Bayesian analysis on meta-analysis of case-control studies accounting for within-study correlation.

Area of Science:

• Biostatistics
• Epidemiology
• Genetics

Background:

• Odds ratios are commonly used in retrospective studies, but existing methods often assume independence within studies, which may be unrealistic.
• Cases and controls within the same study can share common factors, leading to correlation that needs to be addressed for accurate analysis.
• Meta-analyses of case-control studies involve multiple 2x2 tables, requiring methods that can handle correlated data across studies.

Purpose of the Study:

• To extend existing methods for calculating odds ratios to incorporate within-study correlations.
• To develop a Bayesian framework for analyzing multiple 2x2 tables in a meta-analysis setting, considering correlated risks.
• To provide accurate posterior distributions for study-specific and overall odds ratios.

Main Methods:

• Developed exact posterior distribution formulas for odds ratios with within-study prior correlation for single and multiple 2x2 tables.
• Utilized an empirical Bayes approach to estimate hyperparameters, including within-study correlation.
• Extended the methodology to a regression setting for comprehensive analysis.

Main Results:

• Validated the exact posterior distribution formulas through simulation studies.
• Demonstrated the finite sample properties of the inference for the overall odds ratio.
• Applied the method to a twin study on genetic heritability and a meta-analysis on NAT2 acetylation status and colorectal cancer.

Conclusions:

• The proposed Bayesian method accurately accounts for within-study correlations in odds ratio estimation.
• This approach offers improved statistical inference for meta-analyses of case-control studies.
• The findings have implications for understanding genetic heritability and disease risk factors.