External validation of the VTE-PREDICT score for predicting recurrence and bleeding in venous thromboembolism: results from the Registro Informatizado Enfermedad TromboEmbὀlica Registry.

Abstract

Background: The VTE-PREDICT score predicts venous thromboembolism (VTE) recurrence and clinically relevant bleeding (CRB; major and clinically relevant nonmajor bleeding) after acute VTE.

Objectives: We aimed to externally validate the VTE-PREDICT score in the Registro Informatizado Enfermedad TromboEmbὀlica, a prospective registry of patients with VTE.

Methods: Exclusion criteria included enrollment before 2012, active cancer, and anticoagulation other than direct oral anticoagulants, vitamin K antagonists, or low-molecular-weight heparin. VTE recurrence and CRB risks were calculated using VTE-PREDICT for a prediction period of 3 months after the index VTE until the following 1 to 5 years. Predicted risks were then compared with observed risks. C-statistics and calibration plots were assessed.

Results: In total, 17 850 patients (50.3% women) were included in the final analysis, of whom 64.3% had pulmonary embolism. The median age was 67 years (IQR, 52-78). Regarding long-term anticoagulation, 21.8% of patients were treated with a direct oral anticoagulant, 39.9% with a vitamin K antagonist, and 4.8% with low-molecular-weight heparin, whereas 33.6% received no anticoagulant treatment. Cumulative incidences of VTE recurrence and CRB at 1 year were 3.7% (95% CI, 3.4%-4.0%) and 2.6% (95% CI, 2.4%-2.9%), respectively. The c-statistics of VTE-PREDICT for 1 to 5 years varied between 0.70 (95% CI, 0.67-0.72) and 0.73 (95% CI, 0.69-0.76) for VTE recurrence and between 0.65 (95% CI, 0.63-0.67) and 0.67 (95% CI, 0.64-0.70) for CRB. Calibration analysis revealed underestimation of VTE recurrence and overestimation of CRB risk.

Conclusion: VTE-PREDICT showed good discrimination for VTE recurrence and moderate discrimination for CRB, but underestimated the risk of VTE recurrence in high-risk patients.