Interpreting epidemiological evidence: how meta-analysis and causal inference methods are related

Area of Science:

• Epidemiological research methods
• Biostatistics
• Scientific inference

Background:

• Observational epidemiological evidence interpretation uses quantitative meta-analysis and qualitative causal inference.
• Causal inference relies on criteria such as consistency, strength of association, dose-response, and plausibility.

Purpose of the Study:

• To examine the relationship between meta-analysis and causal inference methods.
• To assess meta-analysis's capacity to contribute to causal claims in epidemiology.

Main Methods:

• Qualitative examination of causal inference criteria.
• Quantitative meta-analysis for summarizing effect estimates.
• Analysis of statistical heterogeneity in meta-analysis.

Main Results:

• Meta-analysis provides reproducible weighted averages of effect estimates, improving consistency assessment over rules-of-thumb.
• Statistical heterogeneity does not preclude consistent findings.
• Meta-analysis offers precise estimates for strength of association and dose-response but requires judgment for causal relevance.
• Meta-analysis can identify sources of confounding and bias through heterogeneity assessment.

Conclusions:

• Meta-analysis plays a valuable but limited role in causal inference, enhancing understanding of specific causal criteria.
• Meta-analysis can summarize evidence across different knowledge levels but cannot combine evidence across levels.
• Further research may explore combining evidence across multiple levels of knowledge.