RECOVERY: COVID-19 dexamethasone


RECOVERY Collaborative Group. Preliminary Report. Released June 22 2020

Clinical Question

  • In hospitalised patients with SARS-CoV-2 infection does dexamethasone in additional to usual care, compared to usual care alone, reduce 28-day mortality?


  • In December 2019 the Wuhan province in China became the epicentre of a devastating, highly contagious viral infection, SARS-CoV-2, a novel coronavirus that causes COVID-19 illness
  • By March 2020, the World Health Organisation announced they were deeply concerned by the alarming levels of spread and severity of the disease and characterised COVID-19 as a pandemic
  • To date there are more than 11 million confirmed cases worldwide and 500 000 deaths
  • There are many trials running in COVID-19 patients, with the hope to find effective therapeutic options. To date the trials have been of variable quality or have only released preliminary data with incomplete follow-up, for example ACTT-1 trial for remdesivir
  • Corticosteroids have been used in similar syndromes as COVID-19 including SARS, MERS, influenza and severe community acquired pneumonia. However, there is poor evidence in these cohorts as trials conducted have been underpowered or had methodological flaws
  • The RECOVERY (Randomised Evaluation of COVid-19 thERapY) trial describes an overarching trial design that looks at different treatment options simultaneously in patients with SARS-CoV-2 infection, in an adaptive design. The treatments being studied are:
    • Group A
      • No additional treatment
      • Corticosteroid in the form of dexamethasone administered as an oral (liquid or tablets) or intravenous preparation 6 mg once daily for 10 days. In pregnancy or breastfeeding women, prednisolone 40 mg administered by mouth (or intravenous hydrocortisone 80 mg twice daily) administered instead of dexamethasone
      • Lopinavir 400mg-Ritonavir 100mg by mouth (or nasogastric tube) every 12 hours for 10 days (The arm of this trial has ceased recruitment, as no benefit demonstrated)
      • Hydroxychloroquine (this arm of the trial has ceased recruitment, as no benefit demonstrated)
      • Azithromycin 500mg by mouth (or nasogastric tube) or intravenously once daily for 10 days
    • Group B
      • In addition to these treatments patients can be randomised to receive convalescent plasma or not
    • Group C
      • If there is evidence of a progressive inflammatory state patients can be randomised to receive Tocilizumab or not
  • This review looks at the dexamethasone arm of the RECOVERY trial


  • Unblinded adaptive randomised control trial
  • Following written consent from the patient or next of kin, patients were randomised to dexamethasone 6mg once per day for 10 days (or less if hospital discharge occurred) and compared to patients randomised to usual care alone
  • Patients were hospitalised and were not necessarily in intensive care
  • Patients were allocated using a central web-based randomisation service (without stratification or minimisation)
  • Patients were randomised in a ratio of 2:1:1:1:1 (usual care: dexamethasone: lopinivir/ritonavir: hydroxychloroquine: azithromycin)
  • Sample Size was calculated as the trial progressed, as the 28-day mortality was difficult to predict at the beginning of the pandemic
    • So if mortality was 20%, then 2000 patients needed allocation to the drug and 4000 to usual care to yield a 90% power at 2-sided p = 0.01 to detect an absolute difference of 4% difference in mortality between groups
  • There was no blinding
  • Pre-specified subgroups were age, sex, days since symptom onset, predicted 28-day mortality risk AND level of respiratory support [Mechanical ventilation, O2 only (including NIV), no O2]
  • Hazard ratio from Cox regression used to estimate mortality ratio, and Kaplan-Meier curves constructed to display cumulative 28 day mortality
  • Effects within subgroup categories were compared using a chi-squared test for trend
  • Analysis was by intention to treat
  • The independent Data Monitoring Committee reviewed unblinded study data at intervals of around 2 weeks


  • Multi-centre: 176 NHS hospitals in the UK
  • For this dexamethasone arm, patients were randomised between March 9 2020 and June 8 2020


  • Inclusion:
    • clinically suspected or proven SARS-CoV-2 infection
    • age >18 years (but after May 9 no age limit)
    • patients were permitted to be included if pregnant or breast-feeding
  • Exclusion:
    • a known indication or contraindication to dexamethasone
    • Dexamethasone unavailable in hospital at time of randomisation
  • A total of 2104 patients were randomised to dexamethasone once/day for ten days and compared to 4321 patients randomised to usual care alone.
  • Mean age was 66.1 years, 36% were female, Diabetes in 24%, heart disease in 27%, chronic lung disease 21%
  • 56% had at least one major comorbidity
  • There were 6 pregnant patients
  • Mean age was 1.1 years older in the patients allocated to dexamethasone, otherwise baseline characteristics were similar between groups
    • As age is a powerful contributor to mortality, results are presented with and without adjustment for age


  • Dexamethasone
    • 6mg daily dexamethasone for 10 days: intravenous or oral
    • In pregnant or breastfeeding women, prednisolone 40 mg administered by mouth (or intravenous hydrocortisone 80 mg twice daily)


  • Usual care

Management common to both groups

  • Other management as per caring physicians was not protocolised
  • Use of azithromycin was similar between groups (23% vs 24%). Very few patients received hydroxychloroquine, lopinavir-ritonavir, or IL-6 antagonists, and in similar proportions between groups
  • Remdesivir only became available in the UK after May 26 2020
  • It is not clear from this preliminary data, how many patients in each group were also randomised to group B, ie convalescent serum
  • 1% of patients randomised to dexamethasone also received Tocilizumab. 2% in the usual care group


  • Compliance with treatment: 95% of treatment arm received at least one dose of dexamethasone and 7% of control arm received dexamethasone
  • Primary outcome: 28-Day mortality (dexamethasone vs usual care) – significantly reduced in dexamethasone group
    • Vital status:
      • Age-adjusted mortality 21.6% vs 24.6% [Rate Ratio 0.83; 95% CI 0.74-0.92, p<0.001].
      • When not adjusted for the age imbalance in the baseline the result was similar [Rate Ratio 0.86; 95% CI 0.77-0.96, p=0.006 ]
    • NOTE: rate ratio is closely related to risk ratio and is computed as the ratio of the incidence rate in the dexamethasone group divided by the incidence rate in the usual care group
  • Secondary outcome: (dexamethasone vs usual care)
    • Pre-specified subgroups
      • Mechanically ventilated: 28-day mortality – significantly reduced in dexamethasone group
        • 29% vs 40.7% [Rate Ratio 0.65; 95% CI 0.51-0.82, p<0.001]
        • Mechanically ventilated patients were on average 10 years younger than those not receiving respiratory support and were randomised on average 7 days later after symptom onset than those not receiving support
      • Patients receiving O2: 28-day mortality – significantly reduced in dexamethasone group
        • 21.5% vs 25% [Rate Ratio 0.80; 95% CI 0.70-0.92, p = 0.002]
      • Patients not receiving respiratory support: no significant difference
        • 17% vs 13.2% [Rate Ratio 1.22; 95% CI 0.93-1.61 p=0.14]
      • Time from symptom onset to randomisation influenced outcome. Dexamethasone was associated with a reduction in 28-day mortality among those symptoms for >7 days but not among those with symptoms for <7 days (test for trend p<0.001)
    • Length of hospital stay – significantly reduced in dexamethasone group
      • median 12 days vs 13 days
    • Composite secondary outcome of invasive mechanical ventilation and death – significantly reduced in dexamethasone group
      • Rate Ratio 0.91; 95% CI 0.82-1.00, p=0.049
    • Use of ventilation – significantly reduced in dexamethasone group
      • Rate Ratio 0.76; 95% CI 0.61-0.96, p=0.021
    • Use of renal dialysis or haemofiltration: not reported
    • Documented new major cardiac arrhythmia (including atrial and ventricular arrhythmias): not reported

Authors’ Conclusions

  • Dexamethasone at a dose of 6mg/day given for 10 days reduces 28-day mortality in patients with COVID-19.
  • The effect is best demonstrated in mechanically ventilated patients NNT=8
  • It is also apparent in patients needing oxygen therapy NNT=25
  • It does not look to be helpful (and may be harmful) in patients not requiring mechanical ventilation OR oxygen therapy
  • It is likely more helpful if started > 7 days after symptom onset


  • A remarkable trial conducted rapidly in the setting of a rapidly evolving global pandemic
  • Allocation concealment, intention to treat analysis, near complete follow-up (95%)
  • Robust, clinically meaningful outcomes
  • Relevant pre-specified subgroups
  • This sample is suitable for extrapolation to the global population
  • The subgroups of the ADRENAL trial and APROCCHS trial, who received steroids, with pneumonia and ventilation also had better outcomes, adding support to the subgroup in this trial having a mortality benefit
  • The recent DEXA-ARDS trial also showed a similar benefit


  • Primary outcome only available in 95% (5% of patients had not reached Day 28 and data was extrapolated on their likely outcomes, although there are some inconsistencies in the numbers in this preliminary report)
  • Flaws exist with the adaptive trial design:
    • The trial was unblinded
    • Usual care was not standardised, and rapidly evolved between March and May 2020 in the UK (For instance, noninvasive ventilation moved from being not recommended to being recommended within a few weeks)
    • The sample size was not set a priori, nor the point for an interim analysis. This may increase the chance of a Type 1 error (false positive)
    • The number of treatment arms in the overall trial also increase the chance of a Type 1 error
  • It is unclear as to how patients died
  • Conclusions drawn from sub-group analysis can be spurious or happen by chance, but given the overall results favour dexamethasone and there is a physiological rationale that dexamethasone would benefit sicker patients more, ie those with inflammatory lung complications, the result is credible
  • There is little information in this preliminary report about baseline severity of illness
  • The mortality in the ventilated group is high (40.7%) compared to some countries (although this is compatible with the UK ICNARC data)
    • For example, in Australia, although only a small sample size, the hospital mortality in ventilated COVID-19 patients of 22% according to ANZICS data
    • The capacity strain, ICU admission practices and threshold for ventilation varies enormously between countries- raising some doubt about the external validity of these results
  • The follow-up beyond 28 days may be particularly important to assess harm from steroids due to deconditioning or delayed sepsis for example. In addition, a 28-day follow-up may not be long enough to detect mortality, for example, it is not clear from the data how many patients remained in the ICU at Day 28 and did they in fact die beyond this time point.
  • The trial includes patients with suspected SARS-CoV-2 infection and 15% of patients did not have proven PCR, a small weakness in the scheme of this trial being conducted in an overwhelming pandemic

The Bottom Line

  • This large trial demonstrates a significant benefit from the administration of dexamethasone to hospitalised COVID-19 patients requiring supplemental oxygen or mechanical ventilation
  • Based on this trial, if I had a patient in my ICU with COVID-19 pneumonia requiring oxygen or ventilation, I would administer dexamethasone 6mg daily, for 10 days

External Links


Summary author: Celia Bradford @celiabradford
Summary date: June 27 2020
Peer-review editor: Segun Olusanya @iceman_ex

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.