Estimating Transition Intensity Rate on Interval-censored Data Using Semi-parametric with EM Algorithm Approach.

Area of Science:

• Biostatistics
• Clinical Trials
• Cancer Survivorship

Background:

• Phase IV clinical trials monitor long-term treatment side effects, such as cardiotoxicity in childhood cancer survivors.
• Estimating the incidence of outcomes like cardiotoxicity is crucial but challenging due to interval-censored data from longitudinal patient follow-up.
• Existing parametric models may fail if their underlying assumptions are not met.

Purpose of the Study:

• To propose a novel semi-parametric model for estimating transition intensity rates in an illness-death framework.
• To address the limitations of parametric models in analyzing interval-censored data for long-term adverse event monitoring.

Main Methods:

• Developed a semi-parametric model incorporating a logit relationship for treatment intensities across two groups.
• Employed an Expectation-Maximization (EM) algorithm with profile likelihood for parameter estimation.
• Utilized simulation studies to evaluate the model's performance.

Main Results:

• The proposed semi-parametric model is easy to implement.
• Simulation results indicate the model yields comparable outcomes to traditional parametric models.
• The approach effectively handles interval-censored data in the context of illness-death models.

Conclusions:

• The novel semi-parametric model provides a robust and flexible method for analyzing cardiotoxicity incidence in cancer survivors.
• This approach offers a viable alternative when parametric assumptions are questionable.
• The method facilitates more accurate long-term risk assessment in clinical trial settings.