Li · GeroScience 2026 · prospective cohort study · n=947

DNA methylation-based surrogate markers of C-reactive protein and their associations with health-related traits.

Cited 1 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective longitudinal cohort study assessing biomarker variance, 10-year stability, and mortality risk.

PubMed 41571963 · doi:10.1007/s11357-025-02085-8 · record verified 2026-08-31

What was done

Analyzed blood samples from 947 participants in the Melbourne Collaborative Cohort Study at baseline (1990–1994) and wave 2 (2003–2007). Evaluated five DNA methylation-based CRP (mCRP) surrogate markers (mCRP-Ligthart, mCRP-Wielscher, mCRP-EpiScore, mCRP-GrimAge2, and mCRP-Hillary) alongside high-sensitivity plasma CRP. Stability over ~10 years was measured via intraclass correlation coefficients (ICCs). Associations of wave 2 markers with all-cause mortality (319 deaths) were evaluated using Cox models, and linear regression was used for BMI and age-adjusted PCGrimAge.

What was found

mCRP markers explained between 6.0% (mCRP-Wielscher) and 16.8% (mCRP-GrimAge2) of the variance in plasma CRP. Marker stability across a decade was comparable between plasma CRP and mCRP surrogates, with ICCs ranging from 0.44 (mCRP-GrimAge2) to 0.63 (mCRP-Wielscher). Compared to plasma CRP, mCRP markers showed stronger associations with PCGrimAge, but similar or weaker associations with BMI and mortality. Adjusting for PCGrimAge greatly attenuated mCRP mortality associations, whereas plasma CRP remained strongly associated with mortality even after adjusting for mCRP. Specific hazard ratios and effect estimates were not reported in the abstract.

Why it matters

Clarifies that epigenetic surrogates capture only a modest fraction of circulating CRP variance and that their association with mortality reflects broader biological aging rather than providing superior prognostic utility over directly measured plasma CRP.

Limits

Observational cohort design precludes causal inference. The abstract omits numerical effect sizes, hazard ratios, and confidence intervals for mortality and BMI models. Findings are derived from a single Australian cohort and may not generalize across diverse ancestries.

Cited by