Regression methods for estimating attributable risk in population-based case-control studies: a comparison of...

Area of Science:

• Epidemiology
• Biostatistics
• Public Health Research

Background:

• Estimating attributable risk in case-control studies is crucial for understanding disease causes.
• Existing multivariate methods often use logistic regression, which assumes multiplicative relationships between risk factors.
• Confounding factors require careful adjustment in risk estimation.

Purpose of the Study:

• To propose and evaluate an additive regression model for estimating attributable risk in population-based case-control studies.
• To compare the performance of the additive model against traditional logistic regression models.
• To assess the model's utility in handling multiple exposures and confounding factors.

Main Methods:

• Development of a regression method based on an additive model for risk estimation.
• Application of both additive and logistic regression models to matched case-control data.
• Utilized population-based data from a childhood astrocytoma brain tumor study for empirical validation.

Main Results:

• Both additive and logistic models demonstrated good data fit.
• The additive model provided a more satisfactory estimate of risk attributable to multiple exposures, especially without significant additive interaction.
• Adjusted risk estimates from the additive model summed coherently to the overall joint estimate, unlike the logistic model.

Conclusions:

• The proposed additive regression model offers a valuable alternative for multivariate attributable risk estimation in case-control studies.
• This approach is particularly useful when an additive relationship between covariates is appropriate.
• The additive model facilitates a clearer understanding of the combined impact of multiple risk factors.