Application Of Time-Varying Coefficient Cox Model In Exploring Combination Treatment Of Breast Cancer

There is time independent assumption when Cox model is proposed initially. With further research, time effect is introduced. Thus time-dependent Cox model and time-varying coefficient Cox model is proposed. Also, with the advent of new biomedical technology and the big data being developing in medical industry, the dimension of survival data is growing, the statistical methods to analyze the high-dimensional survival data is becoming more and more important. In this article, we explain the process of coefficient estimation and variable selection of the high dimensional survival data on time-varying Cox model and apply it to real case analysis.In the first chapter, we proposed the existing problem in application of Cox model to high-dimensional survival data. Consider the research status of time-varying coefficient Cox model, we apply adaptive group LASSO with B-splines to establish time-varying coefficient Cox model.In the second chapter, we introduce the method of time-varying coefficient model, B-splines and the penalty function. In third chapter, we demonstrate the algorithm and calculation process of adaptive group LASSO with B-splines.In forth chapter, empirical analysis is carried out on the data of 684 cases of advanced metastatic breast cancer patients. Firstly, the results are obtained by time-varying coefficient model using adaptive group LASSO with B-splines. Then, establish the Cox model with adaptive group LASSO and time-dependent Cox model. Compare and analyse the results of the three models. In the fifth chapter, we explain the results of empirical analysis and give out the problem to be solved in the model and the prospect.In Empirical analysis, these covariates is in final time-varying coefficient Cox model:Dose, ECOG score, pathological stage, changed diameter of target lesion, baseline number of metastatic foci, changed number of metastatic foci, disease progression, history of advance systemic therapy>=5 line, whether on Capecitabine therapy, whether on Taxol therapy, whether on The path for therapy, whether on Herceptin therapy, transfer site, whether distant transfer, the new site of the transfer, changed amount of estradiol. In the time-independent part, these coefficients of covariates is larger:Dose, ECOG score, changed number of metastatic foci, changed diameter of target lesion, history of advance systemic therapy>=5 line, whether on Capecitabine therapy, whether on Taxol therapy, whether on The path for therapy transfer site, whether distant transfer, the new site of the transfer, changed amount of estradiol. The time-varying coefficient of Herceptin therapy was larger when survival time is less than 200 days. With the increase of the survival time, the coefficient gradually decreased to 0. The coefficient of covariates: pathological staging, changed diameter of target lesion, baseline number of metastasis foci, changed number of metastasis lesion, disease progression, history of advance systemic therapy>=5 line, transfer site, whether the distance transfer, the new site of the transfer is negative. The others are positive.The model results, in combination with research of breast cancer medicine, the following conclusions are obtained:1. In combination therapy with drugs, Capecitabine, Taxol and The path for, with the increase of the new drug dose, survival time of advanced metastatic breast cancer patients is siginificantly prolonged.2. Whether Herceptin and new drug used together in the combination therapy is appropriate, and the appropriate new drug dose are both still needs further exploration and research.The model comparison results show that the time-varying Cox model based on adaptive group LASSO with B-splines is more easily interpreted and stable, compared with Cox model with adaptive group LASSO and time-dependent Cox model.