非参数回归方法简介及其在医学研究领域中的应用  

Introduction and application of non-parametric regression method in medical research

作　　者：刘亚航 余勇夫 秦国友 LIU Ya-hang;YU Yong-fu;QIN Guo-you(Department of Biostatistics,School of Public Health,Fudan University,Shanghai 200032,China)

机构地区：[1]复旦大学公共卫生学院生物统计学教研室,上海200032

出　　处：《复旦学报（医学版）》2024年第2期280-284,共5页Fudan University Journal of Medical Sciences

基　　金：国家自然科学基金(82173612);上海市市级科技重大专项(ZD2021CY001)。

摘　　要：本文介绍了非参数回归方法的基本理论,并通过实例数据分析展示其在医学和公共卫生领域的应用,为相关研究提供方法学参考。实例基于某疾病预防控制中心部分慢病管理数据,拟合包含限制性立方样条的Cox比例风险模型及双变量响应模型,探索2型糖尿病人群中血糖均值水平和血糖变异水平对全因死亡的单独和共同作用。结果显示血糖变异水平与全因死亡风险存在非线性关联;在高血糖均值水平下观察到的血糖变异水平与全因死亡的关联比在低血糖均值水平下更强。非参数回归方法可以全面探索连续型暴露因素和结局变量之间复杂的剂量-反应关系,揭示两个连续型暴露因素间的共同作用,可为目标人群制定针对性的干预提供参考依据。该方法在医学和公共卫生研究中有很好的应用和推广价值。This article introduced the basic theory of non-parametric regression and its application in medical and public health research for methodological reference.We conducted Cox proportional hazard models with restricted cubic splines using chronic disease management data from a Center for Disease Control and Prevention.We aimed to explore the separate and combined effects of mean fasting blood glucose level and glucose variability on all-cause mortality among individuals with type 2 diabetes.A non-linear association was observed between glucose variability and the risk of all-cause mortality.The association between glucose variability and all-cause mortality was stronger at higher mean fasting blood glucose levels compared to lower levels.The non-parametric regression methods comprehensively explored dose-response relationships between continuous exposure and outcome,revealing the combined effects of continuous exposures,which provided recommendations for targeted interventions.The method showed promising application value in medical and public health research.

关 键 词：非参数回归 血糖均值 血糖变异 死亡率 

分 类 号：R311[医药卫生—基础医学]