ABSTRACT
As indicated early in Chapter 1, an adaptive design is a design that allows adaptations or modi…cations to some aspects of a trial after its initiation without undermining the validity and integrity of the trial. The adaptations may include, but are not limited to, sample-size re-estimation, early stopping for e¢ cacy or futility, response-adaptive randomization, and dropping inferior treatment groups (Figure 3.1). Adaptive designs usually require unblinding data and invoke a dependent sampling procedure. Therefore, theory behind adaptive design is much more complicated than that behind classic design. Validity and integrity have been strongly debated from statistical, operational, and regulatory perspectives during the past several years. However, despite di¤erent views, most scholars and practitioners believe that adaptive design could prove to be e¢ cient tools for drug development if used properly. The issues of validity and integrity will be discussed in depth in Chapter 18. Many interesting methods for adaptive design have been developed. Vir-
tually all methods can be viewed as some combination of stagewise p-values. The stagewise p-values are obtained based on the subsample from each stage; therefore, they are mutually independent and uniformly distributed over [0, 1] under the null hypothesis. The …rst method uses the same stopping boundaries as a classic group sequential design (Pocock, 1977; O’Brien and Fleming, 1979), and allows stopping for early e¢ cacy or futility. Lan and DeMets (1983) proposed the error spending method (ESM), in which the timing and number of analyses can be changed based on a prespeci-…ed error-spending function. ESM is derived from Brownian motion. The method has been extended to allow for sample-size re-estimation (SSR) (Cui, Hung, and Wang, 1999). It can be viewed as a …xed-weight method (i.e., using …xed weights for z-scores from the …rst and second stages re-
degeneralized this weight method by using the inverse-normal method, in which the z-score is not necessarily taken from a normal endpoint, but from the inverse-normal function of stagewise p-values. Hence, the method can be used for any type of endpoint.
