Related Experiment Videos
False positive rates of randomized phase II designs
1Fred Hutchinson Cancer Research Center, Southwest Oncology Group Statistical Center, Seattle, Washington 98109-1024, USA.
Controlled Clinical Trials
|August 10, 1999
Summary
Randomized Phase II cancer trials can lead to misleading results due to high false-positive rates, potentially misinterpreting early findings as conclusive. These trials should guide Phase III studies, not serve as definitive endpoints.
Area of Science:
- Oncology
- Clinical Trials
- Biostatistics
Background:
- Randomized Phase II designs are increasingly used in cancer research to select treatments for Phase III trials.
- Misapplication of these designs is common, often treating Phase II results as definitive without further evaluation.
Purpose of the Study:
- To quantify the false-positive rates in randomized Phase II selection designs for binary and survival data.
- To highlight the risks of misinterpreting preliminary findings and the inappropriate use of control arms.
Main Methods:
- Simulations were performed for 2-, 3-, and 4-arm selection designs using binary and censored exponential survival data.
- False-positive rates were calculated under the null hypothesis of identical underlying distributions.
Main Results:
- False-positive rates ranged from 20% to over 40% for observing "impressive" differences when no true difference existed.
- The inclusion of control arms and post-hoc hypothesis testing with p-values < 0.05 exacerbate erroneous inferences.
Conclusions:
- Randomized Phase II trial results should be considered preliminary and serve as pilots for Phase III evaluations.
- Caution is advised against treating Phase II findings as conclusive, especially when control arms are involved or post-hoc testing is conducted.