Data-dependent early completion of dose-finding trials for drug-combination

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Model-assisted designs offer advantages in drug combination trials but require pre-set sample sizes and lack statistical basis for early stopping.
• Existing model-assisted designs face limitations in adaptive sample size and early termination without robust statistical grounding.

Purpose of the Study:

• To propose two novel methods for data-dependent early completion of dose-finding trials for drug combinations.
• To address the limitations of fixed sample sizes in model-assisted drug combination trials.
• To introduce methods based on dose retainment probability for efficient trial termination.

Main Methods:

• Developed an early completion method utilizing dose retainment probability based on accumulating trial data.
• Introduced a second method adjusting dose retainment probability with bivariate isotonic regression.
• Evaluated method performance through extensive simulation studies across 12 diverse scenarios.

Main Results:

• Simulation studies demonstrated an average early completion rate of approximately 70%.
• The proposed methods reduced the number of treated patients by an average of 25% compared to planned sample sizes.
• Early completion methods showed comparable accuracy in selecting the maximum tolerated dose combination, with only a ~3% difference from non-early completion methods.

Conclusions:

• The proposed early completion methods demonstrate comparable performance to traditional non-early completion designs.
• These methods provide a statistically sound basis for determining early trial completion, reducing patient burden.
• A framework for calculating dose retainment probability and determining patient numbers for early completion was established.