Abstract
BACKGROUND: Venous thromboembolism (VTE) is a leading cardiovascular disease, yet its etiology is incompletely understood. This study used large-scale, high-throughput aptamer-based proteomics to identify new circulating protein biomarkers and biological pathways for incident VTE. METHODS: We included 4 longitudinal cohorts (the ARIC study [Atherosclerosis Risk in Communities], CHS [Cardiovascular Health Study], MESA [Multi-Ethnic Study of Atherosclerosis], and the HUNT study [Trøndelag Health]) that identified 1371 incident noncancer VTEs among 20 737 participants followed for a maximum of 10 to 29 years. We used the SomaScan to measure baseline plasma levels of ≈5000 to 7000 proteins and examined the prospective relationships between the protein biomarkers and noncancer VTE. We then conducted an external replication of top VTE proteins in 783 incident noncancer VTEs among 39 097 participants in the UKB study (UK Biobank) based on the Olink proteomics platform. We used Cox proportional hazards regression to estimate the association between each protein biomarker and VTE risk. Mendelian randomization (MR) analysis was used to assess the possible causal associations between identified proteins and VTE risk. RESULTS: There were 23 proteins that exceeded a false discovery rate-adjusted P<0.05 (unadjusted P<6.5×10(-4)) in the discovery meta-analysis of ARIC, CHS, and MESA and were replicated in HUNT at (unadjusted) P<0.05. Of these, 15 are new to VTE, and 3 of the 15 (transgelin, sushi, von Willebrand factor type A, EGF and pentraxin domain-containing protein 1, and TIMP4 [metalloproteinase inhibitor 4]) exceeded the Bonferroni corrected significance threshold in HUNT. Sixteen of the 23 top VTE proteins were available on the UKB Olink panel, of which 11 were replicated in the UKB study after Bonferroni correction. MR analysis of the 15 new proteins provided significant evidence for a possible causal role of TIMD4 (T-cell immunoglobulin and mucin domain-containing protein 4) (Bonferroni-corrected P<0.05) and suggestive evidence for TIMP4 and CST3 (cystatin-c) (unadjusted P<0.05) in VTE risk. The direction of association from the MR analyses was opposite of that from the VTE proteomics analysis for TIMP4 and TIMD4 but was consistent for CST3. CONCLUSIONS: We identified several novel plasma proteins for VTE that reflect biological processes outside established VTE pathophysiology, including extracellular matrix regulation, immunity, immune-vascular endothelium interactions, and vascular senescence. Results may provide new modifiable targets to improve VTE risk stratification, prevention, or treatment.