knitr::opts_chunk$set(warning=F, mesage=F, dpi=200, fig.path='figures/', dev=c('png', 'svg'))
library(meta)
library(ggplot2)Statistical analysis was carried out using the
meta (v8.2.1) package (Balduzzi et
al. 2019) in the R version 4.5.2 (2025-10-31)
programming language (R Core Team
2021).
Meta-analysis of proportions was pooled by fitting a random intercept
logistic regression model with the metaprop function to
logit transformed proportions in order to include valid estimates for
studies with very few or no events. Study estimates are shown with
computed Clopper-Pearson 95% confidence intervals. The same pooled
estimates were conducted for subgroups and tested by \(\chi^2\)-test for significant pairwise
differences.
Heterogeneity was assessed by estimating the maximum-likelihood of \(\tau^2\) and quantified with the \(I^2\) index.
Groupwise comparisons of studies were analyzed using odds ratios
between treatment and control group with the metabin
function using a random effects model for the pooled odds ratio using
the Mantel-Haenszel method (Mantel and Haenszel
1959). Heterogeneity was assessed using a restricted
maximum-likelihood estimator of \(\tau^2\).
Publication bias was assessed using a funnelplot of the logit
transformed prevalence and inverse variance and tested using the
metabias function with linear regression test (Egger et al. 1997) and rank correlation test
(Begg and Mazumdar 1994) for
asymmetry.
#d = readxl::read_xlsx('data/extracted.xlsx', sheet=1, na='NA')