Anticoagulation in COVID-19: It Is Time for High-Quality Evidence.

Among the distinguishing features of coronavirus disease-2019 (COVID-19) is a severely increased risk for arterial and venous thrombotic complications. It was rapidly recognized that abnormal coagulation parameters were associated with poor outcomes in patients hospitalized with COVID-19 (1), and prospective cohort studies have shown venous thromboembolism in up to 20% of critically ill patients with COVID-19, a prevalence 3-fold or 4-fold higher than in historical cohorts of critically ill patients with non–COVID-19 viral pneumonia (2). Autopsy studies have also shown microangiopathic thrombosis in patients who died of COVID-19 (3), and there have been numerous case reports of cerebrovascular, coronary, and other micro- and macrovascular thrombotic complications. The increased risk of thrombotic complications in patients with COVID-19 compared with other severe viral infections may be caused by direct invasion of vascular endothelial cells by severe acute respiratory-coronavirus-2 (SARS-CoV-2) (4) on top of activation of the coagulation cascade and platelet aggregation by systemic inflammation (5).

The high risk of thrombotic complications in patients with COVID-19 has led to considerable debate about dosing of anticoagulation in hospitalized patients with COVID-19 (6), and multiple health systems empirically developed and implemented protocols recommending doses of anticoagulation greater than approved and commonly used regimens for prophylaxis of venous thromoboembolism in these medically ill patients.

In this issue of the Journal, Nadkarni et al. (7) describe the results of an observational analysis comparing therapeutic anticoagulation, prophylactic anticoagulation, and no anticoagulation in 4,389 patients hospitalized with COVID-19 at 5 hospitals in New York City (7). They found that, compared with no anticoagulation, both prophylactic and therapeutic anticoagulation were associated with a lower risk of in-hospital mortality and need for intubation. There was no significant difference in these outcomes when comparing prophylactic and therapeutic anticoagulation. The authors had access to a clinically rich electronic health record dataset, used inverse probability treatment weighting with resultant good balance among treatment groups on a large set of measured covariates, and included use of anticoagulation in the model as a time-dependent variable to minimize immortal time bias. Landmark analyses consistent with the primary analysis further reduced the risk of immortal time bias explaining the findings. Although this observational analysis was carefully done, lack of randomization precludes the conclusion that anticoagulation, either prophylactic or therapeutic, caused the observed reduction in mortality and intubation.

Over the past 40 years, dozens of cardiovascular therapies and treatment strategies that were mechanistically promising and supported by observational comparative effectiveness studies showed no benefit or harm in rigorous randomized controlled trials (8). Treatments that show benefit in observational analyses may fail to improve outcomes in randomized controlled trials because of unmeasured confounding or other sources of bias. In the current study, the health system’s treatment pathway called for all admitted patients to be treated with anticoagulation unless they were perceived to be at high risk of bleeding by the treating physician. Reasons patients were not treated with anticoagulation were not described, but it is plausible that physicians did not anticoagulate frail patients, those with histories of severe bleeding or active bleeding, and those with baseline thrombocytopenia or coagulopathy. All these variables may be associated with adverse outcomes in COVID-19, and none was included in the propensity model. Mechanistically promising treatments may fail to improve outcomes in randomized controlled trials because of a failure to understand the risks and benefits of the treatment in complex systems. Regarding anticoagulation in patients with COVID-19, any potential benefit of anticoagulation in reducing thrombotic complications is counterbalanced by increases in bleeding, and this trade-off must be taken into account when making clinical decisions. With systematic adjudication, the authors found that 3% of patients on therapeutic anticoagulation, 1.7% of patients on prophylactic anticoagulation, and 1.9% of patients on no anticoagulation had major bleeding events. Although these data are helpful for assessing the bleeding risk of anticoagulation in patients with COVID-19, caution should be taken when assessing the safety of an anticoagulation strategy without consideration of its effectiveness in preventing ischemic or thrombotic events.

Multiple randomized controlled trials have demonstrated that prophylactic anticoagulation reduces pulmonary embolism in hospitalized medically ill patients, with a stronger effect on critically ill patients (9,10); current guideline recommendations reflect these data (11). The recommendations for prophylactic anticoagulation can likely be extrapolated to patients with COVID-19, and the current observational study supports this extrapolation. However, in light of the predilection of COVID-19 for causing thrombotic events, there is equipoise in the comparison between prophylactic and treatment dose anticoagulation, and rapidly conducted, adequately powered, rigorous randomized controlled trials are urgently needed.

There are many challenges to conducting randomized controlled trials, especially in the face of a pandemic, but they are necessary to determine whether and how treatments affect outcomes. Around the world, health systems were overwhelmed by the first wave of COVID-19 and were forced to scramble to provide patient care and develop empiric care pathways. Enrolling patients into randomized controlled trials was not a priority for most U.S. centers, especially given the cumbersome designs of many traditional trials and need for extensive data collection in parallel to clinical care. However, more than 6 months into the pandemic, critical questions regarding the treatment of hospitalized patients with COVID-19—including the correct dose and type of anticoagulant, the use of biomarkers to identify higher-risk patients, and the duration of therapy—remain unanswered by high-quality data from randomized controlled trials; observational comparative effectiveness analyses are no substitute.

With in-hospital mortality for COVID-19 approaching 20%, a randomized controlled trial would require enrollment of approximately 16,000 patients to have 90% power to detect a 10% relative reduction in in-hospital mortality and approximately 7,000 patients to detect a 15% relative reduction. These numbers are a fraction of the nearly 30,000 patients hospitalized with COVID-19 in the United States as of late September, but many more than can be captured by a single health system. There are 30 active randomized controlled trials worldwide comparing anticoagulation strategies in patients with COVID-19; however, most are single- or oligo-center. With approximately 5,000 people around the world dying of COVID-19 each day, a fairly small relative reduction in mortality would still amount to more than 500 lives saved daily. To enroll patients rapidly into adequately powered randomized controlled trials and advance the field, health systems and investigators will need to cooperate and work together on multicenter trials, which should be simple, pragmatic, and embedded in clinical care.