Estimating time-varying exposure-outcome associations using case-control data: logistic and case-cohort analyses.

Area of Science:

• Epidemiology
• Biostatistics
• Public Health Research

Background:

• Traditional logistic regression in case-control studies limits analysis of time-varying exposure-outcome relationships.
• Estimating time-varying associations is essential for understanding dynamic effects, such as vaccine efficacy over time.

Purpose of the Study:

• To introduce and evaluate two novel approaches for estimating time-varying exposure-outcome associations in case-control studies.
• To enable the analysis of dynamic relationships previously not feasible with standard methods.

Main Methods:

• Method 1: Series of logistic regressions fitted within successive time periods, reusing controls.
• Method 2: Case-control sample treated as a case-cohort study, with controls forming the subcohort.
• Simulations used to validate methods, with extensions for left truncation, frequency matching, and time-varying exposures.

Main Results:

• Both proposed methods accurately estimate time-varying exposure effects using standard case-control data.
• The case-cohort approach avoids ambiguity in control definition, unlike the logistic approach which requires careful control definition within time periods.
• Performance is comparable when controls are optimally utilized in the logistic approach; case-cohort analysis allows continuous time-varying effects.

Conclusions:

• Case-control studies can now be utilized to estimate time-varying exposure-outcome associations, expanding their analytical capabilities.
• The case-cohort analysis provides a more flexible approach, enabling continuous modeling of time-varying associations, unlike the step-function limitation of logistic regression.