Potential misinterpretation of the case-only study to assess gene-environment interaction

Area of Science:

• Epidemiology
• Genetic Epidemiology
• Biostatistics

Background:

• Assessing gene-environment interactions often necessitates novel epidemiologic study designs.
• Case-only designs are proposed as efficient alternatives to traditional case-control studies.

Purpose of the Study:

• To highlight that case-only analyses estimate a different interaction parameter than case-control analyses.
• To quantify the impact of higher disease risk on the interaction parameter estimated by case-only designs.

Main Methods:

• Comparison of interaction parameters estimated by case-only versus case-control analyses.
• Quantification of the impact of high disease penetrance on case-only interaction estimates.

Main Results:

• Case-only analysis estimates departure from multiplicative risk ratios, while case-control estimates departure from multiplicative odds or rate ratios.
• These parameters are approximately equal only when disease risk is low.
• The cross-product ratio from case-only data can be substantially smaller than the odds ratio from case-control data, potentially underestimating interaction.

Conclusions:

• Case-only analyses may underestimate gene-environment interaction effects, particularly in scenarios with high disease risk or highly penetrant genes.
• Explicit definition of multiplicative interaction is crucial to prevent misinterpretation of findings from case-only studies.