It is a great example of a flawed argument in favor of Randomized Clinical Trials during pandemics and against compassionate use.
So, what’s wrong with it? Everything it says about the advantages of RCTs is correct. What it fails to consider is that if you have reasonable data about death rates from the pandemic, that is a priori information that bears on making a judgment about whether a drug works used off label in a non RCT setting.
Consider an extreme example: The disease has been 100% fatal and 100,000 people have died who became infected with it. You administer drug X to 10 of your patients who have come down with the disease. They all recover in 5 days with no serious adverse effects. Methinks, you don’t need an RCT to figure that the probability is high that drug X helps. If you buy the article's argument, you end up letting people die, needlessly.
Bayesian statistics, PLEASE.
Bayesian statistics, PLEASE.
No comments:
Post a Comment