In the RENOVE trial, non-inferiority of extended reduced-dose versus full-dose direct oral anticoagulant treatment to prevent recurrent thromboembolic events at 5-year follow-up was not shown in patients presenting with venous thromboembolism and a high risk of recurrence after anticoagulation discontinuation. Our findings show an unexpectedly low risk of recurrent venous thromboembolism in both groups and a 39% reduction in risk of major and clinically relevant non-major bleeding in the reduced-dose group, resulting in a 33% reduction in net adverse clinical events. 
This study enrolled patients with a strong indication for extended anticoagulation (ie, mostly first or recurrent episode of unprovoked venous thromboembolism) and severe clinical presentation of venous thromboembolism (ie, mostly pulmonary embolism). Male participants were over-represented, probably owing to the study definition of unprovoked venous thromboembolism. Indeed, because there is evidence that women who develop venous thromboembolism during hormonal exposure have a low risk of recurrence,28,29 they were not considered as having unprovoked thrombosis (appendix p 7), which resulted in the inclusion of only five women with venous thromboembolism during hormonal exposure. However, despite the high recurrence risk profile of this study population, the cumulative incidence of recurrent venous thromboembolism (about 2% at 5 years in both groups) was much lower than expected (ie, about 4% at 2 years). As the definition of recurrence used in the current trial was similar to that reported in the AMPLIFY-EXT17 and EINSTEIN-CHOICE trials,18 a plausible explanation for the lower-than-expected recurrence rate could be a good treatment adherence throughout follow-up for more than two-thirds of patients (appendix p 16), together with a low treatment discontinuation rate of 16% at 5 years (figure 1), compared with 14·7% at 1 year in AMPLIFY-EXT and 12·5% at 1 year in EINSTEIN-CHOICE. 
With respect to safety, we found a cumulative incidence of clinically relevant bleeding of 15·2% in the full-dose group at 5 years with a 39% risk reduction in the reduced-dose group. A similar reduction was observed in the incidence of major bleeding. These findings, observed in patients, 28·6% of whom were at high bleeding risk according to the venous thromboembolism bleed score,20 were close to those expected in our key secondary hypothesis.17,18 When evaluating net clinical benefit, the cumulative incidence rates were lower in the reduced-dose group than the full-dose group. Furthermore, we found no between-group differences in risk of arterial thrombotic events, incident cancer, or death, reinforcing the potential safety benefit of a lower dose not outweighed by serious adverse events. 
This study has a number of strengths. First, it reliably establishes that the rates of symptomatic recurrent venous thromboembolic events during anticoagulation were low, whether on reduced-dose or full-dose anticoagulants, even in this population at high recurrence risk after stopping anticoagulation and followed up for a median duration of 3 years. Second, the substantial reduction in bleeding risk with the reduced dose is consistent for both major and clinically relevant non-major bleeding events and, even if expected from real-life studies,30 this finding could not be confirmed in the EINSTEIN-CHOICE17 and AMPLIFIY-EXT18 randomised trials, which were neither designed nor powered for this purpose. Third, the pragmatic inclusion of patients treated with the two most commonly used direct oral anticoagulants in venous thromboembolic disease, namely apixaban and rivaroxaban, allowed indirect comparisons that showed consistent results across the two drugs on both recurrent venous thromboembolic and bleeding events rates (appendix pp 52–54). 
The limitations of this study include the open-label design, which might have influenced the primary outcome. However, there was no rationale to select one direct oral anticoagulant over another, and conducting a double-blind trial with multiple placebos would have challenged recruitment within the expected timeframe and long follow-up. Second, non-inferiority was not confirmed, but whether the full dose was more effective than the reduced dose remains undetermined. Although we cannot exclude that the benefit for major bleeding and clinically relevant non-major bleeding might be offset by the risk of recurrent venous thromboembolism, the low 5-year recurrence rate is reassuring for efficacy (5-year recurrence rates of 2·2% in the reduced-dose group and 1·8% in the full-dose group). In addition, subgroup analyses showed non-heterogeneity in treatment effects and did not identify subgroups for whom dose reduction would be harmful in terms of net clinical benefit (appendix pp 52–54). However, due to the limited power of subgroup analyses, a loss in the net clinical benefit cannot be excluded in some subgroups (eg, in participants with obesity or pulmonary embolism at intermediate–high risk or high risk of death; appendix pp 25–27). These findings might be useful to refine future guidelines and shared decision making processes for patients with venous thromboembolism who need extended anticoagulation. The wide eligibility criteria and the reported study population characteristics, as well as the definition of study outcomes, suggest that our findings might apply to a broader population and have clinically relevant external generalisability. 
In conclusion, although reduction of the direct oral anticoagulant dose in patients with venous thromboembolism requiring extended anticoagulation did not meet the study non-inferiority criteria, the rates of recurrent venous thromboembolism were low in both groups. In the reduced-dose group, clinically relevant bleeding and the composite of recurrent venous thromboembolism or clinically relevant bleeding were lower than in the full-dose group and did not appear to be offset by an increased risk of death or arterial thromboembolic events. Further research will be needed to identify subgroups for which the full dose might remain the preferred option.
