Large-scale plasma proteomic analysis identifies pathophysiological mechanisms and predictive biomarkers for cardiac arrest

Objectives
Cardiac Arrest (CA) is a serious event that threatens life. However, the early proteomic characteristics of CA remain poorly understood. We sought to systematically identify CA-related proteins in a prospective cohort and construct a predictive model.
Methods
The plasma proteomic data were obtained from the UK Biobank cohort and included 2923 plasma proteins and 351 CA cases. The Cox proportional hazards regression model was used to investigate the association between proteins and CA. Locally estimated scatterplot smoothing was used to model the relationship between plasma protein levels and CA progression. Plasma proteins and clinical risk factors significantly associated with CA were used to build Cox proportional hazards regression models to predict the risk of CA events. Model discrimination was quantified by the area under the curve (AUC) of receiver operating characteristic (ROC).
Results
We identified 38 significant CA-related proteins over a 14-year follow-up period. The top notable proteins included GDF15, TNFRSF10B and IGFBP7, which exhibited abnormal levels 14 years prior to diagnosis and demonstrated a consistent upward trend. These CA-related proteins were primarily involved in biological functions such as collagen containing extracellular matrix, humoral immune response, and cytokine activity. Among them, 26 proteins exhibited abnormal fluctuations throughout the entire follow-up period, demonstrating strong predictive performance with an AUC of 0.822 in the test set when combined with demographic indicators to construct a predictive model.
Conclusions
Our study identified plasma protein fluctuation patterns 14 years before CA, providing insights into its pathogenesis and offering opportunities for early warning.