The effect of unbalanced randomization on the progressively censored Savage test

Equal allocation of patients to treatment in a randomized clinical trial may have disadvantages ethically if the new treatment is believed to be at least as beneficial as the standard treatment. Others have considered, in a non-sequential setting, unbalanced randomized designs which allocate fewer patients to the potentially inferior standard treatment. This paper examines unbalanced randomized designs in a sequential comparison of two exponential survival distributions using the progressively censored Savage test. The results reveal no substantial sacrifice in asymptotic power or early stopping properties.