Corticosteroid Timing for COVD-19

Early guidelines reccommended against corticosteroids in COVID-19 but the NEJM RECOVERY trial showed significant benefit (approx 10% absolute reduction in mortality) when dexamethasone was given to patients receiving either oxygen or mechanical ventilation. However earlier onset of dexamethasone (broadly before O2 requirement) was not effective and possibly harmful.

https://www.google.com/url?sa=t&source=web&rct=j&url=https://www.nejm.org/doi/full/10.1056/NEJMoa2021436&ved=2ahUKEwi6seTTrKDuAhWytlkKHXwDCpoQFjAAegQIAxAB&usg=AOvVaw3FO03Y60iQ2a6lJO7pWThK&cshid=1610796885338

Here a new cohort study suggests that the earlier the CS tx (AFTER iCU admission) the better. This is not inconsistent with the Recovery trial but there are disconcerting results which I will highlight.

Also, there still are unresolved issues because “ICU admission” is not a disease variable and may very widely from center to center.

The Recovery trial found that CS tx at onset of oxygen administration was superior to broadly administering CS earlier. However O2 treatment is also not a disease variable as O2 may be given for either low SpO2 or shortness of breath.

Yet even low SPO2 or SOB are late markers for inflammation, the severity of which CS are intended to mitigate.

Very early CS (for example at the onset of symptoms) may be detrimental yet waiting for the patient to require O2 is probably too late. What time is optimal? We do not know.

These studies highlight the strengths and weaknesses of using treatments as time markers. As the second study points out it would be ideal if a biomarker could be identified which indicates a need for CS.

The second point is the danger of providing guidelines without sufficient evidence. WHO recommended against CS in COVID pneumonia. They should have said “We dont know”.

The truth is that in critical care guidelines are commonly reversed even when supported by RCT. So why would anyone promulgate “guidelines” based on opinion derived of speculations. We have to learn from this as, given a 10% abs reduction in mort with CS. many lives may have been lost due to the overconfidence of the thought leaders writing guidelines without sufficient evidence support.

This shows how neccessary it is to avoid blindly following guidelines. There is no substitue for investigating the evidence yourself, the lives of your patients may depend on that effort.

In this regard one bothersome finding in the second cited Obs trial was the finding that patients treated with CS prior to ICU had a MUCH higher mortality. This disappeared after adjustment and they argue this may be due to selection of CS unresponders.

Yet, if early CS acts quickly enough to prevent ICU admission in responders then CS are truely amazing in the speed of their effect in COVID. I dont know of any proof for that but it is true that CS have a very rapid effect on conditions causing severe hypoxemia, like cryptogenic organizing pneumonia. If true, prioritized study of optimal timing of CS is absolutely pivotal.

So it is disconcerting that there was about a 10% (absolute) worse mortality in those treated with CS before ICU admission. Until I saw that anomaly these looked to me to be very robust findings.

Are the adjustments made here valid? Any thoughts?

Hi Lawrence

My interpretation of the RECOVERY RCT findings is that a benefit was demonstrated for use of dexamethasone in patients hospitalized with COVID, and that the overall benefit seemed to be “driven” by patients who were delineated, in a prespecified way, by the fact that they required either oxygen or mechanical ventilation at the time of randomization.

In contrast, authors of the observational study linked above were not examining patients who had been randomized to steroids. Rather, they simply looked back at the treatments that had been received by patients who were admitted to ICU. They observed that those ICU patients who had received steroids “early” (with “early” encompassing those whose steroids were started before their ICU admission OR in the first 48 hours after ICU admission) seemed to do better than those who didn’t receive steroids in ICU or who received them >48 hours after ICU admission.

You are concerned that when the observational study authors further examined the subset of ICU patients who had been treated “early” with steroids, mortality appeared higher among those whose steroids were started prior to their ICU admission, as compared to those whose steroids were started within 48 hours after ICU admission. You are worried that this finding might suggest that it could be harmful to administer steroids prior to ICU admission (?) Secondly, since nonclinical factors can contribute to the timing of ICU admission (e.g., bed availability), using “ICU admission” as a criterion for judging when to administer steroids seems suboptimal. To this end, you are wondering if a biomarker could be identified that might delineate, better than “need for ICU admission,” which patients are likely to benefit from steroids (?)

I think the authors of the observational study are saying that we shouldn’t read too much into their findings regarding the subset of ICU patients whose steroids were started prior to ICU admission. Patients who are so sick that they deteriorate to the point of needing ICU admission in spite of receiving steroids don’t necessarily reflect the risk/benefit ratio for steroids among pre-ICU patients in general. The reason why some patients deteriorate to the point that they need ICU, even after receiving steroids, could be EITHER 1) that the steroids “caused” net harm OR 2) that the steroids simply didn’t provide enough benefit to avert ICU admission (perhaps because the patient was just too sick and on an inevitable downward trajectory). Since patients transferred to ICU (which is the only group examined in this observational study) are most likely being transferred because they need (or are about to need) mechanical ventilation, and since supplemental oxygen usually precedes mechanical ventilation, and since RECOVERY suggests a benefit from steroids among patients requiring supplemental oxygen, wouldn’t option #2 above be the more plausible explanation for the observational study’s findings?

1 Like

Thanks Erin
Yes, I agree.

I brought this to datamethods because of the use of non patient related timing (admission to ICU) in the Cohort study. Having spent much of my life in the ICU, I think that is a measurement of convenience which could have produced the dramatic U shaped mortality curve as an artifact. (Note the RCT demonstrated only a slight increase in mortality in those treated with steroids before requiring oxygen.)

Their speculation that early steroids might be so rapidly effective in some patients as to obviate ICU (so they are not part of the Cohort) as an explanation for the higher mortality of those receiving CS before ICU was the most intriguing.

I remember when CS were shown to be effective for PCP pneumonia in HIV infected patients. We started using them and the response was very exciting as it could be quite rapid.

It may be a pitfall to mingle administrative events (admission to ICU) and real patient data unless one is studying the ICU itself.
The effect leads to speculation because it has a large set of missing cases. (Those who received CS but were not admitted to the ICU). I wonder why such studies are done. Seems like building a radar unit which covers only a portion of the sky and then speculating about the meaning of weather which it reveals. Better to spend the time to cover the sky or perhaps simply deciding not to do the study if the resources are not available.

1 Like

Agree that the purpose of many observational studies seems opaque, especially in situations where results from a large RCT are already available. I suspect that what often happens is that, prior to publication of a large RCT, researchers rush to get funding for observational studies to provide answers, even when it’s very unlikely that such studies will provide actionable results (especially around questions of treatment efficacy). Then, due to delays in analysis and publication, the observational study doesn’t reach a journal until AFTER the definitive RCT is published. This leaves readers scratching their heads as to why the observational study was ever done in the first place…

RCTs are needed to study treatment efficacy and (unfortunately) COVID presents a golden opportunity to study critical care efficacy questions. Maybe the main reason why RECOVERY was able to show a benefit of dexamethasone is that it enrolled a large number of patients, all with the same infection, who presented to hospital at a similar point in their clinical trajectory (i.e., when respiratory compromise set in, often about 10 days after infection onset) (?)

Being able to recruit large numbers of critically ill patients, all infected with the same pathogen, to a clinical trial over a short period of time is probably not a situation that will be encountered outside the context of a pandemic. But RECOVERY shows that, under these “ideal” (yet tragic) conditions, it is possible to identify efficacious critical care therapies and to efficiently discard therapies that don’t provide a clinically important benefit. If I were a researcher, I’d be asking how the key conditions for RECOVERY’s success could be replicated in future critical care trials…

1 Like