Different N's for different analytes? [Regulatives / Guidelines]
Hi everybody and nobody.
Recently I came to similar problem.
For one analyte the minimum sample size was 60 and for another 12.
Plus the analytes requires different sampling times. The first analyte is quite rapid - last sampling at 6h (if I remember it right), the second analyte contrariwise - last sampling 72h.
Because of that, higher number (almost two times) of blood samples from subjects is needed. The idea (not mine) was to have two groups of subjects for minimalize amount of drawn blood.
For example 64 subjects in total. All subjects should be used for the first analyte, but only 1/4 subjects of all (i.e. 16 subjects) for the second analyte (it means 3/4 subjects should end at 6h with blood drawings and 1/4 should continue to 72h).
Of course many things should be done in non-classical way (two different versions of IC, block randomization with length of block 16, ...).
And I'm aware of EMA guideline (1401 - page 14):
Subject accountability
Ideally, all treated subjects should be included in the statistical analysis. However, subjects in a crossover trial who do not provide evaluable data for both of the test and reference products (or who fail to provide evaluable data for the single period in a parallel group trial) should not be included.
The data from all treated subjects should be treated equally. It is not acceptable to have a protocol which specifies that ‘spare’ subjects will be included in the analysis only if needed as replacements for other subjects who have been excluded. It should be planned that all treated subjects should be included in the analysis, even if there are no drop-outs.
So I think there is a conflict. (The end of the story.)
But I'm pointing to statistical correctness of the situation above?
64 subjects (as sample of population) treated and then we have 16 subjects of 64 (so it's sample of sample of population).
Contrarily, one can say we have first 16 subjects as sample of population and then other 48 to have 64 subjects (in total) of population for the "more variable" analyte. - I am not sure about it.
With thought that there is sample of sample of population, is there an alpha inflation?
e.g.:
10 studies (A-J) on 64 subjects sample of population - 9/10 studies confirm BE, alpha=0.10/2=0.05
A confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
B confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
C confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
D confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
E confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
F confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
G confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
H confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
I confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
J failed to prove BE - ten studies on 16 subjects sample of 64 subjects sample - 9/10 studies fail to prove BE - so 1/10 studies confirm BE?
In the sum of 100 studies on sample of sample (9+...+9+1=82), 82/100 studies confirm BE, alpha=0.18/2=0.09
I am really not sure about the principles here. I heard about no such problem from one who could perform it (or maybe couldn't?). x)
Best regards,
zizou
Recently I came to similar problem.
For one analyte the minimum sample size was 60 and for another 12.
Plus the analytes requires different sampling times. The first analyte is quite rapid - last sampling at 6h (if I remember it right), the second analyte contrariwise - last sampling 72h.
Because of that, higher number (almost two times) of blood samples from subjects is needed. The idea (not mine) was to have two groups of subjects for minimalize amount of drawn blood.
For example 64 subjects in total. All subjects should be used for the first analyte, but only 1/4 subjects of all (i.e. 16 subjects) for the second analyte (it means 3/4 subjects should end at 6h with blood drawings and 1/4 should continue to 72h).
Of course many things should be done in non-classical way (two different versions of IC, block randomization with length of block 16, ...).
And I'm aware of EMA guideline (1401 - page 14):
Subject accountability
Ideally, all treated subjects should be included in the statistical analysis. However, subjects in a crossover trial who do not provide evaluable data for both of the test and reference products (or who fail to provide evaluable data for the single period in a parallel group trial) should not be included.
The data from all treated subjects should be treated equally. It is not acceptable to have a protocol which specifies that ‘spare’ subjects will be included in the analysis only if needed as replacements for other subjects who have been excluded. It should be planned that all treated subjects should be included in the analysis, even if there are no drop-outs.
So I think there is a conflict. (The end of the story.)
But I'm pointing to statistical correctness of the situation above?
64 subjects (as sample of population) treated and then we have 16 subjects of 64 (so it's sample of sample of population).
Contrarily, one can say we have first 16 subjects as sample of population and then other 48 to have 64 subjects (in total) of population for the "more variable" analyte. - I am not sure about it.
With thought that there is sample of sample of population, is there an alpha inflation?
e.g.:
10 studies (A-J) on 64 subjects sample of population - 9/10 studies confirm BE, alpha=0.10/2=0.05
A confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
B confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
C confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
D confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
E confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
F confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
G confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
H confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
I confirmed BE - ten (sub)studies on 16 subjects sample of 64 subjects sample - 9/10 studies confirm BE
J failed to prove BE - ten studies on 16 subjects sample of 64 subjects sample - 9/10 studies fail to prove BE - so 1/10 studies confirm BE?
In the sum of 100 studies on sample of sample (9+...+9+1=82), 82/100 studies confirm BE, alpha=0.18/2=0.09
I am really not sure about the principles here. I heard about no such problem from one who could perform it (or maybe couldn't?). x)
Best regards,
zizou
Complete thread:
- Different N's for different analytes? ddubins 2007-12-05 16:16 [Regulatives / Guidelines]
- Different N's for different analytes? Charl 2007-12-06 11:28
- Different N's for different analytes?zizou 2015-12-06 22:47
- Different N's for different analytes? nobody 2015-12-09 09:13
- Different N's for different analytes d_labes 2015-12-08 09:49