[R] meta analysis for sensitivity and specificity

[Viechtbauer, Wolfgang (SP)]

Dear Greg,

I am not sure if I understand your question. If you are asking how to do this in R, then one could use the metafor or meta package for this. The specificity and sensitivity values are proportions, so one would usually meta-analyze them after a logit transformation. But all of the specificity values are equal to 1, so this is pretty pointless. For sensitivity:

dat <- data.frame(pi = c(.66, .55, .76, .40, .46), ni = c(57, 33, 17, 30, 11))
dat$xi <- round(dat$pi * dat$ni)

library(metafor)

dat <- escalc(measure="PLO", xi=xi, ni=ni, data=dat)
res <- rma(yi, vi, data=dat)
res
predict(res, transf=transf.ilogit)

One could also use a logistic mixed-effects model for this:

res <- rma.glmm(measure="PLO", xi=xi, ni=ni, data=dat)
res
predict(res, transf=transf.ilogit)

If you want to analyze the specificity and sensitivity together, then you would want to use a bivariate model. There are some specific packages for this. See the Meta-Analysis Task View (https://cran.r-project.org/web/views/MetaAnalysis.html). I just saw that Michael also replied with the same suggestion (and the note about the mailing list).

Best,
Wolfgang

>-----Original Message-----
>From: R-help [mailto:r-help-bounces using r-project.org] On Behalf Of greg
>holly
>Sent: Thursday, 06 December, 2018 22:38
>To: r-help mailing list
>Subject: [R] meta analysis for sensitivity and specificity
>
>Does anyone know any R library that runs meta-analysis in SAS differently
>for  Sensitivity and Specificity if I have only the following info?
>
>Regards,
>
>Greg
>
>specificity sample_size Sensitivity Sample_size
>1 21 0.66 57
>1 70 0.55 33
>1 19 0.76 17
>1 10 0.4 30
>1 16 0.46 11