General classes of multiple binary regression models in dose-finding problems for combination therapies.

The dose finding problem for single compounds has been generalized to combination therapies and several binary regression models have been proposed in the literature. We propose general tools to evaluate them, in particular dose-free parameterizations of the risks of toxicity and risk ratio functions and plots. New classes of risk functions are also proposed. They generate low dimensional parametric risk functions which are made available to the researcher and commented on for their clinical relevance.