Integration of aggregate data in causally interpretable meta-analysis by inverse weighting.

Area of Science:

• Biostatistics
• Epidemiology
• Health Technology Assessment

Background:

• Causal meta-analysis is challenged by differing effect modifier distributions across trials.
• Transportability methods standardize study results to a target population's case-mix.
• Individual participant data (IPD) is often required for case-mix standardization, posing a practical barrier.

Purpose of the Study:

• To develop novel strategies for causal meta-analysis incorporating aggregate data when IPD is not fully available.
• To extend existing moment-based methods for population-adjusted indirect comparisons.
• To address computational challenges in causal meta-analysis with limited IPD.

Main Methods:

• Developed novel strategies to incorporate aggregate-level data into causal meta-analysis.
• Extended moment-based methods for population-adjusted indirect comparisons.
• Developed computational methods to address M-estimation theory requirements for valid inference.

Main Results:

• Proposed approaches were assessed using simulated data for finite-sample performance.
• The methods were applied to real-world clinical data on psoriasis treatments.
• Investigated the comparative effectiveness of risankizumab versus ustekinumab.

Conclusions:

• The novel strategies enable causal meta-analysis with partially available IPD and aggregate data.
• The developed methods offer a practical solution to data limitations in transportability.
• The study provides insights into the comparative effectiveness of specific psoriasis treatments.