Outcome-dependent sampling with interval-censored failure time data.

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Epidemiologic studies frequently analyze exposure-time relationships with interval-censored failure time data.
• Low failure rates and wide intervals necessitate large cohorts for precise estimation, posing budget challenges.
• Cost-effective sampling designs are crucial for analyzing expensive exposure variables in large cohort studies.

Purpose of the Study:

• To propose an outcome-dependent sampling (ODS) design tailored for interval-censored failure time data.
• To develop an efficient statistical inference procedure for the proposed ODS design.
• To address the challenges of precision and cost in large cohort studies.

Main Methods:

• Developed a novel sieve semiparametric maximum empirical likelihood approach.
• Utilized empirical likelihood and sieve methods to handle infinite-dimensional nuisance parameters.
• Proposed an outcome-dependent sampling (ODS) design for interval-censored data, enriching the sample with informative subjects.

Main Results:

• The proposed ODS design and sieve semiparametric maximum empirical likelihood method demonstrated consistency and asymptotic normality.
• Simulation studies confirmed the method's efficiency and effectiveness in practical scenarios.
• The approach proved more efficient than alternative designs and competing methods.

Conclusions:

• The novel outcome-dependent sampling design and statistical method offer a cost-effective and efficient solution for analyzing interval-censored failure time data in epidemiology.
• This approach enhances precision without requiring excessively large cohorts, making studies more feasible.
• The methodology is applicable to real-world studies, as illustrated by an example from the ARIC study.