022324386
Inclusion Criteria
- Observational studies (cohort, cross-sectional, or case–control) of adults with a history of CIS
- Reported subsequent MS diagnosis status
- Published up to December 31, 2025 in EMBASE, PubMed, or Scopus
- Published in English or Spanish
- ORs reported or raw data available for calculation
Exclusion Criteria
- Full text unavailable
- Missing outcome data
- ORs not reported and raw data unavailable for calculation
- Duplicate publications
- Reviews, conference proceedings, dissertations, ongoing research, or preprints
- Published in languages other than Spanish or English
Arms
| Field | No trial arms — pooled observational studies of adults with CIS |
|---|---|
| N | 9915 |
| Intervention | None (observational data); within each study, candidate predictors were compared between patients subsequently diagnosed with MS and those not diagnosed, then pooled across 72 studies |
| Duration | Median follow-up 37.5 months (IQR 24–60) |
Outcomes
| Outcome | Type | Control | Intervention | HR / OR / RR | P-value |
|---|---|---|---|---|---|
| Diagnosis of MS following a CIS event (per criteria used in each study); pooled odds ratios for candidate predictors using random-effects meta-analysis | Primary | Not applicable (no control arm); ORs compare patients with vs without each predictor/exposure | Not applicable (no intervention; observational predictors/exposures only) | Significant predictors: younger age OR 1.60; multifocal CIS OR 1.55; higher T2 lesion number OR 7.46; periventricular lesions OR 4.08; corpus callosum lesions OR 14.88; infratentorial lesions OR 2.16; spinal cord lesions OR 1.37; gadolinium-enhancing lesions OR 1.91; oligoclonal bands OR 3.57; CSF pleocytosis OR 3.39 | |
| Secondary | 1.6 | 0.0001 | |||
| Secondary | 1.55 | 0.022 | |||
| Secondary | 7.46 | 0.019 | |||
| Secondary | 4.08 | <0.0001 | |||
| Secondary | 14.88 | 0.0156 | |||
| Secondary | 2.16 | 0.0317 | |||
| Secondary | 1.37 | 0.0275 | |||
| Secondary | 1.91 | 0.007 | |||
| Secondary | 3.57 | <0.0001 | |||
| Secondary | 3.39 | 0.0176 | |||
| Secondary | 1.07 | 0.5044 | |||
| Secondary | 1.96 | 0.1820 | |||
| Secondary | 0.71 | 0.2110 | |||
| Secondary | 1.3 | 0.4118 | |||
| Secondary | 1.61 | 0.3667 | |||
| Secondary | 7.2435 | 0.1053 | |||
| Secondary | 4.3032 | 0.0860 |
Subgroup Analysis
Meta-regression (study design, diagnostic criteria, follow-up duration, sample size) identified no significant moderators for younger age or T2 lesion burden. Leave-one-out analyses confirmed robustness for most predictors; multifocal CIS significance was attenuated when the largest study was excluded. Egger's test suggested small-study effects for younger age and oligoclonal bands.
Criticisms
- High between-study heterogeneity for several predictors (I² 76.4% for age, 93.1% for T2 lesion number, 88.1% for smoking)
- MS diagnostic criteria varied across studies (Poser through McDonald 2017), reflecting evolution of criteria over time
- 76.4% of studies had insufficient follow-up and 51.4% lacked confounder adjustment; 36.1% were rated poor quality
- Evidence of small-study effects/publication bias on Egger's test for younger age and oligoclonal bands
- Some analyses (corpus callosum lesions, multifocal CIS, smoking, pleocytosis) based on few studies; multifocal CIS significance attenuated when the largest study was excluded
- Wide confidence intervals for some estimates (e.g., corpus callosum lesions 3.42–64.75, T2 lesions 1.53–36.43)
Funding
The authors have nothing to report.
Based on: CIS-PREDICT (European Journal of Neurology, 2026)
Authors: María Paula Zafra-Sierra, Carolina Ferreira-Atuesta, Daniela Sofía Rodríguez, ..., Saúl Reyes
Citation: Zafra-Sierra MP, Ferreira-Atuesta C, Rodríguez DS, et al. Predictors of Multiple Sclerosis After Clinically Isolated Syndrome: A Systematic Review and Meta-Analysis. Eur J Neurol. 2026;33(8):e70711.
Content summarized and formatted by NeuroTrials.ai.