Exploring the statistical and clinical impact of two interim analyses on the Phase II design with option for direct assignment

Ming Wen An, Sumithra J. Mandrekar, Martin J. Edelman, Daniel J. Sargent

Research output: Contribution to journalArticlepeer-review

2 Scopus citations


Purpose: The primary goal of Phase II clinical trials is to understand better a treatment's safety and efficacy to inform a Phase III go/no-go decision. Many Phase II designs have been proposed, incorporating randomization, interim analyses, adaptation, and patient selection. The Phase II design with an option for direct assignment (i.e. stop randomization and assign all patients to the experimental arm based on a single interim analysis (IA) at 50% accrual) was recently proposed [An et al., 2012]. We discuss this design in the context of existing designs, and extend it from a single-IA to a two-IA design. Methods: We compared the statistical properties and clinical relevance of the direct assignment design with two IA (DAD-2) versus a balanced randomized design with two IA (BRD-2) and a direct assignment design with one IA (DAD-1), over a range of response rate ratios (2.0-3.0). Results: The DAD-2 has minimal loss in power (< 2.2%) and minimal increase in T1ER (< 1.6%) compared to a BRD-2. As many as 80% more patients were treated with experimental vs. control in the DAD-2 than with the BRD-2 (experimental vs. control ratio: 1.8 vs. 1.0), and as many as 64% more in the DAD-2 than with the DAD-1 (1.8 vs. 1.1). We illustrate the DAD-2 using a case study in lung cancer. Conclusion: In the spectrum of Phase II designs, the direct assignment design, especially with two IA, provides a middle ground with desirable statistical properties and likely appeal to both clinicians and patients.

Original languageEnglish
Pages (from-to)157-162
Number of pages6
JournalContemporary Clinical Trials
Issue number2
StatePublished - Jul 2014
Externally publishedYes


  • Biomarker
  • Direct assignment
  • Phase II
  • Two interim analyses


Dive into the research topics of 'Exploring the statistical and clinical impact of two interim analyses on the Phase II design with option for direct assignment'. Together they form a unique fingerprint.

Cite this