完全随机设计和配对设计临床试验研究的统计分析方法及比较  

Comparison of statistical analysis between clinical trials with paired design and simple randomized design

作　　者：王瑞平 陈曦[2] 陈洁[2] 李欣[2] 强燕 付晓伶 李斌[2,4] WANG Rui-ping;CHEN Xi;CHEN Jie;LI Xin;QIANG Yan;FU Xiao-ling;LI Bin(Office of Clinical Research Center,Yueyang Hospital of Integrated Traditional Chinese and Western Medicine,Shanghai University of Traditional Chinese Medicine,Shanghai 200437,China;Department of Dermatology,Yueyang Hospital of Integrated Traditional Chinese and Western Medicine,Shanghai University of Traditional Chinese Medicine,Shanghai 200437,China;Department of Dermatology,Songjiang Hospital,Medical College,Shanghai Jiaotong University,Shanghai 201620,China;Research Institute of Dermatology,Shanghai Academy of Traditional Chinese Medicine,Shanghai 201203,China)

机构地区：[1]上海中医药大学附属岳阳中西医结合医院临床研究中心办公室,上海200437 [2]上海中医药大学附属岳阳中西医结合医院皮肤科,上海200437 [3]上海交通大学医学院附属松江医院皮肤科,上海201620 [4]上海市中医药研究院皮肤病研究所,上海201203

出　　处：《世界临床药物》2019年第11期783-789,794,共8页World Clinical Drug

基　　金：国家重点研发计划“中医药现代化研究”重点专项(2018YFC1705301);上海市松江区科委科技攻关项目(18SJJKGG58)

摘　　要：目的分析临床试验中完全随机设计和配对设计在数据整理和统计分析过程的差异,提高临床研究人员对两种设计方案的认知和研究分析能力。方法基于银屑病"凉血潜阳法"临床研究选取100例对象,分别按照"完全随机设计"和"配对设计"的方案进行数据整理,建立模拟数据库,对不同设计的数据库架构、统计分析指标选择和统计分析等内容进行比较分析。结果完全随机设计通过"组别"变量区分试验组和对照组,配对设计通过在变量名上添加(1,0)区分试验组和对照组,变量个数翻倍。完全随机设计和配对设计在统计分析指标选择、统计描述指标、单因素分析和重复测量数据统计分析、数据展示方式和图形选择等方面均存在区别。结论临床试验研究中,应综合考虑研究设计类型和研究结果的数据分类,选择合适的统计分析模型和指标,改善临床试验中统计分析结果的准确性。Objective To compare the difference of clinical trials with paired design and simple randomized design in the aspect of data arrangement and data analysis,and improve the recognition of clinical trial design and data analysis ability among clinical trial practitioners.Methods Based on a psoriasis clinical trial titled ’Cooling Blood and Latent Yang Method’,100 patients were extracted and rearranged into two simulation databases in accordance with the study plan of ’simple randomized design’ and ’paired design’,then we compare the differences of database structure,statistical index selection and statistical analysis between clinical trials with the paired design and simple randomized design.Results In RCT with simple randomized design,a variable named ’group’ is used to differentiate the treatment group and control group,whereas in RCT with paired design,the number of variables are doubled with each pair added a ’1’ or ’0’ to differentiate the treatment group and control group.Compared with RCT with simple randomized design,the statistical index selection,descriptive analysis,single factor analysis,repeated measurement analysis,data visualization method in RCT with paired design is different.Conclusion In clinical trials,it is critical to contemplate the study design and data type comprehensively before data analysis,and then select the proper statistical analysis model and statistical indexes accordingly could improve the correctness and accuracy of data analysis in clinical trials.

关 键 词：完全随机设计 配对设计 临床试验 比较分析 

分 类 号：R181.2[医药卫生—流行病学]