Semiparametric regression analysis of interval-censored data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Interval-censored data presents unique challenges in survival analysis.
• Existing methods may lack reliability, especially with limited data.
• Accurate regression analysis is crucial for understanding time-to-event data.

Purpose of the Study:

• To develop a semiparametric proportional hazards regression approach for interval-censored data.
• To provide a robust method for estimating regression coefficients and baseline hazards.
• To introduce a reliable variance estimation technique for survival data.

Main Methods:

• A semiparametric approach utilizing an Expectation-Maximization (EM) algorithm.
• Maximizing Cox partial likelihood for regression coefficients and Breslow estimator for baseline hazards.
• Employing Turnbull's algorithm for hazard mass determination and multiple imputation for variance estimation.

Main Results:

• The EM algorithm simplifies due to linear terms for incomplete data.
• Multiple imputation provides reliable variance estimates, outperforming asymptotic methods in smaller datasets.
• The method aligns with standard Cox analysis for right-censored data.

Conclusions:

• The proposed semiparametric method effectively analyzes interval-censored data.
• It offers a robust and reliable alternative for survival data analysis, especially in resource-limited settings.
• The approach is validated through application to breast cancer trial data.