Commensurate prior models with random effects for interval-censored data to accommodate historical controls

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Historical controls can enhance statistical power and reduce sample sizes in rare disease research.
• Existing methods primarily address right-censored survival data, overlooking interval-censored events common in clinical practice.
• Differential assessment timing between historical and clinical data can introduce bias.

Purpose of the Study:

• To propose novel commensurate prior models with random effects for matched interval-censored survival data.
• To address bias arising from differential assessment timing in historical control data.
• To effectively borrow information based on data comparability.

Main Methods:

• Development of commensurate prior models with random effects for interval-censored survival data.
• Utilizing matched data structures to account for comparability.
• Simulation studies to evaluate Type I error control under varying exchangeability assumptions.

Main Results:

• The proposed models effectively control Type I error in simulations, regardless of data exchangeability.
• Demonstrated ability to borrow information based on the degree of comparability between datasets.
• Successful application to real-world clinical trial data (BMT CTN 1101 and 0901).

Conclusions:

• Commensurate prior models offer a robust approach for analyzing interval-censored survival data with historical controls.
• The methods mitigate bias caused by differential assessment timing.
• These models enhance the utility of historical data in rare disease clinical trials.