Abstract

Conversion of slowly accruing conventionally randomized studies to a prerandomized design has apparently been successful in increasing accrual enough so that some of these studies can be completed. Ellenberg (1984) has pointed out some of the ethical dangers of prerandomization. This paper argues that prerandomization must be either unsuccessful or unethical: either conversion to prerandomization will result in no significant increase in the rate of completion of the study or a significant increase in accrual rate will be achieved either at the price of an inadequate attempt to obtain informed consent, at the price of the deceit of patients, or at the price of violations of patient autonomy. The argument of the paper can be sketched as follows: For any given randomized study, either patients prefer one treatment arm to the other or they do not. On the one hand, if they do, then conventional randomization fails. But prerandomization, if done ethically, will fail also. Hence, if prerandomization succeeds in this sort of case, then the trial has been conducted unethically. On the other hand, if patients do not prefer One arm to the other, then prerandomization will succeed. So will conventional randomization. Hence, prerandomization is either unnecessary or unethical. Ellenberg's concerns count as good moral reasons for not prerandomizing if prerandomization is unnecessary. It follows that prerandomization is always wrong