Real-world data emulating randomized controlled trials of non-vitamin K antagonist oral anticoagulants in patients with venous thromboembolism

Abstract

Background: Emulating randomized controlled trials (RCTs) by real-world evidence (RWE) studies would benefit future clinical and regulatory decision-making by balancing the limitations of RCT. We aimed to evaluate whether the findings from RWE studies can support regulatory decisions derived from RCTs of non-vitamin K antagonist oral anticoagulants (NOACs) in patients with venous thromboembolism (VTE).

Methods: Five landmark trials (AMPLIFY, RE-COVER II, Hokusai-VTE, EINSTEIN-DVT, and EINSTEIN-PE) of NOACs were emulated using the South Korean nationwide claims database (January 2012 to August 2020). We applied an active comparator and new-user design to include patients who initiated oral anticoagulants within 28 days from their VTE diagnoses. The prespecified eligibility criteria, exposure (each NOAC, such as apixaban, rivaroxaban, dabigatran, and edoxaban), comparator (conventional therapy, defined as subcutaneous heparin followed by warfarin), and the definition of outcomes from RCTs were emulated as closely as possible in each separate emulation cohort. The primary outcome was identical to each trial, which was defined as recurrent VTE or VTE-related death. The safety outcome was major bleeding. Propensity score matching was conducted to balance 69 covariates between the exposure groups. Effect estimates for outcomes were estimated using the Mantel-Haenszel method and Cox proportional hazards model and subsequently compared with the corresponding RCT estimates.

Results: Compared to trial populations, real-world study populations were older (range: 63-69 years [RWE] vs. 54-59 years [RCT]), with more females (55-60.5% vs. 39-48.3%) and had a higher prevalence of active cancer (4.2-15.4% vs. 2.5-9.5%). The emulated estimates for effectiveness outcomes showed superior effectiveness of NOAC (AMPLIFY: relative risk 0.81, 95% confidence interval 0.70-0.94; RE-COVER II: hazard ratio [HR] 0.60, 0.37-0.96; Hokusai-VTE: 0.49, 0.31-0.78; EINSTEIN-DVT: 0.54, 0.33-0.89; EINSTEIN-PE: 0.50, 0.34-0.74), when contrasted with trials that showed non-inferiority. For safety outcomes, all emulations except for AMPLIFY and EINSTEIN-DVT yielded results consistent with their corresponding RCTs.

Conclusions: This study revealed the feasibility of complementing RCTs with RWE studies by using claims data in patients with VTE. Future studies to consider the different demographic characteristics between RCT and RWE populations are needed.

Keywords: Anticoagulants; Clinical trials; Epidemiologic methods; Factor Xa inhibitors; Venous thromboembolism.

Fig. 1

Cumulative incidence of the effectiveness and safety outcomes in emulation studies. A AMPLIFY emulation B RE-COVER II emulation C Hokusai-VTE emulation D EINSTEIN-DVT emulation E EINSTEIN-PE emulation

Fig. 1

Cumulative incidence of the effectiveness and safety outcomes in emulation studies. A AMPLIFY emulation B RE-COVER II emulation C Hokusai-VTE emulation D EINSTEIN-DVT emulation E EINSTEIN-PE emulation

Fig. 2

Effect estimates and evaluation of agreement between findings from RCTs and corresponding RWE studies. Abbreviations: RCT, randomized controlled trial; RWE, real-world evidence; Std, standardized difference; RA, regulatory agreement; EA, estimate agreement; SD, standardized differences; HR, hazard ratio; CI, confidence interval. Note: AMPLIFY yielded relative risk using the Mantel–Haenszel method; therefore, the corresponding emulation also yielded relative risk using the same method. White circles indicate HR from RWE studies, and black circles indicate HR from RCT. The predefined non-inferiority margin was 1.8 for AMPLIFY and RE-COVER II trial, 1.5 for Hokusai-VTE trial, and 2.0 for EINSTEIN-DVT and EINSTEIN-PE trial