Simultaneous confidence bands for Cox regression from semiparametric random censorship

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Standard Cox regression is widely used for survival analysis.
• Existing methods for confidence bands in survival analysis have limitations.
• Accurate estimation of subject-specific survival curves is crucial in many fields.

Purpose of the Study:

• To develop novel simultaneous confidence bands (SCBs) for subject-specific survival curves.
• To integrate Cox regression with semiparametric random censorship models for improved accuracy.
• To evaluate the performance of the proposed SCBs against existing methods.

Main Methods:

• Utilized Cox regression in conjunction with semiparametric random censorship models.
• Constructed simultaneous confidence bands (SCBs) for individual survival predictions.
• Employed simulation studies to assess empirical coverage and informativeness.

Main Results:

• The proposed SCBs demonstrated correct empirical coverage in simulations.
• The new SCBs were found to be more informative than those based solely on standard Cox regression.
• The methodology was successfully illustrated using two real-world datasets.

Conclusions:

• The combined approach offers a statistically sound and more informative method for subject-specific survival curve estimation.
• The developed SCBs provide reliable confidence intervals for individual survival predictions.
• An extension for handling missing censoring indicators was also proposed, enhancing the method's applicability.