trial designed

trial designed implements the utility-based Bayesian optimal interval (U-BOIN) phase I/II design to identify the optimal biological dose (OBD) for targeted therapy and immunotherapy by jointly modeling toxicity and efficacy with a multinomial-Dirichlet model.

Key Features:

• Utility-Based Approach: Employs a utility function to evaluate the risk–benefit trade-off at each dose level by integrating toxicity and efficacy outcomes.
• Bayesian Framework: Uses a Bayesian optimal interval design that incorporates prior knowledge and continuously updates inference as new data accrue.
• Multinomial-Dirichlet Model: Jointly models toxicity and efficacy outcomes using a multinomial-Dirichlet framework to characterize dose–response relationships.
• Seamless Two-Stage Design: Implements a two-stage design where Stage I explores the dose space using the BOIN approach and Stage II continuously updates posterior utility estimates to guide dose assignment and selection.
• Robustness and Accuracy: Demonstrates robust performance and high accuracy in identifying the OBD based on simulation studies.
• Accommodation of Delayed Efficacy: Extends to settings with delayed efficacy by incorporating short-term endpoints such as immune activity or other biological markers to predict long-term outcomes for real-time decision-making.

Scientific Applications:

• OBD identification in targeted therapy and immunotherapy: Supports identification of the optimal biological dose that balances efficacy and toxicity in targeted agents and immunotherapies.
• Phase I/II dose-finding trials: Provides a framework for integrated toxicity–efficacy dose-finding across phase I/II studies using Bayesian updating.
• Trials with delayed efficacy: Enables incorporation of short-term biomarkers (e.g., immune activity) to inform dose decisions when efficacy manifests late.

Methodology:

Implements a utility-based Bayesian optimal interval (U-BOIN) phase I/II design with a multinomial-Dirichlet joint model of toxicity and efficacy, a seamless two-stage structure (Stage I dose exploration via BOIN; Stage II continuous posterior updating of utility estimates), and incorporation of short-term endpoints (e.g., immune activity) to predict delayed efficacy.