Havasu,
Correct, should had specified it better. 47 meant VX deaths (40 SOC). And yes, that means a 41% mortality reduction.
In regards to what is "good enough" power and what is not: trials normally try to probe what is call the "null hypothesis". Not going into mathematical jargon basically the question is: does the drug work ???
The explanation below lacks of mathematical rigueur but it might help to explain this.
We have what is called a "false positive", that is, we think Vyrologix works and it doesn't. Normally this level is 0.05 (5%). So, is very likely that we think it works and it does (95%).
However, here is the possibility that we have a false-negative: that is, we think Vyrologix does not work but it works !!. This is the Power (also called Type II error) designed into the trials.
From the perspective of patients normally the more critical error (this is not always true) is the false positive. Imagine FDA approves a drug that everybody thinks it helps and it doesn't producing negative clinical outcomes (no treatment, or worse, no treatment with adverse effects).
However, is important as well that we help all those that need help by not abandoning drugs that work resulting in the loss of benefits to patients when the commercialization of drugs that are safe and effective is prohibited.
The p-value and the power are related by parameters in the trial design. Lots of considerations come into play, for example the more patients the better the power of the trial, however at what cost ?? duration ??? adverse effects (more patients in placebo and drug arm with a drug that is not known to work and might produce SAEs) and so on.
Back to the question: FDA normally uses 80%-90% range for this. However, normally there is more leeway for false-negatives than for false-positives due to the obvious consequences. An interesting discussion can be fund below
https://alo.mit.edu/wp-content/uploads/2015/08/FDA_26.pdf
Correct, should had specified it better. 47 meant VX deaths (40 SOC). And yes, that means a 41% mortality reduction.
In regards to what is "good enough" power and what is not: trials normally try to probe what is call the "null hypothesis". Not going into mathematical jargon basically the question is: does the drug work ???
The explanation below lacks of mathematical rigueur but it might help to explain this.
We have what is called a "false positive", that is, we think Vyrologix works and it doesn't. Normally this level is 0.05 (5%). So, is very likely that we think it works and it does (95%).
However, here is the possibility that we have a false-negative: that is, we think Vyrologix does not work but it works !!. This is the Power (also called Type II error) designed into the trials.
From the perspective of patients normally the more critical error (this is not always true) is the false positive. Imagine FDA approves a drug that everybody thinks it helps and it doesn't producing negative clinical outcomes (no treatment, or worse, no treatment with adverse effects).
However, is important as well that we help all those that need help by not abandoning drugs that work resulting in the loss of benefits to patients when the commercialization of drugs that are safe and effective is prohibited.
The p-value and the power are related by parameters in the trial design. Lots of considerations come into play, for example the more patients the better the power of the trial, however at what cost ?? duration ??? adverse effects (more patients in placebo and drug arm with a drug that is not known to work and might produce SAEs) and so on.
Back to the question: FDA normally uses 80%-90% range for this. However, normally there is more leeway for false-negatives than for false-positives due to the obvious consequences. An interesting discussion can be fund below
https://alo.mit.edu/wp-content/uploads/2015/08/FDA_26.pdf
6
