Content area
Full Text
Figure 1. Randomized discontinuation design. PD: Progressive disease. Adapted from [42].
(Figure omitted. See article PDF.)
Phase II trials
Phase II trials allow preliminary assessment of the efficacy of an agent prior to confirmatory testing in Phase III trials. They offer the chance to optimize the dose of an agent (or combination) and obtain a better understanding of its mechanism of action. They need to be efficient and accurate to minimize the cost to patients, as well as the financial investment and time required, and to correctly identify effective (and ineffective) regimens. A recent review of Phase III trials demonstrated that only 47% of these were positive [1], suggesting that current Phase II studies are insufficiently reliable as a screening tool. Thus, there is an urgent need to refine the design and conduct of these trials [2].
Traditional approaches
Traditional approaches to Phase II trial design enroll small numbers of patients in single-arm trials and compare the effect of the drug against 'historical controls'. The end point used to assess efficacy is typically the tumor response rate (RR), as assessed by the standard criteria.
Typical designs involve the specification of a 'no further interest' level of activity/RR, and a 'desired' level of activity. Type I and II error limits are specified and setting suitable error limits enables inefficacious regimens to be screened out while promising regimens are recommended for additional evaluation. Compared with Phase III trials, Phase II studies are usually designed to have low type II errors (10%, or at most 20%) to enable identification of potentially active regimens. However, while Phase III designs are almost always based on a type I error of 5%; in Phase II studies, relatively high type I errors of 10-20% and above are acceptable, as the true activity of a new drug would usually be clarified in a Phase III trial.
Phase II trials typically involve one- or two-stage designs. In a single-stage design, the entire planned sample size of patients needs to be recruited before the study stops and the data are analyzed. Two-stage designs implement early stopping rules, allowing the trial to stop early and (typically) reject the agent if too few responses are seen at the end of the first stage. Having designs with...