Information-Based Composite Likelihood Method for Hybrid Meta-Analysis Integrating Individual Participant Data and...

Area of Science:

• Biostatistics
• Biomedical Research
• Statistical Modeling

Background:

• Meta-analysis is crucial in biomedical research for evaluating treatment effectiveness.
• Conventional meta-analysis uses aggregated data, but combining individual participant data (IPD) and aggregated data (AD) offers potential efficiency gains.
• Integrating IPD and AD in meta-analysis presents statistical challenges.

Purpose of the Study:

• To develop a novel information-based method for hybrid meta-analysis.
• To improve the statistical efficiency of treatment effect estimators by integrating IPD and AD.
• To provide a robust framework for combining diverse data sources in meta-analysis.

Main Methods:

• A novel composite likelihood approach is developed for hybrid meta-analysis.
• The method utilizes all available information from aggregated data, including descriptive statistics.
• It accounts for between-study variability and estimates unknown parameters by maximizing the composite likelihood function.

Main Results:

• The proposed estimators are demonstrated to be consistent and asymptotically normal.
• Simulation studies indicate the method is more efficient than existing meta-analysis techniques.
• The method was successfully applied to clinical trials comparing LDL-C-lowering treatments.

Conclusions:

• The novel composite likelihood method offers an efficient approach for hybrid meta-analysis.
• This technique effectively integrates individual participant data and aggregated data.
• The findings have implications for improving the precision of treatment effect estimates in biomedical research.