Relatore
Alessandro Vagheggini
Research fellow - Dipartimento di Scienze Statistiche Bologna
Abstract
Phase-III clinical trials deal with assessing the efficacy of new drugs on patients; therefore, great care has to be placed on the ethical concern of assigning as much as possible individuals to the better performing treatment(s) still ensuring good statistical performances. Often, these demands could be combined by adopting suitable classes of target allocations which, by depending on the unknown treatment effects, may represent a valid trade-off between ethics and inference.
Response-Adaptive (RA) randomized designs are a class of sequential allocation rules which randomize every assignment and, asymptotically, approach the chosen target. At each step, the unknown parameters as well as the target are estimated, and the probabilities of treatment assignments are computed to force the next allocation to converge to the desired target. Within this framework, we propose a new test statistics based on the treatment allocation proportions, which, coupled with a suitable class of RA rules, is more efficient and uniformly more powerful with respect to the classical Wald test. Moreover, we show how our strategy is able to overcome the duality between ethics and inference, also providing a considerable increase in inferential precision.
Organizzatori
Alessandra Luati
Silvia Cagnone