Improving clinical trial efficiency by biomarker-guided patient selection.

Area of Science:

• Clinical trial design
• Biostatistics
• Pharmacoeconomics

Background:

• Patient markers predict differential treatment response in various therapeutic areas.
• Predictive markers include baseline characteristics and short-term treatment changes.
• Using predictive markers can enhance clinical trial targeting and reduce subject recruitment numbers.

Purpose of the Study:

• To compare the sample sizes required for three different clinical trial designs.
• To evaluate designs based on baseline characteristics or early improvement versus a conventional parallel group design.
• To assess the efficiency of these designs under various realistic scenarios.

Main Methods:

• Simulated data using a model of treatment effect on survival, incorporating primary effects and marker interactions.
• Compared a conventional parallel group design with baseline selection and active run-in designs.
• Evaluated a representative scenario derived from empirical data.

Main Results:

• An active run-in design substantially reduced required sample size when early improvement reliably predicted differential response.
• The baseline selection design was more efficient than the parallel group design but less efficient than the active run-in design.
• The advantage of the baseline selection design was limited in most evaluated scenarios.

Conclusions:

• Active run-in designs can significantly decrease the number of subjects needed for randomized clinical trials.
• Both active run-in and baseline selection designs may face limitations in generalizability.
• Implementation of these designs can present practical difficulties.