A two-stage dose selection strategy in phase I trials with wide dose ranges

At early stages of drug development, the maximum tolerated dose (MTD) must be determined in order to have a range of doses that can be safely administered in humans. This is necessary before the compound can be tested in phase II clinical trials to find the optimal dose in terms of clinical outcome. Although heavily criticized, traditional dose escalation methods are still widely used to estimate the MTD. The recently developed Continual Reassessment Method (CRM) and Logistic Dose Ranging Strategy (LDRS) offer advantages compared to the traditional approach, including improved estimation of the MTD. However, the CRM utilizes a prior distribution that must be specified before having any data from the current trial. The LDRS requires the use of hypothesized data through a seed data set established before the start of the trial. When a wide dose range needs to be tested, the assumed information (prior distribution or seed data set) can have a great impact on dosages used during the trial and on the final estimates of the MTD. This paper combines the LDRS and the traditional dose escalation procedure to suggest a practical two-stage method that provides reliable estimates through relatively easy computation, and yet requires almost no prior information.