Ketamine versus Etomidate | The RSI trial

Some topics just get people excited. The drugs we use during intubation – well, people are going to have an opinion. Etomidate was popular while I trained, but was actually hard to find in the emergency department. Then came worries of adrenal suppression. Was etomidate just changing biochemical outcomes, or was it actually changing patient outcomes? Was etomidate increasing mortality? That debate drove many into the arms of ketamine, but the data has always been questionable. We finally have high quality data, in the form of the RSI trial, but has high quality data ever been the deciding factor in a heated debate?

The paper

The RSI trial: Casey JD, Seitz KP, Driver BE, et al. Ketamine or Etomidate for Tracheal Intubation of Critically Ill Adults. N Engl J Med. 2025 Dec 9. doi: 10.1056/NEJMoa2511420. Epub ahead of print. PMID: 41369227 NCT05277896

The question

Does ketamine result in a lower mortality as compared to etomidate when used as the induction agent for intubation in critically ill adults?

The methods

The RSI trial is a pragmatic, unblinded, multicenter RCT from 6 EDs and 8 ICUs in the United States. 

Patients

Critically ill adult patients undergoing intubation with a medication to induce anesthesia. 

Exclusions: pregnancy, prisoners, trauma, immediate need for randomization that precluded randomization, and if the treating clinician determined that either ketamine or etomidate was either necessary or contraindicated. 

Intervention

Ketamine for induction. They provided a nongram that listed full dose (2mg/kg), intermediate dose (1.5mg/kg), and low dose (1mg/kg) options. Clinicians could choose as they wished. 

Comparison

Etomidate for induction. Full dose was listed as 0.3mg/kg, intermediate as 0.25 mg/kg, and low as 0.2mg/kg.

Shared processes

This was a pragmatic trial. All other patient care decisions were left to the discretion of the treating physician. 

Outcome

The primary outcome was all cause mortality by 28 days. 

There was a single pre-specified secondary outcome: cardiovascular collapse during intubation (a composite of a systolic blood pressure below 65 mm Hg, receipt of a new or increased dose of vasopressors, or cardiac arrest from induction until 2 minutes after intubation).

The results

They enrolled 2365 patients (of 3439 screened), with a median age of 60. About half had sepsis or septic shock, and 22% were receiving vasopressors at the time of enrollment. 55% of the intubations occurred in an emergency department. The groups look very similar at baseline.

There was no difference in the primary outcome, with an all cause mortality at 28 days of 28.1% in the ketamine group and 29.1% in the etomidate group (risk difference adjusted for trial site, −0.8 percentage points; 95% CI −4.5 to 2.9; P = 0.65).

There was more peri-intubation cardiovascular collapse in the ketamine group (22% vs 17%; ARR 5.1% 95% CI, 1.9 to 8.3). The difference was bigger among patients with septic shock (21% vs 31%) and among patients with a high APACHE II score (31% vs 21%). Of course, composites are somewhat difficult to dissect. I will discuss it more below, but it is worth noting that the number of patients with blood pressures less than 65 mmHg was almost identical, and cardiac arrest rates were identical, so the entire difference was driven by receipt of new or increased dose of vasopressors. If the blood pressure numbers were the same, why was one group getting more vasopressors?

My thoughts

Although I am generally not a fan of the New England Journal of Medicine – I definitely don’t count it as a marker of high quality research – it is nice to see emergency medicine research represented. I think this study merits the esteem that some people associate with the journal. This research group is producing some of the most impactful critical care research in the world, and they deserve a ton of praise. Of course, no trial is perfect, and so I will discuss a few minor methodology issues. However, the real question you are going to be faced with is: does the secondary outcome of this trial matter to you when the primary outcome was negative?

There are a few minor issues to consider. I understand the practical reality that leads to allowing clinicians to opt out of a trial if they think one of the agents is either necessary or contraindicated, but I always hate this exclusion criteria. It adds so much subjectivity. The whole point of doing the research is that we don’t know which agent is better. Allowing clinicians to act as if they know better than science just biases the science. It means that no matter what this trial showed, clinicians could still claim that there is a group of patients that definitely needs either ketamine or etomidate, because those were the patients left out of this trial. If there is a real difference between the agents, this exclusion just biases the trial towards a false negative finding. On the other hand, even if there is no real difference, people will be able to claim there is one – just in the patients left out of the trial. 

They excluded all trauma patients here. I am not sure the rationale for that decision, but it does significantly limit generalizability to one of the major populations who face hemodynamic problems during intubation. 

You might quibble with the dosing they used here. I think the pattern of doses is somewhat interesting, and the etomidate group might have a slightly higher proportion of very low doses used. Overall, I think the doses represent standard emergency practice, but they might not represent your specific practice. 

It is a minor change, but I prefer when published manuscripts exactly match the registry. On clinicaltrials.org, they only list a single dose of each medication (the full dose), and say that clinicians will be allowed to choose higher or lower. In the manuscript, they list 3 specific doses, all lower. It isn’t clear when this change was made. I don’t see this affecting my interpretation of the results, but my confidence in studies is always highest when the final manuscript exactly matches the published protocol. 

One confounder to consider: management of adrenal suppression. I can’t see any mention of steroid use in their supplemental material, but if modern intensivists know that etomidate causes adrenal suppression, they might just empirically treat that complication, masking any potential harm. This is a realistic representation of modern medical practice, but what if we just take the same lesson and apply it to ketamine? If we know that ketamine is associated with more hypotension, and we just empirically run vasopressors a little higher when it is used, would we not also negate the ‘harms’ seen here. Would that be any different than giving steroids to the patients who received etomidate?

I have ranted about power calculations in the past, but it is important to continue emphasizing: medical research is almost never designed with scientific goals, or even medical goals, as the priority. These studies are designed with an eye to what can practically be accomplished and what will get published. For example, this study was designed to be able to detect a 5.2% absolute difference in mortality between the two groups! There are essentially no interventions in all of medicine with a 5.2% absolute improvement in mortality. Josh Farkas compares this to PCI for STEMI, where we see a 3% absolute improvement in mortality. In other words, these authors thought that using ketamine instead of etomidate would be twice as valuable as PCI for STEMI! Of course, they didn’t actually think that (I assume). The power calculation is mostly just a joke, but unfortunately it tells you that before the trial was even run, it was not designed in a way that it could possibly demonstrate a real difference between the groups. This trial was basically guaranteed to be negative before they enrolled their first patient. Therefore, even though ketamine was actually better here by 1% – meaning that using ketamine in 100 patients would result in 1 extra patient living as compared to etomidate – there is no way this study was ever going to be able to tell us whether that is a real outcome or just noise. (I expect it is noise, but you just can’t tell.)

In my initial draft of this post, I left this comment as a simple rant about power calculations. However, the more I reflect, the more I wonder whether this might be the most important point in the appraisal of this article. The whole reason this trial was being run was that people were worried that etomidate might be increasing mortality. We then ran a trial that was under-powered from the outset, but does actually show worse mortality in the etomidate group. Considering that that was the hypothesis of the trial, I am now much less convinced that this trial can actually be used as evidence that etomidate doesn’t increase mortality. We are seeing the predicted increase in mortality. We knew the difference wasn’t going to be larger than 5.2%, and this trial was designed such that a difference smaller than 5.2% would be statistically negative. So, these results are the exact results we would expect to see if etomidate does actually result in a small increase in mortality.

I am going to say that again, because it might be the most important point, and I ended up sticking it right in the middle of this article. These results are exactly what you would expect if etomidate causes a small increase in mortality. This trial was designed so that a real 1% increase in mortality would result in a statistically insignificant result, but that difference would clearly be clinically significant. 

Assuming the primary outcome was truly negative, the interpretations of this trial will hinge on the secondary outcome. (Having a single prespecified secondary outcome really increases the strength of a trial, decreasing the risk of false positive secondary outcomes. I would love to see more trials designed like this.) Cardiovascular collapse sounds scary. Ketamine is apparently causing it. If you just read the abstract, you might decide to switch away from ketamine. However, if you actually read the paper, my guess is that you will come to a different conclusion.

At first glance, their definition of cardiovascular collapse seems reasonable. From a blood pressure stand point, you needed to get below 65 mmHg systolic. That is a much more realistic cutoff than many studies that report brief (but ultimately irrelevant) hypotension. Of course, hypotension is just a surrogate outcome. We need to ask whether it resulted in important clinical changes for the patient. This trial was specifically powered for the most important outcome – mortality – and there was no difference. If the blood pressure drops but nothing else changes, does it matter? Is this a tree falling in the woods?

I would be more sold on the importance of the blood pressure changes if mortality was better, but just not statistically so, in the etomidate group. In reality, we see the exact opposite. Mortality was actually 1% better in the ketamine group, despite the “cardiovascular collapse”. “Cardiovascular collapse” sounds a lot less scary if it actually improves survival .

Josh Farkas goes into this a little bit more in his post, but there is prior data that suggests that despite its scary name, the composite surrogate outcome of “cardiovascular collapse” is actually clinically meaningless. In prior data, despite having a much higher rate of “cardiovascular collapse” defined this way, ketamine had lower mortality. (Matchett 2022) The only reason we care about “cardiovascular collapse” is because it seems like it should be related to mortality, but the inverse seems to be true. Considering that this is a secondary outcome in a trial that was either negative or actually favours ketamine for all cause mortality, I think it is very clear that we should be ignoring the “cardiovascular collapse” numbers they report here. 

Of course, critically ill patients die for a lot of reasons. It might be hard to pick out a signal of harm from an induction agent when looking at the noise of 28 day mortality data. However, in the supplement they do provide us with 1 hour mortality, which was identical (0.6% vs 0.5%), and therefore I think it is safe to say that the hemodynamic consequences they report on are clinically irrelevant.

“Cardiovascular collapse”, as used here, is actually a composite outcome, and I am a little confused about what actually happened to these patients. The number of patients in both groups with a blood pressure below 65 mmHg was almost identical, as were cardiac arrest rates. Therefore, the entire difference between the groups was driven by receipt of new or increased doses of vasopressors. However, if the blood pressure numbers were the same, why was one group getting more vasopressors? To me, that sort of undermines the value of this composite, which at first glance seemed very clinically relevant. (It seems to imply that a lot of people were giving vasopressors for drops in blood pressure that may not have been clinically relevant.) 

There is a hint in the data that the “cardiovascular collapse” outcome is biased in this unblinded trial: the etomidate group has 2% more vasopressor use at the time of intubation. In other words, it seems like the clinicians were thinking about blood pressure differently when using these different agents, which could explain the secondary outcome. Additionally, if more etomidate patients were receiving vasopressors before intubation and more ketamine patients received vasopressors after intubation, is there really a difference between these groups?

The other point worth making about the “cardiovascular collapse” outcome is that measurement was limited to 2 minutes after intubation. They don’t provide any information on blood pressures at 5 or 10 minutes. Throughout my career, I have seen a lot of panicked interventions for a single low blood pressure reading, when the appropriate response is often to wait 1 minute and cycle the cuff. I wouldn’t use the wait and see approach with a critically ill peri-intubation patient, but there is a spectrum of patients represented here, and the data simply isn’t granular enough to know what really happened. What was the blood pressure at 5 minutes? Did these patients just get a single dose of push dose pressor, or were they on prolonged drips? Did these patients need vasopressors at all?

In an unblinded pragmatic trial, there are some subtle differences you need to consider when looking at outcomes in just the 2 minutes around intubation. Etomidate and ketamine have very different effects in the first minute of induction. With ketamine, you essentially always have to wait until the paralytic kicks in to start your intubation attempt (like you are supposed to in RSI). However, with etomidate, the deep sedation often allows you to start laryngoscopy before full paralysis. (This difference is probably represented by the 9 second increase in intubation time in the ketamine group). By focusing only on 2 minute outcomes, you introduce the potential for bias. For example, you could imagine etomidate being more sympathetically stimulating just because there is laryngeal manipulation prior to paralysis.

I am already seeing a lot of people talking about this secondary outcome as if it were practice changing, but I think we might be making a subtle logical error. The entire reason for running this trial was that we (or at least some people) thought that etomidate might cause harm. We know etomidate causes adrenal suppression, but we didn’t know if that resulted in real harm (mortality) for patients. Anyone suggesting a change to etomidate based on this trial is saying that the lack of change in mortality proves that the adrenal suppression surrogate is irrelevant.

But shouldn’t that exact same argument apply in reverse? If you came into this trial with concerns about transient hypotension with ketamine, shouldn’t the primary outcome prove to you that that surrogate hypotension is clinically irrelevant? If that argument applies to etomidate, it should definitely apply to ketamine. In fact, even more so, considering that the point estimate for mortality is better in the ketamine group.  

What am I going to do with this data clinically? Well, if you told me that you were going to switch to etomidate, I wouldn’t argue with you. I didn’t have strong opinions before this trial, and I still don’t have strong opinions now. I think this trial demonstrates that these two agents are probably equivalent for important long term outcomes in sick patients. (However, if you really believed that etomidate increased mortality before this trial was published, your belief should probably remain unchanged. This trial demonstrates a very believable 1% increase in mortality with etomidate.)

I think a lot of the “cardiovascular collapse” described here is likely a biased artifact in an unblinded trial. However, I think you should assume that ketamine will drop the blood pressure in sick patients. I also think you should assume that etomidate will drop the blood pressure in sick patients. You should always be prepared for peri-intubation hypotension. As long as you are prepared to resuscitate all of your patients, this secondary outcome is clearly irrelevant.

For the most part, I am currently using ketamine, and I see no reason to change based on this data. I imagine people will make a big deal about the hemodynamic secondary outcomes, but I think this data clearly demonstrates that whatever they are measuring, it doesn’t have any important long term outcomes, so they can be ignored. For all cause mortality, the point estimate favors ketamine. Where I work, everyone is more familiar with ketamine. Changing to etomidate might require extra training. The change would probably increase the risk of errors and miscommunication. My usual dose of ketamine is identical to my usual dose of rocuronium (approximately 1.5 mg/kg), which makes math and communication very easy during a resuscitation. For all of those reasons, ketamine will remain my drug of choice, unless people out there have better arguments in favour of etomidate. If so, leave them below.

Bottom line

This RCT, although underpowered, demonstrates no difference in mortality when etomidate is used instead of ketamine when intubating critically ill adult patients. Although it will be widely discussed, the “cardiovascular collapse” secondary outcome is almost certainly meaningless. 

An alternative bottom line might be: this data is most consistent with ketamine and etomidate having equivalent outcomes, but leaves open the possibility that etomidate results in a small but real increase in mortality, while there is essentially no chance that outcomes are worse with ketamine. Phrased that way, the clinical implications seem more obvious. 

Other FOAMed

EMCrit Wee – The RSI Trial

PulmCrit: Hot take on RSI trial of ketamine vs etomidate

Evidence Triage: It’s the Etomidate vs. Ketamine Showdown

Evidence based medicine is easy

The EBM bibliography

Evidence based medicine resources

EBM deep dives

References

Casey JD, Seitz KP, Driver BE, Gibbs KW, Ginde AA, Trent SA, Russell DW, Muhs AL, Prekker ME, Gaillard JP, Resnick-Ault D, Stewart LJ, Whitson MR, DeMasi SC, Robinson AE, Palakshappa JA, Aggarwal NR, Brainard JC, Douin DJ, Marvi TK, Scott BK, Alber SM, Lyle C, Gandotra S, Van Schaik GW, Lacy AJ, Sherlin KC, Erickson HL, Cain JM, Redman B, Beach LL, Gould B, McIntosh J, Lewis AA, Lloyd BD, Israel TL, Imhoff B, Wang L, Spicer AB, Churpek MM, Rice TW, Self WH, Han JH, Semler MW; RSI Investigators and the Pragmatic Critical Care Research Group. Ketamine or Etomidate for Tracheal Intubation of Critically Ill Adults. N Engl J Med. 2025 Dec 9. doi: 10.1056/NEJMoa2511420. Epub ahead of print. PMID: 41369227

Matchett G, Gasanova I, Riccio CA, Nasir D, Sunna MC, Bravenec BJ, Azizad O, Farrell B, Minhajuddin A, Stewart JW, Liang LW, Moon TS, Fox PE, Ebeling CG, Smith MN, Trousdale D, Ogunnaike BO; EvK Clinical Trial Collaborators. Etomidate versus ketamine for emergency endotracheal intubation: a randomized clinical trial. Intensive Care Med. 2022 Jan;48(1):78-91. doi: 10.1007/s00134-021-06577-x. Epub 2021 Dec 14. PMID: 34904190