Comparative And Analysis Of Mixture Cure Model In R

Cure model is one of the most commonly used models in clinical research.Mixture cure model is often used to explore the effect of a treatment method on the survival time and cure rate of patients in clinical trials.Researchers usually use R to write the corresponding calculation process in analyzing survival data.In the last 20 years,many researchers have created R packages for mixture cure models with different data scenarios,including parametric,semiparametric,or non-parametric methods,clustering or non-clustering data,right-censored or interval censored data,and so on.Although there are many R packages for people to use for the statistical analysis of the mixture cure models,no one has systematically compared and analyzed the advantages and disadvantages of different programs in terms of accuracy and speed.In theory,though,different package approaches correspond to different data and model.However,there are still packages that can work with non-clustered survival data at the same time.This article will focus on four R packages,smcure,GORCure,npcure,and geecure,because they meet all of the conditions above.Although the methods of these three packages correspond to different data censoring conditions,this article will use a simple method to reduce the impact of the different censoring conditions.At the same time,this article will describe a problem discovered in the R package rcure and two functions developed of the R package geecure.In this article,firstly,we introduce the knowledge of mixture cure model.Secondly,we describe in detail the methods of the four R packages we concerned about: smcure,GORCure,npcure,and geecure.Then,we analyze the simulation data of the four methods under different sample sizes,different distribution of covariables,different cure rates and censoring rates,and give the analysis results and compare them.After that,we analyze and compare the four methods again under two sets of real data.Finally,we describe a problem we found in the R package rcure,and develope two functions-predict geecure and plot predict geecure,to extend the drawing capabilities of the R package geecure.