Lancet publication shows higher mortality & arrhythmias in a registry of patients treated with #Hydroxycloroquine .

LIMITATION: This is not randomized data and cannot be used to assess the effectiveness (benefit) of the drug. Safety signal is concerning. https://marlin-prod.literatumonline.com/pb-assets/Lancet/pdfs/S0140673620311806.pdf
Why do we "randomize" where people get assigned one drug or the other by chance alone ? To make sure it is a fair comparison .

For instance, if the #Hydroxycloroquine patients were sicker than the other patients, that might explain at least in part why their mortality was worse.
Sometimes we know the other things that could cause someone to do worse with a drug, that is called a "confounder"

But we DONT KNOW all the things that may cause someone to do worse, & these are called "unidentified confounders"

We can't adjust for "unidentified confounders"
When you assign each patient to get one drug or another by chance (randomize them), it makes sure that both the identified & unidentified confounders are balanced out. This makes it a fair comparison.

We approve drugs based upon these randomized trials not registries like this 1
I will say it again: I have no idea if #hydroxychloriquine is effective, but we need randomized trials to determine its effectiveness, not registries.
If a fraction of these people had been randomized, we would have answered this question by now and could move on.

This is not a science problem, this is a people problem: it is a failure to get organized and collaborate to answer important questions.
Big sample sizes and small p values do not = truth.

If the underlying methodology is not sound, you can be absolutely certain about the wrong conclusion.
You can follow @CMichaelGibson.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: