ABSTRACT
In this chapter, we will study the so-called recursive two-stage adaptive de-
sign (RTAD; Chang, 2006). The recursive approach provides closed forms
for stopping boundaries and adjusted p-values for any K-stage design and
avoids any numerical integration; at the same time it allows for a broad
range of adaptations such as SSR, dropping losers, and changing the num-
ber and timing of analyses without specification of an error-spending func-
tion. The key ideas of the RTAD are (1) a K-stage design (K > 1) can
be constructed using recursive two-stage designs, (2) the conditional error
principle ensures that the recursive process will not inflate type-I error,
and (3) the closed form solutions are obtained through recursively utilizing
the two-stage design solutions for stopping boundary, adjusted p-value, and
conditional power. In this approach, the trial is designed one step ahead at
every interim analysis.
