Classical designs for clinical phase I trials assume that information about
a dose-limiting event (DLE) is available for all the included patients, or
advise not to treat new patients until the information is present. If a DL
E occurs after a lag, however, information at the current time might not be
sufficient to make clear-cut decisions according to these designs. In part
icular, if new patients are available, it is not clear whether to include t
hem in the trial. We suggest a rule that decides on the accrual of each ind
ividual eligible patient. Simulation studies are presented that indicate an
advantage over the standard 'three-at-once' design in the length of the st
udy. Copyright (C) 2001 John Wiley & Sons, Ltd.