Marginal hazards model for case-cohort studies with multiple disease outcomes

Area of Science:

• Epidemiology
• Biostatistics
• Health Research Methodology

Background:

• Case-cohort study designs offer a cost-effective alternative to full cohort studies, particularly for rare diseases.
• A significant advantage is the ability to utilize a single subcohort for investigating multiple diseases or subtypes.
• Simultaneous modeling of event times is crucial for comparing risk factor effects across different disease outcomes.

Purpose of the Study:

• To develop and validate statistical methods for analyzing case-cohort studies with multiple disease outcomes.
• To address the challenge of correlated outcomes within the same subject.
• To extend methods to generalized case-cohort designs applicable to multiple diseases.

Main Methods:

• Application of marginal proportional hazards regression models tailored for case-cohort designs.
• Development of an estimating equation approach for parameter estimation.
• Incorporation of two distinct weighting strategies within the proposed methodology.

Main Results:

• The proposed estimators demonstrate consistency and asymptotic normality.
• Simulation studies confirm the effectiveness of large sample approximations, even in smaller sample sizes.
• The methods were successfully applied to real-world data from the Busselton Health Study.

Conclusions:

• The developed statistical framework effectively handles multiple disease outcomes in case-cohort studies.
• The proposed methods are robust and perform well in various scenarios, including generalized designs.
• This research provides valuable tools for epidemiological studies investigating multiple health outcomes.