Assessing etiological heterogeneity for multinomial outcome with two-phase outcome-dependent sampling design

Area of Science:

• Epidemiology
• Biostatistics
• Genetics

Background:

• Etiological heterogeneity describes diseases arising from distinct causes or exposures.
• Analyzing subtype-specific effects in outcome-dependent sampling requires adjusting for confounding and sampling bias.
• Existing methods may not adequately address these biases or allow subtype comparisons.

Purpose of the Study:

• To develop and validate a statistical method for valid inference on subtype-specific exposure effects.
• To enable formal comparisons of exposure effects across different disease subtypes.
• To address limitations in current approaches for analyzing etiological heterogeneity in complex diseases.

Main Methods:

• Utilized inverse probability weighting (IPW) to fit a multinomial model.
• Applied the IPW approach to two-phase outcome-dependent sampling data.
• Conducted simulations to compare IPW with common regression-based methods for heterogeneity assessment.

Main Results:

• The IPW method provides valid inference for subtype-specific exposure effects and their contrasts.
• IPW demonstrated superior performance compared to common regression methods in simulations.
• The study successfully estimated subtype-specific exposure effects on breast cancer risk.

Conclusions:

• Inverse probability weighting (IPW) offers a robust approach for analyzing etiological heterogeneity.
• This method enhances the accurate estimation of exposure effects across disease subtypes.
• The findings have implications for understanding disease etiology and informing public health strategies.