Highlights
- The SEIZURE score, designed to predict acute seizure risk in encephalitis, demonstrated limited global applicability with a basic AUROC of 0.58 and an advanced AUROC of 0.63.
- The score performed acceptably in Western Europe, specifically in Portugal (Advanced AUROC 0.83), suggesting it may be useful for clinical trial stratification in that region.
- Autoantibody-associated encephalitis and low Glasgow Coma Scale (GCS) scores were strongly associated with inpatient seizures, whereas fever showed no significant predictive value.
- The study underscores the necessity for localized, tailored risk models that account for regional variations in encephalitis etiologies and diagnostic infrastructure.
Introduction: The Clinical Imperative in Encephalitis Management
Encephalitis, characterized by inflammation of the brain parenchyma, remains a complex neurological emergency with high rates of morbidity and mortality. Among its most debilitating complications are acute seizures, which occur in a significant proportion of patients and are known to exacerbate secondary brain injury, prolong hospital stays, and worsen long-term functional outcomes. Identifying patients at the highest risk for these seizures is critical for determining who might benefit from early, targeted anti-seizure medication (ASM) prophylaxis.
To address this clinical need, the SEIZUre Risk in Encephalitis (SEIZURE) score was developed and initially validated in United Kingdom cohorts. The scoring system includes a ‘basic score’—incorporating age, fever, and the Glasgow Coma Scale (GCS)—and an ‘advanced score’ that adds specific etiological data. However, for a clinical prediction tool to be truly transformative, it must demonstrate robust performance across diverse global populations. The recently published international evaluation in BMJ Open sought to test the SEIZURE score’s applicability across 13 countries and all WHO regions.
Study Design and International Scope
This international evaluation study employed a retrospective analysis of 2,032 patients across 19 organizations in 13 countries. The researchers aimed to validate both the basic and advanced iterations of the SEIZURE score. Of the total cohort, 1,324 patients were eligible for SEIZURE score calculations, and 970 were included in the final regression modeling. This massive undertaking involved the Global NeuroResearch Coalition and spanned varied healthcare environments, from high-resource settings in Western Europe and North America to diverse settings in Asia and South America.
The primary outcome was the occurrence of inpatient seizures. The researchers utilized univariate analysis to identify clinical features associated with seizures, followed by multivariable logistic regression and hierarchical clustering. The discriminatory ability of the scores was assessed using the area under the receiver operating characteristic curve (AUROC) with 95% confidence intervals (CIs).
Key Findings: A Tale of Geographic Divergence
Global Discriminatory Performance
The overarching finding of the study was that the SEIZURE score possesses limited discriminatory ability when applied on a global scale. The basic score yielded an AUROC of 0.58 (95% CI 0.55 to 0.62), while the advanced score, which includes etiology, only improved slightly to 0.63 (95% CI 0.60 to 0.66). These values fall below the threshold typically required for a clinical tool to be considered reliable for bedside decision-making in a general population.
Predictors and Associations
Interestingly, the study found that certain factors previously thought to be predictive did not hold up globally. Fever, a core component of the basic score, showed no significant association with inpatient seizures in this international cohort. Conversely, autoantibody-associated encephalitis (such as NMDAR encephalitis), a low GCS at presentation, and presenting with a seizure were frequently associated with subsequent inpatient seizures. These findings suggest that the clinical profile of ‘high-risk’ patients may be more dependent on the underlying biological cause of inflammation than on general systemic symptoms like fever.
The Western European Exception
A striking aspect of the results was the geographic variation in performance. The SEIZURE score performed significantly better in Western Europe (excluding Spain). The highest performance was observed in Portugal, where the basic AUROC reached 0.82 (95% CI 0.69 to 0.94) and the advanced AUROC reached 0.83 (95% CI 0.72 to 0.95). This regional success suggests that the score captures the clinical dynamics of encephalitis in these specific environments more effectively than in others.
Mechanistic Insights: Why Does the Score Fail Globally?
The Role of Aetiological Heterogeneity
One of the primary reasons cited for the score’s poor global performance is the vast difference in encephalitis etiologies across regions. In many parts of the world, infectious causes (such as Japanese encephalitis, malaria, or tuberculosis) predominate, whereas in the UK cohorts where the score was developed, autoimmune and common viral causes (like HSV) are more prevalent. Furthermore, the high percentage of ‘unknown’ etiologies in some regions limits the utility of the advanced score, which relies on definitive diagnosis. The lack of access to advanced diagnostics, such as next-generation sequencing or comprehensive antibody panels, likely contributes to this data gap.
Clinical Predictors: Fever vs. GCS
The failure of fever as a predictor in this global study is noteworthy. While fever is a hallmark of many infectious encephalitides, it may be less common or more transient in autoimmune cases. Additionally, the timing of fever relative to hospital presentation and the use of antipyretics can confound its predictive value. In contrast, GCS proved to be a more stable indicator of neurological compromise and seizure risk across different settings, reflecting the severity of cortical and subcortical involvement.
Expert Commentary: Implications for Clinical Trials and Practice
From a clinical perspective, these results provide a cautionary tale about the ‘universal’ application of risk scores. While the SEIZURE score may not be ready for global implementation, its performance in Western Europe is promising. Experts suggest that in these specific regions, the score could be used to design clinical trials for primary anti-seizure prophylaxis. By identifying a high-risk subgroup, researchers can more efficiently test whether early intervention with ASMs improves long-term outcomes without exposing low-risk patients to unnecessary medication side effects.
However, for the rest of the world, the study highlights a critical need for ‘tailored’ or ‘localized’ scoring systems. A model that works in a region dominated by viral infections may need to prioritize different variables than one used in a region where autoimmune disorders are the leading cause of brain inflammation. Furthermore, the study calls for a global push toward better etiological testing. Without knowing the cause of encephalitis, clinicians are often ‘flying blind,’ making accurate risk prediction nearly impossible.
Conclusion: The Need for Localized Precision
The international evaluation of the SEIZURE score serves as a vital reminder of the complexities inherent in global neurology. While the tool did not meet the threshold for clinical utility in most regions, the Western European results offer a foundation for future prospective trials. Moving forward, the focus must shift toward large-scale, prospective studies that utilize optimized etiological testing and account for regional variations in disease presentation. Until then, clinicians should use the SEIZURE score judiciously, recognizing its limitations outside the specific contexts in which it has shown success. The ultimate goal remains a precision medicine approach where risk is stratified not just by clinical symptoms, but by a deep understanding of the underlying inflammatory process.
References
1. Hughes T, Venkatesan A, et al. International evaluation of the SEIZUre Risk in Encephalitis (SEIZURE) score for predicting acute seizure risk. BMJ Open. 2025;15(12):e099451. doi: 10.1136/bmjopen-2025-099451.
2. Michael BD, et al. The SEIZUre Risk in Encephalitis (SEIZURE) score: a prediction tool for acute seizures. J Clin Med. 2021.
3. Venkatesan A, et al. Diagnosis and management of acute encephalitis: A review. JAMA. 2018;319(6):588-601.
