We compared different procedures for seeking consent to participate in a sham randomised clinical trial and assessed whether refusal is affected by awareness of the severity of outlook. 2035 healthy subjects aged between 20 and 80 years, who visited a scientific exhibition, were enrolled in a hypothetical trial of experimental versus standard therapy, and randomly assigned to groups asked for conventional informed consent or prerandomisation consent. There were four study groups: one-sided informed consent for randomisation (subjects who refused would receive standard treatment); two-sided informed consent for randomisation (subjects who refused could choose between standard and experimental treatment); randomised consent to experimental treatment (subjects who refused would receive standard treatment); and randomised consent to standard treatment (subjects who refused would receive experimental treatment). The refusal rates were 16.2%, 19.9%, 12.1%, and 49.2%, respectively. The perceived severity of the simulated disease affected the refusal rate: the worse the outlook, the lower the refusal rate for informed consent or for consent after randomisation to new treatment, and the higher the refusal rate for consent after randomisation to standard treatment. The prerandomisation design seems to be efficient in a one-sided clinical scenario (eg, a trial of a new drug that would not be given outside the trial) because the refusal rate was substantially lower for prerandomisation to the new treatment than for conventional one-sided informed consent. However, in a two-sided clinical scenario (eg, a trial comparing similar treatments) the prerandomisation design is potentially highly inefficient; the refusal rate was much higher for prerandomisation to standard treatment than for conventional two-sided informed consent.