[Limitations of meta-analysis from published data in epidemiological research]

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Meta-analyses of epidemiological studies are increasingly common for evaluating risk factors with inconsistent findings across studies.
• These meta-analyses often rely on published data, raising questions about data reliability and analytical capabilities.

Purpose of the Study:

• To critically evaluate meta-analyses conducted using published epidemiological data.
• To determine the feasibility of performing qualitative and quantitative dose-response analyses.
• To assess the validity of arguments for meta-analysis in observational studies compared to clinical trials.

Main Methods:

• Discussion of meta-analysis methodologies for published versus experimental/clinical trial data.
• Comparative analysis of meta-analysis justifications in clinical trials and observational studies.
• Examination of challenges in estimating a single pooled risk estimate from published data.
• Illustrative examples from literature to highlight issues with published data meta-analyses.

Main Results:

• Significant challenges exist in obtaining reliable information from meta-analyses of published epidemiological data.
• Performing qualitative and quantitative dose-response analyses may not be consistently achievable.
• Arguments supporting meta-analysis in clinical trials may not directly translate to observational studies.
• Estimating a single, reliable pooled risk estimate presents considerable difficulties.

Conclusions:

• Caution is advised when interpreting meta-analyses based solely on published epidemiological data.
• The reliability of dose-response and pooled risk estimates requires careful scrutiny.
• Further methodological development is needed for robust meta-analyses of observational studies.