[Risk prediction of venous thromboembolism in non-small cell lung cancer patients based on COMPASS-CAT risk assessment model].

To verify the risk prediction efficacies of COMPASS-cancer associated thrombosis (COMPASS-CAT) risk assessment model and the new prediction probability model established based on COMPASS-CAT for venous thromboembolism (VTE) in hospitalized patients with non-small cell lung cancer (NSCLC). We retrospectively collected the clinical data of 373 patients with NSCLC admitted to National Clinical Research Center for Cancer/Cancer Hospital from March 2013 to June 2017. All of them were divided into VTE group (63 cases) and non-VTE group (310 cases) according to VTE occurred or not. According to the COMPASS-CAT risk assessment model, all patients were scored and classified by risk. Chi-square test was used to compare the clinical features between two groups, and multivariate logistic regression analysis was used to evaluate the independent risk factors of VTE in NSCLC patients. Based on the COMPASS-CAT risk assessment model, D-dimer≥1.03 mg/L and hemoglobin<10 g/dl were included to construct a new COMPASS-CAT prediction probability model, the ROC curve was also drawn. We used MedCalc software to compare the difference of ROC curves and analyze the application value of different risk assessment models in predicting VTE risk of NSCLC patients. The incidence of VTE was 16.9% (63/373). The COMPASS-CAT score of VTE group was 6.37±3.40, which was higher than 2.74±2.04 of non-VTE group (<0.001). Univariate analysis showed that the proportion of patients with KPS≤80, COMPASS-CAT≥7, D-dimer≥1.03 mg/L, central venous catheter (CVC), hemoglobin<10 g/L, cardiovascular complications≥2, hyperlipidemia, clinical stages Ⅲ and Ⅳ, KPS≤80 in VTE group was significantly higher than that in non-VTE group (<0.05). Logistic regression analysis showed that D-dimer≥1.03 mg/L, compass-cat score≥7 and hemoglobin <10 g/dL were independent risk factors for VTE. Based on the COMPASS-CAT risk assessment model, a new risk assessment model of COMPASS-CAT was constructed by incorporating the variables of D-dimer ≥1.03 mg/L and hemoglobin <10 g/dl. The area under ROC curve (AUC) of COMPASS-CAT model and new COMPASS-CAT prediction probability model were 0.745 and 0.804, respectively. Compared with COMPASS-CAT model, AUC of new COMPASS-CAT prediction probability model increased by 0.059, with statistically significant difference(=0.007). COMPASS-CAT risk assessment model can predict the risk of VTE in NSCLC patients, and the new COMPASS-CAT prediction probability model constructed by COMPASS-CAT model combined with D-dimer and hemoglobin variables can improve the accuracy of screening VTE risk factors in NSCLC patient.