Testing moderation in network meta-analysis with individual participant data.

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Meta-analytic methods are crucial for estimating intervention effectiveness and comparative effectiveness.
• Network meta-analysis (NMA) integrates direct and indirect treatment comparisons.
• Existing NMA methods primarily focus on main effects.

Purpose of the Study:

• To extend network meta-analysis (NMA) methods for examining moderator effects.
• To investigate the impact of individual participant data (IPD) versus aggregate data on detecting moderator effects.
• To develop a generalized multilevel model for NMA.

Main Methods:

• Developed a generalized multilevel model for NMA incorporating within-trial and between-trial heterogeneity.
• Included participant-level covariates in the NMA framework.
• Proposed a new NMA diagram and defined homogeneity and consistency across trials.
• Conducted a simulation study to assess power for detecting main and moderator effects.

Main Results:

• Power to detect moderator effects is substantially greater with individual participant data (IPD) compared to aggregate data NMA.
• The generalized multilevel model effectively accounts for heterogeneity and participant-level covariates.
• Simulation results demonstrated the model's utility in assessing power for main and moderator effects.

Conclusions:

• Extending NMA to investigate moderator effects is feasible and valuable.
• Individual participant data (IPD) offers superior sensitivity for detecting moderator effects in NMA.
• The proposed generalized multilevel model provides a robust framework for advanced NMA.