S35.1: Multiple comparisons and modelling in dose finding: a unified approach

✍ Scribed by Frank Bretz; Jose Pinheiro; Mike Branson

• Publisher: John Wiley and Sons
• Year: 2004
• Tongue: English
• Weight: 73 KB
• Volume: 46
• Category: Article
• ISSN: 0323-3847
• DOI: 10.1002/bimj.200490127

✦ Synopsis

The analysis of dose response studies has long been divided according to two major strategies: multiple comparison procedures and model-based approaches. The model-based approach assumes a functional relationship between the response and the dose, taken as a quantitative factor, according to a pre-specified parametric model. The fitted model is then used to estimate an adequate dose to achieve a desired response. Such an approach provides flexibility in investigating the effect of doses not used in the actual study. However, the validity of its conclusions will highly depend on the correct choice of the a priori unknown dose-response model. This creates a dilemma within the regulated environment in which drug development takes place, since it is required to have the analysis methods (including the choice of the dose-response model) defined prior to the study. Multiple comparison procedures, on the other hand, regard the dose as a qualitative factor and make very few, if any, assumptions about the underlying dose-response model. The primary goal is often to identify the minimum effective dose that is statistically significant and produces a clinically relevant effect. One approach is to evaluate the significance of contrasts between different dose levels, while preserving the familywise error rate. Such procedures are relatively robust to the underlying dose-response shape, but they are not designed for extrapolation of information beyond the observed dose levels. In this talk, we describe a unified strategy to the analysis of data from dose-response studies which combines multiple comparison and modeling techniques. We assume the existence of several candidate parametric models and use multiple comparison techniques to choose the one most likely to represent the true underlying dose-response curve. Such a procedure allows the selection of the most adequate dose-response model within the candidate set, while preserving the familywise error rate. The selected model is then used to provide inference on adequate doses, as described above. The methods will be illustrated with data from a phase II dosefinding study.