Detection of multiple change points in a Weibull accelerated failure time model using sequential testing

Area of Science:

• Biostatistics
• Survival Analysis
• Cancer Epidemiology

Background:

• Cancer diagnosis and treatment advancements alter disease incidence and mortality rates.
• Traditional survival analysis methods often fail to account for these evolving distributional changes.
• Change point problems in survival analysis address shifts in time-ordered data, especially with censoring or truncation.

Purpose of the Study:

• To propose a novel sequential testing approach for detecting multiple change points.
• To apply this method within the flexible Weibull accelerated failure time (AFT) model.
• To develop a procedure that infers the number of change points directly from data, rather than requiring it as a prior input.

Main Methods:

• Developed a sequential testing procedure for change point detection.
• Utilized the Weibull accelerated failure time (AFT) model, chosen for its flexibility with hazard rates and reparameterization capabilities.
• Employed a simulation study to validate the method's performance.
• Applied the method to real-world cancer data.

Main Results:

• The proposed sequential testing method accurately detects multiple change points in survival data.
• The method effectively estimates the parameters of the Weibull AFT model.
• Simulation studies confirmed the accuracy of change point detection and model estimation.
• Real data applications demonstrated the practical utility in identifying hazard rate shifts.

Conclusions:

• The developed sequential testing approach provides an accurate and data-driven method for identifying multiple change points in survival data.
• The Weibull AFT model is suitable for analyzing time-ordered cancer data with evolving hazard rates.
• This methodology enhances survival analysis by accounting for distributional shifts, offering improved insights into cancer progression and treatment efficacy.