Imbalanced randomization sequence before conducting the study [Design Issues]
Hello Mohammed Yehia,
While it is a tradition to aim for balance I don't think there is a strong scientific argument against imbalance at the design stage. Many, many E+S trials are actually designed this way, with for example relatively few subjects allocated to placebo arms.
If you have heard a regulator saying this isn't possible then it isn't possible but that would be dictated by taste, not science. You can achieve full type I error control and justify a power calculation.
I am leaning towards thinking it can be done, but of course you may ask why anyone would aim for an imbalance of one. If it is simply a matter of "we can afford maximum 33 subjects" then you can surely say that scientifically you are squeezing the max out of the data by including as many subjects as possible, but power should still be decent whatever that means.
If on the other hand you have failed to source enough Test or Ref units so that IMP availability creates a bottleneck then certainly the whole ordeal could be kind of dodgy from the outset perhaps and power could suffer a lot?
So let us hear the actual background, please.
❝ Is it possible to make imbalanced Randomization sequence before entering the study?
❝
❝ For example, using 33 subjects for 2 treatment, 2 sequence, 4 period cross over study.
While it is a tradition to aim for balance I don't think there is a strong scientific argument against imbalance at the design stage. Many, many E+S trials are actually designed this way, with for example relatively few subjects allocated to placebo arms.
If you have heard a regulator saying this isn't possible then it isn't possible but that would be dictated by taste, not science. You can achieve full type I error control and justify a power calculation.
I am leaning towards thinking it can be done, but of course you may ask why anyone would aim for an imbalance of one. If it is simply a matter of "we can afford maximum 33 subjects" then you can surely say that scientifically you are squeezing the max out of the data by including as many subjects as possible, but power should still be decent whatever that means.
If on the other hand you have failed to source enough Test or Ref units so that IMP availability creates a bottleneck then certainly the whole ordeal could be kind of dodgy from the outset perhaps and power could suffer a lot?
So let us hear the actual background, please.
—
Pass or fail!
ElMaestro
Pass or fail!
ElMaestro
Complete thread:
- Imbalanced randomization sequence before conducting the study Mohamed Yehia 2018-01-11 16:19 [Design Issues]
- Imbalanced randomization sequence before conducting the studyElMaestro 2018-01-11 17:10
- Imbalanced randomization sequence before conducting the study Mohamed Yehia 2018-01-11 20:36
- Imbalanced randomization sequence before conducting the studyElMaestro 2018-01-11 17:10