剂量反应关系Meta分析在R软件中的实现  

Implementation of dose-response for meta-analysis in R software

作　　者：周权[1] 李卉[2] 罗美玲[3] 薛静[4] 黎敏[5] 

机构地区：[1]常德市第一人民医院科教科,湖南常德415003 [2]常德市第一人民医院肿瘤科,湖南常德415003 [3]长沙市疾病预防控制中心计划免疫科,湖南长沙410001 [4]中南大学公共卫生学院流行病与卫生统计系,湖南长沙410078 [5]南昌大学第二附属医院神经内科,江西南昌330006

出　　处：《现代预防医学》2015年第5期789-794,共6页Modern Preventive Medicine

摘　　要：目的介绍如何利用R软件中实现剂量反应关系的Meta分析。方法拟通过实例数据的演示，在R软件中，先利用Metafor程序包进行探索性二分类Meta分析，在得出意义结果的基础上再利用dosresmeta程序包进行剂量反应的Meta分析，并绘制非线性（和）或线性剂量反应曲线结果R软件分析结果显示：饮酒是结直肠癌的危险因素，剂量反应关系的非线性检验P值处于临界水平（0．046）。非线性剂量反应模型显示：相对于最低饮酒量（0g/天），每日饮酒20g、40g、60g发生结直肠癌的相对危险度为：1（0．93～1．1）、1．3（1．13～1．4）、1．5（1．26～1．9）；线性剂量反应模型显示：在最低饮酒量（0g／天）的基础上，每增加12g，发生结直肠癌的危险增加8％，非线性或线性剂量反应曲线图形基本相似。结论R软件可以实现剂量反应关系的Meta分析，功能强大，实用性强。Objective To implement the dose-response for meta-analysis in R software. Methods The code was illustrated with examples of alcohol intake and colorectal cancer. Firstly, the metafor package was used to explore binary classification meta analysis. Under the condition of a significant RR value, we conducted the dose-response meta-analysis with dosresmeta program and drew the linear and nonlinear dose-response curves. Results The results analyzed by R show that alcohol intake is a risk factor for colorectal cancer, and the P value for non-linearity test of dose-response relationship is 0.046. The non-linearity model shows that compared with non-drinkers, the pooled RR values of colorectal cancer were 1 （0.93-1.1）, 1.3 （1.13-1.4）, 1.5 （1.26-1.9） for 20, 40, 60gram/day, respectively; The linearity model shows that the risk of colorectal cancer increased 8% with a 12-g/day increase for the alcohol intake. There was no significant difference between the linear and nonlinear dose-response curves. Conclusion Dose-response meta-analysis can be implemented in R software, and it is powerful and practical.

关 键 词：剂量反应关系 META分析 系统评价 R软件 

分 类 号：R195[医药卫生—卫生统计学]