5.2. Key Statistical Aspects of Toxicology Studies

The key statistical aspects of the design and analysis of toxicology studies to be discussed are randomization, power evaluation and data analysis.

5.2.1. Randomization

Since proper randomization is the foundation of valid statistical inference, this chapter starts with the randomization methods for two commonly used designs in toxicology, namely, the parallel design and Latin square design.

Parallel designs include studies with one factor, two factors, one-factor with repeated measurements, and more. In a parallel design, separate groups of animals are assigned to combinations of factor levels. For example, in a one-factor design with four dose groups, there would be four different groups of animals, each receiving one of the four doses. For a two-factor design with four dose groups and two routes of delivery, there would be eight different groups of animals each receiving one of the eight combinations of the four doses and two routes.

Latin square designs are efficient for incorporating three factors each with the same number of levels. For example, a basic 4 × 4 Latin square design can accommodate four animals, each receiving one of four treatments in four dosing periods. This design allows for the evaluation of the treatment, animal, and time effects. However, one has to assume that all two- or three-way interactions are not significant.

The random assignment of animals to the groups for the parallel and Latin square designs is discussed in Section 5.3.

5.2.2. Power Evaluation

Power evaluation characterizes the strength of an inferential test. It is a function of the size of the change to be detected, the variability, the Type I error rate, and the sample size. Although certain sample sizes in standard toxicology studies are commonly accepted and based, in part, on regulatory guidance, many toxicology studies targeted at special endpoints merit an assessment of the power and sample size. An example of this is the evaluation of QT prolongation in large animal toxicology studies with four treatment groups and three or four animals per group for each sex. The QT interval is a measure of the time between the start of the Q wave and the end of the T wave in the heart's electrical cycle. The details of the study design and statistical tests are described in Section 5.4. The evaluation of power is performed in a two-factor analysis of variance (ANOVA) framework by simulation.

5.2.3. Data Analysis

For each well-designed study, the statistical hypotheses and tests are defined in the protocol. The most commonly collected data for toxicology studies is body weight. A body weight change, either a gain or loss, is important for monitoring the well-being of the animal and toxicities of a compound. Section 5.5 describes statistical methods used in the analysis of body weight data in toxicology studies based on a one-factor ANOVA model with repeated measures.

To save space, some SAS code has been shortened and some output is not shown. The complete SAS code and data sets used in this book are available on the book's companion Web site at http://support.sas.com/publishing/bbu/companion site/60622.html.