A weighting approach to causal effects and additive interaction in case-control studies: marginal structural linear...

Area of Science:

• Epidemiology
• Biostatistics
• Genetic Epidemiology

Background:

• Estimating additive interaction in case-control studies traditionally relies on logistic regression.
• Logistic regression models may be misspecified if the underlying data-generating model is linear.
• This misspecification can lead to inaccurate estimates of interaction effects.

Purpose of the Study:

• To propose an alternative method for estimating additive interaction in case-control studies.
• To address the limitations of logistic regression in capturing additive interaction.
• To enable robust estimation of interaction measures like relative excess risk due to interaction.

Main Methods:

• Utilized an inverse probability of treatment weighting (IPTW) approach for causal effects and additive interaction.
• Assumed no unmeasured confounding for the validity of the causal inference.
• Fitted a marginal structural linear odds model, specifying exposure models conditional on covariates rather than outcome models.

Main Results:

• The proposed IPTW approach allows estimation of additive interaction without relying on outcome conditional modeling assumptions.
• This method is applicable to dichotomous exposures and case-control data.
• Demonstrated the assessment of additive interaction between genetic and environmental factors in a case-control study.

Conclusions:

• The inverse probability of treatment weighting approach provides a more robust method for estimating additive interaction in case-control studies.
• This method overcomes the potential misspecification issues associated with traditional logistic regression models.
• The approach facilitates reliable estimation of interaction effects, crucial for understanding complex etiological relationships.