|
Abstract:
|
With increasing competition in oncology drug development and emerging of new entities allowing more combination therapies, research demand for efficient and precise methods in identifying the most tolerable doses reaches historical high. While the problem is an order of magnitude harder since there is a whole frontier of dose combinations acceptable from safety perspective instead of one single dose in monotherapy dose escalation. We propose a two-stage, Bayesian method based, split cohort design (TBSC design), where an initial rule based escalation (i.e. 3+3-like along the diagonal of dose combination) is followed by an adaptive stage where a continuously updated Bayesian logistic regression is used to suggest the dose combinations on the location of the acceptable-risk-frontier.In addition, our proposed method can incorporate a) concurrently running mono-trials, b) insert new doses if deemed required, c) incorporate historical information in case one of the entity is well-known and d) initiate several dose cohorts in parallel to speed up trial completion. The proposed method is compared to others used in industry in terms of precision and trial speed.
|
