The ‘Roid’ to Recovery? REMAP-CAP @ St Emlyn’s

This week JAMA published three important trials on the use of steroids in patients with severe COVID-19, with an additional cherry on top; a subsequent inclusive meta-analysis (totalling 7 RCTs and 1703 patients) from the World Health Organisation. A bumper day for steroid fans and a celebration of collaborative science. We recommend you read them all. And the accompanying editorial. Maybe get yourself a coffee first….

Here in the UK, we continue to recruit to the REMAP-CAP trial and so were naturally excited to see this first output focussing on the steroid domain. Would the results be different to those from RECOVERY? How would the authors present the results of an adaptive platform trial, using Bayesian analysis? You can read the abstract below as summary. But like always, we enjoy a deep dive into these things..

What type of study is this?

The REMAP-CAP trial is a Randomised, Embedded, Multi-factorial, Adaptive Platform Trial for Community-Acquired Pneumonia. It was set up in advance of a future pandemic as a way of testing multiple treatments collaboratively and rapidly. The trial has sites in over 250 countries with roughly half of those in the UK. This is a trial that can continue to run over a long period of time, during which new treatments can be added or removed as evidence moves and changes – a novel and commendable concept. We are all for evidence based agility at St Emlyns, as recently highlighted. This will be the first publication on an intervention domain from the REMAP-CAP group, but we can expect to see more publications as the pandemic progresses.

REMAP-CAP is one of many studies prioritised for activity and support by the NIHR urgent public health panel. There are over 130 UK sites recruiting to the trial. You can learn more about REMAP-CAP design in the video below.

In this paper the authors specifically look at the effectiveness of early intravenous hydrocortisone in patients requiring ICU admission, compared to usual care.

Tell me about the patients.

Patients recruited to this domain of REMAP-CAP had to be >18, with presumed or confirmed SARS CoV 2 infection and requiring admission to the ICU for provision of cardiovascular or respiratory organ support, defined as any vasopressor/ionotropic support and any invasive, non invasive or high flow oxygen support (>30l with FiO2 >0.4), respectively. In general, the investigators only excluded patients who were expected to imminently die, or those who had already been enrolled within the trial. Within this particular domain, additional exclusions were time based (>36h on ICU), systemic corticosteroid use and hypersensitivity.

Embedded, multifactorial, adaptive…. what actually happened to these patients?

Well, interestingly this has and will continue to change over the course of the pandemic. This is the nature of an adaptive trial. REMAP-CAP has thus far permitted randomisation of patients to the following groups, though not all of these groups were available in all centres. As this is a multifactorial trial patients can be randomised into more than one treatment arm.

COVID-19 treatment domains (from ICNARC site information):

  • Antivirals 
    Four interventions: (1) no antiviral; (2) lopinavir/ritonavir; (3) hydroxychloroquine; (4) lopinavir/ritonavir and hydroxychloroquine combination – Interventions including hydroxychloroquine are no longer open for recruitment
  • immunomodulation
    Five interventions: (1) no immune modulation (2) interferon beta 1a (3) interleukin 1 receptor antagonist [anakinra] (4 and 5) interleukin 6 inhibition [tocilizumab and sarilumab]
  • Immunoglobulin therapy
    Two interventions: (1) no immunoglobulin; (2) convalescent plasma
  • Therapeutic anticoagulation
    Two interventions: (1) local standard venous thromboprophylaxis (2) therapeutic anticoagulation with intravenous unfractionated heparin or subcutaneous low molecular weight heparin
  • Vitamin C
    Two interventions: (1) no vitamin C; (2) vitamin C
  • Statin therapy
    Two interventions: (1) no statin therapy (2) simvastatin
  • Corticosteroids (three interventions: (1) no steroids; (2) steroids for 7 days (3) steroids if shock present) – no longer open for recruitment

That’s clearly a very large number of treatment arms. When compared to our traditional view of single intervention trials, this may appear excessive, but remember that this is not that sort of trial. The adaptive/platform design allows multiple interventions to be simultaneously and concurrently tested. This takes a little time to get your head around, but it does have a number of significant advantages in a pandemic.

  • Sample size does not need to be predetermined, but instead is monitored by the data committee
  • Nearly every patient is eligible for at least one arm of the trial making recruitment rapid and inclusive
  • Different sites can offer a different range of interventions dependent on their local resources/capabilities/funding
  • Sites can rapidly set up with simple interventions whilst preparing for more challenging interventions over time
  • Interventions can be withdrawn when evidence confirms or refutes their effectiveness
  • A common outcome between interventions allows easy comparison between the effectiveness of the interventions

But that’s not all in this paper, right?

No. This article reports on the steroid domain only. Patients in this study were randomised to either usual care (no hydrocortisone) OR to a fixed 7-day course of hydrocortisone (50mg every 6 hours mostly, although we should highlight that 2 patients got 100mg every 6h) OR to 50mg every 6 hours whilst in a shocked state, with the option to stop steroid once vasopressors had been off for >24h. Do remember though, that all these patients could also be receiving other randomised interventions as above, if the site was active, the patient eligible and consenting.

The primary end point was organ support–free days (days alive and free of ICU-based respiratory or cardiovascular support) within 21 days.
Patients who died were assigned –1 day. The primary analysis was a bayesian cumulative logistic model that included all patients enrolled with severe COVID-19, adjusting for age, sex, site, region, time, assignment to interventions within other domains, and domain and intervention eligibility. Superiority was defined as the posterior probability of an odds ratio greater than 1 (threshold for trial conclusion of superiority >99%).

Although the trial was designed with no maximum sample size, calculations were performed using trial simulations of the adaptive design rules. The authors suggest that if both hydrocortisone arms could reach an odds ratio of 1.75 for improved outcome, a sample of 500 patients would provide 90% power.

Tell me about the results.

1165 patients were screened, with almost half ineligible (a mixture of exclusions, clinician preference and declined consent). Always interesting to see that 7% of patients approached on the ICU weren’t actually receiving any organ support, which perhaps highlights the differences in continental definitions of critical care and reminds us to be cautious about generalising. A total of 403 patients were entered into the trial and randomised within the steroid domain, of which 379 completed follow up. Arguably that’s quite a small sample considering the sheer number of hospitals and the volume of patients – but remember this is just one arm of the study and the trial was stopped early, as below. Recruiting around 400 critically ill patients with the same disease in just over 3.5 months is also no mean feat.

Only 80% of these patients had confirmed SARS-CoV-2 infection. Important to know. Baseline characteristics were mostly well balanced between intervention groups; The average age was 60, the majority (71%) were male and mean BMI around 30. There are some variations however. Median APACHE 2 score, rates of invasive mechanical ventilation and vasopressor use at baseline were all lower in the usual care/no steroid arm. Also worth mentioning that 15% of patients in the no hydrocortisone arm did actually receive systemic corticosteroids, with a median duration of 2 days (IQR 2-6).

Regarding the primary outcome, the median organ support–free days were 0 (IQR, –1 to 15), 0 (IQR, –1 to 13), and 0 (–1 to 11) days. Not particularly uplifting. However, subcomponents of this primary outcome showed more variation between groups, with mortality rates of 30%, 26%, and 33% and a median of 11.5, 9.5, and 6 median organ support–free days among survivors, for the fixed dose, shock dependent and no hydrocortisone arms respectively.

I’ll bet there is more to it than that…

Well, you don’t use the words Bayesian in your methods section unless you plan to go to town on the outcomes. The authors next step (all preplanned and above board) was to increase the sample size by including covariate data from all patients recruited to REMAP-CAP and admitted to ICU requiring organ support (taking N to 576). They subsequently performed a comparative analysis by describing the odds of increased organ support-free days in hydrocortisone groups, compared to a reference population receiving no hydrocortisone. The median adjusted odds ratios were 1.43 (95% credible interval, 0.91-2.27) and 1.22 (95% credible interval, 0.76-1.94) for fixed dose and shock-dependent hydrocortisone, respectively. We think this means that the odds of increased organ support-free days are likely higher in both groups, but the credible intervals highlight that the odds may not be different at all. The authors report this latter interpretation more formally, by describing the percentage chance of superiority for fixed dose hydrocortisone as 93% and shock dependent hydrocortisone as 80%, compared to no hydrocortisone. We’d be lying if we said we understood exactly how they came to these percentage estimates. Bayesian analysis is complex, but is increasingly used to tackle challenges with low frequency outcomes or incidence rates. Here are some explanations from a stats company, a statistician and the International Society for Bayesian Analysis (no less….). But personally, I like this diagram from an open access data science community platform for a short sharp summary. Prior beliefs before the study followed by evidence from the trial, inform the subsequent distribution of posterior beliefs (results).

Any other outcomes of note?

Secondary outcomes were looked at using a similar Bayesian analysis, with less impressive results – the adjusted odds ratios for inhospital mortality were 1.03 (95% CrI 0.53 to 1.95) and 1.10 (95% CrI 0.58 to 2.11) for fixed dose and shock dependent hydrocortisone respectively. Even Bayesian thinking couldn’t save these outcomes, with a probability of superiority below 65% for both options. However, table 3 shows a collation of fairly convincing results on organ support measures; days free of cardiovascular and respiratory support were all increased by any hydrocortisone regimen, with the probability of superiority >85% consistently and approaching 98% for some. Of all patients free of mechanical ventilation at baseline admission, progression to IMV, ECMO or death occurred in 46% with a fixed dose regimen, 60% with a shock dependent dose and a whopping 77% with no hydrocortisone. Interesting, if not statistically significant, stuff. More than 1 serious adverse event was reported in 4 (3%), 5 (3%), and 1 (1%) patients in the fixed-dose, shock-dependent, and no hydrocortisone groups, respectively.

There is loads more to discuss, which we can’t possibly do justice to here. You can listen to an excellent podcast led by Rob MacSweeney with the lead authors on the link below.

Any big problems?

All trials have flaws and REMAP-CAP is not perfect either. It’s an open label trial such that everyone knows what treatments are being given as an intervention. This may introduce bias into the outcomes and also promote clinicians to break protocol if they have strong views on whether patients might benefit from certain treatments. For many of the therapies that’s not a problem in the UK, as they were only available as part of a trial (e.g. convalescent plasma). But for other well known treatments such as steroids, that may not be the case. Clinicians can influence the decision as to whether patients are permitted to be randomised to certain therapies which again may influence the patient cohort and thus bias the results.

The trial has no sample size calculation as in a new disease it is not really possible to identify a predicted effect size (and it ended earlier than anticipated). The arms of the trial are ordinarily closed by the data monitoring committee based on scheduled interim analyses, but on this occasion the early termination required a different approach and a rather different analyses that has left us with a degree of uncertainty as to the effectiveness of hydrocortisone. We need to remind you also that 15% of patients in the non hydrocortisone group did in fact receive it and that this trial did not achieve it’s predicted effect size, or power, based on the reporting in this paper. However, many of these issues identified should really bias the effect towards the null. The fact that the ‘no hydrocortisone’ group had lower APACHE scores, lower rates of cardiovascular organ support and received some hydrocortisone would be more likely to reduce any effect size from the intervention, rather than amplify it. Despite this, the authors still identify a high percentage probability for superiority of hydrocortisone, compared to routine care.

What does this all mean in context?

Well, the results appear to mean that hydrocortisone is probably beneficial. The fixed dose regimen also seemed to outperform the ‘shock dependent’ regimen, without any particularly identifiable increase in adverse events. Good news for everyone. This also ties in with the results of the other trials reported in JAMA and the previously discussed RECOVERY trial.

Speaking of which, did you get time to read the meta-analysis? If not, the take away was very positive – a summary odds ratio of 0.66 [95% CI, 0.53-0.82]; P < .001 for mortality, based on a fixed-effect meta-analysis, with low risk of bias and little inconsistency (I2 = 15.6%). More good news.

Further break down of results between types of corticosteroid had us interested though. Might some of this be explained by the difference between the relative mineralocorticoid vs. glucocorticoid effects of these two drugs as postulated in the paper? Perhaps it is more influenced by overall recruitment figures, with the RECOVERY trial accounting for 59.1% of all participants. Always more questions, but these ones are small fry really. The bottom line appears to be that corticosteroids reduce the risk of death in severe COVID-19.

Speaking of bottom lines…

For us, there are three important outcomes here. Firstly it’s that platform adaptive trials like REMAP-CAP and RECOVERY are an essential component for pandemic responses and we should have them in place for the next one. This is particularly important now, as we are in a phase of mild complacency believing that lightening has just struck (therefore is unlikely to do so again) and we are long overdue the next flu pandemic.

Secondly this trial builds on previous research supporting the use of corticosteroids in severe COVID-19, which is how we do science. Study, evaluate, repeat until confident. We think we have probably reached such levels of confidence with this intervention that it can be considered evidence based care. Bravo to all those involved and thanks to all patients who generously contributed to these studies, in the hope of improving future care.

Lastly, this trial and others tell us that not all steroids are the same in severe COVID-19, but that most of them probably do some good. This is helpful for resource poor areas which may have limited access to different drugs and also leaves us with food for thought on future work.

The last word

If you have the time to read through all trials published today, and in particular the meta-analysis then it is increasingly clear that steroids have an important part to play in the treatment of COVID-19 patients. However, as the accompanying editorial describes very well, we still have work to do on refining many aspects such as dosage, duration, patient selection, drug selection and more. Individualised medicine remains key and trial data like this should not bulldoze over that.

Are these the ‘roids you’re looking for? Maybe they are. But make sure you consider everything about the individual in front of you. There are often good reasons why a population based approach does not necessarily suit an individual, particularly a critically ill one. I am sure that’s what you would want and expect, from your healthcare professional.

Best wishes

Dan and Simon.

References and further reading

  1. St Emlyn’s COVID-19 blogs
  2. St Emlyn’s COVID-19 podcasts
  3. Simon Carley, “Dexamethasone, COVID-19 and the RECOVERY trial. St Emlyn’s,” in St.Emlyn’s, June 28, 2020,
  4. Horby et al. Effect of Dexamethasone in Hospitalized Patients with COVID-19: Preliminary Report. NEJM July 17, 2020 DOI: 10.1056/NEJMoa2021436
  5. Simon Carley, “The RECOVERY platform trial: No benefit to Hydroxychloroquine in Covid-19. St Emlyn’s,” in St.Emlyn’s, June 6, 2020,
  6. REMAP-CAP trial site
  8. REMAP-CAP at
  9. Effect of Dexamethasone on Days Alive and Ventilator-Free in Patients With Moderate or Severe Acute Respiratory Distress Syndrome and COVID-19. The CoDEX Randomized Clinical Trial. JAMA. Published online September 2, 2020. doi:10.1001/jama.2020.17021
  10. Association Between Administration of Systemic Corticosteroids and Mortality Among Critically Ill Patients With COVID-19 A Meta-analysis The WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group JAMA. Published online September 2, 2020. doi:10.1001/jama.2020.17023
  11. Effect of Hydrocortisone on 21-Day Mortality or Respiratory Support Among Critically Ill Patients With COVID-19 A Randomized Clinical Trial Pierre-François Dequin, MD, PhD et al for the CAPE COVID Trial Group and the CRICS-TriGGERSep JAMA. Published online September 2, 2020. doi:10.1001/jama.2020.16761
  12. Effect of Hydrocortisone on Mortality and Organ Support in Patients With Severe COVID-19 The REMAP-CAP COVID-19 Corticosteroid Domain Randomized Clinical Trial The Writing Committee for the REMAP-CAP Investigators JAMA. Published online September 2, 2020. doi:10.1001/jama.2020.17022
  13. Corticosteroids in COVID-19 ARDS Evidence and Hope During the Pandemic Hallie C. Prescott, MD, MSc; Todd W. Rice, MD, MSc JAMA. Published online September 2, 2020. doi:10.1001/jama.2020.16747
  14. Evidence-based medicine and COVID-19: what to believe and when to change. Carley S, Horner D, Body R, Mackway-Jones K.
  15. Urgent Public Health COVID-19 Studies
  16. STATA: What is Bayesian analysis
  17. Bayesian Analysis: A Practical Approach to Interpret Clinical Trials and Create Clinical Practice Guideline John A. Bittl and Yulei HeOriginally published10 Aug 2017
  18. What is Bayesian Analysis? Kate Cowles, Rob Kass, and Tony O’Hagan
  19. Bayesian Statistics explained to Beginners in Simple English.

Cite this article as: Dan Horner, "The ‘Roid’ to Recovery? REMAP-CAP @ St Emlyn’s," in St.Emlyn's, September 3, 2020,

Thanks so much for following. Viva la #FOAMed

Scroll to Top