Ketamine, Etomidate, and the inconvenient habit of evidence ruining a perfectly good assumption
Ketamine is brilliant.
There. I've said it.
It provides anaesthesia. It provides analgesia. It preserves respiratory drive better than many alternatives. It has bronchodilatory properties. It can make somebody's blood pressure go up, which is something of a novelty amongst drugs we use to render people unconscious. I'm looking at you propofol...
And, perhaps most importantly, it means I can say the phrase “dissociative anaesthesia”, which makes me sound considerably more intelligent than I actually am.
Ketamine has therefore developed something of a reputation.
Got a critically unwell patient?
Hypotensive?
Peri-arrest?
Generally held together by approximately four remaining catecholamines and the goodwill of the intensive care registrar?
Answer? Ketamine.
At least, that is often the thought process.
But is ketamine actually the haemodynamically stable induction agent we think it is?
A recent systematic review and meta-analysis in the British Journal of Anaesthesia suggests the answer might be:
Not necessarily.
And unfortunately, things are about to get considerably more complicated.
Spectacles on.
Welcome back to Under The Lens!
If you're new here, Under the Lens is where I take an interesting medical paper, stare suspiciously at it for far too long, and eventually decide whether I actually believe it.
The rules remain simple.
What did they ask?
What did they do?
What did they find?
What did I like?
What made me squint?
And most importantly, Would I actually do anything differently because of it?
Today's victim is:
Ketamine versus alternate agents to prevent postinduction haemodynamic instability during tracheal intubation of critically ill adults: a systematic review and meta-analysis
by Loughnan and colleagues.
A catchy title. Really rolls off the tongue.
The paper in one minute
The authors performed a prospectively registered systematic review and meta-analysis looking at randomised controlled trials of adult emergency tracheal intubation in ICU or Emergency Department settings.
They found 10 trials involving 4,673 patients.
Nine compared ketamine with etomidate.
One compared ketamine with a midazolam and sufentanil regimen.
Their primary outcome was post-induction haemodynamic instability.
And the headline?
Compared with etomidate, ketamine increased the risk of post-induction haemodynamic instability (RR 1.28, 95% CI 1.13–1.45).
In absolute terms, that's approximately 44 additional episodes per 1,000 intubations.
Among studies judged to be at low risk of bias, the result was similar (RR 1.32, 95% CI 1.15–1.52). The authors rated that evidence as high certainty.
Oh.
That wasn't what ketamine's PR department wanted.
But before we dramatically throw every vial of ketamine into the nearest sharps bin, there is quite a lot to unpack.
First, why did we think ketamine was so stable anyway?
2-(2-chlorophenyl)-2-(methylamino)-cyclohexanone (known to its friends as ketamine) has sympathomimetic effects. Specifically it's an indirect sympathomimetic agent.
Now, sympathomimetic is a really fun word to say. And it essentially means (usually) a drug which can induce the sympathetic nervous system. Yay, pharmacology and physiology!
Ketamine does this by increasing catecholamine, and dopamine activity in particular. By increasing the circulating catecholamines and sympathetic tone, there is better maintenance of heart rate, vascular tone and blood pressure.
Compare that to many other induction agents, (sleepy milk aka propofol in particular) which have the inconvenient habit of causing vasodilation, myocardial depression, or both. This is generally not ideal when your patient's blood pressure is already 82/something concerning.
Hence, ketamine's reputation.
Etomidate, however, has its own trick.
It produces relatively little cardiovascular depression. Which makes it rather attractive in haemodynamically unstable patients.
Unfortunately, etomidate also inhibits adrenal steroidogenesis.
Medicine never gives us nice things without a catch.
This has led to longstanding concern about adrenal suppression, particularly in critically ill and septic patients, and therefore an understandable desire to find something equally haemodynamically stable without poking the adrenal glands.
Ketamine looked rather appealing.
There is just one small physiological inconvenience.
Ketamine's sympathomimetic effect partly relies upon there being some sympathetic reserve left to stimulate. And critically ill patients occasionally arrive having already spent every catecholamine they own. In profoundly catecholamine-depleted states, ketamine's direct negative inotropic effects may become more apparent.
So perhaps the statement “Ketamine raises blood pressure” needs an asterisk.
A fairly sizeable asterisk.
What did they actually do?
Methodologically, there's plenty here that I like.
The review was prospectively registered with PROSPERO and conducted in accordance with PRISMA. Big ticks in the systematic review world.
They searched six databases and three trial registries, included only randomised controlled trials, used duplicate screening and extraction, assessed risk of bias, performed random-effects meta-analysis and formally assessed certainty using GRADE. They also used Hartung–Knapp adjustment, which is a particularly sensible approach when dealing with random-effects meta-analysis and relatively small numbers of studies.
That sentence has almost certainly caused at least three readers to close the browser.
Come back.
The statistics get more interesting, I promise.
The authors identified ten eligible trials containing 4,673 patients. Importantly, they also obtained pre-publication data from the recently completed RSI trial, allowing that large study to be incorporated before its full publication had even entered the normal literature-search pipeline.
And that trial matters.
A lot.
Enter the enormous elephant in the forest plot
The recently published RSI trial randomised 2,365 critically ill adults to ketamine or etomidate for emergency tracheal intubation.
It found no meaningful mortality difference (28.1% vs 29.1%).
But cardiovascular collapse occurred in 22.1% with ketamine versus 17.0% with etomidate.
Now look at the meta-analysis forest plot. The Casey/RSI study contributes 60.9% of the weight.
Sixty. Point. Nine. Percent.
So although we can accurately say:
“A meta-analysis of eight ketamine-vs-etomidate trials found…”
we should also recognise that this isn't eight similarly sized studies holding hands and collectively arriving at an answer.
One enormous study has arrived at the party, eaten most of the buffet and is now controlling the Spotify playlist.
That isn't necessarily bad.
Large, well-conducted trials should carry substantial weight. And importantly, the older studies generally point in a reasonably consistent direction, with statistical heterogeneity reported as I² = 0% for the ketamine-versus-etomidate analysis.
So this isn't simply one massive study dragging seven dissenting trials kicking and screaming across the line.
But it does mean that this meta-analysis should probably be interpreted as:
“The previous evidence plus one very large new trial now gives us considerably more confidence.”
rather than:
“Ten independent studies have suddenly discovered something completely new.”
So what actually went wrong with ketamine?
Here's where things become particularly interesting.
The pooled primary outcome suggests increased haemodynamic instability. But what does haemodynamic instability actually mean?
Because that phrase can encompass rather a lot.
Across the included trials, definitions ranged from things as dramatic as cardiac arrest to things as comparatively common as systolic blood pressure below 90 mmHg. Event rates therefore varied from around 2.3% to 43.8% depending partly on how each study defined the outcome.
This is an important point.
A cardiac arrest and a transient systolic blood pressure of 88 are both haemodynamic events.
They are not, I would suggest, quite the same afternoon.
So the authors sensibly looked at the individual components.
Compared with etomidate:
Post-induction hypotension: RR 1.09 [0.84–1.40]
Cardiac arrest: RR 1.14 [0.79–1.65]
Neither convincingly different. But:
Early vasopressor requirement: RR 1.35 [1.17–1.56]
Aha.
The increased composite risk appears largely driven by more early vasopressor use. Which is interesting.
But also considerably less dramatic than stating “ketamine causes cardiovascular collapse.”
Those are not interchangeable statements.
What I liked
1) They asked an extremely relevant question
Emergency intubation is one of those moments where tiny pharmacological decisions can suddenly matter enormously.
These patients often have limited physiological reserve, hypovolaemia, vasoplegia, cardiac dysfunction, hypoxaemia, acidosis, and the inconvenient burden of being extremely unwell.
Then we give them an induction drug, remove their sympathetic drive and introduce positive-pressure ventilation.
What could possibly go wrong?
Choosing an induction agent is therefore not academic trivia (though it might be in my Christmas quiz). This matters.
2) Only randomised trials
For a question like this, that's really valuable.
There is enormous potential for confounding by indication in observational data.
The sickest, most hypotensive patient may be more likely to receive ketamine specifically because they are sick and hypotensive. Then if they subsequently collapse, ketamine gets blamed for something it may not have caused.
Randomisation substantially reduces that problem.
3) The methods are generally strong
Just like father-PRISMA wanted this review ticked multiple boxes.
Prospective registration.
Broad search.
Trial registries.
Duplicate review processes.
Risk-of-bias assessment.
GRADE.
Sensitivity analyses.
Random-effects modelling.
And they didn't simply stop once they found a statistically significant forest plot and declare victory. They explored comparator effects, individual components, risk of bias, dose and post-induction blood pressure.
That is good systematic-review behaviour.
4) The result survives the better quality studies
That might sound weird but it will soon make more sense.
Four low-risk-of-bias studies containing 3,777 patients produced a very similar estimate (RR 1.32 [1.15–1.52]).
That's reassuring.
If the effect disappeared the moment we removed poorer-quality studies, I'd be considerably less impressed.
It doesn't.
What made me squint?
Now for my favourite section.
1) "Haemodynamic instability" is doing a lot of work
As mentioned earlier, the studies didn't all define the primary outcome identically.
Some used blood-pressure thresholds.
Some vasopressor requirements.
Some composites.
Some included cardiac arrest.
That creates clinical heterogeneity, even if the mathematical heterogeneity statistic cheerfully reports I² = 0%.
And this is a lovely critical-appraisal lesson.
I² = 0 does not mean the studies are identical. It means the observed effect estimates were statistically consistent beyond what would be expected through sampling variation.
You can still have important differences in populations, drugs, doses, co-induction, vasopressor practice ,and outcome definitions.
Statistics, sadly, cannot magically make clinically different things identical simply by producing a pleasing zero. Though it would make for a great score on Pointless.
2) The headline outcome seems mainly to be a vasopressor story
This is probably my biggest interpretative caveat.
Ketamine increased the composite outcome versus etomidate. But when separated out, the difference was primarily attributable to early vasopressor requirement.
There wasn't convincing evidence for more cardiac arrest. There wasn't convincing evidence for more hypotension using the pooled definitions.
Now, needing a vasopressor is not irrelevant.
Far from it.
Peri-intubation cardiovascular instability is associated with poor outcomes. I think most clinicians would generally prefer their patient's circulation not to disappear immediately after securing an airway. But I can't speak for everyone...
Importantly, vasopressor administration is partly a clinician behaviour. As such, they all have different thresholds and different departments have different practices.
Some will start noradrenaline before induction.
Some will give a push-dose metaraminol.
Some will watch those millimetres of mercury fall towards the Earth's core before conceding defeat.
The authors themselves discuss this. The large RSI trial used a particularly stringent hypotension threshold of systolic BP <65 mmHg, meaning clinicians may quite reasonably have given vasopressors before patients ever reached the formal hypotension endpoint.
So I don't think the correct takeaway is:
“Ketamine makes patients crash.”
I think it's closer to:
“Compared with etomidate, patients given ketamine seem more likely to require circulatory support around intubation.”
Less exciting headline. More useful interpretation.
3) One very large trial dominates the answer
Back to our forest-plot elephant.
The RSI trial accounts for 60.9% of the ketamine-vs-etomidate pooled estimate.
Again, this is not inherently bad.
It was large.
It was randomised.
It deserves weight.
But it means this review's conclusions have changed substantially because the evidence base has changed substantially, not because meta-analysis has uncovered some hidden truth that individual studies couldn't see.
Now I know we're currently in the squinting part of the article, but I actually think that's a strength of the paper. This is exactly what systematic reviews should do. Our overall estimate should be updated when important new evidence arrives.
But it's worth understanding what's under the bonnet. That's the squinting part.
4) "Ketamine might be better than other agents" deserves an enormous asterisk
I know what you're thinking.
Is asterisk his word of the week? Partly yes, it's a great word and has its own symbol! It's also very prevalent in this paper.
Against etomidate, ketamine looks worse.
Against the non-etomidate comparator (RR 0.51 [0.29–0.89]) ketamine looks better.
Hence the statistically significant interaction according to comparator.
Great.
Except…
How many studies provided that non-etomidate comparison? One.
How many patients? 80.
And what was the comparator? A regimen involving midazolam and sufentanil.
So I would be extremely cautious with any sentence resembling:
“Ketamine is superior to alternative induction agents.”
Alternative agent? Maybe.
Alternative agents? Hold your horses.
This tells us almost nothing about propofol or -sevoflurane. It tells us almost nothing about other commonly used approaches.
And in places where etomidate isn't routinely available (including much of UK practice) that's quite an important limitation.
The authors acknowledge this lack of comparative evidence themselves.
Their conclusion is appropriately cautious.
Mine would probably be even more so.
5) Who decides what drug was actually given?
Another subtle issue is that “ketamine induction” isn't necessarily one universally reproducible intervention.
Dose matters. Co-induction matters. Opioids matter. Pre-existing vasopressors matter. Patient physiology matters. Speed of administration may matter.
And real-world induction is very rarely simply insert ketamine → observe blood pressure.
Interestingly, a subsequent correspondence in the BJA has already specifically raised the issue of heterogeneity in ketamine-based induction technique, which reinforces the importance of not treating ketamine as a completely uniform intervention across trials.
The meta-analysis looked for a dose-response relationship and didn't demonstrate one, but the available evidence is limited.
So the clinically important question may eventually become less 'Ketamine or etomidate?' and more 'which drug, at what dose, in which physiological phenotype, with what haemodynamic optimisation?'
Which is annoyingly less catchy.
6) And now, conflicts of interest
This deserves careful handling.
Conflict of interest ≠ scientific misconduct.
And I think internet discussions about research sometimes forget this.
Matthew Semler is an author of this systematic review.
He was also chief investigator of the RSI trial, the enormous ketamine-vs-etomidate trial that contributes the majority of the statistical weight to the main result. The systematic review also obtained the RSI trial data before publication directly from its investigators.
The authors openly declare this. They also declare that Semler has received compensation from Baxter Healthcare, DynaMedex and Reprieve Cardiovascular, which they state is unrelated to this work. Another review author, Alastair Brown, is chief investigator of an ongoing trial comparing ketamine with standard care.
So.
Does this invalidate the meta-analysis?
No. Not remotely.
The review itself reports no specific funding, the protocol was prospectively registered, the methodology is transparent, and the RSI study's principal funding came from PCORI and the US National Heart, Lung, and Blood Institute.
But is the overlap worth knowing about?
Absolutely.
Researchers naturally develop views about questions they spend years investigating.
That's human. It doesn't require money. It doesn't require nefarious moustache-twirling. It doesn't mean somebody manipulated anything.
But when one of the review authors led the trial providing over half of the weight in your principal comparison, I think that's relevant context for the reader.
My response isn't:
“Aha! Conflict! Bin the paper!”
It's:
“Interesting. Spectacles remain on.”
And frankly, transparent declaration is exactly what we want researchers to do.
7) Mortality stubbornly refuses to care about our favourite induction drug
Perhaps the most humbling result.
Mortality: RR 0.96 [0.86–1.08].
No convincing difference.
Likewise, there was no significant difference in ventilator-free days, ITU-free days or vasopressor-free days.
We have a plausible short-term haemodynamic difference. But we don't currently have convincing evidence that choosing ketamine rather than etomidate meaningfully changes longer-term patient outcomes.
Perhaps peri-intubation instability simply isn't large enough to alter those outcomes.
Perhaps modern rescue treatment mitigates it.
Perhaps the studies remain underpowered for smaller but important differences.
Or perhaps the drug used for sixty seconds of a weeks-long critical illness simply isn't the main determinant of whether somebody survives.
The audacity.
What does this actually mean at the end of the bed?
This is where I think the paper becomes genuinely useful.
It challenges a heuristic.
And heuristics are incredibly useful in medicine.
They let our brains make rapid decisions without reconstructing pharmacology from first principles at 3am.
But they can become dangerous when they mutate into unquestioned truths.
One of those truths has often sounded like “Ketamine is the haemodynamically stable option.”
This evidence suggests that's too simplistic.
Compared with etomidate, the evidence now looks reasonably convincing that ketamine results in more peri-intubation haemodynamic instability, particularly a greater requirement for early vasopressor support.
That doesn't make ketamine bad. It doesn't make etomidate good for everyone. And it definitely doesn't mean that an unstable patient's induction can be reduced to selecting the correct vial.
Because if somebody is critically unwell enough that you're deeply worried about cardiovascular collapse during intubation, perhaps the most important intervention happens before the induction drug enters the cannula.
Recognising the shock. Optimising haemodynamics. Having vasopressors ready. Considering the dose. Planning the airway. Anticipating the effects of positive-pressure ventilation. And accepting that no induction agent comes with a complimentary haemodynamic force field.
Essentially, the job of a good anaesthetist!
Would this change my practice?
Has it changed how I think about ketamine?
Yes.
I think that's the strongest compliment I can give this paper.
Before reading it, I probably held some version of “Ketamine = cardiovascularly friendly.”
After reading it, I'd add considerably more punctuation:
“Ketamine = cardiovascularly friendly… sometimes… depending on compared with what, in whom, at what dose, and how many endogenous catecholamines remain?”
Less marketable. Probably more accurate.
Would I stop using ketamine in haemodynamically unstable patients?
No.
Would I assume ketamine is going to protect a profoundly sick patient from peri-intubation hypotension simply because it says ketamine on the vial?
Also no.
And in a healthcare system where etomidate may not be readily available, this paper leaves a rather important unanswered question...
What should we use instead?
Because the evidence against non-etomidate comparators consists essentially of one small trial.
So for many UK clinicians, this review may be excellent at disturbing a comfortable assumption without providing an immediately usable replacement.
How very scientific.
The Four Eyed Scorecard
Clinical interest - 5/5 Glasses
I love this question.
Emergency induction is common, consequential and full of strongly held beliefs.
A paper capable of making anaesthetists and intensivists argue about ketamine is also essentially guaranteed entertainment.
Methodological strength - 4/5 Glasses
As already mentioned, I think father-PRISMA would give a pat on the back for this paper.
Prospectively registered.
Randomised trials only.
Broad search strategy.
Appropriate risk-of-bias assessment.
GRADE.
Sensitivity analyses.
Thoughtful statistical methods.
I lose one pair of spectacles mainly because harmonising quite variable definitions of haemodynamic instability remains intrinsically messy, however carefully you do it.
Results - 4/5 Glasses
The ketamine-vs-etomidate result is fairly convincing and remains present in low risk-of-bias trials.
But the headline composite requires interpretation.
The difference is largely driven by early vasopressor use, and the evidence against non-etomidate alternatives is extremely thin.
Applicability - 3/5 Glasses
For clinicians who routinely choose between ketamine and etomidate?
Very relevant.
For those of us practising somewhere etomidate is rarely sitting conveniently in the drug cupboard?
Slightly awkward.
The paper tells us something important about ketamine. It tells us considerably less about what we should actually replace it with.
Practice changing potential - 3/5 Glasses
Would I stop using ketamine?
No.
Would I stop describing it as inherently haemodynamically protective?
Probably.
Would I have the vasopressor ready anyway?
Definitely.
Although, to be fair, I probably should have been doing that anyway.
The Four Eyed Verdict - 4/5 Glasses
I really like this paper.
It is a well-conducted systematic review incorporating an important new randomised trial and it challenges a clinically relevant assumption.
The signal against etomidate is difficult to ignore.
But the interesting bit isn't “ketamine bad”
It's:
“Ketamine isn't magically haemodynamically safe.”
The increased composite instability is largely about early vasopressor requirement.
We don't have convincing evidence of increased cardiac arrest. We don't have convincing evidence of worse mortality. And we have remarkably little good evidence telling us how ketamine compares with induction agents other than etomidate.
The author overlap with the influential RSI trial is worth recognising, but it's transparently declared and isn't a reason to dismiss the work.
So.
Would I read it?
Absolutely.
Would I change my understanding of ketamine because of it?
Yes.
Would I ceremonially dispose of every ketamine ampoule in the department?
No. That would upset pharmacy.
Will I now become even more irritating when somebody says “use ketamine, it's cardiovascularly stable”?
Almost certainly.
And that, I think, makes it worth putting Under the Lens.
Four Eyed love,
Dr Steff
Found this interesting? Of course you did! Read more of my stuff here:
