Stopping when the experimental regimen does not appear to help.

We discuss a procedure that offers the possibility of reducing the number of patients assigned to experimental regimens which are no more effective than a standard regimen. The procedure is useful in advanced cancer trials in which the length of the accrual period might be more than twice the median survival time for the standard regimen. The procedure is intuitively appealing and offers a 50 per cent chance of stopping accrual to ineffective therapies, yet is associated with a loss of power of less than 0.02.