The decision to discontinue or continue anticoagulation for venous thromboembolism (VTE) after initial treatment is challenging because individual risks for recurrence and bleeding are heterogeneous. Therefore, one study investigated the development and external validation of models to predict the 5-year risk of recurrence and bleeding in patients with VTE without cancer who completed at least 3 months of initial treatment to estimate the individual absolute benefits and harms of prolonged anticoagulation.
The biggest challenge in treating patients with VTE is deciding on the duration of anticoagulant therapy. Primary treatment of VTE consists of three months of anticoagulation [2–4]. A primary treatment duration of more than three months, but with a limited duration, is not recommended, as this only delays the recurrence of thrombosis until treatment is discontinued [5]. The guidelines recommend weighing the risks of recurrent VTE and bleeding when making this decision. Anticoagulant treatment is effective in reducing the risk of recurrence of VTE but is associated with an annual risk of major bleeding of up to 2% [4]. Currently, the risk of VTE recurrence is estimated by categorizing patients as having unprovoked VTE, VTE triggered by minor or severe transient risk factors, or persistent risk factors [2]. Deciding on the duration of anticoagulant treatment in individual patients nevertheless remains difficult for several reasons: first, the risk of recurrence and bleeding varies among patients, even within the groups mentioned above. For example, there may be patients at high risk of bleeding in whom the benefits of prolonged treatment still outweigh the risks of bleeding because the risk of VTE recurrence is very high. Second, the guidelines do not provide recommendations on how exactly to assess and weigh the risks of recurrent VTE and bleeding [2–4]. Treatment decisions are based primarily on the presence of provoking factors as a way to categorize recurrence risk rather than absolute bleeding risk. However, bleeding can have a significant negative impact on quality of life and can also lead to mortality.
To improve clinical decision making, well-functioning models are needed to individually estimate the absolute risks of VTE recurrence and bleeding. Previous studies have shown that medical decision making can be personalized based on estimates of individual absolute treatment effects [6–8]. Such estimates can be obtained by combining predicted individual absolute risks with relative treatment effects from studies. Therefore, a recently published study [1] addressed the development and external validation of models for the prediction of (i) recurrent VTE; and (ii) major bleeding and clinically relevant non-major bleeding (CRNMB) within five years in patients with VTE without active cancer who have completed at least three months of primary anticoagulant treatment.
Results patient population
A total of 15 141 patients (mean age ± standard deviation 57.1 ± 15.8 years, 41% female, 69% with unprovoked VTE, and 49% with extended anticoagulation) were included in the combined data set for model derivation. In this population, 220 recurrences and 169 competing non-VTE-related deaths occurred during the median follow-up of 191 days (interquartile range: [IQR] 44-446 days). In addition, 737 bleeding events and 145 competing non-bleeding deaths occurred during a median follow-up of 189 days (IQR: 42-372 days).
Model derivation and validation studies
Competing risk-adjusted models were derived to predict recurrent VTE and clinically relevant bleeding (nonsevere and severe) using 14 readily available patient characteristics. Models were derived from combined individual patient data from the bleeding risk study, Hokusai-VTE, PREFER-VTE, RE-MEDY, and RE-SONATE (n=15 141, 220 recurrences, 189 bleeding events). External validity was assessed using the Danish VTE Cohort, EINSTEIN-CHOICE, GARFIELD-VTE, MEGA, and Tromsø studies (n=59 257, 2283 recurrences, 3335 bleeding events). Absolute treatment effects were estimated by combining the models with hazard ratios from trials and meta-analyses.
Primary end points were time to first recurrent VTE (recurrence model) and time to first bleeding event (bleeding model). Recurrent VTE was defined as objectively confirmed fatal or nonfatal recurrent deep vein thrombosis (DVT) or pulmonary embolism (PE) or a death to which PE contributed or could not be excluded. To capture all clinically relevant bleeding events, bleeding was defined as a composite of major bleeding and CRNMB according to the International Society on Thrombosis and Haemostasis [9,10].
Individual net benefit of extended anticoagulant treatment.
To estimate the absolute risks of treatment, the models can be combined with hazard ratios (HRs) for recurrent VTE and bleeding from trials and cohorts for different extended treatment strategies. The individual treatment effect within five years was calculated by subtracting the absolute risks with prolonged anticoagulation from the untreated risk, using data from the Danish VTE Cohort. Subsequently, the net benefit was calculated assuming that the risk of VTE recurrence and bleeding were equal.
Model derivation
The internal C statistic for the recurrent VTE model ranged from 0.51 to 0.79; overall, 0.68 (95% CI: 0.65-0.72). With bootstrapping, this value was 0.68 (95% CI: 0.62-0.73). The C statistic of 0.51 was observed in the REMEDY trial, in which all patients received extended anticoagulation and the range of predicted risks was limited. Internal C statistics for the bleeding model ranged from 0.65 to 0.73; overall, they were 0.69 (95% CI: 0.67-0.72) in both the main analysis and bootstrapping. The pooled coefficients derived from the internal-external cross-validation are similar to those derived in the total population for both models, although the coefficients from the internal-external cross-validation were smaller. The C statistics found from the internal-external cross-validation were also in the same range as in the overall population, although the pooled estimate for the recurrent VTE model was slightly lower.
External validation
C-statistics for the recurrent VTE model ranged from 0.48 (0.45-0.52) to 0.71 (0.66-0.77). Calibration plots showed agreement between predicted and observed risk for up to five years of follow-up. For GARFIELD-VTE, the predicted risks did not match the observed risks. For the bleeding risk score, C statistics ranged from 0.61 (0.54-0.67) to 0.68 (0.65-0.70). For both models, predicted risks were higher than observed risks in patients with higher predicted risks in the Danish VTE cohort, GARFIELD-VTE, and Tromsø study. The calibration plot for the bleeding risk score in the Tromsø study reflects a very small number of outcome events. When restricted to patients without or with prolonged anticoagulant treatment for recurrent VTE and bleeding, the predicted risks were higher and more homogeneous. Cohort studies were found to somewhat underestimate risks for recurrent VTE, whereas study populations (EINSTEIN-CHOICE and the combined model development population) somewhat overestimated risks.
Comparison with existing risk scores
In the overall population, discrimination of VTE PREDICT risk scores after adjustment for the effect of prolonged anticoagulation is comparable to other existing risk scores for recurrent VTE and bleeding. When restricted to a subset of patients without extended treatment (recurrent VTE) and to patients with extended treatment (bleeding), the point estimates of the pooled C-statistics for VTE-PREDICT were highest [0.61 (95% CI: 0.58-0.63) for recurrent VTE; 0.63 (95% CI: 0.61-0.64) for bleeding.
Individually predicted absolute benefits and harms
The absolute risks for recurrent VTE and bleeding within five years ranged from 3.8% to 19.1% for recurrent VTE and from 1.3% to 19.0% for bleeding. In the Danish VTE cohort, with prolonged treatment with full-dose direct oral anticoagulant (DOAC), the absolute predicted reduction in the risk of recurrent VTE is greater than the increase in the risk of bleeding in 77% of patients. Figure 1 [1] shows an example of how the VTE-PREDICT risk score can be used in clinical practice.
The VTE-PREDICT risk score, assessed using data from 15 141 VTE patients, estimated the absolute risk of recurrence and clinically relevant bleeding with and without prolonged anticoagulation. External validation in various clinical settings (n=59 257) showed good calibration up to five years. The VTE-PREDICT risk score can be used to estimate the benefits and harms of prolonged anticoagulant treatment in individual patients with VTE without cancer. The interactive calculator, available at https://vtepredict.com, facilitates use and supports shared decision making in clinical practice.
Literature:
- de Winter MA, et al: Recurrent venous thromboembolism and bleeding with extended anticoagulation: the VTE-PREDICT risk score. EurHeartJ 2023.
https://doi.org/10.1093/eurheartj/ehac776 - Ortel TL, Neumann I, Ageno W, et al: American Society of Hematology 2020 guidelines for management of venous thromboembolism: treatment of deep vein thrombosis and pulmonary embolism. Blood Adv 2020;4:4693-4738. https://doi.org/10.1182/bloodadvances.2020001830
- Kearon C, Akl EA, Ornelas J, et al: Antithrombotic therapy for VTE disease. Chest 2016;149: 315-352. https://doi.org/10.1016/j.chest.2015.11.026
- Konstantinides SV, Meyer G, Becattini C, et al: 2019 ESC guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European Respiratory Society (ERS). Eur Heart J 2019;41: 543-603. https://doi.org/10.1093/eurheartj/ehz405.
- Boutitie F, Pinede L, Schulman S, et al: Influence of preceding length of anticoagulant treatment and initial presentation of venous thromboembolism on risk of recurrence after stopping treatment: analysis of individual participants’ data from seven trials. BMJ 2011;342: d3036.
https://doi.org/10.1136/bmj.d3036. - Dorresteijn JAN, Visseren FLJ, Ridker PM, et al: Estimating treatment effects for individual patients based on the results of randomised clinical trials. BMJ 2011;343: d5888. https://doi.org/10.1136/bmj.d5888.
- Stam-Slob MC, Connolly SJ, van der Graaf Y, et al: Individual treatment effect estimation of 2 doses of dabigatran on stroke and major bleeding in atrial fibrillation. Circulation 2019; 139: 2846-2856. https://doi.org/10.1161/CIRCULATIONAHA.118.035266.
- de Vries TI, Eikelboom JW, Bosch J, et al: Estimating individual lifetime benefit and bleeding risk of adding rivaroxaban to aspirin for patients with stable cardiovascular disease: results from the COMPASS trial. Eur Heart J 2019;40: 3771-3778a. https://doi.org/10.1093/eurheartj/ehz404.
- Kaatz S, Ahmad D, Spyropoulos AC, Schulman S: Definition of clinically relevant non-major bleeding in studies of anticoagulants in atrial fibrillation and venous thromboembolic disease in non-surgical patients: communication from the SSC of the ISTH. J Thromb Haemost 2015;13:2119-2126. https://doi.org/10.1111/jth.13140
- Schulman S, Kearon C: Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost 2005;3:692-694. https://doi.org/10.1111/j.1538-7836.2005.01204.x.
CARDIOVASC 2023; 22(1): 42-43
