Abstract
INTRODUCTION: Venous thromboembolism(VTE) is a leading cause of death in patients with lung cancer. Furthermore, hospitalization of patients with advanced lung cancer for VTE treatment represents a major economic burden on the national public health resources. Therefore, we performed this prospective study to identify clinical biomarkers for the early identification of VTE in lung cancer patients. METHODS: This prospective study enrolled 158 patients with confirmed lung cancer, including 27 who were diagnosed with VTE within six months of the follow-up after lung cancer diagnosis. Multivariate logistic regression analysis was used to evaluate the diagnostic performancese of all the relevant clinical features and laboratory indicators in identifying lung cancer patients with a higher risk of VTE. A novel risk prediction model was constructed consisting of five clinical variables with the best diagnostic performances and was validated using the receiver operation characteristic(ROC) curves. The diagnostic performances of the new risk prediction model was also compared with the Khorana risk score (KRS) and the Padua risk score (PRS). RESULTS: The VTE group of lung cancer patients (n = 27) showed significantly higher serum levels of fibrin degradation products (FDP), D-dimer, thrombomodulin (TM), thrombin-antithrombin-complex (TAT), α2-plasmin inhibitor-plasmin Complex (PIC), and tissue plasminogen activator-plasminogen activator inhibitor complex (t-PAIC) compared to those in the non-VTE group (n = 131). ROC curve analyses showed that the diagnostic efficacy of the new VTE risk prediction model with TM ≥ 9.75 TU/ml, TAT ≥ 2.25ng/ml, t-PAIC ≥ 7.35ng/ml, history of VTE, and ECOG PS score ≥ 2 was superior than the KRS and the PRS in the early identification of lung cancer patients with a higher risk of VTE. CONCLUSIONS: The new risk prediction model showed significantly high diagnostic efficacy in the early identification of lung cancer patients with a high risk of VTE. The diagnostic efficacy of the new risk prediction model was higher than the KRS and the PRS in this cohort of lung cancer patients.