The RECOVERY Trial
The Randomised Evaluation of COVID-19 therapy (RECOVERY) trial is a randomized, controlled, open-label, adaptive, platform trial comparing a range of possible treatments with usual care in patients hospitalized with COVID-19 in the UK. The treatments being evaluated within this trial include low-dose dexamethasone (now only recruiting children following the publication of the findings in adults), azithromycin, tocilizumab and convalescent plasma.
In the dexamethasone arm of the trial patients hospitalized with COVID-19 were randomised to receive usual care (n=4321) or usual care plus dexamethasone (n= 2104). The primary end point of the trial was 28-day mortality. Overall it was shown that 24.6% of patients that were randomised to receive usual care died within 28 days compared to 21.6% of patients who received usual care plus dexamethasone (p< 0.001). However a pre-specified analysis of 28-day mortality by level of respiratory support at baseline showed that there was an absolute reduction in mortality rates of 11.7% (29.0% vs 40.7%) in patients receiving invasive mechanical ventilation – this equates to a number needed to treat (NNT) of 9 to produce one additional survivor at 28 days. Similarly a reduction of 3.5% (21.5% vs 25%) was seen in patients receiving oxygen but not on ventilation at baseline – this equates to an NNT of 29. However no significant reduction was seen in patients not receiving respiratory support at baseline. In this final subgroup the study did not rule out the possibility that dexamethasone could increase 28-day mortality rates [17% vs 13.2%, RR 1.22 (95%CI: 0.93 to 1.61)]
As the trial report has only been made available in pre-publication form a lot of the detail we would ideally like is not yet available. Methodological detail around statistical adjustment for use of interim analyses of findings and the final outcome in patients who were still hospitalised when the results were censored would be of interest. Similarly additional clinical insight into outcomes in particular patient subgroups including diabetes and older patients and further analyses of relationship between timing of treatment initiation and symptoms would be welcome. The trial was not set up to explore the implications of dose, course length or type of corticosteroid administered but in light of these results these would seem worthy of additional research.
A recent BMJ editorial (link – https://www.bmj.com/content/370/bmj.m2648 ) stressed the importance of getting patient selection and the timing of treatment initiation right as it would appear that dexamethasone primarily acts by counteracting an immune system over-reaction in the second phase of the disease in the minority of patients that develop severe disease. The implications of dexamethasone’s potentially detrimental effects on viral replication rates during the first phase are still unclear.