A Bayesian Joint Model Of Longitudinal Data And Survival Data And Its Application In Clinical Research Of Alzheimer’s Disease

In the clinical research of Alzheimer’s disease(AD),patients are usually repeatedly observed,and the corresponding longitudinal data and survival data are collected.There is often a correlation between longitudinal and survival data,which would be ignored in a separate analysis,resulting in large estimation bias.In exploring the relevant structures of longitudinal data and survival data,the researchers introduced a joint model of longitudinal data and survival data.The standard joint model generally adopted mixedeffect model to model the longitudinal process,while Cox proportional risk model used to model the survival process and adopted maximum likelihood method to estimate the parameters.In order to better fit the longitudinal trajectory and process the skewed data to obtain a more accurate estimate,a joint model of skew-normal longitudinal data and survival data is established in this paper,and the Bayesian method is used to estimate the parameters.This paper studies the joint model of skew-normal longitudinal data and survival data,and the model is applied to the clinical study of Alzheimer’s disease:(1)The joint model of longitudinal data and survival data is established,and the parameters of the model are estimated based on Bayesian method.(2)In order to better fit the relationship between response variables and time,the unknown smooth function is modeled by B spline method.(3)Considering the data characteristics,under the condition that the longitudinal response variables obey the skew-normal distribution,the parameters of longitudinal data and survival data are estimated based on Bayesian method.(4)The simulation study of the biased normal bayesian joint model shows that the deviation and root mean square error of the estimated value of the parameter of interest under the biased normal error assumption are small,and the proposed method in this paper is flexible and efficient.(5)The analysis of clinical research on alzheimer’s disease showed that for every 1-point increase in neuropsychological score ADAS-Cog13,the risk of conversion to AD increased by 10.19%,and for every 1-point increase in longitudinal RAVLT imme score,the corresponding risk decreased by 8.02%.Among the survival covariates,APOE4 was an important risk factor,and the joint model with ADAS-Cog13 and RAVLT imme as longitudinal response variables respectively indicated that the risk of AD transformation by APOE4 gene increased by 118.76% and 107.86%,respectively,for each additional APOE 4 gene.The joint model of skew-normal longitudinal data and survival data established in this paper can better fit the data characteristics and obtain reasonable and accurate parameter estimation.