2
$\begingroup$

I have been asked to perform the randomisation for a small RCT using simple randomisation. There are 100 individuals to randomise to one of two arms (control vs intervention). The researchers have not specified any stratifying variables, so the options are limited on being able to use something like minimization.

The researchers have asked if additional participants come forward can they also be randomised at the point of entry (rolling recruitment) after the initial 100 have been recruited and randomised (using a pre-generated randomisation list, n=100). Is this possible as the balance is not guaranteed, unlike a stratified method like minimization? Is there an alternative technique please?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Even though extending the study in this way is safe if strict masking procedures are in place (i.e., the person doing this does not have access to treatment-stratified study outcomes), it may raise a suspicion from a reviewer, especially if using a traditional frequentist approach. It's much safer to use blocked randomization to make sure than any imbalance is inconsequential. For example, randomly choose 4-participant or 6-participants at random and generate the randomization so that each block (whether 4 or 6) is balanced on allocations.

With Bayesian approaches you don't even need to do a sample size calculation except for the budget manager.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thank you, Professor Harrell for your very helpful response! A quick followup question, if there were say an additional 5 participants and we have chosen a 6-participant block, does that matter or do we accept that this is the best possible. $\endgroup$
    – ReadBeard
    Commented Mar 9, 2022 at 12:45
  • 1
    $\begingroup$ In general you generate way more blocks than you need just in case, and many studies fall short of randomizing the intended number. $\endgroup$
    – Frank Harrell
    Commented Mar 9, 2022 at 13:05
  • $\begingroup$ Thanks again, Professor Harrell. $\endgroup$
    – ReadBeard
    Commented Mar 9, 2022 at 13:59

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.