The majority of phase III clinical trials, despite being conducted after promising phase II trials, are "negative," with the new therapy determined in the end to be too toxic or insufficiently efficacious. One explanation is the heterogeneity of the populations participating in various phases of development, which results in an erroneous estimation of the toxicity and thus a diluted therapeutic effect. This may lead to termination of evaluation of a therapy, even if a sub-population, defined by a particular characteristic, may stand to benefit from it. In this thesis, we propose a close examination of the methodological aspects of phase II trials which would permit improved early identification of toxic therapies and of responsive populations, so that phase III trials may be designed only with the best targeted populations in mind. We present as well a new phase II clinical trial methodology which we have developed to take into account trial population heterogeneity and its importance in current clinical practice. With this method, drug development is less often stopped for the entire phase II population and less non sensitive patients are exposed to toxic drugs in the second part of phase II trials, and next in phase III trials.
