Development and validation of a predictive score for venous thromboembolism in newly diagnosed non-small cell lung cancer.

INTRODUCTION

The risk of venous thromboembolism (VTE) varies among tumour types, and different cancer type-specific risks for VTE prediction remain undefined. We aimed to establish a prediction model for non-small lung cancer (NSCLC)-associated VTE.

MATERIALS AND METHODS

We analysed data from a prospective cohort of patients with newly diagnosed NSCLC. We then developed a VTE risk prediction model using data of patients who were recruited from 2013 to 2017 (n = 602, development cohort) and validated this model using date of patients recruited from 2018 to 2019 (n = 412, validation cohort). The cumulative 6 months VTE incidence observed in both cohorts was calculated.

RESULTS

The parameters in this new model included Eastern Cooperative Oncology Group (ECOG) performance status ≥2 (1 point), EGFR mutation (-1 point), neutrophil count ≥7.5 × 10/L (2 points), hemoglobin <115 g/L (1 point), CEA ≥5.0 ng/mL (2 points), and D-dimer level ≥1400 ng/mL (4 points). The cross-validated concordance indices of the model in the development and validation cohorts were 0.779 and 0.853, respectively. Furthermore, the areas under the curve in the two cohorts were 0.7563 (95% confidence interval [CI]: 0.6856-0.8129, P < 0.001) and 0.8211 (95% CI: 0.7451-0.8765, P < 0.001) for development and validation cohorts, respectively.

CONCLUSIONS

The new VTE risk prediction model incorporated patient characteristics, laboratory values, and oncogenic status, and was able to stratify patients at high risk of VTE in newly diagnosed NSCLC within 6 months of diagnosis.