Weighted least-squares regression with competing risks data.

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• The semiparametric accelerated failure time (AFT) model is widely used in survival analysis for its interpretability and connection to linear models.
• Analyzing competing risks data, where multiple distinct causes of failure are possible, presents unique challenges not fully addressed by standard AFT models.
• Existing methods for competing risks often lack the simplicity and direct interpretation of linear regression approaches.

Purpose of the Study:

• To develop a novel least-squares (LS) linear regression approach for analyzing cause-specific subdistribution functions in the presence of competing risks.
• To adapt conventional LS equations to handle data incompleteness inherent in competing risks scenarios.
• To extend the methodology for risk prediction and analysis in clustered competing risks settings.

Main Methods:

• Proposed a modified least-squares (LS) linear regression model tailored for cause-specific subdistribution functions.
• The LS equation was adapted to account for data incompleteness under competing risks.
• The methodology was extended to address clustered competing risks and risk prediction.

Main Results:

• The proposed LS estimators were demonstrated to be consistent and asymptotically normal.
• Consistent estimation of the variance-covariance matrix was achieved.
• Simulation studies confirmed the method's ability to provide rapid and valid statistical inferences and predictions.

Conclusions:

• The developed LS linear regression method provides a simple and effective tool for analyzing competing risks data.
• The approach is suitable for risk prediction and analysis in both standard and clustered competing risks scenarios.
• Application to oncology datasets highlights its practical utility in routine clinical data analysis.