Blood-Based DNA Methylation Models Improve Short-term Cardiovascular Risk Stratification in Individuals With Type 2 Diabetes

Objective: Cardiovascular complications are the leading cause of death in type 2 diabetes mellitus (T2DM). We investigated whether blood-based DNA methylation (DNAm) biomarkers could improve risk stratification for major adverse cardiovascular events (MACE) in individuals with T2DM. Research design and methods: Epigenome-wide DNAm was profiled in peripheral blood mononuclear cells obtained at baseline from 305 T2DM participants of the Thiazolidinediones Or Sulphonylureas and Cardiovascular Accidents. Intervention Trial (TOSCA.IT) study, including 81 who experienced MACE within 5 years (T2DMACE) and 224 who did not (T2D). The population was randomly divided into training and internal test sets for model development and evaluation. Differential DNAm analyses were performed under five frameworks differing by covariate adjustment. Binomial least absolute shrinkage and selection operator (LASSO) logistic regression with stability selection was used to generate DNAm-based models to stratify individuals at higher risk of incident MACE over 5 years. Results: Differential DNAm analysis identified 23,839, 21,529, 21,500, 12,114, and 7,101 differentially methylated CpG sites across the five analytical frameworks. LASSO logistic regression was used to develop five DNAm-based models. Stability selection retained 22-26 CpG predictors per model. When evaluated in the internal test cohort, the five models showed high discriminative performance, with receiver operating characteristic area under the curve values ranging from 0.915 to 0.937 and area under the precision-recall curve values from 0.866 to 0.940. Several CpGs were independently associated with either increased or decreased risk of MACE, including 12 CpGs identified across all of the models. Conclusions: The DNAm-based models we have generated may improve risk stratification for short-term cardiovascular events in individuals with T2DM, providing a promising approach to refine identification of those at increased risk of MACE.