How to predict recurrent venous thromboembolism and bleeding? A review of recent advances and their implications

After the first venous thromboembolism (VTE), anticoagulant treatment duration should be based on the balance between the risk of recurrence and bleeding. However, this decision is challenging on the individual level. Prediction models that accurately estimate these risks may help selecting patients that would benefit from either short or indefinite anticoagulant treatment. Currently, 17 models to predict VTE recurrence and 15 models to predict bleeding in VTE patients have been proposed. In addition, 7 models to predict bleeding in anticoagulated patients, mostly for atrial fibrillation, have been evaluated for use in VTE patients. Sex, age, type, and location of the index event and D‑dimer levels were the most often included predictors of recurrent VTE, whereas age, history of (major) bleeding, active malignancy, antiplatelet therapy, anemia, and renal insufficiency were most often used for the prediction of bleeding. In this review, we provide a summary of these models and their performance. Notably, these models are rarely used in clinical practice and none of them is incorporated in current guidelines due to insufficient accuracy or insufficient validation. Moreover, evidence supporting the value of implementing these models is still lacking. Before these models can be used in routine care, further refinement may be required, and their added value and feasibility should be proven in both management and implementation studies.