Related Experiment Video
Updated: Aug 8, 2026
06:28
E-Patient Counseling Trial (E-PACO): Computer Based Education versus Nurse Counseling for Patients to Prepare for Colonoscopy
Published on: August 1, 2019
Internal pilot studies II: comparison of various procedures
D M Zucker1, J T Wittes, O Schabenberger
1Department of Statistics, Hebrew University, Mount Scopus, Jerusalem, Israel. mszucker@mscc.huji.ac.il
Statistics in Medicine
|December 28, 1999
Summary
This study compares statistical designs for clinical trials that allow sample size adjustments. The Stein design is best for its original purpose, but other methods offer more power for detecting smaller effects.
Area of Science:
- Biostatistics
- Clinical Trial Design
- Statistical Power
Background:
- The two-stage adaptive design allows for sample size recalculation based on interim variance estimates.
- Renewed interest in Stein's 1945 design has led to modifications by Wittes & Brittain and Gould & Shih.
Purpose of the Study:
- To compare the statistical power of different two-stage designs.
- To evaluate procedures controlling the conditional type I error rate.
Main Methods:
- Comparison of Stein's original test, an idealized Wittes-Brittain test, and an approximate-optimal test.
- Evaluation of Gould & Shih's procedure and a 'second segment' procedure for type I error control.
Main Results:
- Stein's test is unexpectedly optimal under its original assumptions.
- Approximate-optimal and Wittes-Brittain procedures show superior power for smaller treatment differences.
- The 'second segment' procedure is more powerful when extensive post-interim data collection is expected.
Conclusions:
- The choice of two-stage design depends on the specific goals and expected data collection.
- Modified designs offer advantages in power for detecting subtle treatment effects.
