A formula is given for the calculation of the statistical power of paired intervention trials in which the units of randomization are groups rather than individuals. Factors affecting the power of such trials are also investigated and an example is given where the power of group randomization is compared to that of individual randomization.