Collaborative inference for treatment effect with distributed data-sharing management in multicenter studies.

Area of Science:

• Biostatistics
• Clinical Trials
• Data Science

Background:

• Multicenter clinical studies face data sharing barriers due to distributed data sources.
• Merging data for centralized analysis is time-consuming and complex, especially with propensity score modeling.
• Existing methods lack thorough investigation for incorporating complex modeling in meta-analyses.

Purpose of the Study:

• To propose a new collaborative inference framework that avoids merging subject-level raw data from multiple sites.
• To enhance data privacy and reduce sensitivity to data distribution imbalances in multicenter studies.
• To enable efficient statistical analysis using only shared summary statistics.

Main Methods:

• Developed a distributed inference framework for collaborative analysis.
• The method requires sharing only summary statistics, not raw subject-level data.
• Theoretical analysis and numerical simulations were used to validate the approach.

Main Results:

• The proposed distributed inference approach shows minimal loss of statistical power compared to centralized methods.
• The framework offers maximal data privacy protection and is robust to unbalanced data distributions.
• Algorithms and large-sample properties for the distributed method were established.

Conclusions:

• The novel distributed inference framework provides an efficient and privacy-preserving alternative for multicenter clinical trial data analysis.
• This method is particularly beneficial when propensity score modeling is involved.
• The approach was validated through simulations and applied to a study on basal insulin's effect on post-transplantation diabetes mellitus.