clinical trial interpretation

[stator]

Hi!
given two randomised clinical trials, the first compare treatment A vs B; the second compare treatment A vs B vs C. after enrolment, the investigators decided to combine the arms A and B of the two studies and they have made one single comparison(A vs B), in the other hand they compare the arms B and C of the second trial. they 've made the two comparisons with an alpha level =0.05 each. my question, is the incorporation of the arm B of the second trial in theses two comparisons is making them dependant? if it's the case, how should they deal with the alpha-inflation problem?
thx

 

[hlsmith]

Can you rephrase the question but put subscripts or better define the groups. Such as A vs. newB(e.g. B2 + C2). I want to make sure I understand your question before writing an answer.

Thx.

 

[stator]

Thank you hlsmith.
Two trials : 1 and 2
Trial 1, has two arms : A1 vs B1
Trial 2, has three arms: A2 vs B2 vs C2
Patients in the arms A1 ans A2 were treated with the same treatment A.
Patients in the arms B1 ans B2 were treated with the same treatment B.
patients in the arm C2 had the treatment C.

the first comparison A1+A2 vs B1+B2, level of significance for the primary endpoint =0.05.
second comparison: B2 vs C, level of significance for the primary endpoint =0.05.
I'm concerned by the alpha-inflation problem given that the arm B2 was used in the two comparisons, is't really an issue in this particular situation?

 

[hlsmith]

Yes, I agree. If both of these comparisons are a part of the same set of analyses (experiment), I would just take both p-values and multiply them by 2 (Bonferroni Correction). Obviously the comparison groupings are a little tweaked, but this will remove the threat.

If wanted a more sophisticated approach could be used, but the Bonferroni approach is well known and is considered a cautious alternative.

P.S., Thanks for rewriting your post. It drastically helped in its interpretation, I don't think I would have guessed that is exactly what you were trying to convey in the first post.

 

[stator]

thank you for your answer, actually, the investigators have changed the statistical plan after they have looked at the data!!so i'm really concerned about the validity of their results...
thx

 

[hlsmith]

Sorry to hear. That is why i ask investigators to blind me to the assignment at times. Let us know if you have any more questions.