1
$\begingroup$

if we were designing a clinical trial to compare an old and new treatment and there was strong a priori evidence of a treatment gender interaction (ie new treatment does better than old in men, worse in women) would one possible way to investigate this be to have 1 (control) arm for the old treatment and 2 new (active) treatment arms for the new treatment, 1 for just men and 1 for just women ie 3 arms in total and compare both new treatment arms to the control arm ? Would we need to control the gender proportions in the control arm in some way or is it just a suboptimal idea ? what other approaches could there be ?

$\endgroup$

1 Answer 1

0
$\begingroup$

I think that the only correct way to address this would be stratified randomization (ie creating two separate randomization lists, one for men, and one for women).

A good article on this is Kernan et al, J Clin Epidemiol 1999.

If you wish to have an even number of men and women enrolled, you might with to include only a fraction of men (which are often over-represented in some fields, eg cardiology), but this could create some issues with external validity.

$\endgroup$
1
  • $\begingroup$ many thanks, Joe for the helpful answer and journal article, it seems to confirm that gender specific arms are a very non standard idea for dealing with the issue. I'll suggest stratified randomization. $\endgroup$ Jul 30, 2018 at 8:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.