A hands-on approach for fitting long-term survival models under the GAMLSS framework

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Traditional survival models often fail with clinical data containing patients insusceptible to events.
• Ignoring non-susceptible patients leads to inadequate and potentially biased survival estimations.
• The presence of a 'cured' fraction necessitates specialized modeling techniques.

Purpose of the Study:

• To apply the generalized additive models for location, scale, and shape (GAMLSS) framework to long-term survival analysis.
• To develop a flexible statistical model that accommodates competing risks and a cured fraction.
• To demonstrate the utility of the `gamlss` package in R for fitting these complex survival models.

Main Methods:

• Utilized the GAMLSS framework for flexible statistical modeling.
• Modeled the number of competing causes of the event of interest using a negative binomial distribution.
• Parameterized the model based on the cured fraction, linking it to relevant covariates.

Main Results:

• The proposed GAMLSS approach effectively handles long-term survival data with a cured fraction.
• Several existing survival models are shown to be special cases of this generalized framework.
• The `gamlss` R package provides a robust platform for implementing and analyzing these models.

Conclusions:

• GAMLSS offers a powerful and flexible methodology for fitting long-term survival models, particularly when a cured fraction is present.
• The negative binomial distribution for competing risks enhances model adaptability.
• This framework provides a unified approach to modeling survival data with non-susceptible individuals.