Extending the Case-Control Design to Longitudinal Data: Stratified Sampling Based on Repeated Binary Outcomes

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Data Analysis

Background:

• Existing case-control sampling methods may not be optimal for longitudinal studies with retrospective exposure data collection.
• Cost constraints often limit the size of subsamples that can be analyzed in secondary analyses of cohort studies.
• Longitudinal binary response data requires specialized sampling strategies for efficient analysis.

Purpose of the Study:

• To develop generalized case-control sampling designs for longitudinal studies with retrospective exposure data.
• To introduce novel stratified outcome-dependent sampling strategies to address cost limitations.
• To enhance the precision of parameter estimates for time-varying and time-fixed covariates in longitudinal analyses.

Main Methods:

• Proposed a novel class of stratified outcome-dependent sampling designs for longitudinal binary response data.
• Created distinct strata based on the occurrence of the event of interest during follow-up (never, sometimes, always).
• Utilized an imputation-based estimation procedure for estimating baseline covariate associations.

Main Results:

• Subjects exhibiting response variation (event occurring at some but not all follow-up times) are highly informative for time-varying exposures.
• Sampling all subjects with response variability can yield highly precise parameter estimates, comparable to full cohort analysis when exposure varies within individuals.
• Baseline covariate associations can be estimated with very high precision irrespective of the sampling design using the proposed imputation method.

Conclusions:

• Stratified outcome-dependent sampling designs offer flexible and efficient methods for analyzing longitudinal data with retrospective exposure assessment.
• Focusing sampling on individuals with response variability significantly enhances the precision of parameter estimates for time-varying exposures.
• The proposed methods provide a cost-effective approach to maximizing information extraction from longitudinal cohort studies.