Traditionally, oncology Phase I dose escalation trials focus on identifying a maximum tolerated dose and on the generation of safety evidence. Here, a Bayesian Logistic Regression Model is a popular and widely used method to guide dose escalation.

We currently experience a huge paradigm shift in oncology related to dose optimization (“Project Optimus”). With the emergence of modern cancer therapies, the rationale of simply moving forward with the maximum tolerated dose up to a registrational trial could lead to suboptimal dose levels and is not accepted anymore. With this, it is of interest to generate more evidence across a broad dose range already during the earliest phases of clinical development to allow for a better characterization of activity and tolerability. To achieve this, the FDA recommends in their draft guidance to consider “add[ing] more patients to dose-level cohorts in a dose-finding trial […]”. In dose escalation trials, this concept is known as back-fill cohorts, where additional patients are recruited on lower dose levels while dose escalation is ongoing at higher doses. While back-fill cohorts help generating more evidence already during dose finding, this approach comes with various statistical and clinical challenges that need to be addressed.

In this presentation, we will discuss some of these challenges that occur when back-fill cohorts are implemented in dose-finding trials guided by a Bayesian logistic regression model. This includes possible conditions on when such cohorts should be opened and how this additional data can be included in the statistical analyses. Additionally, we will present operating characteristics from trial simulations that investigated the impact of enrolling back-fill cohorts during dose escalation in trials with monotherapy and/or combination therapy arms. Special focus lies on the risk of exposing a high numbers of patients to overly toxic doses.

Treatment of cancer has rapidly evolved over time in quite dramatic ways, for example from chemotherapies, targeted therapies to immunotherapies and chimeric antigen receptor T-cells. Nonetheless, the basic design of early phase I trials in oncology still follows pre-dominantly a dose-escalation design monitoring the safety of the first treatment cycle only. With toxicities occurring at later stages beyond the first cycle and the need to treat patients over multiple cycles, the focus on the first treatment cycle only is becoming a limitation in nowadays multi-cycle treatment therapies. Here we introduce a multi-cycle time-to-event model allowing guidance of dose-escalation trials studying multi-cycle therapies. The challenge lies in balancing the need to monitor safety of longer treatment periods with the need to continuously enroll patients safely. We introduce in this work a multi-cycle time to event model which is formulated as an extension to existing approaches like the escalation with overdose control principle. The model is motivated from a drug development project and evaluated in a simulation study.

In Phase I/II dose-finding trials, the objective is to find the Optimal Biological Dose (OBD), a dose that is both safe and shows sufficient activity that maximises some optimality criterion based on safety and activity. In cancer treatment is typically given over several cycles complicating the identification of the OBD as both toxicity and activity outcomes may occur at any point throughout the follow up of multiple cycles. In this work we present and assess the Joint TITE-CRM, a model-based design for late onset toxicities and activity based on the well-known TITE-CRM. It is found to be superior to both currently available alternative designs that account for late onset bivariate outcomes; a model-assisted method and a bivariate survival design, as well as being both intuitive and computationally feasible.