Genetic Risk Scoring for Predicting Multiple Sclerosis After Optic Neuritis
Optic neuritis is an inflammatory disorder of the optic nerve that typically presents with subacute visual loss, impaired colour vision, and pain during eye movement. Although it is frequently associated with multiple sclerosis, optic neuritis is clinically heterogeneous and may also arise from neuromyelitis optica spectrum disorder, sarcoidosis, vasculitis, infection, or other immune-mediated conditions. This distinction is clinically consequential: vision in multiple-sclerosis-associated optic neuritis often recovers substantially, whereas some non-multiple-sclerosis forms can cause rapid and irreversible axonal injury unless immunosuppressive treatment is initiated promptly. At the first presentation, however, the underlying cause may remain uncertain while clinicians await magnetic resonance imaging, serological testing, and cerebrospinal-fluid analysis. Loginovic and colleagues therefore investigated whether inherited susceptibility to multiple sclerosis could help stratify patients presenting with otherwise undifferentiated optic neuritis. Their study addresses an important decision point at which improved risk estimation could potentially support both acute ophthalmological management and earlier neurological follow-up.
Construction of a Multiple Sclerosis Genetic Risk Score
The investigators developed a multiple sclerosis genetic risk score, or MS-GRS, by combining risk variants identified through previous genome-wide association studies. The non-HLA component was based on 307 single-nucleotide polymorphisms outside the extended human leukocyte antigen region, weighted according to the logarithm of their published odds ratios. Because HLA variation contributes substantially to multiple sclerosis susceptibility and includes non-additive interactions, the researchers separately constructed an HLA score incorporating eight imputed HLA alleles and two regional variants. The final MS-GRS was obtained by summing the HLA and non-HLA components, thereby representing genetic susceptibility as a continuous quantitative measure rather than a binary classification. This design reflects the polygenic architecture of multiple sclerosis: no single common variant is sufficient for prediction, but the aggregate burden of many susceptibility alleles can distinguish groups with different average risks. Importantly, the score estimates predisposition rather than diagnosis and cannot determine whether an individual patient currently has multiple sclerosis.
Study Design and Population-Based Analysis
The primary analysis used genetic and longitudinal health-record data from 483,506 unrelated UK Biobank participants with suitable genotype and phenotype information. The researchers identified 2,369 individuals with multiple sclerosis and 687 with optic neuritis. Among the optic neuritis cases, 545 had no known multiple sclerosis diagnosis at their first recorded presentation and were therefore classified as having undifferentiated optic neuritis. During a median cumulative follow-up of 18.4 years, 124 of these 545 individuals—approximately 22.8%—were subsequently diagnosed with multiple sclerosis, with a median interval of 3.8 years between optic neuritis and diagnosis. The investigators assessed the MS-GRS both as a discriminator between established multiple sclerosis cases and controls and as a predictor of multiple-sclerosis-free survival after optic neuritis. Their survival model incorporated the genetic score together with sex and whether optic neuritis had occurred between 18 and 50 years of age, variables selected through Cox proportional-hazards modelling. The participant flow chart on page 2 of the article illustrates the quality-control process and the final diagnostic groups included in the analysis.
Genetic Burden Distinguished Multiple Sclerosis from Control Groups
The MS-GRS was significantly higher among participants with multiple sclerosis than among unaffected controls. In the UK Biobank, the full score produced a receiver-operating-characteristic area under the curve of 0.721 for distinguishing multiple sclerosis cases from healthy controls. When the score was combined with demographic variables, deprivation index, and genetic principal components, the area under the curve increased to 0.752. The distributions shown in the violin plots on page 4 demonstrate a graded pattern: controls had the lowest average score, participants with optic neuritis but no subsequent multiple sclerosis had an intermediate score, and those with multiple sclerosis—whether or not they had optic neuritis—had the highest scores. This pattern supports a degree of shared genetic architecture between optic neuritis and multiple sclerosis, while also indicating that isolated optic neuritis is biologically heterogeneous. Nevertheless, the overlap between distributions remained considerable, meaning that the MS-GRS is unsuitable as a stand-alone diagnostic test. Its principal value lies in modifying prior probability when interpreted alongside clinical and demographic information.
Prediction of Multiple Sclerosis Following Optic Neuritis
Within the group presenting with undifferentiated optic neuritis, each one-standard-deviation increase in the MS-GRS was associated with a 29% increase in the adjusted hazard of a future multiple sclerosis diagnosis, corresponding to a hazard ratio of 1.29 with a 95% confidence interval of 1.07–1.55. Female sex was independently associated with approximately twice the hazard, while optic neuritis onset between 18 and 50 years of age was associated with a hazard ratio of 2.43 relative to onset outside this age range. When the combined model was used to divide patients into quartiles of predicted risk, the differences were clinically substantial. By the end of follow-up, approximately 3.6% of individuals in the lowest-risk quartile had developed multiple sclerosis, compared with 14.7% in the second quartile, 31.6% in the third, and 41.2% in the highest-risk quartile. The Kaplan–Meier curves on page 6 show the progressive separation of these groups over time, while the calibration plots on page 5 indicate that predicted probabilities broadly corresponded to observed outcomes at 5-, 10-, and 20-year horizons.
External Validation and Potential Clinical Relevance
A major strength of the study was replication in two independent healthcare-linked genetic cohorts: Geisinger in the United States and FinnGen in Finland. Among participants with undifferentiated optic neuritis, 16.8% of the Geisinger cohort and 37.8% of the FinnGen cohort subsequently developed multiple sclerosis, illustrating substantial differences in baseline incidence and ascertainment across populations. After adjustment for these prevalence differences, the UK Biobank-trained model retained meaningful risk stratification. In Geisinger, multiple sclerosis developed in approximately 6.7% of the lowest-risk quartile and 30.6% of the highest-risk quartile; corresponding values in FinnGen were 10.2% and 60.7%. The genetic model also modestly improved time-dependent discrimination compared with models containing age and sex alone. Clinically, such stratification could help identify patients requiring expedited multiple sclerosis assessment, closer imaging surveillance, lumbar puncture, or early consideration of disease-modifying therapy. Conversely, a low predicted multiple sclerosis risk might increase suspicion of alternative inflammatory causes of optic neuritis. However, the model should complement—not replace—MRI findings, antibody testing, optical coherence tomography, neurological examination, and specialist judgment.
Limitations and the Path Toward Precision Neuro-Ophthalmology
Despite its promise, the model is not yet ready for routine clinical deployment. Most participants were of European ancestry, and polygenic scores developed predominantly in European populations may perform less accurately in other ancestry groups, potentially worsening existing disparities in genomic medicine. The UK Biobank also has recognised healthy-volunteer and age-selection biases, and diagnoses were derived largely from coded healthcare records rather than uniform prospective neuro-ophthalmological assessment. Information on brain MRI lesions, Epstein–Barr virus serostatus, contemporaneous vitamin D levels, and detailed optical coherence tomography measurements was unavailable or incomplete. Some optic neuritis cases may therefore have been misclassified, while the associations with age and sex may have been partially overfitted to the derivation cohort. Prospective studies must determine whether adding the MS-GRS to current diagnostic pathways improves treatment decisions, visual outcomes, time to multiple sclerosis diagnosis, and cost-effectiveness. The study nevertheless provides a compelling proof of concept: combining polygenic susceptibility with a small number of clinically accessible variables can generate materially different risk estimates after optic neuritis and may contribute to a more individualised model of neuro-inflammatory disease management.
Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.
References:
International Multiple Sclerosis Genetics Consortium., MultipleMS Consortium. Locus for severity implicates CNS resilience in progression of multiple sclerosis. Nature 619, 323–331 (2023). https://doi.org/10.1038/s41586-023-06250-x
