Confidence intervals for effect parameters common in cancer epidemiology

Area of Science:

• Epidemiology
• Biostatistics
• Cancer Research

Background:

• Cancer epidemiology relies on effect parameters like odds ratios and rate ratios.
• Accurate confidence intervals are crucial for interpreting these effect measures.

Purpose of the Study:

• To review approximate confidence intervals for key effect parameters in cancer epidemiology.
• To discuss methods suitable for both crude and adjusted analyses.

Main Methods:

• Review of approximate confidence intervals based on efficient scores.
• Description of Cornfield's method and Mantel-Haenszel estimators.
• Discussion of methods for stratified analysis with confounding factors.

Main Results:

• Approximate confidence intervals offer computational feasibility and near-nominal coverage rates.
• Methods are applicable to odds ratios (case-control) and rate ratios/differences (cohort studies).
• Mantel-Haenszel based intervals are suitable for summary measures in stratified analyses.

Conclusions:

• Approximate confidence intervals are practical and reliable tools in cancer epidemiology.
• The reviewed methods support accurate estimation of effect parameters, even with confounding.
• Recent developments enhance the utility of these statistical approaches.