The continual reassessment method provides a new approach for estimating the maximum tolerated dose in a phase I clinical trial. In this paper we investigate a potential stopping rule for this method. We indicate certain situations in which the finally chosen dose level can be predicted with certainty after having included a number of patients strictly less than the projected sample size. There is therefore a clear gain when the only purpose of the trial is to identify a target dose for future experimentation. When prediction with certainty is not possible, we can still estimate the probability that the current level will be the same as all subsequent recommended levels.