Effect of tailoring anticoagulant treatment duration by applying a recurrence risk prediction model in patients with venous thromboembolism compared to usual care: A randomized controlled trial

Background Patients with unprovoked (i.e., without the presence of apparent transient risk factors such as recent surgery) venous thromboembolism (VTE) are at risk of recurrence if anticoagulants are stopped after 3-6 months, yet their risk remains heterogeneous. Thus, prolonging anticoagulant treatment should be considered in high-risk patients, whereas stopping is likely preferred in those with a low predicted risk. The Vienna Prediction Model (VPM) could aid clinicians in estimating this risk, yet its clinical effects and external validity are currently unknown. The aim of this study was to investigate the clinical impact of this model on reducing recurrence risk in patients with unprovoked VTE, compared to usual care. Methods and findings In a randomized controlled trial, the decision to prolong or stop anticoagulant treatment was guided by predicted recurrence risk using the VPM (n = 441) , which was compared with usual care (n = 442). Patients with unprovoked VTE were recruited from local thrombosis services in the Netherlands (in Utrecht, Harderwijk, Ede, Amersfoort, Zwolle, Hilversum, Rotterdam, Deventer, and Enschede) between 22 July 2011 and 30 November 2015, with 24-month follow-up complete for all patients by early 2018. The primary outcome was recurrent VTE during 24 months of follow-up. Secondary outcomes included major bleeding and clinically relevant non-major (CRNM) bleeding. In the total study population of 883 patients, mean age was 55 years, and 507 (57.4%) were men. A total of 96 recurrent VTE events (10.9%) were observed, 46 in the intervention arm and 50 in the control arm (risk ratio 0.92, 95% CI 0.63-1.35, p =0.67). Major bleeding occurred in 4 patients, 2 in each treatment arm, whereas CRNM bleeding occurred in 20 patients (12 in intervention arm versus 8 in control arm). The VPM showed good discriminative power (c-statistic 0.76, 95% CI 0.69-0.83) and moderate to good calibration, notably at the lower spectrum of predicted risk. For instance, in 284 patients with a predicted risk of >2% to 4%, the observed rate of recurrence was 2.5% (95% CI 0.7% to 4.3%). The main limitation of this study is that it did not enroll the pre-planned number of 750 patients in each study arm due to declining recruitment rate. Conclusions Our results show that application of the VPM in all patients with unprovoked VTE is unlikely to reduce overall recurrence risk. Yet, in those with a low predicted risk of recurrence, the observed rate was also low, s