Using case description information to reduce sensitivity to bias for the attributable fraction among the exposed

Area of Science:

• Epidemiology
• Biostatistics
• Cancer Research

Background:

• The attributable fraction among the exposed (AFe) quantifies preventable disease cases.
• Estimating AFe is challenging due to potential hidden biases in statistical inference.

Purpose of the Study:

• To propose a novel statistical approach for estimating AFe that minimizes sensitivity to hidden bias.
• To leverage case-specific information, such as cancer subtypes, to enhance AFe estimation.

Main Methods:

• Development of a new statistical inference method for AFe.
• Utilizing case description information (e.g., cancer subtype) to reduce bias.
• Evaluation through asymptotic tools, design sensitivity, simulation studies, and a case study.

Main Results:

• The proposed method demonstrates reduced sensitivity to hidden bias in AFe estimation.
• Case studies, including alcohol and breast cancer risk, show the utility of incorporating cancer subtype data.
• A sensitivity parameter addresses potential selection bias introduced by using case definition information.

Conclusions:

• Leveraging cancer subtype information offers a promising strategy to improve the accuracy of attributable fraction estimation.
• The new method provides a more robust approach to statistical inference for AFe, accounting for potential biases.
• This research contributes to better understanding exposure-specific disease risks and prevention strategies.