Construction and verification of a nomogram to predict the probability of venous thromboembolism in lung cancer patients.

Background: Lung cancer (LC) patients account for 20% of all cancer-related venous thromboembolism (VTE) events, which is the second leading cause of mortality in these patients. However, there are no LC-specific risk scores for VTE prediction. Therefore, we considered that an LC-specific nomogram would more accurately predict VTE probability for these patients than the current widely used VTE risk scores. Methods: A total of 676 patients from the Guizhou Provincial People's Hospital (January 2016 to September 2021) were included in this study, of which 169 LC patients who developed VTE were time-matched with 507 (1:3 ratio) LC patients without VTE. These patients were randomly divided at a 2:1 ratio to form primary (451) and validation (225) cohorts. The accuracy of six VTE risk scores was assessed by producing area under the receiver operating characteristic (ROC) curves (AUC). A multivariate analysis was employed to select predictive features, which were then used to construct a nomogram for VTE prediction. Results: Among the scoring methods, COMPASS-CAT had the highest AUC (0.799). The multivariate analyses revealed that acute infection, bed rest (>3 days), D-dimer level >1.47 μg/mL, adenocarcinoma, and carcinoembryonic antigen (CEA) >10.935 ng/mL were independent predictors of VTE risk for LC patients. The nomogram constructed using these factors enabled VTE prediction with a concordance index of 0.882 and an AUC of 0.894. Conclusion: This article describes the construction and validation of a new nomogram for VTE prediction in LC patients, which has a predictive performance that is higher than any of the widely used conventional risk assessment tools.