So I am making a thread on serotonin trials in covid because keeps coming up courtesy of @farid__jalali and the fact that about 70years of respiratory physiology recognise it as pulmonary vasoconstrictor

acidosis, serotonin, hypoxia, hypercarbia btw all = pulmonary vasonconstriction. Serotonin also activates platelets and makes people look all hypermetabolic and agitated (as does sepsis and catecholamines btw to be clear) so we all get the rationale.

So I am sympathetic to it as a modifiable segment of covid because also some studies have shown some elevated serotonin levels (not unique to covid either btw, it's an endogenous inflammatory signal). Then someone showed me the fluvoxamine trial RCT and I was like

oo fun that's clever, they properly randomised, found that fluvoxamine prevented clinical deterioration in 100% of patients, 0.009 p-value. It's such a selective serotonin antagonist, it must be that good right? OMG THE TRIAL WILL MAKE YOU CRY

A pause whilst I upload all my evidence of how outraged I am that this was put across as a large brilliant RCT by JAMA. (And it is great, and real world applicable, and well intended, and clever, but the interpretation and conclusion from those who read it is not!)

So basically this was outpatient RCT with roughly 70 in each arm either fluvoxamine or placebo. Primary outcomes were shortness of breath (self reported) , oxygen sats below 92% or hospitalisation for above. Secondary was death for example.

Baseline characteristics

All good except double the asthma in the fluvoxamine group. For those less familiar with this drug it's non serotonin effect is on S1R which affects calcium signalling, and is an endoplasmic reticulum shock protein that tbh is more exciting to me than serotonin.

Crucially it sometimes part of complex therapy for asthma

Anyhow. Six patients in the placebo arm had clinical deterioration either shortness of breath or Sats < 92. None in placebo. I worked out 6/70 or whatever is roughly 0.085 background risk, so to miss it 70 x in the fluvoxamine arm is

0.915^70 which is just 0.002 or 2% (rough stats to orientate ourselves) so I was like hmm okay that's fair. Were the six people who deteriorated asthmatics though?

Well...! This is the bit that is driving me nutty. There is a line that addresses the trial couldn't be absolutely sure that fluvoxamine was driving the difference because a) six is small number anyway and b) they couldn't be sure about baseline oxygen Sats

So let's look at baseline oxygen sats (pre entry to trial) whilst median of both groups is 97% you see the distribution is not. Those who clinically deteriorated in orange. Some even started with sats of 92%

Six is a small number so let's read through who the patients are. I argue your honour that three of the patients used to drive the analysis don't even count let alone this 0.009 p value. (So cross!)

So the first two don't even get admitted to hospital! Literally go get checked out. Oh what an awful adverse event. What we don't know is did they have asthma also? They are young...

Here we are, third one. Chronic respiratory failure. Would have had maybe one evening dose of fluvoxamine in other arm and still ended up the next morning with sats of 75% in hospital. Not even short of breath! That is problem with dyspnoea as an outcome

So then the adverse events in fluvoxamine arm. They still get SoB btw just no sats less than 92.

Plus we can't even prove any of this was serotonin mediated. Other things I'm not sure about are something complicated about missing data which I think they might have included (someone smart will be able to tell me)

And also this complicated outcome model of numbering various strange outcomes that I also thing is flawed. I must point out all trials are flawed and many just as badly or worse but people are changing practice on them.

Really bad press releases giving us this kind of info but when you go teaming you realise the problems. Patients who count as bad outcomes in the analysis that start with the outcome you're looking for is really bad!

I still like serotonin as an option this trial just doesn't confirm it works to me at all (and it's still nicely set up, it's the presentation and inference that's flawed!) we all need to get more literate. And get bigger numbers! And objective measures.

* sorry for all the typos I was just being a bit enthusiastically annoyed!