Comparison of pooled risk estimates for adverse effects from different observational study designs: methodological...

Area of Science:

• Epidemiology
• Biostatistics
• Pharmacovigilance

Background:

• Diverse study designs, including case-control and cohort studies, are employed to evaluate adverse effects.
• A need exists to compare risk estimates derived from meta-analyses of case-control studies against those from other observational study designs.

Purpose of the Study:

• To ascertain whether risk estimates from meta-analyses of case-control studies differ from those derived from other study designs.
• To investigate potential biases in the estimation of adverse effects based on study methodology.

Main Methods:

• Conducted comprehensive literature searches across 10 databases, supplemented by reference checking, expert consultation, and hand-searching of journals and conference proceedings.
• Included meta-analyses where pooled relative measures of adverse effects (odds ratios or risk ratios) from case-control studies could be directly compared with pooled estimates from other observational study types.

Main Results:

• Analyzed 82 meta-analyses, finding overlapping 95% confidence intervals for pooled harm estimates across study designs in 78% of cases.
• Identified 23 instances of discrepant findings; 70% originated from significantly elevated pooled estimates in case-control studies.
• Observed funnel plot asymmetry suggesting higher risk estimates from case-control studies, with cohort/cross-sectional studies yielding pooled odds ratios 0.94 times lower than case-control studies.

Conclusions:

• Meta-analyses of case-control studies tend to yield slightly higher estimates of harm compared to meta-analyses of other observational studies.
• Potential confounding factors, such as variations in drug dosage, duration, and study populations, cannot be excluded when comparing between different study designs.