Author: Dr Matthew Durie, Dr Vinodh Nanjayya, and Dr Aidan Burrell
Peer reviewer: A/Prof Chris Nickson
RECOVERY Collaborative Group, Horby P, Lim WS, et al. Dexamethasone in Hospitalized Patients with Covid-19 – Preliminary Report [published online ahead of print, 2020 Jul 17]. N Engl J Med. 2020;10.1056/NEJMoa2021436. doi:10.1056/NEJMoa2021436
Following a pre-print release of results in June 2020, the peer-reviewed results from the United Kingdom (UK) RECOVERY dexamethasone trial were published on July 17, 2020, suggesting a mortality benefit with the use of dexamethasone for patients hospitalised with COVID-19 who required supplemental oxygen or additional respiratory supports.
What was the design?
RECOVERY is a multicentre, open-label, randomised controlled trial conducted within 176 National Health Service (NHS) hospitals in the UK. As with other “platform” trials, dexamethasone was one of several therapies to which eligible patients could be randomised. In this arm of the trial, 2104 patients hospitalised with COVID-19 were randomised to receive dexamethasone, 6mg daily, IV or oral for up to 10 days in addition to routine care, compared with 4321 patients randomised to receive usual care alone. Patients who deteriorated could be further randomised to receive a second-tier therapy such as convalescent plasma or tocilizumab. Study investigators, but not treating clinicians, were blinded to the allocation.
The primary outcome was mortality at 28 days and was adequately powered to detect an absolute reduction in mortality of 4%, assuming baseline mortality of 20% in the standard care group.
What were the results?
At enrolment, 60% of patients were receiving supplemental oxygen, 16% were invasively ventilated, while 24% were hospitalised but not receiving any additional respiratory support.
Overall, patients randomised to receive dexamethasone had a 28-day mortality of 22.9% compared with 25.7% in the standard care group (age-adjusted rate ratio 0.83, 95% CI 0.75 to 0.93). The benefit was higher for patients requiring invasive ventilation (mortality rate 29.3% vs 41.4%, rate ratio of 0.64 (95% CI 0.51 to 0.81), favouring dexamethasone). In other words, for intubated patients, dexamethasone was associated with a relative risk reduction for mortality of between 19% to 49%, while for patients requiring supplemental oxygen, the relative risk reduction was smaller, between 6% to 28%.
Dexamethasone did not appear to reduce mortality if patients were not receiving oxygen, with a mortality of 17.8% among those received dexamethasone, compared with 14.0% for those who received usual care (rate ratio 1.19, 95% CI 0.91 to 1.55).
Dexamethasone was also associated with a lower chance of progressing to invasive ventilation (for those not already intubated) and a shorter duration of hospital admission by 1 day (12 vs 13 days).
Why are the results ‘preliminary’?
When the results were first released, not all patients had reached 28-day follow up. In the published paper, over 99% of patients now have a 28-day outcome known, strengthening the findings. However, the authors still describe the results as preliminary, and state they plan further analyses at 6 months, as well as additional endpoints including cause-specific mortality, the requirement of renal replacement therapy, significant cardiac arrhythmias and duration of ventilation.
What are the strengths of the study?
This study is a very large, mostly well designed and rapidly conducted randomised controlled trial with a high rate of follow-up. The researchers have chosen clinically relevant endpoints and included relevant subgroups that were prespecified. The pragmatic design with broad inclusion criteria makes it generalisable to other settings, and the standard care group mortality in the trial is similar to that reported independently for hospitalised patients with COVID-19 across the UK (Docherty AB. et al. 2020), suggesting there was no enrolment bias.
What are the weaknesses?
The major weakness of the trial is the risk of bias, particularly the absence of a placebo or blinding of treating clinicians. Clinicians may change, often unconsciously, how they manage patients when they are aware the patient is receiving a study drug, potentially exaggerating an effect size. We know from the ADRENAL study that steroids reduce vasopressor use, and this may make patients “look better” in the short term (Venkatesh et al. 2018). This begs the question – did the clinicians continue treatment longer in the intervention arm patients due to a short-term improvement post dexamethasone, thereby effecting the primary outcome? There is also a risk of bias in how patients were screened for the study, with little explanation given (other than stating the study including 15% of all hospitalised patients with COVID-19 at the time). The lack of standardisation of other elements of care adds to the risk of clinician bias, but also reflects the pragmatic design of the trial, and the challenges of conducting urgent research during an ongoing pandemic and stressed health care system.
The second concern relates to the treatment effect size. Steroids have been one of the most extensively studied interventions in ICU and such dramatic responses have not been seen in other large trials of sick ICU patients (e.g. ADRENAL). It is worth noting that a recent small unblinded RCT of 277 intubated patients with moderate to severe non-COVID-19 ARDS also showed 15% absolute reduction in 60-day mortality with dexamethasone compared with standard care, albeit in higher dose (21% vs 36%, 95% CI -26% to -5%; Villar J et al. 2020). However, it is rare, if ever, that interventions initiated in ICU have demonstrated such a large treatment effect and therefore the results carry a degree of implausibility. However, it is equally rare that such a large trial has been performed of a single condition, and most (if not all) other major steroid trials have studied heterogenous syndromes (e.g. ARDS or shock).
Thirdly, while consistent with the UK context (as above), the mortality rate for invasively ventilated patients in the standard care group (41%) was twice that reported for invasively ventilated patients with COVID-19 in Australian ICUs at approximately 21% (ANZICS & SPRINT-SARI combined report to March 31). This is not necessarily a weakness but limits how the results from this study can be applied outside of the UK in the Australian context.
Fourthly, there are limitations to the subgroup analyses. Those receiving invasive mechanical ventilation at randomisation were on average 10 years younger than those not on any supplemental oxygen, and only 2% of patients in the invasively ventilated group were aged over 80, compared with 35% of patients in the no oxygen group. This raises two concerns. First, mortality from COIVD-19 increases with age, yet patients in the UK who are invasively ventilated (i.e. sicker) are younger than those who were not (Docherty AB. et al. 2020), suggesting that age influences the decision to offer additional respiratory support. Secondly, in the supplementary appendix, age < 70 is associated with benefit from dexamethasone, while age > 70 is not. Both age and degree of respiratory support were pre-defined subgroups, yet the authors have focused on degree of respiratory support rather than age in their paper. It is likely that age and degree of respiratory support confound each other, and from the data provided, one cannot exclude the possibility that age affects response to dexamethasone, at least partially independently from degree of respiratory support. It is hard therefore to conclude that all invasively ventilated patients (especially those aged over 70) will benefit from dexamethasone, while all those not on oxygen (especially those aged less than 70) will not.
Morris T et al. 2020 further explores the statistical methods and potential limitations of the study, particularly regarding the subgroup analyses.
Other weaknesses include:
- Despite randomisation there was a baseline age difference of 1.1-years between the two groups (which the authors attribute to chance). This was unexpected but is unlikely to have introduced Type 1 (false positive) error as patients in the dexamethasone group (who had better outcomes) were the older cohort. They report age-adjusted rate ratios for the outcomes.
- The authors do not explicitly state the proportion of patients still hospitalised at 28-days, however, inferring from the results it appears that approximately 10% of patients may remain in hospital.
- Additionally, the day 28 primary outcome of this preliminary trial is shorter than most ICU trials (usually day 90), and important later complications of steroids, such as sepsis and critical illness myopathy may have been missed. Subsequent analyses should address this concern.
- Treatment crossover was 8% of patients in the standard-care group receiving dexamethasone, and 5% of those in the dexamethasone not receiving it. Both figures are greater than the absolute effect size of -2.8%.
- The number of patients who did not have positive PCR for SARS-CoV-2 (12%) is also greater than the absolute effect size. If these patients did not have COVID-19 and responded differently to the therapy, it could have biased the result. The authors state a post-hoc analysis of patients with PCR proven infection had consistent findings, but these results are not included in the publication.
- The authors do not adequately describe how these second-line treatments were applied, and these have the potential to bias the result, as only deteriorating patients were eligible for randomisation to a second therapy. An inferior first-line treatment could result in a greater proportion of patients in that group receiving a second treatment (with either beneficial or harmful effect) compared with the other group.
What does this mean for an intensivist?
Despite some limitations in this study, dexamethasone appears to reduce mortality for patients with COVID-19 who require respiratory support, and the benefit seems greater for patients with higher respiratory support requirements. Questions remain around the use of dexamethasone for patients with less severe COVID-19 who do not require oxygen therapy, and the potential harm associated with its use in this group cannot be excluded.
Results of other large trials comparing steroids in hospitalised patients with COVID-19 are awaited to see if this result can be reproduced.
- RECOVERY Collaborative Group, Horby P, Lim WS, et al. Dexamethasone in Hospitalized Patients with Covid-19 – Preliminary Report [published online ahead of print, 2020 Jul 17]. N Engl J Med. 2020;10.1056/NEJMoa2021436. doi:10.1056/NEJMoa2021436
- Normand ST. The RECOVERY platform [published online ahead of print, 2020 Jul 21]. N Engl J Med. 2020;10.1056/NEJMe2025674. doi:10.1056/NEJMe2025674
- Docherty AB, Harrison EM, Green CA, et al. Features of 20 133 UK patients in hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: prospective observational cohort study. BMJ. 2020;369:m1985. Published 2020 May 22. doi:10.1136/bmj.m1985
- Venkatesh B, Finfer S, Cohen J, et al. Adjunctive Glucocorticoid Therapy in Patients with Septic Shock. N Engl J Med. 2018;378(9):797-808. doi:10.1056/NEJMoa1705835
- Villar J, Ferrando C, Martínez D, et al. Dexamethasone treatment for the acute respiratory distress syndrome: a multicentre, randomised controlled trial. Lancet Respir Med. 2020;8(3):267-276. doi:10.1016/S2213-2600(19)30417-5
- ANZICS Centre for Outcome and Resource Evaluation and Monash University SPRINT-SARI Australia Combined report on COVID-19 admissions to Australian and New Zealand ICUs. 2020 Jun. [accessed Jul 22, 2020]. Available at URL: https://www.anzics.com.au/wp-content/uploads/2020/06/ANZICS-CORE-SPRINT-SARI-combined-report.pdf
- Morris T, Dahly D, Hood K, & Gates S. (2020). Statistical review of Effect of Dexamethasone in Hospitalized Patients with COVID-19 – Preliminary Report (Version 1.0). http://doi.org/10.5281/zenodo.3928540