Model-Based Adaptive Optimal Design (MBAOD) Improves Combination Dose Finding Designs: an Example in Oncology.

Area of Science:

• Clinical trial design
• Pharmacology
• Biostatistics

Background:

• Phase 1 combination therapy trials are complex due to multiple agents.
• Determining optimal dosing regimens requires advanced methodologies.
• Neutropenia is a key toxicity to manage in cancer drug development.

Purpose of the Study:

• To exemplify and evaluate model-based adaptive optimal design (MBAOD) for a phase 1 combination therapy trial.
• To optimize the dosing regimen of paclitaxel and a hypothetical new compound for phase 2 studies.
• To target a 33% probability of grade 4 neutropenia while maximizing efficacy.

Main Methods:

• MBAOD was applied to a simulated phase 1 trial involving paclitaxel and a novel compound.
• Dose optimization considered both drug amounts and the dosing schedule of the new drug.
• Exploration of various starting conditions, search paths, and stopping criteria, including the "3+3 rule".

Main Results:

• MBAOD demonstrated flexibility in modifying doses and schedules during the trial.
• The "3+3 rule" was conservative and safe but limited efficacy selection ( <21%).
• Removing the "3+3 rule" improved performance (>67% maximal efficacy selection) with similar toxicity.

Conclusions:

• MBAOD is a promising and flexible tool for dose-finding studies in combination cancer therapy.
• Adaptive designs can improve the efficiency and outcomes of early-phase combination trials.
• MBAOD facilitates the integration of complex clinical protocol requirements into trial design.