A Two-Stage Method to Minimize the Expected Sample Size of a Single-Arm Study Under the Alternative Hypothesis

• Activity Number: 385 - SPEED: Statistical Methods and Applications in Medical Research, Risk Analysis, and Marketing Part 1
• Type: Contributed
• Date/Time: Wednesday, August 10, 2022 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Xiaobo Zhong* and Qian Li
• Companies: Bristol Myers Squibb and Biostatistics and Strategic Consulting
• Keywords: Clinical Trial; Design; OPTIMAR; Power; Type I error; Sample Size

Abstract

Often the evaluation of new therapeutic drugs adopt single-arm design with binary endpoints, such as overall response rate (ORR) in oncology. Although the planned sample size (SS) varies by purpose, these studies typically enroll < 100 patients. Like randomized and controlled trials, the single-arm study can use multi-stages design embedded in interim looks for adapting SS or early stopping. We propose a two-stage design, named OPTIMAR (OPTIMum under Alternative Response) to optimize a study by minimizing the expected SS under alternative response (AR). OPTIMAR includes 4 important pre-specified design parameters: SS for interim (N1) and final analyses (N), two critical values (CV1 and CV2), where CV1< CV2. The decision-making rules are Step 1: At N1, compare the observed ORR with CVs. We recommend stopping the trial immediately and claim failure if ORR< CV1 or claim success if ORR>CV2. Otherwise, keep enrolling patients until N is reached; Step 2: Conduct final analysis at N patients based on 95% one-sided confidence limit only. Simulation studies showed that OPTIMAR can control type I error well and requires smaller expected SS under AR compared to Simon’s 2-stage optimal design.