Original Article
Cure models in analyzing long-term survivors
Abstract
Introduction: If in the process of surviving data analysis, we are confronted with a high percentage of censors, caused when the study comes to an end, and if the time of survey is long enough, some percentage of the population might have long-term survival, as a result of which we are to make careful use of cure models. These models are categorized based on mixture and non-mixture cure models. Following the publication of Chen [1999] article and the submission of a procedure based on latent variable distribution in recent years, non-mixture or promotion time cure model have come to attention.
Purpose: In this article, Poisson and compound Poisson models are considered for latent variable distribution based on which the cure rate is estimated.
Methods: Model parameters were estimated using Bayesian approach, and to compare the models fitness, Deviance Information Criteria (DIC) was used. The applicability of the model has been shown on some stomach cancer data.
Conclusions: According to DIC, Poisson and compound Poisson cure models had a better fitting in comparison with the typical Weibull survival model.
Purpose: In this article, Poisson and compound Poisson models are considered for latent variable distribution based on which the cure rate is estimated.
Methods: Model parameters were estimated using Bayesian approach, and to compare the models fitness, Deviance Information Criteria (DIC) was used. The applicability of the model has been shown on some stomach cancer data.
Conclusions: According to DIC, Poisson and compound Poisson cure models had a better fitting in comparison with the typical Weibull survival model.
