Chapter 2
Assessing Toxicities in a Clinical
TTrial: Bayesian Inference for
Multivariate Ordinal Data
2.1 Overview
We address modeling and inference for data that include ordinal outcomes nested
within categories. The data format can alternatively be seen as multivariate ordinal
data with each dimension of the multivariate outcome corresponding to one level of a
categorical variable. The motivating application is to model adverse event (toxicity)
data in clinical trials. Toxicity type and severity are usually recorded as categorical
and ordinal outcome, respectively. In a randomized phase III study, in addition to the
efficacy of the study agent, investigators and regulators are also interested in learning
about the toxicity profile of the study agent. Traditionally, simple descriptive statis-
tics such as cross-tabulations have been provided. However, this purely descriptive
approach fails to offer an in-depth understanding of how the treatment affects both
the toxicity type and the severity associated with a specific type of toxicity.
The multinomial probit (MNP) model (Aitchison and Bennett, 1970) and the
More intriguing information
1. The name is absent2. The name is absent
3. Wirtschaftslage und Reformprozesse in Estland, Lettland, und Litauen: Bericht 2001
4. The name is absent
5. Restructuring of industrial economies in countries in transition: Experience of Ukraine
6. The Role of Evidence in Establishing Trust in Repositories
7. Regulation of the Electricity Industry in Bolivia: Its Impact on Access to the Poor, Prices and Quality
8. LIMITS OF PUBLIC POLICY EDUCATION
9. The name is absent
10. NVESTIGATING LEXICAL ACQUISITION PATTERNS: CONTEXT AND COGNITION