Can Genetics Predict How Severe Multiple Sclerosis Will Become? A Reality Check from Real-World Clinical Data
Most genetic work in multiple sclerosis (MS) has focused on susceptibility—why one person develops MS and another does not—and that effort has now mapped hundreds of loci linked to risk. What people living with MS and their clinicians often want, though, is a different kind of answer: how fast will disability accumulate, and can we predict that early enough to change decisions? Two major progression/severity genome-wide association studies (GWAS) recently raised hopes by reporting genetic variants potentially tied to long-term disability—most notably rs10191329A as the first genome-wide significant “severity” single-nucleotide variant (SNV). Kreft and colleagues set out to test whether those headline GWAS findings actually hold up in a deeply phenotyped, real-world clinical cohort, and whether they are ready for individual-level prognostication.
The real-world cohort and what the team measured
The investigators leveraged the South Wales MS Registry, a prospective dataset initiated in 1985 with standardized longitudinal clinical capture, including neurologist-assessed Expanded Disability Status Scale (EDSS) scores obtained during in-person examinations. They analyzed 1,455 people with MS (excluding non-European ancestry to match prior GWAS validation populations), performed genotyping (multiple arrays), imputed genome-wide variants, and extracted rs10191329 plus additional candidate progression variants. Disability was quantified not only with raw EDSS milestones, but also with the age-related MS severity score (ARMSS), which adjusts disability for patient age to better compare trajectories across individuals.
The headline result: rs10191329A did not predict disability severity here
Across analyses designed to mirror prior GWAS designs—including a subgroup of 277 “older, longstanding MS” participants comparable to the International MS Genetics Consortium (IMSGC) severity cohort—rs10191329 genotype showed no meaningful association with ARMSS, age at onset, or sex distribution (illustrated in Figures 1–2). In the full 1,455-person cohort, regression modeling again found rs10191329A unable to predict ARMSS once relevant covariates were considered, despite similar allele frequencies to the discovery work. Importantly, the observed effect direction in this dataset trended opposite to the original report and was small, underscoring how fragile “severity signals” can be when moved from discovery settings to routine-care cohorts.
Long-term disability milestones and “extremes” analyses also came up negative
If a variant truly accelerates progression, it should show up when you track time-to-event outcomes that matter clinically: EDSS 4 (walking limitation), EDSS 6 (need for a walking aid), EDSS 8 (wheelchair/bed restriction), and conversion to secondary progressive MS (SPMS). Survival analyses (Kaplan–Meier and Cox models) found no significant differences by rs10191329 genotype for any of these endpoints (Figure 3). The authors then pushed harder with an “extremes of outcome” strategy: propensity-score matching 35 homozygous risk (AA) carriers to 140 homozygous non-risk (CC) carriers on sex, age at onset, disease course, and treatment exposure. Even under that design—built specifically to increase comparability—there were still no differences in early/late ARMSS or in time to EDSS milestones and SPMS.
No signal in relapse behavior or lesion localization either
A practical clinical hope is that a severity genotype might correlate with inflammatory activity (relapses) or with where disease “hits” in the nervous system—features that clinicians already use to judge aggressiveness. In relapse-onset MS, rs10191329A carriage was not associated with annualized relapse rate, time to a second clinical relapse, incomplete recovery after the first relapse (“fixed disability”), or the anatomical localization of onset/relapses when categorized by standard clinical domains (Figure 4). Taken together, the negative results suggest that—at least in this cohort—rs10191329 does not meaningfully tag a relapse-driven pathway to later disability, nor does it map cleanly onto recognizable clinical phenotypes.
What did replicate: two “suggestive” variants and the age-at-onset effect of HLA-DRB1*1501
While rs10191329 failed replication, the study did reproduce modest associations for two suggestive MSBase severity variants: rs7289446G (and its perfect-LD partner rs1207401) and rs868824C, each showing links to disability outcomes with relatively small effect sizes (Table 2). The authors also tested broader polygenic signals: a weighted genetic risk score (wGRS) built from established MS susceptibility loci (non-HLA plus HLA burden) did not associate with ARMSS, and a severity-focused wGRS based on a small set of severity SNVs likewise showed no relationship to ARMSS or EDSS milestone timing. Finally, they examined HLA-DRB1*1501 (via proxy rs3135388): it was not associated with long-term disability measures here, but it was associated with younger age at onset (Table 3), aligning with prior literature that HLA influences when MS starts more reliably than how fast disability progresses.
What this means for patients, clinicians, and MS genetics going forward
The most clinically consequential takeaway is straightforward: today, single-variant genotyping—especially rs10191329—does not appear ready to guide individualized MS management in routine practice, at least not on the basis of disability forecasting. The paper also explains why replication is hard: disability metrics like EDSS are nonlinear and weighted toward ambulation; measurement density and treatment eras vary across decades; and the homozygous risk genotype is rare, limiting power even in sizable cohorts. Yet the broader scientific message is optimistic: severity GWAS signals may still illuminate biology (potentially CNS resilience/neurodegeneration pathways), but they must be stress-tested across independent, representative cohorts with standardized outcomes before being translated into counseling, therapy selection, or “precision MS” risk communication.
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:
Kreft, K. L., Uzochukwu, E., Loveless, S., Willis, M., Wynford‐Thomas, R., Harding, K. E., ... & Robertson, N. P. (2024). Relevance of multiple sclerosis severity genotype in predicting disease course: a real‐world cohort. Annals of Neurology, 95(3), 459-470.
