Combination Therapy in Oncology Clinical Development

Organizers: Freda Cooner (Amgen), Qi Jiang (Seagen), Anastasia Ivanova (UNC)
Chair: Anastasia Ivanova (UNC)
Vice Chair: Freda Cooner (Amgen)

Miao Zang (Sanofi)
Ruitao Lin (MD Anderson)
Pooja Saha (UNC)
Inna Perevozskaya (GSK)


Title: Phase I Oncology Trial Design of Dose Escalation for Novel-Novel Combination Therapy
Speaker: Miao Zang (Sanofi)

In dose-finding oncology studies, the objective is to determine the maximum tolerated dose of the experiment agent, defined as the highest dose with an acceptable dose-limiting toxicity rate. Many methods are proposed for studies to find the dose of a single agent, in either monotherapy or in combination with a backbone therapy with an established dose. Dose finding for trials to find the doses of more than one agent is more difficult than that for conventional single-agent trials because of potentially complicated drug–drug interactions. Single agent dose escalation model is assuming that as dose increases toxicities increase accordingly. However, when combining several agents, such assumption is not applicable. A one-dimensional order of the doses of the experimental agents does not exist, implying a complicated multi-dimensional grid-search might be needed to find the right dose, At the same time, the dose escalation (or de-escalation) decision in single agent trial assumed an ordered dose sequence. Without a pre-specified ordered sequence, it is not easy to set the dose decision rule for novel-novel combination therapy. Recently, many designs have been proposed for drug combination trials. Most of these methods consider dose levels in discrete space. For example, Riviere et al. proposed a method by modeling the relationship between the logit of combination therapy toxicity and the logit of single agent toxicity 2014. Another type of approach is still considering the dose levels in continuous space. An example is the interaction model from Novartis group. Several simulation studies were conducted to assess the performance of both methods. Eventually, Riviere’s model was selected and applied to a Phase 1b/2, open-label, multicenter dose-escalation and dose expansion study of a novel-novel drug combination trial for adult participants with relapsed/refractory solid tumors. The real trial data is going to be presented in the end.

Title: BOIN Combination Design for Dose Finding in Early-Phase Oncology Trials
Speaker: Ruitao Lin (MD Anderson)

In the era of precision medicine, drug combinations have been widely introduced in clinical trials in order to improve treatment efficacy. In contrast to single-agent phase I trials which usually assume a monotone dose-toxicity relationship, dose-finding studies for combination therapy face many challenges in selecting the MTD among the range of possible combinations. To tackle the challenges in combination trials, numerous phase I drug-combination trial designs have been proposed using different statistical tools. In this talk, we will revisit the Bayesian optimal interval combination (BOIN COMB) design for dose finding of a single MTD in drug-combination trials. The dose-finding rule and implementation details will be illustrated using examples. We conduct a comprehensive Monto Carlo study to investigate the performance of BOIN COMB from the perspectives of accuracy, safety. For a fair comparison, we use a novel random scenario generator to simulate various toxicity scenarios randomly and objectively for drug-combination trials. To promote the use of novel trial designs in real applications, in the last part of the talk, we will list and compare some software/programs for implementing the drug-combination designs.

Title: Model-based dose-finding for single- and multi-agent immunotherapy trials
Speaker: Pooja Saha (UNC)

In many phase 1 oncology trials of immunotherapies, no dose-limiting toxicities are observed and the maximum tolerated dose cannot be identified. In these settings, dose-finding can be guided by a biomarker of response rather than the occurrences of dose-limiting toxicity. The recommended phase 2 dose can be defined as the dose with mean response equal to a pre-specified value of a continuous response biomarker. To target the mean of a continuous biomarker, we build on the idea of the continual reassessment method and the quasi-Bernoulli likelihood. We extend the design to a problem of finding the recommended phase 2 dose combination in a trial with multiple immunotherapies.

Title: Phase I/II seamless designs
Speaker: Inna Perevozskaya (GSK)

Phase I dose-escalation in oncology has seen a dramatic uptake of innovative design usage over the past decade. Many companies are adopting a model-based approach versus traditional 3+3 designs, but even these improved methods cannot dramatically increase precision of MTD finding due to their limited sample size and restrictive dose exploration.  There is added on complexity of evaluating multiple drugs (combinations), multiple grades of toxicity or simultaneous assessment of target engagement and toxicity in a single study. In other words, Phase 1 oncology trials of today are very different from small, MTD-focused studies of a single drug with dichotomous endpoint of the past. They are larger, often include substantial expansion cohorts, and their objectives are more complex than simple MTD determination as a single dose to take forward. Such goals are better addressed by incorporating a seamless Phase 2-like extension into initial dose-escalation phase and using combined data along with quantitative decision making to improve the probability of success on further development.