ABSTRACT
The classical multiple-arm design with the strong familywise type-I error
control is a single-stage design with several active arms and a common
control arm, in which Dunnett’s test (1955, 1964) based on multivariate
normal distribution is most used for the multiple comparisons. To improve
the design, recently, multiple-arm adaptive designs have been studied. A
typical multiple-arm confirmatory adaptive design (often called drop-the-
loser, drop-arm, or pick-the-winner design or adaptive dose-finding design,
or phase-II/III seamless design) consists of two stages: a selection stage and
a confirmation stage. For the selection stage, a randomized parallel design
with several doses and a placebo group is employed for selection of doses.
After the best dose (the winner) is chosen, the patients of the selected dose
group and placebo group will enter the confirmation stage. New patients
will be recruited and randomized to receive the selected dose or placebo.
The final analysis is performed with the cumulative data of patients from
both stages (Chang 2011; Maca, et al., 2006).