When Genetic Risk Meets Childhood Obesity: What UK Biobank Reveals About Gene–Environment Interactions in Multiple Sclerosis

Multiple sclerosis (MS) is a complex disease: it does not arise from a single “MS gene” or a single exposure, but from a layered architecture in which inherited susceptibility and environmental inputs jointly shape risk. Jacobs and colleagues used the UK Biobank—one of the largest deeply genotyped population cohorts—to ask a precise question: does a person’s genome-wide genetic liability modify the impact of established early-life environmental risk factors for MS?

Rather than focusing only on the well-known HLA region (major histocompatibility complex, MHC), they explicitly tested whether polygenic risk outside HLA also participates in gene–environment interplay, a plausible contributor to the “missing risk” that remains even after major GWAS discoveries.

Study Design in Plain Scientific Terms: Cases, Controls, and Carefully Timed Exposures
The authors identified ~2,250 MS cases and ~486,000 controls in UK Biobank (ascertained via ICD codes, self-report, primary care records, and death registers), excluding individuals diagnosed before age 20 to reduce ambiguity about whether exposures preceded disease onset.

They focused on early-life/adolescent factors to further mitigate reverse causation: for example, childhood body size at age 10 (a proxy for childhood obesity), smoking before age 20, and timing of puberty-related measures like age at menarche.

Each exposure was tested in multivariable logistic regression models adjusted for major confounders (age, sex where appropriate, ethnicity, birth latitude, and socioeconomic deprivation), and statistical stringency was applied with Bonferroni correction across 10 exposures.

Which Environmental Factors “Show Up” in This Cohort?
In this dataset, three early-life factors emerged with strong evidence of association with MS: higher childhood body size at age 10, smoking before age 20, and earlier age at menarche (in women).

The effect sizes are modest but consistent with MS epidemiology: for example, participants reporting being “plumper” at age 10 had higher odds of MS than those reporting being “thinner,” and early smoking carried increased odds as well (Table 1).

A key visual summary is the forest plot (Figure 1, page showing the plot) where these signals stand out against several exposures that do not show convincing associations here (e.g., month of birth, breastfeeding, maternal smoking), underscoring that even large cohorts do not necessarily reproduce every previously reported relationship—often because of measurement differences, selection, or recall limitations.

Building Polygenic Risk Scores: Separating the MHC From the Rest of the Genome
To quantify inherited susceptibility, Jacobs et al. constructed 64 polygenic risk scores (PRS) using a standard clumping-and-thresholding approach with weights from the largest MS GWAS available at the time, then chose the best-performing models in a training subset and evaluated them in a held-out testing subset.

Importantly, they created PRS including the MHC region (PRS_MHC) and excluding it (PRS_non-MHC) to see whether any interaction signal was merely “HLA in disguise.” The strongest PRS were clearly associated with MS in the validation cohort, with PRS_MHC explaining more variance than PRS_non-MHC—as expected given the outsized effect of HLA—but both retained significant predictive value (Figure 2 and accompanying results).

They also showed a monotonic increase in MS odds across PRS deciles and reasonable discrimination (ROC curves; Figure 3), but they did not find relationships between PRS and proxies of disease course such as age at first report or claiming disability benefits.

The Central Result: Childhood Obesity and Polygenic Risk Interact
The study’s headline finding is a statistically robust additive interaction between childhood body size and polygenic risk, observed both when the PRS includes the MHC and when it excludes the MHC.

The authors quantify additive interaction using the attributable proportion due to interaction (AP)—a metric asking, in effect, “how much of the combined effect exceeds what we would expect by simply summing the separate effects?” They report AP ≈ 0.17 for PRS×childhood body size on both PRS definitions, with confidence intervals excluding zero and p-values surviving multiple testing (Table 2; Figure 4A).

In contrast, they found little evidence for multiplicative interaction, and the suggestive signal between PRS_MHC and age at menarche did not clear their multiple-testing bar (Table 2).

Conceptually, the result implies that genetic background can amplify the adverse association of childhood adiposity with MS risk, consistent with a model in which early metabolic/immune perturbations interact with many small-effect immune variants—not solely HLA.

A Secondary Genetic Insight: Polygenic Background May Modulate HLA-DRB1*15:01
Beyond environment, the authors also explored whether non-MHC polygenic risk modifies the effect of the high-impact HLA allele DRB1*15:01, analogous to “polygenic modulation” seen in other disease contexts.

They observed evidence of additive interaction between DRB1*15:01 carriage and PRS_non-MHC (AP reported around 0.24 with strong statistical support; Figure 4B), but again not on the multiplicative scale.

If replicated, this supports a biologically intuitive framing: HLA-mediated antigen presentation may be necessary but not sufficient, and the broader immune-genetic background may condition how strongly that risk allele “expresses” at the phenotype level.

Interpretation, Limitations, and Why This Matters for Prevention Science
The authors are careful about what these analyses do—and do not—prove. Statistical interaction is not automatically biological interaction, and UK Biobank brings familiar caveats: retrospective self-reported exposures (notably childhood body size), potential misclassification of MS status, limited ancestral diversity (predominantly White European), and selection/collider bias from the volunteer cohort structure.

They also emphasize that splitting UK Biobank into training/testing subsets does not replace true external replication.

Still, the work is valuable because it operationalizes an actionable idea: preventive interventions in rare, complex diseases are hard to power, but risk stratification—identifying those most likely to benefit—can make trials more feasible.

If childhood obesity truly has a stronger effect among those with high polygenic susceptibility, then targeting obesity prevention and metabolic health early in life might yield the largest marginal risk reduction precisely in that genetically enriched subgroup—an approach that aligns with precision prevention rather than generic population messaging.

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:
Jacobs, B. M., Noyce, A. J., Bestwick, J., Belete, D., Giovannoni, G., & Dobson, R. (2021). Gene-environment interactions in multiple sclerosis: a UK Biobank study. Neurology: Neuroimmunology & Neuroinflammation, 8(4), e1007.