Fig. 1

Risk of the attained power falling below threshold values (77%, 75%, or 73%) due to sample size imbalance as the number of units randomized (total sample size) increases in a parallel, two-arm, individually randomized trial using unrestricted randomization. Results were obtained under the condition that the power is 80% when the sample sizes are equal in the two arms. The risk of low attained power can be substantial with small sample sizes but decreases rapidly as the sample size increases. (The lack of smoothness is a consequence of discretization on the probabilities of obtaining allocations with different sample sizes.)