Early stopping of a randomised trial for efficacy: a medical, ethical and regulatory challenge

Abstract

In 2017, the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) trial demonstrated the efficacy of a combination of lowdose rivaroxaban 2.5 mg two times per day and aspirin 100 mg one time per day in reducing ischaemic events in a broad population of highrisk patients with chronic coronary syndromes and/or peripheral arterial disease. The European Society of Cardiology therefore recommends this combination therapy for longterm secondary prevention in highrisk patients. COMPASS was prematurely terminated after a mean followup of 23 months due to a clear benefit in the combination arm. This decision, although ethically justified, raises important questions both in the context of this specific study and also in the context of clinical research in general. Dagenais et al led a post hoc analysis of the COMPASS trial to assess the effect of switching study treatment combination of lowdose rivaroxaban and aspirin to nonstudy aspirin after early stopping of the trial. Patients included in the rivaroxaban alone arm were not considered in the present analysis. Patients were followed from the time of the early stopping until last contact, which could be the final followup of the proton pump inhibitor randomisation, in August 2018, or the beginning of the longterm openlabel extension (LTOLE) study during which patients received the combination of rivaroxaban and aspirin for 3 years or until the medication was available on the market. This LTOLE study began after regulatory and ethical approval in different countries. Median followup duration was 1.02 years (0.74, 1.11). The main outcomes were similar to the main COMPASS analysis: a composite of cardiovascular (CV) death, myocardial infarction (MI), stroke, its individual components and limb events. Events were verified using a computerprogrammed algorithm or were adjudicated if the algorithm did not confirm the event. After discontinuing lowdose rivaroxaban and switching to nonstudy aspirin, patients had similar rates of the composite outcome, MI and CV death, or limb events, but experienced a higher risk of stroke (HR 1.74, 95% CI=1.05 to 2.87). Most of the excess strokes occurred during the first 6 months, with very few events in the first month after stopping rivaroxaban. The main message of this study is that discontinuing lowdose rivaroxaban in patients previously randomised to the combination treatment translated into a loss of CV protection and even increased the risk of stroke. Cessation, whether temporarily or permanently, of antithrombotic therapy is a reallife clinical challenge that cardiologists have to face frequently. In several populationbased studies, 4 discontinuation of aspirin was associated with an increased risk of CV events such as CV death, hospitalisation for MI or stroke. This was observed for patient both in secondary prevention and also, and interestingly, in primary prevention where the benefit of aspirin is yet to prove. This excess of CV events, including allcause mortality, was also confirmed when withdrawing oral anticoagulants in patients with atrial fibrillation. It remains however unclear whether discontinuation of antithrombotic is associated with a thrombotic rebound (ie, a loss of benefit) with ‘catchingup’ effect or a nonprolongation of the benefit, where crude event rate is similar between groups. The present paper by Dagenais et al brings more evidence to this particular issue, while focusing this time on the withdrawal of lowdose anticoagulant for vascular protection in patients with sinus rhythm. For the composite outcome of CV death, MI or stroke, the benefit of the combination treatment may not have disappeared in a relatively short period of time after withdrawal of lowdose rivaroxaban but rather was ‘not sustained’. Indeed, just before discontinuing combination therapy and switching to nonstudy aspirin, patients from the combination arm were at lower risk of event, but the event rate afterwards was similar. Thus, the benefit gained by patients under combination therapy was not lost but rather diminished by discontinuing lowdose rivaroxaban. On the contrary, the authors report an excess risk of stroke that could be seen, this time, as a thrombotic rebound effect. However, this higher risk of stroke should be interpreted with caution. Overall, patients previously randomised to the combination therapy had a significant excess risk of allcause stroke when discontinuing lowdose rivaroxaban, but a similar risk of ischaemic or uncertain stroke (39 vs 24, HR 1.62 (0.97 to 2.69)), although close to statistical significance. Thus, the excess risk of stroke seems to be weighted by three haemorrhagic strokes, all occurring in the group of patients previously randomised to lowdose rivaroxaban. These events are scarce, but may question the hypothesis of a rebound thrombotic effect after discontinuing lowdose rivaroxaban rather than a nonprolongation of cerebrovascular protection. Furthermore, ischaemic events did not appear before the first month after discontinuing lowdose rivaroxaban, and gradually arisen, similar to what has been previously described in large registries data of recurrent ischaemic events in stable vascular patients under aspirin only. These findings are also important in daily clinical practice in situations where stopping lowdose rivaroxaban may be indicated: surgery —whether urgent or scheduled, percutaneous procedure or haemorrhagic event. Based on the results from this study, discontinuing lowdose rivaroxaban in a short period of time may be safe and feasible, with no apparent excess of event, provided it is resumed as soon as possible. Beside from adding evidence to the field of antithrombotic management of stable vascular patients, this study raises an important and understudied question regarding posttrial access to medication that was proven effective in randomised trial. Even if we observed important progress in ethical control and regulation of patients involved in human clinical research over the last decades, this question remains unresolved. Yet, this important concept was introduced in the Declaration of Helsinki, but only in 2000, and states that ‘in advance of a clinical trial, sponsors, researchers and host country governments should make provisions for posttrial access for all participants who still need an intervention identified as beneficial in the trial. This information must also be disclosed to participants during the informed consent process’. FACT (French Alliance for Cardiovascular Trials), DHUFIRE, Hopital BichatClaude Bernard, Paris, France Université de Paris, Assistance Publique—Hôpitaux de Paris, Paris, France