Question: Funnel plot for meta-analysis of Bland-Altman studies using R
1
gravatar for sanders.m
2.2 years ago by
sanders.m10
sanders.m10 wrote:

I would like to make a funnel plot for a meta-analysis of Bland-Altman studies, with R regression test for symmetrie. On the x-axis the mean difference (bias) between two methods should be presented, on the y-axis the standard error (SD/√N).

However, the scripts described as an example online are based on data from 2x2 tables (outcome 1 and 2 for group 1 and 2), from which Relative Risk can be calculated by R. The problem is that my data consist of 1 bias (difference between both groups) and 1 SD of the bias.

Example:

library(metafor)

### load BCG vaccine data
dat <- get(data(dat.bcg))

show(dat.bcg)

### calculate log relative risks and corresponding sampling variances
dat <- escalc(measure="RR", ai=tpos, bi=tneg, ci=cpos, di=cneg, data=dat)

### random-effects model
res <- rma(yi, vi, data=dat)

### standard funnel plot
funnel(res)

### classical Egger test 
regtest(res, model="lm")

I do not know how to translate this script to a script for my situation. I have thought about several solutions:

Solution 1: Insert x-coordinates (bias) and y- coordinates (standard error) in R. After plotting the dots, R could perhaps calculate Egger, based on the plot. I do not know if R is able plot x- and y-coordinates and to calculate Egger based on x- and y-coordinates instead of raw data. I already tried as follows, but it didn’t work.

library(metafor)

mean.e<-c(-0.4, 0.3, 2.3))
sd.e<-c(1.55, 0.68, 3.22)

### fit random-effects model
res <- rma(yi=mean.e, vi=sd.e)

### classical Egger test
regtest(res, model="lm")

### standard funnel plot
funnel(res)

Solution 2: Make a script and insert the following data: mean(=bias).experimental, sd.experimental, n.experimental and mean.control, sd.control, n.control. As the data presented by the individual studies are already a bias (=mean difference) between two methods, there is no data available for a control group. This problem can be solved by adding extreme values for a virtual control group (mean 0, sd 0.0001, n 1000). Due to the extreme values the control group data has no influence on the outcome. So the outcome is equal to the experimental data, just how I want it. I hope R is able to calculate the bias (between experimental and virtual control group) and standard error and present it as a funnel plot, with Egger regression test. However, I do not know how to write such a script.

I hope somebody is able to help me, or could give me some good advice.

Thank you in advance!

Greetings Margot Sanders

R • 703 views
ADD COMMENTlink modified 2.2 years ago by zx87549.9k • written 2.2 years ago by sanders.m10
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1127 users visited in the last hour
_