Thread by @GidMK

The new viral claim is that cannabis can help with COVID-19, due to a few lab-bench studies

I'm interested to know what people think the likelihood of this panning out is (poll in next tweet)

On average, what proportion of compounds that are found to do something in the lab end up having a benefit in actual clinical trials of human beings?

Now, this is a complex question, but it's also enlightening, because the real answer is that it kind of depends

Of the compounds that make it to human trials, around 1 in 5 are successfully licensed. That doesn't mean they 'work', exactly, but that they are approved for use for some condition at some point

This varies WIDELY by discipline. For cancer, less than 4% of drugs are successful, but for infectious disease it's more like 1 in 4 that work

So, we could say 1 in 4, but there's a catch - not all drugs that are tested in a lab make it into human testing at all!

Unfortunately, figures for what proportion of things that work in a petri dish eventually make it into human testing is murky, and it's hard to get a good estimate of the number

I've seen the number 1 in 100 bandied around, but let's assume it is a bit higher, and say that it's really 1 in 20. At that rate, we'd be looking at about 1% of all drugs that work in a petri dish that eventually get licensed to treat human beings

Getting back to the original tweet, this casts the framing in an entirely new light. At best, these compounds have a 25% chance of working - at worse, it's more like 0.25%

On top of that, this is just the likelihood that the drug will be licensed at some point for ANY indication. The likelihood that this specific claim will be true? Probably quite a bit lower