Re: Dose Escalation Methods in Phase I Cancer Clinical Trials

Extract

The review by Le Tourneau et al. ( 1 ) missed some important references that contradict a number of their main findings. For example, they found that the widely used standard 3 + 3 design is simple to implement and safe. It is simple, no argument. However, several studies ( 2–6 ) have shown that the standard design is not safe. Worse, given that the purpose of a phase I clinical trial is to estimate the maximum tolerated dose to be used in further studies, the standard method is very inaccurate in terms of dose level estimation. Reiner et al. ( 2 ) showed that over a wide set of possibilities, the standard design incorrectly estimated the maximum tolerated dose in more than 66% of clinical trials. The standard design is known to both overdose and underdose more frequently than modern designs while failing to accurately indicate effective doses.

We ( 3 ) have pointed out that a major feature of the standard design is that it is a “memoryless” design, that is, the selected dose level is based on observation from at most six patients at one dose level and all previous information concerning other dose levels, and the distribution of toxic effects is ignored. Although it is widely thought that the standard design can select a dose level and come to a conclusion by using few patients, that information comes at a high cost in that the probability of an incorrect recommendation is unacceptably high ( 2 ).