Predicting Multiple Sclerosis Risk: Integrating Polygenic and Environmental Scores to Enable Prevention
In their 2022 mini-review in Frontiers in Neurology (doi: 10.3389/fneur.2021.761973), Hone and colleagues frame MS risk prediction as a practical necessity for prevention science rather than as an abstract modeling exercise. Because MS incidence is relatively low at the population level, prevention trials face a statistical and logistical bottleneck: either recruit extremely large cohorts or selectively enroll people whose baseline risk is meaningfully higher than average. The central promise of risk prediction—built from genetic and environmental information—is therefore trial enrichment: concentrating future preventive interventions in groups where enough MS outcomes will occur to detect benefit within feasible sample sizes and timelines.
The Genetic Architecture of MS: Strong Signals, Distributed Liability
The review emphasizes that MS susceptibility is polygenic: risk is shaped by many variants of small effect, with a notable exception in the major histocompatibility complex (MHC) region, where HLA alleles (particularly HLA-DRB1*15:01) have outsized influence. This architecture naturally invites quantitative aggregation—moving beyond single “risk alleles” toward a genome-wide liability perspective. Yet the same biology that makes genetic discovery successful (many loci contributing incremental risk) also complicates individual prediction: a person’s genotype can be statistically informative while still being insufficiently decisive to function as a standalone clinical classifier.
Polygenic Risk Scores: What They Capture—and What They Do Not
Polygenic risk scores (PRS) sum risk across variants to produce a single, individual-level measure. Across published efforts reviewed in the paper, PRS typically show moderate case–control discrimination, with reported AUCs spanning roughly 0.52 to 0.8 depending on cohort design, variant inclusion, and whether the MHC is modeled (the MHC often improves performance). The authors highlight a key interpretive point: AUC reflects relative separation between cases and controls, not the absolute probability that a given person will develop MS. In other words, a PRS can rank individuals by susceptibility while still providing limited actionable certainty for any one person—especially in a low-prevalence disease.
Environmental Risk Scores: The Needed Counterweight to Genetics
The review notes that multiple environmental exposures contribute to MS susceptibility, including low serum vitamin D, several dimensions of Epstein–Barr virus (EBV) exposure and immune response (e.g., infectious mononucleosis history, higher anti-EBV antibody titers, EBV seropositivity), childhood obesity, and smoking, alongside other proposed factors such as head injury, solvent inhalation, and shift work. Importantly, some environmental effects appear to be potentiated by genetic background—particularly by HLA-DRB1*15:01—underscoring that MS risk is not simply “genes plus environment,” but often “genes through environment” (and vice versa). Environmental risk scores (ERS), in principle, allow these exposures to be combined into a parallel quantitative axis that can complement PRS.
Hybrid Models and Reported Performance: Incremental Gains, Persistent Gaps
A major empirical takeaway in the article is that hybrid scores—PRS combined with ERS and/or covariates—often outperform PRS alone, with some cohorts reporting AUC improvements when environmental measures (and standard covariates) are integrated. The paper summarizes multiple studies comparing approaches, including large-scale biobank analyses where adding covariates and principal components (to manage ancestry structure) can materially shift performance metrics. Still, even when headline AUCs appear “impressive,” the authors caution that such numbers can obscure the real-world meaning of a test. The practical question is not only whether cases tend to score higher than controls, but whether a thresholded “high-risk” label would correctly identify future MS in enough people to justify downstream action.
The Core Practical Barrier: Low Prevalence and the PPV Reality Check
The review is particularly direct about why MS prediction is difficult to translate clinically: in low-prevalence settings, even tests with high sensitivity and specificity can yield low positive predictive value (PPV). The authors illustrate this with a concrete thresholding example in which a PRS cut-off captures all MS cases but also flags many healthy individuals, producing a PPV that remains modest because the base rate of MS is small. This is not merely a statistical footnote; it determines ethical and clinical feasibility. If most “high-risk” people will never develop MS, then intensive surveillance, anxiety, or preventive immunomodulation becomes hard to justify—whereas the same risk tool may still be very useful for research enrichment, where the goal is to increase event rates in a trial rather than to diagnose individuals.
Opportunities Ahead: From Risk Stratification to Prevention-Ready Cohorts
Hone et al. ultimately position MS risk scores as enabling infrastructure for prevention trials: tools to identify strata where disease incidence is higher and where interventions can be tested efficiently. They also outline why progress is plausible: expanding GWAS sample sizes, better exposure measurement (particularly for EBV and vitamin D), clearer modeling of gene–environment interaction, and rigorous validation across cohorts and ancestries. A realistic near-term endpoint is not a consumer-facing “MS prediction test,” but well-calibrated stratification pipelines that can recruit prevention-ready cohorts, support pragmatic screening strategies, and help answer a high-value question for the field: when risk is quantified using both inherited and acquired factors, which preventive levers meaningfully lower the probability of MS onset, and in whom?
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:
Hone, L., Giovannoni, G., Dobson, R., & Jacobs, B. M. (2022). Predicting multiple sclerosis: challenges and opportunities. Frontiers in Neurology, 12, 761973.
