Semiparametric inference for a two-stage outcome-dependent sampling design with interval-censored failure time data

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Interval-censored failure time data presents unique analytical challenges.
• Existing sampling designs may not fully optimize efficiency for such data.
• Accurate inference is crucial for understanding disease progression and treatment efficacy.

Purpose of the Study:

• To propose a novel two-stage outcome-dependent sampling design for interval-censored failure time outcomes.
• To enhance study efficiency by utilizing observed failure times to inform second-stage sampling.
• To develop a robust inference procedure for the proposed design.

Main Methods:

• A two-stage outcome-dependent sampling strategy was developed.
• A sieve semiparametric maximum pseudo likelihood procedure was employed for inference.
• The method utilizes all available data from the two-stage design.
• Asymptotic properties of the regression parameter estimator were established.

Main Results:

• The proposed design and inference procedure are statistically consistent and asymptotically normal.
• A consistent estimator for the asymptotic variance was derived.
• Simulation studies confirmed the practical utility and superior efficiency compared to existing methods.
• The approach was successfully applied to a phase 3 HIV vaccine trial.

Conclusions:

• The proposed two-stage outcome-dependent sampling design offers a more efficient approach for studies with interval-censored failure time data.
• The developed sieve semiparametric maximum pseudo likelihood method provides reliable inference.
• This methodology has significant implications for optimizing resource allocation and enhancing statistical power in clinical research.