Interval estimation of the attributable risk in case-control studies with matched pairs

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Attributable risk (AR) quantifies preventable cases by eliminating a risk factor.
• Assessing AR impact is crucial for community health and intervention strategies.
• Accurate interval estimators are vital for reliable AR estimation in case-control studies.

Purpose of the Study:

• Develop and evaluate novel interval estimators for AR in matched case-control studies.
• Compare the performance of five different interval estimators.
• Identify the most reliable and efficient estimator for various epidemiological scenarios.

Main Methods:

• Considered five asymptotic interval estimators for AR.
• Included Wald's statistic, logarithmic transformations (log(x), log(1-x)), logit transformation, and a novel quadratic equation estimator.
• Evaluated estimator performance using coverage probability and average length in simulations.

Main Results:

• The quadratic equation estimator demonstrated consistent coverage probability and superior efficiency.
• Logarithmic transformation log(1-x) showed good coverage but lower efficiency than Wald's.
• Logarithmic transformations log(x) and log(x/(1-x)) can be misleading for low OR/PE; logit is efficient for high OR/PE.

Conclusions:

• The quadratic equation estimator is recommended for general use when OR and PE ranges are unknown.
• The logit transformation estimator is recommended for situations with high OR and PE (>=4 and >=0.50, respectively).