6 min read
3 min read
The development of innovative oncology therapies has evolved dramatically, moving beyond traditional chemotherapies to include immunotherapies, targeted agents, radiotherapies, and cell and gene therapies. As these new modalities emerge, the limitations of conventional dose-escalation methods— such as the widely used 3+3 design—have become increasingly apparent. In particular, there is growing uncertainty about the appropriateness of rule-based algorithms for determining the recommended Phase 2 dose (RP2D) in this rapidly advancing landscape.
This shift in thinking aligns with the FDA Oncology Center of Excellence’s 2023 initiative, Project Optimus, which emphasizes identifying an optimal biological dose rather than simply the maximum tolerated dose (MTD). To meet these evolving regulatory expectations, Bayesian study designs have the potential to advance complex decision algorithms and provide a flexible alternative for early-phase oncology research.
👉 Limitations of Traditional Dose-Escalation Designs
Despite their limitations, rule-based study designs such as the 3+3 remain common in Phase 1 oncology trials due to their perceived simplicity, ease of implementation, and general acceptance by clinical sites. These designs often require smaller sample sizes and are familiar to regulatory agencies and investigators. However, their rigidity can result in missed opportunities to gain more comprehensive insights into a new therapy’s pharmacological and safety profiles.
For example, in a typical 3+3 design, if two or more of six participants experience a dose-limiting toxicity (DLT), the dose level is deemed unsafe and further escalation is halted. This approach not only restricts the number of evaluable participants at each dose level but also restricts re-escalation, potentially leading to suboptimal dose selection and only limited exploration of the dose-response relationship.
👉 Advantages of Bayesian Study Designs
Bayesian adaptive designs offer a more flexible, data-informed framework that enables better decision-making during dose escalation. These designs allow for:
- Dose escalation, de-escalation, and re-escalation, with decisions guided by a continuously updated posterior probability distribution.
- Integration of multiple data types beyond DLTs, including pharmacokinetics (PK), pharmacodynamics, immune responses, and other biomarkers.
- Cohort size flexibility, allowing more patients to be treated at a dose level when supported by safety and model-based evidence.
- Learning across dose levels, where information gathered at lower doses can be used to inform the evaluation and model-based estimates for higher dose levels.
With sufficient sample size and by leveraging accumulating data during the trial, Bayesian designs reduce the risk of overdosing.1
However, these benefits hinge on thoughtful implementation. Overly rigid constraints on cohort size or dose escalation rules diminish the advantages of a Bayesian framework, potentially reducing it to a 3+3 equivalent.
The additional nuance of parameters such as PK, immunology and efficacy does not occur by default in any study design. Their inclusion must be prespecified within the study design, with appropriately dynamic decision criteria and/or multi-parametric models planned to account for this added complexity.
👉 Incorporating Complexity in Dose Optimization
Traditional study designs often focus narrowly on acute toxicities, typically identified as DLTs, during the first cycle of treatment. Yet for many novel oncology agents, especially those with immunologic or delayed effects, this time frame may not sufficiently capture the full therapeutic and safety profile. Incorporating extended follow-up and additional parameters such as drug accumulation, cytokine release, immune cell activation, or delayed toxicities can provide important additional information to inform dose selection.
Bayesian designs are one option to accommodate complexity, with growing research to support the use. When structured to include multidimensional decision criteria, these designs have the potential for efficient dose optimization, especially in the context of biologics, immunotherapies, and consideration of PK profiles2
👉 Veristat’s Recommendations for Early Oncology Trials
To fully realize the benefits of Bayesian approaches in early-phase oncology trials, Veristat recommends that sponsors and investigators consider the following:
- Align study objectives with key research questions: Define the parameters critical for dose selection, including therapeutic index, PK and pharmacodynamic characteristics, and relevant biomarkers. These should be embedded into the decision criteria.
- Extend observation periods: Move beyond single-cycle evaluation to capture the full safety and efficacy profile of novel agents, particularly those expected to have delayed effects or cumulative toxicity.
- Define success multidimensionally: Combine safety, efficacy, PK, and immunologic data into an integrated decision framework to guide dose optimization.
👉 Conclusion
Early phase Bayesian designs represent a strategic advantage in early-phase oncology trials. With increasing emphasis on dose optimization over MTD, sponsors who invest in adaptive, model-informed trial designs will be better positioned to bring safer, more effective therapies to patients. At Veristat, we support our partners in implementing flexible and scientifically rigorous early-phase strategies tailored to the unique challenges of oncology drug development.
At Veristat, we are ready to help you navigate this complexity. Whether you are refining a dose-escalation strategy or designing a first-in-human study, our experts are here to support you in implementing innovative, data-driven approaches that accelerate your path to success. Reach out to us to learn more.
Reach out to us to learn more:
Reference:
1. Zhou H, Yuang Y, Nie L. Accuracy, Safety, and Reliability of Novel Phase I Trial Designs. Statistics in Clinical Cancer Research, 24(18); 4357-64; doi 10.1158/1078-0432.CCR-18-0168.
2. Chen X, He R, Chen X, Jiang L, Wang F. Optimizing dose-schedule regimens with Bayesian adaptive designs: opportunities and challenges. Frontiers in Pharmacology, 2023; doi 10.3389/fphar.2023.1261312.