Sherlock,
I know a little about clinical trial design and is actually not complicated.
What we want to achieve is statistical significance but there is no universal minimum number of people required to achieve it. The required sample size depends on effect size (how big the treatment effect is), how “noisy” the data is and study design (if it is placebo controlled, etc.). Of course, FDA has minimum requirements (p=0.05 and/or power 80%-90%).
In our case, being a Small Phase II trial, where we are kind of signal finding, or finding out if Leronlimab might work, the required number of people is typically 40–100 total. The key here is what effect is expected, if the effect is “medium” this number is adequate. If the expected/achieved effect is strong a small number of patients can reach statistical significance.
For example, think that you have 20 people in a study and 18 show a strong response to a treatment. This is a small number of people, but the result is likely statistically significant. Now, let's assume that there are 15 people that have some response. Well, this might mean something, but the FDA would tell us: Let's do a trial with 100’s of people to confirm the drug is doing something (the numbers are for example purposes only).
Also, the outcomes are important (what one is measuring), for continuous outcomes one needs fewer people, for binary outcomes (e.g., remission yes/no) more people and for survival outcomes, even more people.
All this can easily calculated with a formula or a statistics application but this is the gist of powering a trial.
I know a little about clinical trial design and is actually not complicated.
What we want to achieve is statistical significance but there is no universal minimum number of people required to achieve it. The required sample size depends on effect size (how big the treatment effect is), how “noisy” the data is and study design (if it is placebo controlled, etc.). Of course, FDA has minimum requirements (p=0.05 and/or power 80%-90%).
In our case, being a Small Phase II trial, where we are kind of signal finding, or finding out if Leronlimab might work, the required number of people is typically 40–100 total. The key here is what effect is expected, if the effect is “medium” this number is adequate. If the expected/achieved effect is strong a small number of patients can reach statistical significance.
For example, think that you have 20 people in a study and 18 show a strong response to a treatment. This is a small number of people, but the result is likely statistically significant. Now, let's assume that there are 15 people that have some response. Well, this might mean something, but the FDA would tell us: Let's do a trial with 100’s of people to confirm the drug is doing something (the numbers are for example purposes only).
Also, the outcomes are important (what one is measuring), for continuous outcomes one needs fewer people, for binary outcomes (e.g., remission yes/no) more people and for survival outcomes, even more people.
All this can easily calculated with a formula or a statistics application but this is the gist of powering a trial.
10
