多状态竞争风险模型在2型糖尿病与痛风发病关系研究中的应用

发布时间：2018-03-01 10:42

本文关键词： 多状态 Markov 模型竞争风险 2 型糖尿病痛风发病风险　出处：《暨南大学》2015年硕士论文　论文类型：学位论文

【摘要】：目的应用多状态竞争风险模型深入探讨2型糖尿病与痛风在独立发病与合并发病的状态下发病风险随时间变化规律,阐明两病种随时间的交互作用关系;通过比较多状态竞争风险模型、logistic回归模型、Kaplan-Meier法生存分析,探讨何种模型更适宜处理多结局的纵向研究资料,为其他类似问题的解决提供方法学借鉴。方法1.采用分层整群抽样获得原始队列样本人群,结合广州市社区户籍居民的回顾性调查,利用健康档案补充完善发病时间等信息,建立2型糖尿病与痛风发病的历史性队列;2.采用经典多因素logistic回归分析,分析T2DM与痛风在逐渐调整协变量的个数下两者一方对另一方发病的影响;3.采用Kaplan-Meier法估计累积发生函数,比较健康状态→T2DM和健康状态→痛风、T2DM→T2DM+痛风和痛风→T2DM+痛风的累积发病曲线差别;4.引入多状态竞争风险模型,估算健康状态、T2DM、痛风、T2DM+痛风四种状态之间的累积转移风险和转移概率,利用参数回归模型对累积转移风险进行拟合,确定参数分布后,进一步求得各状态之间的转移风险函数并进行比较。结果多状态竞争风险模型结合logistic累积分布模型拟合得到的转移风险函数结果较好。通四个过程的转移风险函数曲线及峰值比较,可以看出2型糖尿病与痛风间的关系为前期相互促进,后期则相互竞争。四过程的转移风险函数显示:60岁前,痛风促进T2DM的发病风险,由于痛风的存在,使T2DM发病峰值提前10年出现,60岁后发病风险快速下降;68岁前,T2DM促进痛风的发病风险,由于T2DM的后期竞争作用使痛风发病高峰提前11年出现,68岁后进入快速下降期。多状态竞争风险模型分别与logistic回归模型、KaplanMeier生存分析比较后发现logistic回归模型仅能发现两病的患病关联,Kaplan-Meier法无法准确估计存在竞争风险和多状态的生存分析。结论多状态竞争风险模型分析显示,2型糖尿病与痛风随时间的交互作用关系为前期呈现暨南大学硕士学位论文多状态竞争风险模型在2型糖尿病与痛风发病关系研究中的应用相互促进的协同作用、后期则转为相互竞争的抑制作用。糖尿病或痛风的存在,会使另一病种的发病高峰提前10-11年出现。对于多结局的纵向资料,多状态竞争风险模型能够分析不同结局间随时间的转化概率,在探索不同结局间交互作用关系及变化方面更具优势。
[Abstract]:Objective to explore the relationship between type 2 diabetes mellitus and gout with time by using multi-state competitive risk model. By comparing the survival analysis of Kaplan-Meier method with the logistic regression model of multi-state competitive risk model, this paper discusses which model is more suitable for dealing with the longitudinal research data of multiple outcomes. Methods 1. Using stratified cluster sampling to obtain the original cohort sample population, combining with the retrospective investigation of community residents in Guangzhou, using health files to supplement and perfect the information of onset time, etc. The history cohort of type 2 diabetes mellitus and gout was established. The classical multivariate logistic regression analysis was used to analyze the influence of one side of T2DM and gout on the incidence of type 2 diabetes mellitus and gout. 3. The cumulative occurrence function was estimated by Kaplan-Meier method. Relatively healthy state. 鈫扵2DM and health status. 鈫扜out T2DM. 鈫扵2DM gout and gout. 鈫扵he difference of cumulative incidence curve of T2DM. 4. Introducing multi-state competitive risk model to estimate the cumulative transfer risk and transfer probability between the four states of T2DMand T2DM. The cumulative transfer risk was fitted by the parametric regression model. After determining the distribution of parameters, Results the transfer risk function fitted by the multi-state competition risk model combined with the logistic cumulative distribution model is better. The transfer risk function of the four processes is better. Number curve and peak value comparison, It can be seen that the relationship between type 2 diabetes and gout promotes each other in the early stage and competes with each other in the later stage. The transfer risk function of the four processes shows that before the age of 60, gout promotes the risk of T2DM, due to the existence of gout. The risk of onset of T2DM was decreased rapidly after 60 years of age, and the risk of gout was promoted by T2DM before 68 years of age. Due to the late competitive effect of T2DM, the onset peak of gout appeared 11 years earlier and entered a rapid decline period after 68 years old. Compared with Kaplan Meier survival analysis of logistic regression model, multi-state competition risk model found that logistic regression model can only be found. Kaplan-Meier method can not accurately estimate the existence of competitive risk and multi-state survival analysis. Conclusion Multi-state competitive risk model analysis shows that the interaction between type 2 diabetes mellitus and gout with time is preexisting. Application of Multi-state competitive risk Model in the study of the relationship between Type 2 Diabetes Mellitus and Gout. Later on, it turns to competitive inhibition. The presence of diabetes or gout can lead to a peak of another disease occurring 10-11 years earlier. Multi-state competitive risk model can analyze the probability of transition between different endings over time, and it has more advantages in exploring the interaction and change among different endings.
【学位授予单位】：暨南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R587.1;R589.7

【相似文献】