Enhancing generalizability and efficiency in clinical trials through dynamic information borrowing for both...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmaceutical Research

Background:

• External information enhances clinical trial validity and efficiency.
• Borrowing information in both experimental and control arms presents unique challenges.
• Existing research often focuses on control arm information borrowing only.

Purpose of the Study:

• To evaluate information borrowing methods in two-arm clinical trials.
• To compare the operating characteristics of test-then-pool, normalized power prior, calibrated power prior, and elastic prior methods.
• To identify optimal methods for different levels of data heterogeneity.

Main Methods:

• Applied four distinct information borrowing methods: test-then-pool, normalized power prior, calibrated power prior, and elastic prior.
• Utilized a two-arm information borrowing framework.
• Conducted simulation studies under most and least desirable scenarios of information borrowing.

Main Results:

• All methods controlled type I error rates within 5% in desirable scenarios.
• Normalized power prior, calibrated power prior, and elastic prior increased power to 95% in desirable scenarios.
• In least desirable scenarios, type I error rates exceeded 20% for these priors, with power decreasing to ~80%.

Conclusions:

• Normalized power prior, calibrated power prior, and elastic prior are effective with minimal data heterogeneity.
• The test-then-pool approach is a more robust choice when significant discrepancies exist between historical and current data.
• Findings guide the selection of appropriate information borrowing strategies in clinical trials.