Related Experiment Videos
Analysis of pilot and early phase studies with small sample sizes
Weichung Joseph Shih1, Pamela A Ohman-Strickland, Yong Lin
1Division of Biometrics, School of Public Health and Cancer Institute of New Jersey, University of Medicine and Dentistry of New Jersey, 683 Hoes Lane West, P.O. Box 9, Piscataway, NJ 08854 USA. shihwj@umdnj.edu
Statistics in Medicine
|June 15, 2004
Summary
This study proposes new statistical methods for analyzing small clinical trials. The focus is on testing individual treatment effects rather than group averages, using a mixture-distribution approach for better small sample analysis.
Area of Science:
- Biostatistics
- Clinical Trial Design
Background:
- Limited attention has been given to statistical methodologies for small sample studies like pilot or proof-of-concept trials.
- The Institute of Medicine (IOM) recently held a workshop to address methodologies for clinical trials with few participants.
Purpose of the Study:
- To propose novel statistical approaches for analyzing small pilot studies.
- To shift the hypothesis testing paradigm in small trials from group effects to individual subject effects.
Main Methods:
- Proposing multiple test procedures for individuals with sufficient observations.
- Developing and focusing on a mixture-distribution approach for individuals with one or more observations.
- Presenting likelihood ratio tests for mixture models in one- or two-group settings.
Main Results:
- The proposed mixture-distribution approach offers a viable statistical framework for small pilot studies.
- Likelihood ratio tests are presented for the developed mixture models.
- Demonstration of the methods through practical examples.
Conclusions:
- The mixture-distribution approach provides a robust statistical methodology for analyzing small clinical trials.
- Testing for individual treatment effects is more appropriate for pilot studies than traditional group-level analyses.
- The proposed methods enhance the statistical rigor of small sample research.