2
mydat <- data.frame(stage1 = c(4, 3, 2, 1), n1 = c(10, 40, 30, 20))
mydat
  stage1 n1
1      4 10
2      3 40
3      2 30
4      1 20

I have a simple data set with 4 studies (1 per row) and I have the number of events (stage1) and the sample size (n1) from each study. At stage1 of the study, 4 out of 10 people from study #1 have the disease. In comparison, only 3/40 people from study #2 have the disease, etc. 

library(meta)
metaprop(mydat$stage1, mydat$n1)

I use the metaprop function to perform a meta-analysis of the single proportions from the 4 studies. However, suppose each study reports other effect estimates at a later stage. 

mydat2 = data.frame(stage1 = c(4, 3, 2, 1), n1 = c(10, 20, 30, 40), stage2 = c(7, 5, 3, 4), n2 = c(10, 20, 30, 40))
mydat2
  stage1 n1 stage2 n2
1      4 10      7 10
2      3 20      5 20
3      2 30      3 30
4      1 40      4 40

So here, at stage2, a total of 7 people out of 10 possible people now have the disease from study #1. How exactly would I take that into account so that the correlation is adjusted correctly? Should I make stage and study indicators in my data.frame: 

> mydat3 = data.frame(event = c(mydat2$stage1, mydat2$stage2), n = c(mydat2$n1, mydat2$n2), stage = c(rep(1, 4), rep(2, 4)), study = c(rep(c(1, 2, 3, 4), 2)))
> mydat3
  event  n stage study
1     4 10     1     1
2     3 20     1     2
3     2 30     1     3
4     1 40     1     4
5     7 10     2     1
6     5 20     2     2
7     3 30     2     3
8     4 40     2     4

But how would I include that in my metaprop function call? I'm also open to using other functions (not just metaprop).

Adrian
  • 9,229
  • 24
  • 74
  • 132

1 Answers1

0

Here a solution using generalized linear effects model. 2.4 is the odds ratio of an increase in prevalence for a unit increase of stage.

fit <- glmer(cbind(event, n-event) ~ (1|study) + stage, data = mydat3,         
family=binomial(link = "logit"))
summary(fit)
es <- fixef(fit)
confint <- confint(fit)[-1,]
exp(cbind(es, confint))

                     es       2.5 %    97.5 %
 (Intercept) 0.05082079 0.006362861 0.3405915
 stage       2.39413175 1.001923435 6.1305560
paoloeusebi
  • 1,056
  • 8
  • 19