I love @Michael_Harhay 's fascinating paper about adult critical care trials. https://www.atsjournals.org/doi/suppl/10.1164/rccm.201401-0056CP I did some manual calculations trying to ask how often trials that aimed at mortality were nominally positive (p<0.05) and secondarily whether they were truly positive (not

overturned by subsequent trials). In their Table E3, I count 28 (counting liberally, so including vasopressor comparisons or intensive insulin). Of those, if I'm reading them right, 2 were positive, and 1 was "Jedi mind trick" positive. The two positives were levosimendan and

selective decontamination. Levosimendan was subsequently overturned with the LEOPARDS trial. Selective decontamination may well work and just hasn't caught on in USA. I'll admit that. The NMB trial (ACCURASYS), which required Jedi mind tricks, was overturned by ROSE.

So 1/28 (<4%) are actually positive. But that isn't even the whole story. Several of the negative trials had net harm. Now, admittedly, most of the trials were underpowered, and when you let them have other endpoints, the success rates were better. But crucially,

the low power only matters because miracle drugs are so vanishingly rare. So, to @timothygirard 's point: this is why we do clinical trials.

We do them because we honestly don't know and we need to find out. We do it as an expression of respect and obligation to the patients we serve and to society at large.