ADAPTING REJECTION BOUNDARIES WHEN THE TARGETED ACCRUAL IS NOT ATTAINED IN 2-STAGE PHASE II CLINICAL TRIALS by

Although using the 2-stage designs proposed in the literature has certain advantages such as limiting the expected number of people exposed to inactive agents while retaining overall size and power, the practical application of these methods can be challenging in a cooperative setting because of the difficulty in obtaining the exact sample sizes as demanded by these procedures. Chen and Ng proposed a flexible design that is nearly optimal but allows the accrual of each stage to vary by 8 patients. This was a significant advance; however, the possibility of overaccruing or underaccruing still exists. In addition, the calculation of the rejection boundaries using their methods is not a trivial task, so statisticians typically must rely on the tables presented in their manuscript when designing a clinical trial. The methods discussed in this paper allow for the adjustment of rejection boundaries while retaining reasonably good operating characteristics for a design when the realized sample size falls outside the range prescribed by Chen and Ng or misses a targeted accrual as specified by another procedure (such as Simon). Procedures are available in SAS for implementing these methods.