Related Experiment Videos
Weighted p-value adjustments for animal carcinogenicity trend test
1Division of Biometry and Risk Assessment, National Center for Toxicological Research, Food and Drug Administration, Jefferson, Arkansas 72079, USA. jchen@nctr.fda.gov
Biometrics
|July 6, 2000
Summary
This study introduces weighted p-value adjustments for animal carcinogenicity experiments, improving the detection of dose-related tumor trends. Weighted methods enhance statistical power for critical tumor endpoints, unlike traditional methods.
Area of Science:
- Toxicology
- Statistical analysis
- Carcinogenesis research
Background:
- Animal carcinogenicity studies typically analyze 10-30 tumor sites individually.
- Standard p-value adjustments for multiple comparisons can reduce statistical power to detect dose effects.
- Controlling overall Type I error rate or familywise error rate (FWE) is crucial but challenging.
Purpose of the Study:
- To propose and evaluate weighted p-value adjustment methods for analyzing tumor data in carcinogenicity experiments.
- To enhance the power of detecting dose-related trends, especially for critical tumor endpoints.
- To compare the performance of weighted adjustments against unweighted methods using simulations and real-world data.
Main Methods:
- Developed two weighted adjustment methods: weighted p adjustment and weighted alpha adjustment.
- Classified tumors into Class A (critical endpoints) and Class B (less critical) for differential weighting.
- Employed Monte Carlo simulations to assess FWE control and statistical power.
- Applied the modified poly-3 test, an NTP standard, to analyze a 2-year carcinogenicity study dataset.
Main Results:
- Both weighted adjustment methods effectively controlled the familywise error rate (FWE).
- Weighted adjustments increased statistical power for Class A tumors (critical endpoints) and decreased it for Class B tumors.
- Analysis of a National Toxicology Program (NTP) dataset revealed a significant dose-related trend for a rare tumor when using weighted adjustments, which was missed by unweighted analysis.
Conclusions:
- Weighted adjustments offer a more powerful approach for analyzing carcinogenicity data, particularly when prioritizing specific tumor types.
- The proposed methods provide a valuable tool for re-evaluating existing carcinogenicity data and improving the sensitivity of future studies.
- Differential weighting based on tumor rarity or biological significance can lead to more meaningful discoveries in toxicology research.