BOIN-ET: Bayesian optimal interval design for dose finding based on both efficacy and toxicity outcomes

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Phase I oncology trials aim to find optimal doses (OD) balancing toxicity and efficacy for later trials.
• Traditional dose-finding methods often assume monotonic relationships between dose, toxicity, and efficacy, which may not hold for novel therapies.
• Molecular targeted agents, cytostatic drugs, and immunotherapies may exhibit non-monotonic dose-response relationships.

Purpose of the Study:

• To extend the Bayesian optimal interval (BOIN) design to incorporate both toxicity and efficacy outcomes for optimal dose identification.
• To introduce a new nonparametric design, BOIN-ET, that does not rely on model-based assumptions.
• To evaluate the performance of BOIN-ET against existing model-based methods.

Main Methods:

• The study proposes the Bayesian optimal interval with efficacy and toxicity (BOIN-ET) design.
• BOIN-ET is a nonparametric approach, avoiding assumptions about dose-response relationships.
• A simulation study was conducted to compare BOIN-ET with model-based designs.

Main Results:

• The BOIN-ET design demonstrated advantages in selecting the correct optimal dose.
• BOIN-ET showed improved patient allocation to the optimal dose across various realistic scenarios.
• Simulations indicated superior performance in terms of both correct OD identification and patient distribution.

Conclusions:

• The BOIN-ET design offers a robust and flexible approach for phase I oncology dose-finding.
• This method is particularly valuable for novel therapeutics where traditional dose-response assumptions may not apply.
• BOIN-ET enhances the efficiency of early-phase clinical trials by improving optimal dose selection and patient allocation.