Application of Causal Inference Methods to Pooled Longitudinal Non- Randomized Studies: A Methodological Systematic...

Area of Science:

• Medical research methodology
• Epidemiology
• Biostatistics

Background:

• Observational data are crucial for real-world medical insights but require careful methods for causal inference.
• Individual-level patient data meta-analyses (IPD-MAs) are valuable but their methodological rigor for causal estimation needs assessment.
• Previous evaluations highlighted potential limitations in the application of causal inference techniques in medical research.

Approach:

• A systematic review evaluated the methodology and reporting quality of IPD-MAs published in 2009, 2014, and 2019.
• Over 16,000 titles and abstracts were screened, with 29 full-text articles included in the final analysis.
• The review focused on the implementation of causal inference methods and the quality of reporting within these studies.

Key Points:

• Causal methodologies were infrequently employed in the analyzed IPD-MAs.
• Only three of 29 studies utilized quasi-experimental methods; none used G-methods for time-varying confounding.
• Reporting quality was generally poor, hindering the interpretation and trustworthiness of findings.

Conclusions:

• There is a need for enhanced collaboration between clinicians and methodologists to improve causal inference in IPD-MAs.
• A proposed checklist for reporting guidelines aims to enhance the quality and transparency of IPD-MAs.
• Strengthening causal methodology and reporting in IPD-MAs can improve the reliability of evidence for health policy decisions.