Global comparison of radiation and chemotherapy dose-response curves with a test for interaction

In the analysis of the effects of radiation or drugs on clonogenic survival data of mammalian cells, it is often advantageous to compare entire dose-response curves generated under different experimental conditions rather than to conduct single-dose comparisons. We propose a two-stage method for the global comparison of such curves. The first stage consists of individual fits of a flexible model to the dose-response sequences. The second stage treats the fitted coefficients as data and analyzes them jointly using a multivariate analysis of variance. For dose-response models of the type commonly used in the analysis of clonogenic survival data, this method allows the definition of a statistical test for interaction among treatments. That is, a test that the combination of agents produces a dose-response curve which is not what would be expected if the agents were acting multiplicatively (additively in the log cell survival scale). Analyses of cell survival curves by this method for experiments on iododeoxyuridine-mediated radiosensitization and on chemosensitization of bleomycin cytotoxicity in a human bladder cancer cell line (647V) are presented.