Randomised block design is more powerful than minimisationBMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7178.263 (Published 23 January 1999) Cite this as: BMJ 1999;318:263
EDITOR—I agree with Treasure and MacRae that the greatest determinant of the power of a study is its design.1 Minimisation improves power by comparing similar groups but there is a similar technique—randomised block design—which has even more power.2
In this design subjects are assigned to blocks on the basis of their characteristics—for example age, sex, or number of pack years smoked—and then randomly allocated to a treatment group within their block. In this way like is matched with like, and not only will the mean value of each confounding factor be similar in each group but the distribution will be identical in each treatment group. This means that the situation cannot arise in which two groups have the same average age but one comprises middle aged people and the other comprises half younger people and half older people.
The randomised block design also enables interactions of treatment with the blocked variables to be analysed; for example, is the treatment more effective in the elderly? The major disadvantage of this design is the same as in minimisation: assignment to a block becomes a major undertaking. However, this may be offset by the greatly increased power of the design, which results in smaller, cheaper trials.