Cell-free plasma telomere length correlated with the risk of cardiovascular events using machine learning classifiers.

This retrospective study explored the association between circulating cell-free plasma telomere length (cf-TL) and coronary artery disease (CAD) and heart failure (HF). Data from 518 participants were collected, including clinical and laboratory data. cf-TL was measured in plasma samples and machine learning (ML) classification models were developed to differentiate between CAD, HF and control conditions. Our results showed that cf-TL was significantly prolonged in HF patients compared to controls, but no significant difference was observed between CAD patients and controls. Additionally, cf-TL was significantly correlated with nitric oxide metabolites (NOx) and flow-mediated dilation (FMD), suggesting a potential link with endothelial function. To avoid data leakage and ensure the model captured only relationships relevant to the research question, we utilized a temporal data split, holding out the last year's data for testing (n = 81) and using the remaining data for training (n = 324) and validation (n = 109). The ML models using four variables achieved an area under the curve (AUC) of 0.795 in the validation dataset and 0.717 in the test dataset for CAD classification, and 0.829 in the validation dataset and 0.806 in the test dataset for HF classification. SHAP analysis revealed that cf-TL had minimal impact on the predictions of the CAD model, as indicated by consistently low SHAP values, whereas in the HF model, cf-TL exhibited a broader range of SHAP values, indicating a greater contribution to the model's classification. These findings suggest that cf-TL may play a more prominent role in HF pathophysiology and could serve as a valuable biomarker for predicting HF risk. Further studies are warranted to explore cf-TL's diagnostic and prognostic potential across different cardiovascular diseases.