Abstract
This paper discusses, from statistical viewpoint, about the methodology to examine whether substances have threshold on tumorgenecity or genotoxicity based on experiment data. When we try to discriminate models with threshold from those without threshold through in vivo experiment, the necessary sample size n will be too huge to be practical. In vitro assays seem to be more promising. Current methodologies utilized in in vitro assays, however, are not suitable for our purpose because they are involved in the determination of the positivity/negativity of test chemicals, while the methodologies we require are those for estimating dose-response relationships. Collaborative studies among toxicologists and statisticians are desirable to establish reasonable methodologies.
