Unit information prior for incorporating real-world evidence into randomized controlled trials

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Health Services Research

Background:

• Randomized controlled trials (RCTs) are the gold standard for treatment effect inference.
• Integrating real-world evidence (RWE) from observational studies can enhance RCT results.
• Existing methods for RWE integration face challenges with bias and data requirements.

Purpose of the Study:

• To extend the unit information prior (UIP) technique for synthesizing non-randomized evidence into current RCTs.
• To demonstrate the UIP's ability to use summary statistics from observational studies.
• To improve the statistical efficiency and interpretation of treatment effect estimation.

Main Methods:

• The unit information prior (UIP) method was extended to incorporate data from multiple historical observational studies.
• The approach requires only summary statistics from published observational studies.
• Weighting schemes were employed within the UIP to mitigate potential biases in real-world evidence.

Main Results:

• Extensive numerical experiments demonstrated improved statistical efficiency in treatment effect estimation across various outcome types.
• The UIP method showed robust performance in simulations.
• The approach effectively borrows information from historical datasets.

Conclusions:

• The extended unit information prior (UIP) offers a valuable method for integrating real-world evidence into randomized controlled trials.
• The UIP improves statistical efficiency and provides clear interpretations for treatment effect estimation.
• The approach is practical and demonstrated utility in a real-world COVID-19 treatment trial.