Bayesian Network Meta-Analysis with Missing Participants • rnmamod

Official Git repository of rnmamod R package

In R Views November 2021: “Top 40” New CRAN Packages

In Medium Peek-a-boo: Evidence synthesis using the rnmamod R package

In the YouTube channel of ESMARConf2023

Contributors

Loukia Spineli, Chrysostomos Kalyvas, Katerina Papadimitropoulou

Website

The website of rnmamod currently includes two short tutorials (Description of the network and Perform network meta-analysis). There is also a structured reference list to facilitate access to the documentation of the functions available in the package.

Description

The rnmamod package performs one-stage fixed-effect or random-effects Bayesian network meta-analysis (NMA) while adjusting for missing participant outcome data (MOD) using the pattern-mixture model. In the case of two interventions, the package performs one-stage Bayesian pairwise meta-analysis. Standard aggregate data NMA is a special case of NMA with MOD, where either all studies or no studies have MOD available, respectively.

The package handles data inputs in arm-based format: - binary data (effect size: odds ratio, risk ratio, and risk difference)
- continuous data (effect size: raw and standardised mean differences, ratio of means).

The rnmamod package comprises a suite of all necessary models for estimation and prediction of the intervention effect, and evaluation of the consistency assumption locally and globally. Also includes a rich suite of visualisation tools to aid in interpretation of the results and preparation of NMA manuscript submission. Missing participant outcome data are addressed in all models of the package.

Getting started

Run the following code to install and load the package from CRAN:

install.packages("rnmamod")
library(rnmamod)

or run the following code to install the development version of the package:

install.packages("devtools")
devtools::install_github("LoukiaSpin/rnmamod")

Example

Baker et al. (2009) comprising 21 trials comparing seven pharmacologic interventions with each other and placebo in chronic obstructive pulmonary disease (COPD) patients. The exacerbation of COPD (harmful outcome) is the analysed binary outcome.

head(nma.baker2009)
#>                 study t1 t2 t3 t4 r1 r2 r3 r4 m1 m2 m3 m4  n1  n2 n3 n4
#> Llewellyn-Jones, 1996  1  4 NA NA  3  0 NA NA  1  0 NA NA   8   8 NA NA
#>        Paggiaro, 1998  1  4 NA NA 51 45 NA NA 27 19 NA NA 139 142 NA NA
#>          Mahler, 1999  1  7 NA NA 47 28 NA NA 23  9 NA NA 143 135 NA NA
#>        Casaburi, 2000  1  8 NA NA 41 45 NA NA 18 12 NA NA 191 279 NA NA
#>       van Noord, 2000  1  7 NA NA 18 11 NA NA  8  7 NA NA  50  47 NA NA
#>         Rennard, 2001  1  7 NA NA 41 38 NA NA 29 22 NA NA 135 132 NA NA

Network plot

Create the network plot using the netplot function:

# The names of the interventions in the order they appear in the dataset
interv_names <- c("placebo", "budesonide", "budesonide plus formoterol", "fluticasone", "fluticasone plus salmeterol", "formoterol", "salmeterol", "tiotropium")

netplot(data = nma.baker2009, drug_names = interv_names, show_multi = TRUE, edge_label_cex = 1)

Perform Bayesian random-effects network meta-analysis

The following code performs a Bayesian random-effects network meta-analysis under the missing at random assumption while using an intervention-specific informative missingness odds ratio (assumption = "IDE-ARM") in the logarithmic scale:

res <- run_model(data = nma.baker2009,
                 measure = "OR",
                 model = "RE",
                 assumption = "IDE-ARM",
                 heter_prior = list("halfnormal", 0, 1),
                 mean_misspar = c(0, 0), 
                 var_misspar = 1,
                 D = 0,
                 n_chains = 3,
                 n_iter = 10000,
                 n_burnin = 1000,
                 n_thin = 1)

League table

Illustrate all possible pairwise comparisons of the interventions using a league heatmap. Interventions are sorted in decreasing order by their posterior mean SUCRA (surface under the cumulative ranking) value in the main diagonal:

league_heatmap(full1 = res, drug_names1 = interv_names)

Rankogram with SUCRA curves

The following code presents the hierarchy of the interventions in the network using integrated rankograms and SUCRA curves:

rankosucra_plot(full1 = res, drug_names1 = interv_names)

Funding source

The development of the rnmamod package was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft) (grant no. SP 1664/1-3 and SP 1664/2-1)

rnmamod: Bayesian network meta-analysis R package