stratified randomization in clinical trial design

Question

we have 5 treatment groups (blank control, placebo, 3 doses drugs) and want to divide the 160 patients into 20, 20, 40, 40, 40 (1:1:2:2:2) for the 5 treatment groups. So this is like 8 blocks randomization.

But we also want to control gender and marriage status, thus we have 4 strata: male married, male unmarried, female married, female unmarried

Since the patients are to be recruited are hard to predict how many in each strata, in reality how we can balance the strata when doing the randomization?

Answer 1

The method you are looking for is minimisation which is breifly explained in the Wikipedia entry As each participant enters the study you calculate for all you variables for which you wish to minimise how unbalanced the current allocations are. You then allocate the next participant to minimise that imbalance. In more refined methods you allocate that pariticpant with high probability to the relevant arm to avoid the allocation being predictable in advance.