Hypothesis

Genetically predicted overweight accelerates epigenetic aging, as measured by DunedinPACE, primarily through a chronic low‑grade inflammatory pathway involving IL‑6‑driven NLRP3 inflammasome activation in adipose tissue. This causal effect is confounded by survival bias and operates partly via frailty as a mediator, such that unadjusted MR estimates overstate the direct impact of BMI on aging.

Mechanistic Rationale

• Adipose tissue expansion recruits macrophages that secrete IL‑6, activating the NLRP3 inflammasome and increasing systemic IL‑6 and CRP levels [1].
• Persistent IL‑6 signaling promotes SASP (senescence‑associated secretory phenotype) in endothelial and immune cells, accelerating epigenetic clocks such as DunedinPACE [2].
• Frailty shares inflammatory pathways with overweight; elevated IL‑6 predicts both higher frailty index and faster epigenetic aging, positioning frailty as a mediator rather than a pure confounder [3][4].
• Survival bias in elderly cohorts removes individuals with high genetic BMI risk who succumb earlier, distorting MR estimates [5].

Testable Predictions

1. In a multivariable MR model that includes genetically predicted IL‑6 levels as a covariate, the causal effect of BMI on DunedinPACE will attenuate by ≥40% compared to univariable MR.
2. The indirect effect of BMI on DunedinPACE mediated through IL‑6 (estimated via product‑of‑coefficients) will be significant and account for the majority of the total effect.
3. Adjusting for time‑varying frailty in a g‑formula analysis will further reduce the estimated BMI effect, indicating partial mediation through frailty.
4. Sensitivity analyses using survival‑weighted MR (inverse probability of censoring weighting) will yield effect estimates closer to those from the g‑formula, correcting for survival bias.

Study Design

• Data source: UK Biobank participants with baseline BMI, genotype, circulating IL‑6/CRP, frailty index, and longitudinal DunedinPACE measurements (baseline and follow‑up at ~5‑year intervals).
• Step 1 – Multivariable MR: Use BMI‑associated SNPs as instruments; include IL‑6‑associated SNPs as a second exposure. Estimate direct (BMI→DunedinPACE) and indirect (BMI→IL‑6→DunedinPACE) effects via multivariable inverse‑variance weighting.
• Step 2 – Mediation quantification: Apply the product‑of‑coefficients method to obtain the proportion mediated by IL‑6, bootstrapping confidence intervals.
• Step 3 – g‑formula: Model time‑varying BMI, IL‑6, frailty, and comorbidities as inputs to a parametric g‑formula that simulates interventions (e.g., 1‑unit BMI reduction) on DunedinPACE trajectories, adjusting for time‑dependent confounding and censoring.
• Step 4 – Survival weighting: Compute inverse probability of censoring weights based on baseline health and genetic risk to re‑weight MR estimates, comparing them to g‑formula outputs.

Potential Pitfalls and Mitigations

• Weak instrument bias: Verify F‑statistics >10 for BMI and IL‑6 instruments; use LIML or modal‑based estimation if needed.
• Horizontal pleiotropy: Conduct MR‑Egger and weighted median analyses; employ MR‑PRESSO to detect and correct outlier SNPs.
• Measurement error in frailty: Use frailty index derived from deficit accumulation, validated against clinical outcomes.
• Limited IL‑6 genetic variance: Combine IL‑6 SNPs with cis‑acting eQTLs for IL6R to strengthen instrument strength.

If the multivariable MR shows a substantial attenuation of the BMI effect after accounting for IL‑6, and the g‑formula reveals a smaller, mediated effect once frailty and time‑varying confounding are addressed, the hypothesis will be supported. Conversely, if BMI’s effect on DunedinPACE remains robust and unmediated, the inflammatory pathway hypothesis will be falsified, prompting exploration of alternative mechanisms (e.g., lipid‑induced oxidative stress or hemodynamic stress).