Article Text

Systematic review
Incidence and determinants of seizures in multiple sclerosis: a meta-analysis of randomised clinical trials
Free
  1. Valeria Pozzilli1,2,
  2. Shalom Haggiag3,
  3. Massimiliano Di Filippo4,
  4. Fioravante Capone1,2,
  5. Vincenzo Di Lazzaro1,2,
  6. Carla Tortorella3,
  7. Claudio Gasperini3,
  8. Luca Prosperini3
  1. 1 Unit of Neurology, Neurophysiology and Neurobiology, Department of Medicine and Surgery, Campus Bio-Medico University, Roma, Lazio, Italy
  2. 2 Fondazione Policlinico Universitario Campus Bio-Medico, Rome, Italy
  3. 3 MS Centre, Department of Neurosciences, San Camillo Forlanini Hospital, Roma, Italy
  4. 4 Section of Neurology, Department of Medicine and Surgery, University of Perugia, Perugia, Italy
  1. Correspondence to Dr Luca Prosperini, MS Centre, Dept. of Neurosciences, San Camillo Forlanini Hospital, Roma, Italy; luca.prosperini{at}gmail.com

Abstract

Background Seizures are reported to be more prevalent in individuals with multiple sclerosis (MS) compared with the general population. Existing data predominantly originate from population-based studies, which introduce variability in methodologies and are vulnerable to selection and reporting biases.

Methods This meta-analysis aims to assess the incidence of seizures in patients participating in randomised clinical trials and to identify potential contributing factors. Data were extracted from 60 articles published from 1993 to 2022. The pooled effect size, representing the incidence rate of seizure events, was estimated using a random-effect model. Metaregression was employed to explore factors influencing the pooled effect size.

Results The meta-analysis included data from 53 535 patients and 120 seizure events in a median follow-up of 2 years. The pooled incidence rate of seizures was 68.0 per 100 000 patient-years, significantly higher than the general population rate of 34.6. Generalised tonic-clonic seizures were the most common type reported, although there was a high risk of misclassification for focal seizures with secondary generalisation. Disease progression, longer disease duration, higher disability levels and lower brain volume were associated with a higher incidence of seizures. Particularly, sphingosine-1-phosphate receptor (S1PR) modulators exhibited a 2.45-fold increased risk of seizures compared with placebo or comparators, with a risk difference of 20.5 events per 100 000 patient-years.

Conclusions Patients with MS face a nearly twofold higher seizure risk compared with the general population. This risk appears to be associated not only with disease burden but also with S1PR modulators. Our findings underscore epilepsy as a significant comorbidity in MS and emphasise the necessity for further research into its triggers, preventive measures and treatment strategies.

Data availability statement

Data are available upon reasonable request. Data presented in this article will be made available by a reasonable request of a qualified investigator (requests should be sent to the corresponding author: luca.prosperini@gmail.com).

Statistics from Altmetric.com

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.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Population studies hint at heightened seizure prevalence in multiple sclerosis (MS) but lack consistency. Our meta-analysis, focusing on 63 randomised clinical trials and 53 535patients with MS, offers a rigorous and standardised assessment.

WHAT THIS STUDY ADDS

  • Our meta-analysis reveals a significantly increased seizure incidence in patients with MS (68.0 per 100 000 patient-years). Factors like disease progression, longer duration, higher disability and treatment with sphingosine-1-phosphate receptor modulators increases the risk.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • These findings highlight the significant epilepsy comorbidity in MS, guiding clinical management and policy considerations. Understanding risk factors, including treatment-related ones, informs better patient care.

Introduction

Seizures are more prevalent in individuals with multiple sclerosis (MS) compared with the general population.1 A recent meta-analysis has revealed that patients with MS have a threefold higher risk of developing epilepsy and twice the odds of having prevalent epilepsy compared with individuals without MS.2 While the exact underlying pathophysiological mechanism remains partially understood, the prevailing hypothesis suggests that the presence of juxtacortical3 or cortical4 lesions, and brain atrophy,5 significantly contributes to this comorbidity.

Patients with a more pronounced disability and a longer disease duration appear to be at a heightened risk of experiencing seizures, indicating a connection between the severity of MS and the occurrence of epilepsy.1 6 A population-based study conducted in Sweden found notable differences in epilepsy incidence across various MS phenotypes. The cumulative incidence stood at 2.2% in patients with relapsing-remitting (RR), in contrast to an incidence of 5.5% in those with primary (PP) and secondary progressive (SP) courses.1 However, seizures can manifest even in the early stages, such as clinically isolated syndrome, which marks the initial demyelinating event before other MS symptoms emerge.7 8 Seizures can take on different forms, including focal, focal to bilateral tonic-clonic or, more rarely, tonic-clonic of unknown origin.9

Existing data concerning the association between epilepsy and MS originate from hospital-based observations10 11 and population-based studies, but the latter ones are susceptible to selection bias and information inaccuracies. The interpretation and comparison of these studies are further complicated due to the considerable variability in epilepsy definitions and assessment methodologies.2 In an effort to overcome the inherent limitations of the aforementioned studies, we have opted to analyse aggregated data from randomised clinical trials (RCTs). Typically, these trials exclude patients with concurrent medical conditions to avoid noise in safety data analysis.12 This approach provides a more dependable count of relevant adverse events (AEs) and offers robust data on MRI and MS medications, two aspects that have been comparatively underexplored in relation to seizure occurrence in MS. The objective of our meta-analysis is twofold: (1) to determine the incidence of seizures in patients participating in RCTs, where seizures are recorded as AEs, and (2) to identify potential factors contributing to a higher incidence of seizures.

Methods

Study design and registration

Our meta-analysis with metaregression adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement13 and is registered on Open Science Framework database (https://doi.org/10.17605/OSF.IO/AM53H).

Search strategy

PubMed and Google Scholar were searched for articles published without time restrictions by selecting the filters ‘Clinical Trial, Phase III’ and ‘Adult: 19+years’ and by using the keyword ‘multiple sclerosis’. One reviewer ran the search strategy and screened the initial titles after removing duplicates. Two authors independently examined each potential relevant article, using the following criteria as defined by the PI(E)COS model14: (1) population, adult persons affected by any form of MS; (2) intervention, not applicable; (3) comparison, not applicable; (iv) outcomes, new-onset seizure or epilepsy; and (4) setting, experimental setting. We excluded articles not written in English, and those reporting study results with a sample size of <100 or a follow-up of <6 months. We selected only phase III RCTs because they are large enough to demonstrate even uncommon AEs (ie, ≥1/1000 and <1/100). Lastly, we excluded RCTs with potential epileptogenic drugs, for example, dalfampridine or similar compounds.

Data extraction

Two authors independently extracted data, with discrepancies resolved by a third author. The following information were extracted from each article: trial name, year, total sample size (ie, the total number of patients who were randomised), study duration, mean age, mean disease duration (ie, years elapsed from the clinical onset), predominant MS subtype (clinically isolated syndrome, RR, PP or SP), mean Expanded Disability Status Scale (EDSS) score, mean lesion volume on T2-hyperintense sequence, mean normalised brain volume and number of seizures or epilepsy events (outcome of interest).

We conducted a thorough review of the included articles, including tables with AE data and, when available, appendices. We also cross-referenced information with the ClinicalTrials.gov website for consistency.

Statistical analysis

We calculated the pooled effect size of included studies on the incidence rate of seizure or epilepsy as the number of events per patient-years in the whole RCT samples (ie, regardless of arm allocation) by a random-effect model based on an empirical Bayes estimate. Patient-years were calculated as the sample size multiplied by RCT duration expressed in years. Study estimates were converted to double arcsine transformed incidence rates. This approach offers some advantages as more accurate when the events of interest are rare, avoiding to either exclude RCTs with zero events or add arbitrary increments; moreover, it has variance-stabilising property.15

We fitted univariable and multivariable metaregression equations to explore which variables (moderators) influenced the pooled effect size. Normality assumption was checked for all variables entered in models and for their residuals; in case of violation of normality assumption, moderators were log-transformed.

Between-study heterogeneity and inconsistency were expressed as Cochran’s Q and I2, respectively. The risk of publication bias was assessed by Kendall’s τ Egger’s Z test of asymmetry. Regression coefficients (β) with their corresponding 95% CIs were also reported for moderators. The coefficient of determination (R2) was used to assess the goodness of fit for each weighted model.

Two-tailed p<0.05 were considered significant.

Results

Search findings

In total, we identified 257 articles, out of which 60 (S01–S60) fulfilled the eligibility criteria for this meta-analysis (see figure 1). The full references for the included articles can be found in online supplemental appendix. The selected articles presented data from 63 RCTs, with 3 articles reporting outcomes from studies with identical designs (S47, S53 and S60). The majority of these RCTs were conducted to investigate the safety and efficacy of 21 disease-modifying treatments compared with placebo (k=38) (S01–S07, S10–S15, S18–S21, S24–S26, S30, S31, S33–S35, S37–S39, S41, S42, S45, S48–S50, S56 and S57) or active comparators, including interferon beta (k=10) (S22, S27, S28, S40, S43, S47, S51 and S52), teriflunomide (k=5) (S53, S55 and S60) and dimethyl fumarate (k=1) (S59), in terms of relapse rate and/or disability progression.

Supplemental material

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-analyses flow chart for study selection.13

The remaining RCTs encompassed various study designs. These included dose/interval comparison studies for different interferon beta formulations (k=2) (S08 and S09), non-inferiority studies comparing interferon beta with glatiramer acetate (k=2) (S16 and S17) and azathioprine versus interferon beta (k=1) (S36). Additionally, there were studies examining dose comparisons of glatiramer acetate 40 mg vs 20 mg (k=1) (S23), a comparison between generic and ‘branded’ glatiramer acetate (k=1) (S44) and a dosing interval comparison of natalizumab administered every 6 weeks versus every 4 weeks (k=1) (S58). One study focused on exploring the combined effect of interferon beta and glatiramer acetate versus each drug administered individually (k=1) (S32). Among the included studies, 24 RCTs featured multiple arms, including low and high doses of experimental drugs compared with placebo or comparators (S01, S03, S05, S17, S20-S31, S34, S35, S38, S39, S41, S42, S45 and S58), with one study incorporating glatiramer acetate as an active reference comparator (S29). Further details about the included RCTs are presented in table 1.

Table 1

Summary of randomised clinical trials selected for meta-analysis and meta-regressions

None of the selected RCTs excluded patients with comorbid epilepsy, whereas ‘uncontrolled epilepsy’ was an exclusion criterion only in the BEYOND trial (S17).

Participants

The pooled data derived from 63 RCTs encompassed 53 535 patients, with a median follow-up duration of 2 years (range: 0.75–3 years), corresponding to a cumulative 100 469 patient-years. A total of 120 patients experienced epileptic seizure events, resulting in a pooled incidence rate of 68.0 (95% CIs: 49.1 to 86.9) per 100 000 patient-years. Notably, there was substantial heterogeneity across the included studies (Q62=159.8, p<0.001; I2=54.4%), while the probability of publication bias was found to be low (Kendall’s τ=0.05, p=0.58; Egger’s Z = –1.09, p=0.27).

Comparatively, patients with MS exhibited a 1.96-fold elevated risk of experiencing incident seizures in contrast to the general population aged 18 years and above. This latter group displayed an incidence rate of 34.6 (95% CIs: 28.4 to 42.2) per 100 000 patient-years, as elucidated by a recent meta-analysis that comprehensively synthesised the prevalence and incidence of epilepsy from internationally published studies.16

Based on the most recent classification provided by the International League Against Epilepsy (ILAE),16 the most frequent seizure type reported across the analysed RCTs was generalised tonic-clonic seizures (n=47, 40.8%), followed by focal seizures (n=9, 7.5%) and focal to bilateral tonic-clonic seizures (n=5, 4.2%). Instances of status epilepticus were noted in eight cases (7.0%), though without any specific subtype definition. Additionally, 49 cases (40.8%) were categorised using generic terms such as ‘epilepsy’ or ‘seizure’. See table 2 for further details regarding the types of seizures reported in the selected RCTs.

Table 2

Types of seizure events (n=120) reported in selected randomised clinical trials (k=63)

Reporting of adverse events

The reporting of AEs varied among the articles, often being documented only if their occurrence exceeded specific thresholds, such as >2%, >3%, >5% or >10% of participants. Alternatively, AEs were reported if they led to treatment discontinuation or were graded as ‘serious’ according to the International Conference on Harmonisation Good Clinical Practice (ICH-GCP) guidelines. Among the analysed RCTs, a total of 52 were registered on ClinicalTrials.gov (S12-S60), but results were not posted for eight of them (S12, S14, S15, S17, S24, S36 and S57–S59). The remaining 11 RCTs, all conducted before the year 2000, were not listed in the database (S01–S11).

It is worth noting that a standardised framework for AE reporting was not available prior to the year 2010. However, following 2010, AEs were often coded using the Medical Dictionary for Regulatory Activities and/or the Common Terminology Criteria for Adverse Events classification systems in 18 RCTs. Certain articles concentrated solely on events deemed of particular interest for their investigational products. This included categories like infusion-related reactions, infections and neoplasms. Notably, three RCTs investigating sphingosine-1-phosphate receptor (S1PR) modulators specifically identified seizures as AEs of special interest (EXPAND, INFORMS and OPTIMUM) (S45, S50 and S55).

Determinants of incident seizures

As mentioned earlier, the reporting of events of interest showed an upward trend over the years (k=63, β=0.001, p=0.003). Higher incidence rates of seizures were associated with studies that enrolled patients with progressive disease courses (k=63, β=0.020, p=0.038), longer time since clinical onset (k=63, β=0.010, p=0.046), greater EDSS scores (k=63, β=0.031, p=0.035) and lower normalised brain volume (k=25, β =–0.542, p=0.008). These estimates remained consistent even after accounting for the calendar year (data not shown). Age and T2 lesion volume did not demonstrate an effect on the pooled effect size (table 3).

Table 3

Metaregressions exploring the effect of demographic, clinical and MRI variables (moderators) on the rate of seizure events in randomised clinical trials on multiple sclerosis

Of particular note, among the total of 120 seizure events, nearly half (n=55) occurred within RCTs involving S1PR modulators, whereas the remaining events were distributed fairly evenly among the other RCTs. In light of this, a further meta-analysis was conducted on eight RCTs involving fingolimod, ozanimod, ponesimod and siponimod) (S21, S22, S38, S45, S50–S52 and S55). Unlike the main analysis, where we estimated the incidence rate of seizure occurring within each RCT (regardless of treatment allocation), in this supplemental meta-analysis, we made a comparison of seizure event rates in investigational therapy arms versus the control arms, that is, either placebo or comparators (interferon beta or teriflunomide). For this supplemental analysis, the low-dose and high-dose arms of S1PR modulators were merged. The results indicated a significantly heightened risk of incident seizures in patients allocated to S1PR modulators compared with placebo or comparators (see figure 2), yielding an arcsine square root transformed risk difference of 0.031 (95% CIs 0.013–0.049), corresponding to a back-transformed risk difference of 20.5 (95% CIs 1.7 to 39.3) events per 100 000 patient-years. The risk ratio for incident seizures was estimated at 2.45 (95% CIs 1.43 to 4.21, p=0.008).

Figure 2

Forest plot showing the effect size, expressed as arcsine square root risk difference, for comparison of seizure event rates in the investigational therapy arms (fingolimod, ozanimod, ponesimod and siponimod) versus the control arm (ie, either placebo or comparators—interferon beta or teriflunomide) in randomised clinical trials with sphingosine-1-phosphate receptor modulators (k=8). RCT, randomised clinical trial; RE, random effects; S1PR, sphingosine-1-phosphate receptor.

A leave-one-out sensitivity analysis confirmed the robustness of this finding, with risk difference ranging from 14.7 to 27.8 per 100 000 patient-years and risk ratios ranging from 2.15 to 2.99. The pooled arcsine square root transformed risk difference remained consistently significant even after iteratively excluding individual RCTs, substantiating that this finding was not driven by any single study (all p<0.05).

Discussion

Incidence of seizures in multiple sclerosis

Our comprehensive meta-analysis, encompassing AEs reported in 63 RCTs spanning from 1993 to 2022, substantiates that individuals diagnosed with MS exhibit a heightened incidence of seizures compared with the general population, with an almost twofold elevated risk. This finding, however, is somewhat lower than the outcome reported by a recent meta-analysis conducted by Kuntz et al.2 Their study revealed an approximate threefold increased relative risk, derived from the pooling of observational studies, which are often subject to ascertainment bias and exhibit greater heterogeneity when compared with rigorously conducted RCTs.

Our investigation further clarifies that according to the ILAE classification,17 generalised tonic-clonic seizures are the most common seizure type in MS, followed by focal seizures and focal to tonic-clonic seizures. Notably, the existing body of literature on epilepsy in the context of MS suggests that the predominant seizure types have a focal origin9 11 or are characterised as focal to bilateral tonic-clonic.9 The variance between these findings and our results could be attributed to the varied terminology employed in RCTs to describe seizures. In many instances, terms like ‘convulsion’, ‘seizure’ or ‘epilepsy’ are used, which lack clarity in terms of their semiology. Additionally, focal seizures with secondary generalisation might be missed, leading to misclassification as generalised tonic-clonic seizures.

Determinants of seizures in MS

We reconfirm that the incidence of seizures is notably higher among patients with a progressive phenotype, longer disease duration and higher EDSS scores. These findings align with other studies that have identified a connection between higher disability scores and epilepsy,6 18 19 as well as an association between the cumulative incidence of epilepsy and progressive disease.1 19

While seizures can sometimes present as the initial manifestation of MS, this occurs in a minority of cases.9 Notably, a multicentre study involving patients with MS and seizures found that the most common disease course at the time of the first seizure was SPMS.20 Interestingly, we did not observe any correlation between incident seizures and age or total T2 lesion load, whereas our findings indicate more frequent incident seizures in patients with smaller normalised brain volume, consistent with the observations of a longitudinal study by Calabrese et al 18 which noted accelerated cortical atrophy in patients with epilepsy.

It is plausible that the specific localisation of demyelinating lesions, rather than their total volume or count, contributes to the epileptogenic effects of MS.18 Intracortical lesions (detectable only through specific MRI sequences not universally available and not employed in RCTs) have been identified as a significant factor in epilepsy related to MS.4 Subcortical lesions, especially in the cerebellum and basal ganglia, have been correlated to epileptogenesis in other conditions of lesion-related epilepsy, thus suggesting the impairment of inhibitory control mechanisms that results in a ‘network disorder’ underlying seizure occurrence.21 Loss of neural network resilience is considered another crucial step preceding seizure emergence,22 and a recent study showed that network reorganisation, together with high lesion load and altered integrity of mesio-temporal grey matter structures, correlates with a greater propensity for epilepsy occurrence in MS.23

On the other hand, we should also consider a prevailing neurodegenerative process affecting grey matter as playing a crucial role in triggering seizures. Cortical demyelination is a prominent feature in patients with SPMS and is characterised by a predominant population of activated microglia, possibly driven by leptomeningeal inflammation.24 Proliferation and activation of microglia that occurs in MS-related progressive neurodegenerative state may also contribute to the development of seizures, a phenomenon demonstrated in animal models of acquired epilepsy.25

However, despite the efforts made to elucidate the mechanisms underlying epileptogenesis, the pathogenesis of seizures in MS remains poorly understood. The extensive involvement of cortical grey matter, as seen in patients with MS, seems to conflict with the relatively low frequency of unprovoked seizures,26 especially as compared with other neurological diseases causing structural epilepsy. In this regard, two different, not mutually exclusive, hypotheses have been raised: cortical damage does not play a central role in MS-related epileptogenesis; some patients with MS are somehow ‘resilient’ to seizures.26 27

Lastly, we cannot rule out a bidirectional relationship between disease burden and epilepsy, such that poor seizure control results in disability worsening and increased brain atrophy rate.19 28 Epilepsy constitutes indeed a significant aspect of the disease burden in MS, leading to compromised quality of life and cognitive function.29 In light of the above, the pharmacological control of epilepsy in patients with MS can be regarded as a way to prevent disability accumulation and brain tissue loss.

Association between risk of seizure with specific disease-modifying treatments

The novel aspect of our meta-analysis is the identification of a 2.45-fold heightened risk of incident seizures associated with S1PR modulators in comparison with placebo or active comparators. In line with the findings from our meta-regression model, a higher incidence of seizures was found in the EXPAND study (S50) that recruited only patients with SPMS, a disease course featured by long disease duration, severe disability and marked brain atrophy. In contrast to our findings, a recent meta-analysis concludes for no evidence of an association between DMTs and seizure risk.30 This discrepancy mainly stems from differences in eligibility criteria, data extraction methodology and statistical approach to handle studies reporting zero events. However, despite such differences, even the aforementioned meta-analysis found a trend towards a higher seizure risk for siponimod versus placebo.30

A cautious warning about the epileptogenic potential of S1PR modulators has already been raised in paediatric MS and transplant settings.31 In the 2-year, double-blind, double-dummy, active-controlled, parallel-group, multicentre RCT evaluating the safety and efficacy of fingolimod versus once weekly intramuscular interferon beta−1a in paediatric patients with MS (PARADIGMS), seizures were reported in 5.6% of fingolimod-treated patients and 0.9% of interferon beta-treated patients.32 Similarly, in the renal transplant population, the risk of seizures was slightly elevated in the fingolimod 2.5 mg and 5 mg groups compared with mycophenolate mofetil group.33

Mechanisms underlying such a higher incidence of seizures under treatment with S1PR modulators still need to be unravelled. Few preclinical studies suggested enhanced excitability of sensory neurons mediated by activation of types 1 and 3 S1PR by their S1P ligand.34 35 On the other hand, S1PR modulators and specifically fingolimod have consistently demonstrated to exert disease-modifying antiepileptic effects in experimental models (eg, pentylenetetrazol-induced kindling, lithium chloride-pilocarpine injection and kainic acid-induced status epilepticus) through different mechanisms.36 Reasons underlying such contrasting results might depend on several factors. First of all, modelling epilepsy with experimental procedures could not fully reproduce the complexity of S1P-mediated supraspinal circuit modifications. Second, the short-term and long-term effects of S1P signalling on brain function might somehow be different, also depending on the developmental stage. Although the pharmacological effects of S1PR modulators might in part counteract some elements of the pathogenic cascade underlying epileptogenesis,36 the net effect on neuronal excitability of S1P might be more difficult to predict, because of both circuit complexity and the wide distribution of S1PRs in the central nervous system (CNS).37 By using electrophysiological recordings coupled with transgenic breeding strategies, it has been demonstrated that a type 1 S1PR agonist depolarises resting membrane potential in type B, but not type A, somatostatin neurons in the central nucleus of the amygdala.38 Similarly, exposure to a type 1 S1PR agonist produced a significant increase in the excitability of some, but not all, small-diameter sensory neurons in the dorsal root ganglia.39 The demonstration that such effects (known to push neurons closer to the threshold for action potential firing) are only seen in specific neuronal subtypes, can lead to the hypothesis that the cellular physiology of CNS cells might be differentially influenced by S1PR modulators. Type 1 S1PR activation might indeed influence neuronal intrinsic excitability through an effect on membrane ion channels, including KCNQ potassium channels, transient receptor potential channels and voltage-gated chloride channels.38 Investigation on the effects of S1PR modulators on neuronal excitability and the potential role of glial cells in such events deserves ad hoc studies carried out in different cell types, ideally in experimental models of neuroinflammation.

Strengths and limitations

Our quantitative synthesis drew on data extracted from RCTs, distinguishing our approach from previous meta-analyses that pooled data from observational studies, which are susceptible to selection bias and inaccuracies in information reporting. While it is true that experimental data are generally less applicable to real-world clinical settings than observational data, RCTs boast exceptional internal validity, yielding precise measurements of treatment efficacy and toxicity under ideal conditions.40 RCTs, given their exclusion of individuals with comorbidities and medically unstable conditions,41 are less likely to inflate estimates of the effect size of interest. Nonetheless, we cannot rule out that patients with comorbid epilepsy were enrolled in RCTs, because none of the selected studies considered it among the exclusion criteria and the impossibility obtain information on concomitant medications indicating comorbid epilepsy.

Furthermore, when interpreting our findings, it is important to recognise the significant diversity in AEs reporting across the selected RCTs, irrespective of whether the events were deemed drug related.42 Despite the Consolidated Standards of Reporting Trials (CONSORT) recommendations, which have been extended to include more detailed guidelines for reporting harms (CONSORT Harms) in 2004 and 2022,43 the variability in safety data reporting remains substantial. Over time, the reporting of AEs increased, which could potentially lead to underestimation in earlier RCTs and overestimation in more recent ones. Additionally, some articles omitted mentioning AEs unless they occurred in more than 2%, 5% or 10% of participants, while others exclusively reported AEs that exhibited differential frequencies between treatments or focused solely on serious AEs or severe reactions (S41, S45, S47 and S56). To address the potential under-reporting bias stemming from these inconsistencies, we supplemented our analysis with safety data extracted from the clinicaltrials.gov website, where such information are systematically reported.

While our statistical methodology is robust, our findings cannot be applied directly to individuals due to the risk of ecological fallacy.44 Validating ecological hypotheses at the individual level would require individual-level analysis, which is unfeasible due to unavailable raw RCT datasets. Although establishing definitive causal connections is challenging due to the risk of ecological fallacy, our results are consistent with plausible biological mechanisms.

Conclusions

Epilepsy represents a significant comorbidity in the context of MS, warranting intensified efforts to uncover its underlying triggers and enhance strategies for both prevention and treatment.45 Our comprehensive meta-analysis underscores that MS poses a notable risk factor for seizures, with a nearly twofold heightened risk compared with the general population. Elevated incidence rates were observed in cases of longer disease duration, higher levels of disability and a progressive disease phenotype. Additionally, a correlation was established between seizures and brain atrophy, although no such correlation was identified with total T2 lesion load. We are aware that the pathophysiology of seizures in MS remains not fully elucidated and establishing a causal connection goes beyond our study aim. Notably, among the various treatments evaluated in the analysed RCTs, S1PR modulators were linked to a more than twofold increased risk of seizures. Further investigations are imperative to unveil the underlying mechanisms behind this association.

Data availability statement

Data are available upon reasonable request. Data presented in this article will be made available by a reasonable request of a qualified investigator (requests should be sent to the corresponding author: luca.prosperini@gmail.com).

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors Conception and design of the study and drafting a significant portion of the manuscript/figures: VP, SH, MDF, FC, VDL, CT, CG. Acquisition and analysis of data and revision of manuscript content: VP, FC, CT, LP. Supervision and drafting the final version of the manuscript: SH, VDL, CT, CG, LP. Statistical analysis: LP. Guarantor: LP.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests Potential conflicts of interest: nothing to report relevant to the present study. Financial disclosures: VP: no conflicts of interests. SH: travel funding and/or speaker honoraria from Biogen, CSL Behring, Novartis, Roche and Sanofi. MDF: advisory boards and steering committees for and received speaker or writing honoraria, research support and funding for travelling from Alexion, Bristol-Mayer Squibb, Bayer, Biogen, Horizon, Janssen, Merck, Mylan, Novartis, Roche, Sanofi, Siemens Healthineers, Teva and Viatris. FC: travel grants and/or speaking honoraria from Biogen, Merck, Roche and Sanofi and research grants from Merck. VDL: no conflicts of interests. CT: honoraria for speaking and travel grants from Almirall, Bayer, Biogen, Merck, Novartis, Roche, Sanofi and Teva. CG: honoraria for speaking and travel grants from Biogen, Merck, Novartis, Roche, Sanofi, Teva and Viatris. LP: personal fees and non-financial support from Biogen, Bristol-Mayer Squibb, Merck, Novartis, Roche, Sanofi and Viatris.

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

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles