DNA methylation GrimAge version 2.
Level 3 - non-randomized controlled study
Multi-cohort observational validation study assessing mortality and morbidity prediction
PubMed 36516495 · doi:10.18632/aging.204434
What was done
The authors developed DNAm GrimAge version 2, trained on individuals aged 40 to 92 years, incorporating two new DNA methylation-based surrogate estimators of plasma proteins: log-transformed high-sensitivity C-reactive protein (logCRP) and log-transformed hemoglobin A1C (logA1C). They evaluated GrimAge2 across 13,399 blood samples from nine cohort studies, testing its performance after age and sex adjustment for predicting all-cause mortality across racial and ethnic groups, associations with age-related conditions (coronary heart disease, lung function via FEV1, computed tomography-measured fatty liver disease), applicability to younger individuals, and utility in saliva samples.
What was found
After adjustment for age and sex, GrimAge2 outperformed the original GrimAge in predicting mortality across multiple racial and ethnic groups (meta P = 3.6x10^-167 versus P = 2.6x10^-144). GrimAge2 was inversely correlated with lung function (FEV1 correlation = -0.31, P = 1.1x10^-136). DNAm logCRP positively correlated with morbidity count (P = 1.3x10^-54), and DNAm logA1C was strongly associated with type 2 diabetes (P = 5.8x10^-155). DNAm PAI-1 outperformed GrimAge2 in correlating with triglycerides (correlation = 0.34, P = 9.6x10^-267) and visceral fat (correlation = 0.41, P = 4.7x10^-41). GrimAge2 was also reported to apply to younger individuals and to track metabolic syndrome markers in saliva.
Why it matters
GrimAge version 2 refines epigenetic mortality risk prediction by integrating DNA methylation proxies for systemic inflammation and glycemic regulation across diverse cohorts.
Limits
The abstract reports extreme P-values and select bivariate correlations but omits effect sizes (such as hazard ratios or predictive accuracy metrics like AUC or C-indices) for mortality prediction. Training was limited to individuals aged 40 to 92, and the observational design cannot establish whether modifying these methylation markers alters clinical outcomes.
Cited by
- supports GrimAge is particularly effective at predicting cardiovascular disease mortality risk.