True DisCoVeRy of COVID-19 Disease Burden Versus Speculated Antiviral Cardiovascular Risk Requires a Control Group

(See the Major Article by Terzić et al. on pages 382–91.) Imagine if data on wild-type severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viral infection was submitted to a regulatory agency for safety review: the attendant coronavirus disease 2019 (COVID-19) would prompt safety alerts and red flags. Beyond the respiratory disease, its direct and indirect sequelae are multisystem in nature, with attendant cardiovascular risks, particularly manifest in immunologically naïve populations. The literature is now rife with cases of autonomic insufficiency, bradycardia, multiple types of tachycardias, thromboembolic events, and myocarditis resulting from the circulating virus and host response themselves. Given that background, the study of any COVID-19 therapeutic must be gauged in terms of its risk profile in the COVID-19 population by accounting for the cardiovascular risk of COVID-19 disease, along with preexisting patient comorbidities, always including a randomized controlled trial (RCT) [1]. Analyses of randomized groups of treated and control groups are essential to mitigate the risk of misattributing the effects of the disease itself—COVID-19—from the safety and treatment effects of the investigational agent. The fields of cardiology and infectious disease have led the forefront of evidence-based medicine, yielding some of the clearest answers to pressing clinical questions from well-conducted RCTs. In cardiology studies, the hard outcomes under study are frequently “MACE”—major adverse cardiovascular events, with a clinical trial design testing whether a therapy lowers MACE (efficacy), is neutral with respect to MACE (inefficacy/futility), and with stopping rules if there is a signal for increase in MACE (harm). The very presence of a cardiac adverse event (CAE) fulfilling MACE criteria is not interpreted in isolation but rather within its context and with reference to relative rates in the treated and control groups… or at least such was the case before 2019 “B.C.” (the times Before COVID). Yet for multiple reasons, a loss of objectivity has seemingly occurred, with anecdotal reports posted on social media and uncontrolled observational reports published as evidence of definitive safety risk of a therapy, without the needed context of the risk attributable to the underlying disease, COVID-19 itself. Nevertheless, the question of cardiovascular risk is particularly relevant a priori with respect to adenosine-like small-molecule antivirals, and it is imperative that each adenosine-analog be evaluated for risks of activating adenosine receptors, a family of G-protein coupled receptors; incidental activation of adenosine receptors could trigger an adenosine-like bradycardic event, and therefore sound, scientific evidence is key; the question should not presuppose the answer as there is selectivity to the adenosine receptor, and each adenosine-like analog will have its unique co-efficient of binding and activation (if any). The route and rate of dosing are salient. Adenosine itself is used to diagnostic and therapeutic benefit for paroxysmal supraventricular tachycardias when delivered with a rapid intravenous push to generate a brief sinus arrest, whereas when infused intravenously at a slower rate, adenosine can be used for vasodilatory challenge, with lower risk of bradycardia, cardiac conduction blocks, or pauses. For these reasons, adenosine-like medications are not delivered by intravenous push, but rather via alternate means such as slow infusion (eg remdesivir) or orally (eg, VV116 [2]). Like other investigational drugs, adenosine-like antivirals are assessed first for safety in phase I investigations typically in healthy volunteers where the natural risk of cardiac complications are low. By the time of a phase 3 trial, however, the antiviral studies are in patients either with, or at risk of, the study disease and subject both to their own intrinsic medical risk from their comorbidities, the risk from the viral disease itself, and any additional risk from the investigational drug. The relative burden of cardiovascular complications must be weighed in the treated versus untreated control group of an RCT. Presence of cardiac adverse events (CAE) in a treated group, when present but at a lower incidence rate than in the untreated control group, may reflect efficacy instead of risk. Unfortunately, the term “adverse event” in a clinical trial is often poorly understood and by itself does not imply mechanistic causality. In blinded RCTs, adverse events are routinely adjudicated by an allocation-blinded investigator as being causal even when the patient was only provided a placebo. Thus, it is the relative rate of adverse events in the groups randomized to treatment versus the untreated group that will inform whether there is a safety difference. In the current issue of Clinical Infectious Diseases, Terzić and colleagues report cardiac safety data from the remdesivir-treated and standard-of-care arms from DisCoVeRy, an open-label, phase 3 RCT of several agents in patients hospitalized with COVID-19 between March 2020 and January 2021 at 48 European sites [3, 4]. The stimulus for this assessment was prompted by innumerable uncontrolled case reports hypothecating safety concerns for remdesivir based on the premise that remdesivir is an intravenously administered 1′-cyano-substituted adenosine-like analog prodrug. However, it is notable that remdesivir's nucleotide base is a pyrollotriazine, not a purine [5]. DisCoVeRy [4] enrolled patients with relative (SpO2 ≤ 94%) or absolute hypoxemia, up to and including invasive mechanical ventilation, with randomization and analysis stratified by COVID-19 severity. Although DisCoVeRy was randomized, it was neither placebo-controlled nor blinded. Although such a design would be preferable, DisCoVeRy was well-conducted and remains informative, with adequate distribution between remdesivir-treated patients (n = 429) and those randomized to standard-of-care (n = 428). The authors examined the modified intention-to-treat cohorts of patients who received at least 1 dose of remdesivir and their corresponding standard-of-care controls. The primary outcome under study was time to first cardiac adverse event (CAE) through trial day 29. Importantly, the authors adopted a stringent definition of CAEs examining only those that were considered truly of cardiac origin. For example, dyspnea is an ambiguous term, particularly for a respiratory viral disease, and thus was not considered a CAE if due to desaturation due to COVID-19; similarly pulmonary progression of COVID-19, cardiac arrest immediately following unsuccessful liberation from a ventilator for acute respiratory distress syndrome, and creatine kinase due to COVID-19-rhabdomyolysis with normal EKG and troponin were appropriately excluded. Furthermore, the authors adjusted their analysis according to COVID-19 severity at randomization, as that was accounted for during randomization. The authors found no cardiac safety concern with hazard ratio for CAE of exactly unity, with statistics that are about as null as achievable (hazard ratio [HR] 1.0, 95% confidence interval [CI]: .7–1.5, P = .98). Secondary analyses were performed for the number of bradycardic AEs; the numbers were matched at 8 individuals in the remdesivir cohort and 8 in the placebo cohort. Beyond the primary time-dependent Cox analysis, sensitivity analyses including competing-risk models provided internal consistency of the results. This important trial dispelling any general CAE from remdesivir emphasizes the value of the scientific method and correct interpretation of AEs. Although DisCoVeRy was not blinded, we must remember that in randomized, blinded RCTs AEs can even be attributed to placebo, as the term AE merely highlight plausibility. To discern causality, the comparison between a drug requires systematically collected and rigorously evaluated safety outcome data obtained from a valid comparator group. This was also achieved by ACTT-1, a double-blinded RCT, which found no safety difference between remdesivir and placebo in over 1 thousand patients hospitalized for COVID-19 pneumonia [6]. Whereas case reports, series, and observational cohorts remain valuable in bringing a potential issue to attention, they offer a Janus-like risk when the resultant social-media cacophony suppresses and distracts attention from rigorous safety data sets collected according to the scientific method. Taken to the extreme, uncontrolled data can lead to the absurd result of misattributing the pathophysiology of the disease itself to a safe and effective therapy. DisCoVeRy included patients hospitalized for COVID-19 with hypoxemia. Many patients with mild or moderate COVID-19 do not even require telemetry monitoring, and 16.5% of patients in PINETREE RCT received 1 or more infusions at home [7]. Together, these randomized controlled data from DisCoVeRy augment the existing safety data to provide re-assurance that the decision for telemetry monitoring should be determined based upon the severity of COVID-19 and the patient's intrinsic cardiovascular status. It should also dispel apprehension and overcome impediments to the indicated use of remdesivir, a safe antiviral therapy that reduces mortality in patients hospitalized for COVID-19 [8, 9, 10].