Related Experiment Videos
Comparison of two approaches to analyzing correlated binary data in developmental toxicity studies
1Center for Biostatistics and Epidemiology, Pennsylvania State University, Hershey 17033, USA.
Teratology
|November 1, 1995
Summary
This study introduces new statistical methods for analyzing correlated developmental toxicity data, focusing on chemical exposure effects across multiple outcomes in animal models. The findings highlight the importance of choosing appropriate statistical models for accurate effect homogeneity assessment.
Area of Science:
- Toxicology
- Biostatistics
- Developmental Biology
Background:
- Correlated developmental toxicity data, with multiple measures per unit (e.g., rat pups), requires specialized statistical analysis.
- Assessing the homogeneity of chemical exposure effects across different outcomes (e.g., malformations, ossification) is crucial.
Purpose of the Study:
- To present recently developed statistical methodology for analyzing correlated developmental toxicity data.
- To assess the homogeneity of chemical exposure effects across different outcomes within clusters (within-cluster effects) versus between clusters (between-cluster effects).
Main Methods:
- Comparison of cluster-specific models (interpreting parameters as within-cluster effects) and population-averaged models (interpreting parameters as group differences).
- Illustration using data from two developmental toxicity studies: di(2-ethylhexyl)phthalate exposure in mice and phenytoin exposure in rats.
Main Results:
- Population-averaged models, common in developmental toxicity literature, may ignore significant cluster variation due to genetic and environmental factors.
- Confidence interval-based inference for effect homogeneity is dependent on the chosen statistical model class (cluster-specific vs. population-averaged).
Conclusions:
- The choice of statistical model significantly impacts the inference of effect homogeneity in developmental toxicity studies.
- Accurate analysis of correlated developmental toxicity data requires careful consideration of within- and between-cluster effects.