Stratified randomization

Steve Simon

2004-08-24

A writer on the IRB Discussion Forum asked a question about how to explain stratified randomization in an informed consent document.

An example of stratified randomization is to have one randomization table for men and a separate randomization table for women. Stratified randomization is an acknowledgement that the randomization process is not perfect. You would not expect 100 flips of a coin to yield exactly 50 heads and 50 tails, so you should not expect that simple randomization will lead to perfect covariate balance.

The CONSORT statement has some nice definitions of stratified randomization and similar methods at www.consort-statement.org/table38b.htm. If you ever have to create a randomization table yourself, I have a nice tutorial on the web at www.childrensmercy.org/stats/plan/random.asp.

Understanding controlled trials: Baseline imbalance in randomised controlled trials. Roberts C, Torgerson DJ. BMJ 1999: 319(7203); 185-. [Full text] [PDF]

You can find an earlier version of this page on my original website.