Caution When Comparing the Impact of Corticosteroids in COVID-19.

To the Editor:

Meta-analyses have the advantage of providing a summary effect estimate when there are a number of methodologically homogenous studies that examine the same intervention in the same study populations. Forty-five days after the publication of the Recovery trial, a meta-analysis of eight randomized trials (7,184 participants) was published, and a corresponding WHO guideline panel subsequently recommended systemic corticosteroids in patients with severe and critical COVID-19 (strong recommendation, based on moderate certainty evidence).

In contrast, the systematic review and metanalysis published in CHEST (March 2021) by Cano et al includes only one randomized controlled trial and 72 observational studies, analyzed together, with a search end date of July 22, 2020. Of the observational studies, only four reported outcomes of propensity score-matched populations. Even with adjustment or propensity-matching, observational studies are subjected to residual cofounding (imbalances in baseline characteristics and post-baseline time-dependent patient differences that influence the decision to prescribe corticosteroids) and other sources of bias. Given this, the Cochrane handbook specifically discourages meta-analysts from pooling randomized controlled trials and observational studies together given, the method heterogeneity. Based on their analysis, the authors of this review concluded that “the potential role for corticosteroids as an immunomodulatory agent in COVID-19 needs to be explored further in clinical trials.”

The authors also combined studies without a predefined treatment protocol with preregistered trials that used an established and explicit corticosteroid protocol. Unsurprisingly, Cano et al were unable to generate conclusions around optimal dosing, indication, and timing of corticosteroids across studies.

Nevertheless, despite the limitations in their analysis, we agree with Cano et al that there is a need for further studies that will examine corticosteroids in COVID-19. Specifically, more data are needed that evaluate the impact of type of corticosteroid, timing of initiation, dose, mode of administration, duration, and dose tapering on patient-important outcomes. Further exploration of laboratory parameters of oxygenation and inflammation and how they may be incorporated into corticosteroid treatment protocols would also be important. The MEDEAS trial (Methylprednisolone vs. Dexamethasone in COVID-19 Pneumonia trial, ClinicalTrials.gov Identifier: NCT04636671) will address this issue by comparing the RECOVERY randomized controlled trial protocol to a protocol similar to the one investigated in an Italian prospective observational study.