Analysis of case-cohort designs.

Area of Science:

• Epidemiology
• Biostatistics
• Occupational Health

Background:

• The case-cohort design is optimal for analyzing time-to-failure data in large cohorts with rare events.
• It involves collecting covariate data from all failures and a sample of censored observations.
• This design offers greater flexibility than nested case-control designs but is underutilized due to perceived analytical complexity.

Purpose of the Study:

• To illustrate the computation of a simple variance estimator for case-cohort studies.
• To discuss model-fitting techniques using SAS for case-cohort data.
• To compare the efficiency and analysis of case-cohort designs versus nested case-control designs.

Main Methods:

• Demonstration of variance estimation and model fitting techniques in SAS.
• Consideration of three distinct weighting methods for analysis.
• Comparison via simulation with a nested case-control design using an occupational exposure study of nickel refinery workers.

Main Results:

• The study illustrates practical computation of variance estimators and model fitting in SAS for case-cohort designs.
• Case-cohort sampling proved more efficient than a comparable nested case-control design in the simulated occupational exposure study.

Conclusions:

• The case-cohort design, despite perceived complexity, is a powerful and efficient tool for epidemiological research, particularly for rare outcomes.
• The methods presented simplify the analysis, encouraging wider adoption of this valuable study design.
• The findings support the superior efficiency of case-cohort sampling in specific scenarios compared to nested case-control designs.