使用 metaprop（R 中的元包）计算精确比例

Question

我正在对接受特定药物插管的患者进行荟萃分析。具体来说，我有 17 项研究。在每一项研究中，都有一部分患者接受插管。不同亚群有不同的插管风险。

我的问题是，当我绘制森林图时，比例似乎不正确。

library(meta)
studies <- c("Cong 2015", "Gangathimmaiah 2017", "Parsch 2017", "Hopper 2015", 
              "Isbister 2016", "Kowalski", "Riddell 2017", "Burnett 2012",
              "Burnett 2015", "Cole 2016", "Cole 2017", "Hollis 2017", "Iwanicki 2014", 
              "Keseg 2014", "Olives 2016", "Scaggs 2016", "Schepke 2015")
obs <- c(0, 3, 0, 0, 0, 0, 2, 2, 14, 25, 28, 10, 16, 8, 85, 0, 2)
denom <- c(18, 21, 22, 32, 49, 5, 23, 12, 49, 64, 49, 38, 32, 35, 135, 7, 52)
setting <- c("AMT", "AMT", "AMT", "ED", "ED", "ED", "ED", "EMS", "EMS", 
             "EMS", "EMS", "EMS", "EMS", "EMS", "EMS", "EMS", "EMS")
m1 <- metaprop(obs, denom, studies, comb.random=FALSE, byvar=setting, 
               bylab="Setting", byseparator=": ")
forest(m1, print.tau2 = FALSE, col.by="black", text.fixed = "Total",
       text.fixed.w = "Subtotal", rightcols = c("effect","ci"), 
       leftlabs=c("Study","Intubated","Total"))

当我绘制此图时，第一个子组显示 61 名插管患者中的 3 名。但是比例显示为0.09，其实应该是0.05。

我做错了什么？

Answer 1

您应该使用 metaprop 的 method="GLMM" 参数：

library(meta)
library(metafor)

m1 <- metaprop(obs, denom, studies, comb.random=F, byvar=setting, 
               bylab="Setting", byseparator=": ")
m2 <- metaprop(obs, denom, studies, comb.random=F, byvar=setting, 
               bylab="Setting", byseparator=": ", method="GLMM")
m2$w.fixed <- m1$w.fixed

forest(m2, print.tau2 = FALSE, col.by="black", text.fixed = "Total",
       text.fixed.w = "Subtotal", rightcols = c("effect","ci"), 
       leftlabs=c("Study","Intubated","Total"))

使用 metaprop（R 中的元包）计算精确比例

Calculate exact proportion using metaprop (meta package in R)

r

statistics

forestplot