Local continual reassessment methods for dose finding and optimization in drug-combination trials

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Optimizing cancer drug combinations is challenging due to small sample sizes and vast dose ranges.
• Existing model-based designs require extensive parameter elicitation and can be unstable with sparse data or model misspecification.

Purpose of the Study:

• To develop robust and efficient designs for exploring dose combinations in early-phase cancer clinical trials.
• To reduce the complexity of model calibration and improve design stability.

Main Methods:

• Proposed local, underparameterized models for dose exploration, building on the partial ordering continual reassessment method.
• Developed local data-based continual reassessment method (LD-CR) designs for identifying maximum tolerated dose combinations (MTD) and optimal biological dose combinations (OBD).
• LD-CR designs model only local data from neighboring dose combinations, enhancing flexibility and stability.

Main Results:

• Simulation studies demonstrated competitive performance of LD-CR designs for MTD identification compared to existing methods.
• LD-CR designs showed advantages over model-based methods for OBD optimization.
• The local modeling approach proved flexible for characterizing local dose-exploration spaces.

Conclusions:

• Local data-based continual reassessment method designs offer a robust and efficient alternative for dose-finding in combination cancer therapies.
• This approach simplifies model calibration and enhances design stability, particularly beneficial in early drug development.
• The method shows promise for both MTD and OBD identification, outperforming existing model-based strategies in certain aspects.