Two-Phase, Generalized Case-Control Designs for the Study of Quantitative Longitudinal Outcomes

Area of Science:

• Epidemiologic study design
• Biostatistics
• Longitudinal data analysis

Background:

• Longitudinal studies often face challenges with costly exposure data collection.
• Existing designs may not efficiently utilize available phase 1 data for phase 2 exposure assessment.

Purpose of the Study:

• To propose a general class of two-phase epidemiologic study designs for quantitative longitudinal data.
• To enable efficient collection of exposure data (e.g., biomarkers) on a subset of subjects in phase 2.
• To facilitate valid and efficient inferences for associations between exposures and outcomes.

Main Methods:

• Summarizing longitudinal outcomes using linear regression for each subject.
• Defining sampling strata based on estimated regression intercept or slope distributions.
• Employing stratified sampling from strata defined by intercepts, slopes, or a mixture.
• Utilizing ascertainment-corrected maximum likelihood and multiple-imputation for estimation.

Main Results:

• Samples with extreme intercept values reduce variance for time-fixed exposure associations.
• Samples enriched with extreme slope values reduce variance for time-varying exposure associations, including time interactions.
• The proposed methods allow for valid and efficient statistical inferences.

Conclusions:

• The developed two-phase design offers a flexible and efficient approach for longitudinal studies with subset exposure data.
• This methodology is applicable to various research areas, including genetic associations with health outcomes.
• The study provides a framework for optimizing resource allocation in complex epidemiologic research.