COVID-19 has disrupted every aspect of life, including how, when, and where research is published. Early on in the pandemic, experts warned that the use of corticosteroids could worsen disease based on data from prior epidemics, and multiple clinical trials subsequently began to examine the effects of corticosteroids in COVID-19. Synthesizing these data poses a challenge to clinicians but a welcome opportunity for a meta-analysis to answer the clinical question. A group from the World Health Organization (WHO) did just that and recently published a prospective meta-analysis of randomized clinical trials comparing corticosteroids to no corticosteroids for the treatment of critically ill patients with COVID-19.
Investigators searched multiple trial registries and identified nine potential trials. Seven trials participated in formal analysis, as one potential trial failed to submit data and another had insufficient recruitment. When the RECOVERY trial was published in June 2020, the six other trials were stopped early at variable timepoints in recruitment. Only data from mechanically ventilated patients were included from the RECOVERY trial; the remaining six trials measured outcomes from ventilated and non-ventilated critically ill patients. Altogether, 1703 patients (median age 60 years, 29 percent female) were randomized (678 in the corticosteroid group, 1025 in the usual care/placebo group). Corticosteroid therapy included dexamethasone (n=459), hydrocortisone (n=195), and methylprednisolone (n=95). Convalescent plasma was the most common additional therapy in both arms and few patients received remdesivir (n=7). A total of 222/678 (33 percent) people in the corticosteroid group died compared to 425/1025 (41 percent) in the control group (NNT=12). In a meta-analysis using a random effects analysis the odds ratio (OR) for mortality was 0.7 (95% CI 0.48-1.01). In subgroup analysis, both patients requiring and those not requiring mechanical ventilation appeared to benefit from corticosteroid therapy (28-day mortality OR 0.69, 95% CI 0.55-0.86 and OR 0.41, 95% CI 0.19-0.88, respectively). Only dexamethasone was associated with a reduced all-cause mortality at 28 days as compared to control (3 trials, OR 0.64, 95% CI 0.50-0.82). Hydrocortisone and methylprednisolone were not associated with a lower 28-day all-cause mortality (3 trials, OR 0.69, 95% CI 0.43-1.12 and 1 trial, OR 0.91, 95% CI 0.29-2.87, respectively). Serious adverse events related to corticosteroid therapy were rare but unrecorded in the largest trial.
This meta-analysis provides some evidence of benefit for the use of steroids among critically ill adults but has a few notable limitations. Six of seven trials stopped early, which could easily underestimate the harms of intervention, particularly as RECOVERY did not report serious adverse events. Trials included heterogeneous populations and relatively few critically ill patients not requiring mechanical ventilation. In the paper’s primary outcome, the authors performed meta-analysis using a fixed effects model. This model gave confidence intervals that did not cross one and an effect estimate with good precision. However, a random effects analysis would have been more appropriate because of the variability in populations and interventions. Although the benefit using the random effects model was not statistically significant, that does not mean there is no benefit. In fact, the dexamethasone analysis showed a clinical benefit. It appears corticosteroids provide a clinically meaningful reduction in 28-day mortality for critically ill patients with COVID-19 but there are still many unanswered questions about which populations would be most likely to benefit as well as important details of administration.
For more information, see the open-access topic COVID-19 (Novel Coronavirus) in DynaMed®.