Testing for measured gene-environment interaction: problems with the use of cross-product terms and a regression...

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Gene-environment interaction (G × E) is crucial in understanding complex diseases.
• Regression models with product terms are commonly used to assess G × E effects.
• Existing methods may lead to erroneous conclusions, especially with polymorphic genotypes.

Purpose of the Study:

• To evaluate the appropriateness of product terms for modeling G × E interactions.
• To identify limitations of current regression approaches for specific genotype models.
• To propose an accurate method for assessing G × E effects.

Main Methods:

• Analysis of regression models with product interaction terms.
• Comparison of interaction assessment for binary versus three-category genotypes.
• Development and presentation of a reparameterized regression equation.

Main Results:

• Product terms accurately characterize G × E interaction for binary genotypes.
• Product terms can yield false positive and false negative results for three-category genotypes.
• The proposed reparameterized equation accurately captures G × E interaction effects.

Conclusions:

• Standard product term modeling is inappropriate for three-category genotypes.
• A reparameterized regression equation provides a more accurate assessment of G × E interactions.
• Recommendations are provided for interpreting G × E interactions based on genotype models.