Combining Aggregate Data and Individual Patient Data in Model-Based Meta-Analysis: An Illustrative Case Study of...

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Pharmacometrics

Background:

• Model-based meta-analysis (MBMA) integrates aggregate data (AD) for increased statistical power.
• Access to individual patient data (IPD) is often limited, posing challenges for detailed covariate analysis.
• Aggregate covariate data or published stratified results are typically used to probe predictive covariates.

Purpose of the Study:

• To quantify the benefits of accessing individual patient data (IPD) in model-based meta-analysis (MBMA).
• To compare the performance of MBMAs with and without IPD under different covariate scenarios.
• To evaluate the impact of varying ratios of IPD versus AD studies and stratified AD studies.

Main Methods:

• A three-step approach was used to assess the benefits of IPD in MBMA.
• Two scenarios were explored: MBMA with and without IPD (no covariates), and MBMA with and without IPD (with a predictive covariate).
• Performance was evaluated based on different ratios of IPD studies to AD studies and covariate-stratified AD studies.

Main Results:

• The benefit of IPD over AD was not evident in models without covariates.
• Including stratified IPD led to improved performance in covariate models.
• The performance evaluation considered various ratios of IPD/AD studies and stratified AD studies.

Conclusions:

• Individual patient data (IPD) provides limited advantages in model-based meta-analysis (MBMA) when covariates are not considered.
• Stratified individual patient data (IPD) significantly enhances the performance of covariate models in MBMA.
• The study highlights the importance of data stratification for improving covariate analysis in meta-analytic models.