Remdesivir for COVID-19 in 2024 and Beyond: Checking the Expiry Date of the Milk Carton

(See the Major Article by Mozaffari et al. on pages 63–71.) The pace with which evidence was generated during the coronavirus disease 2019 (COVID-19) pandemic was exceptional. Two of the most impactful therapies, systemic corticosteroids (mainly dexamethasone [1]) and remdesivir [2] were ultimately demonstrated to reduce mortality in randomized controlled trials (RCTs). Differential efficacy by the severity of illness was suggested by an apparent increased mortality risk in patients who received corticosteroids for mild disease (as defined by requirement for oxygen [3]) and in those requiring invasive mechanical ventilation or extracorporeal membrane oxygenation (IMV/ECMO) who received remdesivir [2]. Much of how we use steroids and remdesivir in the current era is informed by these RCTs. Yet these RCTs took place before the availability of primary vaccination and boosters, before most of the population had natural immunity, before the addition of adjunctive immunosuppression with interleukin 6 and/or Janus kinase (JAK) inhibition [4], and before the evolution of the virus through the Omicron lineage and beyond. With so much change in this disease, one needs to wonder if the initial evidence should come with an expiry date [5]. While it remains likely that dexamethasone and remdesivir reduce mortality rates in the current era, the absolute effect size is almost certainly smaller than in the preimmunity era. Whether remdesivir is of any benefit to patients admitted for reasons other than COVID-19, a phenomenon that is much more common in the current era, is a question no RCT has ever addressed. With no large RCTs on the horizon, we are left with a reliance on observational evidence to inform the discussion while remaining cognizant of the limitations of any “real-world evidence” [6]. Such limitations include, among others, selection bias, survivor bias (“immortal time”), competing risks, confounding by indication, and residual confounding (particularly due to a lack of granularity in source data). Along these lines, the industry-associated article by Mozaffari et al [7] attempts to demonstrate the benefit of remdesivir in a more modern postvaccine, post-Omicron (December 2021–April 2023) insurance claims data set. Using propensity score matching and eligibility criteria to address confounding by indication and selection bias, and a landmark analysis to try and mitigate immortal time bias, the authors find an association between remdesivir use and reduced mortality rate in patients receiving dexamethasone across all levels of oxygen support from none through to IMV/ECMO. Does this mean that all patients who are receiving dexamethasone should receive remdesivir regardless of oxygen requirements? Let us first address one elephant in the room: if corticosteroids like dexamethasone likely increase the mortality rate in patients who do not require oxygen [1, 3], why were >40% of the patients in this cohort on dexamethasone without oxygen? Is this an issue with validity of the oxygen exposure data, an issue with the quality of practice, or something else? Whatever the reason, dexamethasone in patients not requiring oxygen should not represent a standard of care. So it becomes challenging to know what to make of this population. Putting that aside, there are several reasons why the findings in this observational study may overstate the benefits of remdesivir. First, while the landmark analysis does reduce immortal time bias, it also excludes people who were well enough to go home (who were more likely untreated) and those who were sick enough to receive tocilizumab or baricitinib (who were more likely treated) or to die (difficult to predict the direction of bias). Second, while the propensity score attempts to get at issues of confounding by indication, we will never truly know why some patients received remdesivir and some did not. If the decision was informed by prognosis or comorbidity (eg, advanced cancer or dementia), it is unlikely most data sets have the granularity required to detect that. Third, patients who received remdesivir later than 2 days into the admission were excluded. So, patients whose illness progressed from baseline would only be included if they continued to receive dexamethasone monotherapy despite deterioration. Why would the physicians not start remdesivir in a patient with COVID-19 who was getting worse? If that decision has anything to do with overall prognosis, it is another source of bias in favor of remdesivir. Mozaffari et al [7] have attempted to make the most of the observational data and used advanced techniques to try and reduce or mitigate bias. They have provided the best estimate available under the specific analytic assumptions they have made. Nonetheless, we have seen that the best “target trial emulations” may arrive at estimates that are statistically “close enough” to the registrational RCTs but could still lead to opposite regulatory decisions (see, for example, sacubitril-valsartan) [8]. With no further RCTs on the horizon, how can we use remdesivir in the modern era? Both the observational and RCT-level evidence shows that if you have COVID pneumonia with a real requirement for oxygen, you should receive dexamethasone, remdesivir, and potentially an interleukin 6 or JAK inhibitor. What should we do with patients with symptomatic COVID-19 but without hypoxemia? It is probable that remdesivir still shortens the duration of illness and it may reduce the mortality risk, but the benefit is almost certainly less striking than in the initial nonimmune population in the RCTs. Cost-effectiveness would be unclear in a population where many have now had >5 vaccines and/or recovered from COVID-19 multiple times. If patients who are initially not on oxygen deteriorate, neither the RCTs nor the current article tell us whether adding remdesivir improves outcomes at that point, but it seems like a very reasonable approach. Finally, whether remdesivir has a mortality benefit in patients on IMV/ECMO remains unclear. The mortality rate was higher in the RCTs, but these patients would now receive dexamethasone and tocilizumab or baricitinib, which was not a standard of care during the initial remdesivir RCTs, and these sickest patients were also excluded in this article. Unfortunately, the COVID-19 evidence is like an unlabeled milk carton in the refrigerator: without knowing the expiry date, it can be hard to know whether to drink it. In these cases, we need to use judgment, prior knowledge, and our noses. Data availability. None. Financial support. None.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.