Meta-Analyses of Observational Studies: Principles, Challenges, and Recommendations
Observational studies play a crucial role in pediatric research by assessing disease burden, real-world effectiveness and safety, risk factors, and rare outcomes. They are particularly important for questions for which randomized controlled trials would be unethical, impractical, or not well aligned with the clinical question. The incorporation of observational studies into meta-analyses allows for greater synthesis of observational data; however, conducting these evidence syntheses introduces unique methodological challenges. These include increased heterogeneity, residual confounding, and a higher risk of within- and between-study biases, which can compromise the validity of pooled estimates. This article outlines key considerations for conducting meta-analyses of observational studies. Guidance is provided on protocol development, defining inclusion criteria, selecting appropriate study designs, extracting effect estimates, and managing confounding and heterogeneity. Using a published meta-analysis on the association between particulate matter exposure and childhood asthma, we illustrate practical challenges and strategies for interpretation.
Authors
- Alastair Fung (ORCID: https://orcid.org/0000-0001-5615-4947)
- Elise Peterson Lu (ORCID: https://orcid.org/0009-0000-8998-9895)
- David D’Arienzo (ORCID: https://orcid.org/0000-0002-9601-8533)
- Nassr Nama (ORCID: https://orcid.org/0000-0002-0199-3599)
Institutions
- Western University (CA)
- Seattle Children's Hospital (US)
- University of Toronto (CA)
- University of Washington (US)
- Montreal Children's Hospital (CA)
- London Health Sciences Centre (CA)
- Hospital for Sick Children (CA)
- Seattle Children's Hospital (US)
- McGill University (CA)
Publication Details
- Journal
- Hospital Pediatrics
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1542/hpeds.2026-009342
- Primary Topic
- Meta-analysis and systematic reviews
- Type
- article
- Field-Weighted Citation Impact
- 0.00