Statistical Assessment of Toxicologic Pathology Studies

Toxicologic pathologists employ animal models to help understand and predict responses in humans, in terms of toxic and pharmacologic effects in particular. This use remains both the major tool for biomedical advances and a frequent source of significant controversy. Application of statistics to data sets from toxicologic pathology studies involves hypothesis testing, model fitting, and reduction of dimensionality. This chapter reviews statistical approaches to processing and analyzing data (i.e., descriptive statistics, confidence intervals, parametric and non-parametric methods, multiple range tests, handling of outliers) as well as specific applications (e.g., Bayesian statistics) and other experimental design and statistical methods. A set of decision trees is included to assist in choosing the appropriate methods for specific study designs. These analyses are used to establish the statistical nature and probe the biological relevance of the data being generated, and also to design and analyze experiments accordingly.