bnma: A package for network meta analysis using Bayesian methods

Description

A package for running Bayesian network meta analysis

Details

Network meta-analysis or mixed treatment comparison (MTC) is a method that allows simultaneous comparison of more than two treatments. We use a Bayesian approach to combine both direct and indirect evidence as in Dias et al. 2013a. This package is a user friendly application that can run network meta analysis models without having to code a JAGS model. The program takes the input data and transforms it to a suitable format of analysis, generates a JAGS model and reasonable initial values and runs the model through the rjags package. The focus of this package was inclusion of multinomial response and various options for adding covariates and/or baseline risks effects. Also, while sampling, the package uses Gelman-Rubin convergence criteria to decide whether to continue sampling or not. Furthermore, package includes different models such as contrast based models and unrelated mean effects (UME) model and nodesplitting model to test for inconsistency.

References

A.J. Franchini, S. Dias, A.E. Ades, J.P. Jansen, N.J. Welton (2012), Accounting for correlation in network meta-analysis with multi-arm trials, Research Synthesis Methods 3(2):142-160. doi:10.1002/jrsm.1049

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013b), Heterogeneity-Subgroups, Meta-Regression, Bias, and Bias-Adjustment, Medical Decision Making 33(5):618-640. doi:10.1177/0272989X13485157

S. Dias, N.J. Welton, D.M. Caldwellb, A.E. Ades (2010), Checking consistency in mixed treatment, Statistics in Medicine 29(7-8, Sp. Iss. SI): 932-944. doi:10.1002/sim.3767

S. Dias, N.J. Welton, A.J. Sutton, D.M. Caldwell, G. Lu, and A.E. Ades (2013), Evidence synthesis for decision making 4: inconsistency in networks of evidence based on randomized controlled trials, Medical Decision Making 33(5):641-656. doi:10.1177/0272989X12455847

C.H. Schmid, T.A. Trikalinos, I. Olkin (2014), Bayesian network meta-analysis for unordered categorical outcomes with incomplete data, Research Synthesis Methods 5(2):162-185. doi:10.1002/jrsm.1103

A. Gelman, D.B. Rubin (1992), Inference from iterative simulation using multiple sequences, Statistical Science 7(4):457-472. doi:10.1214/ss/1177011136

D.J. Spiegelhalter, N.G. Best, and B.P. Carlin (1998), Bayesian deviance, the effective nunmber of parameters, and the comparison of arbitrarily complex models, Technical report, MRC Biostatistics Unit, Cambridge, UK.

F.A. Achana, N.J. Cooper, S. Dias, G. Lu, S.J.C. Rice, D. Kendrick, A.J. Sutton (2012), Extending methods for investigating the relationship between treatment effect and baseline risk from pairwise meta-analysis to network meta-analysis, Statistics in Medicine 32(5):752-771. doi:10.1002/sim.5539

F.A. Achana, N.J. Cooper, S. Bujkiewicz, S.J. Hubbard, D. Kendrick, D.R. Jones, A.J. Sutton (2014), Network meta-analysis of multiple outcomes measures accounting for borrowing of information across outcomes, BMC Medical Research Methodology 14:92. doi:10.1186/1471-2288-14-92

G. Salanti, A.E. Ades, J.P.A. Ioannidisa (2011), Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial, Journal of Clinical Epidemiology 64(2):163-171. doi:10.1016/j.jclinepi.2010.03.016

G. van Valkenhoef, G. Lu, B. de Brock, H. Hillege, A.E. Ades, and N.J. Welton (2012), Automating network meta-analysis, Research Synthesis Methods 3(4):285-299. doi:10.1002/jrsm.1054

N.J. Cooper, A.J. Sutton, D. Morris, A.E. Ades, N.J. Welton (2009), Addressing between-study heterogeneity and inconsistency in mixed treatment comparisons: Application to stroke prevention treatments in individuals with non-rheumatic atrial fibrillation, Statistics in Medicine 28:1861-1881. doi:10.1002/sim.3594

W. Viechtbauer (2010), Conducting meta-analyses in R with the metafor package, Journal of Statistical Software, 36(3):1-48. doi:10.18637/jss.v036.i03

See Also

network.data, network.run

Beta blockers to prevent mortality after myocardial infarction

Description

A dataset of 22 trials investigating beta blockers versus control to prevent mortality after myocardial infarction. Control is coded as 1 and beta blocker treatment is coded as 2.

Usage

blocker

Format

A list of Outcomes, Treat, Study, and N.

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Find deviance statistics such as DIC and pD.

Description

Calculates deviance statistics. This function automatically called in contrast.network.run and the deviance statistics are stored after sampling is finished.

Usage

calculate.contrast.deviance(result)

Arguments

Value

References

A.J. Franchini, S. Dias, A.E. Ades, J.P. Jansen, N.J. Welton (2012), Accounting for correlation in network meta-analysis with multi-arm trials, Research Synthesis Methods 3(2):142-160. doi:10.1002/jrsm.1049

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Examples

network <- with(parkinsons_contrast, {
 contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)
calculate.contrast.deviance(result)

Find deviance statistics such as DIC and pD.

Description

Calculates deviance statistics. This function automatically called in network.run and the deviance statistics are stored after sampling is finished.

Usage

calculate.deviance(result)

Arguments

Value

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Examples

#parkinsons
network <- with(parkinsons, {
 network.data(Outcomes, Study, Treat, SE = SE, response = "normal")
})

result <- network.run(network)
calculate.deviance(result)

Trials of low dose and high dose statins for cardiovascular disease vs. placebo

Description

A dataset of 17 studies investigating dosage of statin for cardiovascular disease. There are two treatments and a placebo. High dose statin is coded as 3, low dose statin as 2, and placebo is coded as 1 and treated as a baseline treatment. Outcomes are reported as three mutually exclusive unordered outcomes. First column of the outcome is the patients who are still alive (ALIVE). Second column is fatal non-cardiovascular disease (FnCVD). And, the last column is fatal cardiovascular disease (FCVD).

Usage

cardiovascular

Format

A list of Outcomes, Treat, Study, and N

References

C.H. Schmid, T.A. Trikalinos, I. Olkin (2014), Bayesian network meta-analysis for unordered categorical outcomes with incomplete data, Research Synthesis Methods 5(2):162-185. doi:10.1002/jrsm.1103

Trials of certolizumab pegol (CZP) for the treatment of rheumatoid arthritis in patients

Description

A dataset of 12 trials for investigating CZP for the treatment for those who had failed on disease-modifying antirheumatic drugs, including methotrexate (MTX). Data provides the number of patients who have improved and there are 6 different treatments with placebo. Mean disease duration (years) is provided as a covariate.

Usage

certolizumab

Format

A list of Outcomes, Treat, Study, N, covariate, and Treat.order

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013b), Heterogeneity-Subgroups, Meta-Regression, Bias, and Bias-Adjustment, Medical Decision Making 33(5):618-640. doi:10.1177/0272989X13485157

Make a network object for contrast-level data containing data, priors, and a JAGS model file

Description

This is similar to the function network.data, except it uses contrast-level data instead of arms-level data. Contrast-level format uses treatment differences relative to the control arm. Note that in two arm trials there is only one contrast value per trial, but in three arm trials there are two contrast values relative to the control arm.

Usage

contrast.network.data(
  Outcomes,
  Treat,
  SE,
  na,
  V = NULL,
  type = "random",
  rank.preference = "higher",
  mean.d = 0,
  prec.d = 1e-04,
  hy.prior = list("dunif", 0, 100)
)

Arguments

Value

Creates list of variables that are used to run the model using contrast.network.run

References

A.J. Franchini, S. Dias, A.E. Ades, J.P. Jansen, N.J. Welton (2012), Accounting for correlation in network meta-analysis with multi-arm trials, Research Synthesis Methods 3(2):142-160. doi:10.1002/jrsm.1049

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Examples

network <- with(parkinsons_contrast, {
 contrast.network.data(Outcomes, Treat, SE, na, V)
})

Make a contrast network deviance plot

Description

This makes a contrasrt network deviance plot which plots residual deviance (resdev_study) vs. all study.

Usage

contrast.network.deviance.plot(result)

Arguments

Value

None

Examples

network <- with(parkinsons_contrast, {
contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)
contrast.network.deviance.plot(result)

Make a leverage plot

Description

This function makes a leverage vs. square root of residual deviance plot

Usage

contrast.network.leverage.plot(result)

Arguments

Value

None

Examples

network <- with(parkinsons_contrast, {
contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)
contrast.network.leverage.plot(result)

Run the model using the network object

Description

This is similar to the function network.run, except it uses contrast-level data instead of arms-level data.

Usage

contrast.network.run(
  network,
  inits = NULL,
  n.chains = 3,
  max.run = 1e+05,
  setsize = 10000,
  n.run = 50000,
  conv.limit = 1.05,
  extra.pars.save = NULL
)

Arguments

Value

Examples

network <- with(parkinsons_contrast, {
 contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)

Draws network graph using igraph package

Description

This function draws network graph using igraph package

Usage

draw.network.graph(network, label.dist = 2)

Arguments

Value

None

Examples

#cardiovascular
network <- with(thrombolytic, {
 network.data(Outcomes, Study, Treat, N=N, response = "binomial")
})
draw.network.graph(network)

Generate autocorrelation diagnostics using coda package

Description

This function generates autocorrelation diagnostics using coda package. User can specify lags and parameters to display. Note that to display extra parameters that are not saved, user needs to first specify parameters in extra.pars.save parameter in network.run function.

Usage

network.autocorr.diag(
  result,
  lags = c(0, 1, 5, 10, 50),
  extra.pars = NULL,
  only.pars = NULL
)

Arguments

Value

Returns autocorrelation diagnostics

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.autocorr.diag(result, only.pars = "d")

Generate autocorrelation plot using coda package

Description

This function plots autocorrelation using coda package.

Usage

network.autocorr.plot(result, extra.pars = NULL, only.pars = NULL)

Arguments

Value

None

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.autocorr.plot(result)

Make a covariate plot

Description

This function makes a covariate plot of how the relative effect changes as the covariate value changes. User needs to specify one base treatment and one comparison treatment to make this plot (base category and comparison category is also needed for multinomial). The function uses the relative.effects to calculate the correct relative effect. 2.5%, median, and 97.5% C.I. are drawn.

Usage

network.covariate.plot(
  result,
  base.treatment = NULL,
  comparison.treatment = NULL,
  base.category = NULL,
  comparison.category = NULL,
  covariate.name = NULL
)

Arguments

Value

None

Examples

########### certolizumab (with covariate)
network <- with(certolizumab, {
 network.data(Outcomes, Study, Treat, N=N, response="binomial", Treat.order,
 covariate = covariate, covariate.type = "continuous", hy.prior = list("dhnorm", 0, 9.77))
})

result <- network.run(network)
network.covariate.plot(result, base.treatment = "Placebo", comparison.treatment = "CZP",
covariate.name = "Disease Duration")

Create a treatment cumulative rank plot

Description

This function creates a treatment cumulative rank plot. Rank preference can be specified by the rank.preference parameter in network.data

Usage

network.cumrank.tx.plot(
  result,
  txnames = NULL,
  catnames = NULL,
  legend.position = c(1, 1)
)

Arguments

Value

None

See Also

rank.tx

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.cumrank.tx.plot(result, txnames = c("control", "beta blocker"))

Make a network object containing data, priors, and a JAGS model file

Description

This function makes a network object that can be used to run network meta-analysis using network.run. User needs to specify Outcomes, Study, Treat, N or SE, and response. Prior parameters are filled in automatically based on the data type if not specified. The input data should be arm-level so that we have observations for each treatment in each study. The input data is preprocessed to fit the format necessary to run model in JAGS.

Usage

network.data(
  Outcomes = NULL,
  Study = NULL,
  Treat = NULL,
  N = NULL,
  SE = NULL,
  response = NULL,
  Treat.order = NULL,
  type = "random",
  rank.preference = "higher",
  baseline = "none",
  baseline.risk = "independent",
  covariate = NULL,
  covariate.type = NULL,
  covariate.model = NULL,
  mean.d = NULL,
  prec.d = NULL,
  mean.Eta = NULL,
  prec.Eta = NULL,
  hy.prior.Eta = NULL,
  mean.bl = NULL,
  prec.bl = NULL,
  hy.prior.bl = NULL,
  mean.cov = NULL,
  prec.cov = NULL,
  hy.prior.cov = NULL,
  hy.prior = NULL,
  mean.A = NULL,
  prec.A = NULL,
  Z = NULL,
  Z_bl = NULL
)

Arguments

Value

Creates list of variables that are used to run the model using network.run

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

F.A. Achana, N.J. Cooper, S. Dias, G. Lu, S.J.C. Rice, D. Kendrick, A.J. Sutton (2012), Extending methods for investigating the relationship between treatment effect and baseline risk from pairwise meta-analysis to network meta-analysis, Statistics in Medicine 32(5):752-771. doi:10.1002/sim.5539

N.J. Cooper, A.J. Sutton, D. Morris, A.E. Ades, N.J. Welton (2009), Addressing between-study heterogeneity and inconsistency in mixed treatment comparisons: Application to stroke prevention treatments in individuals with non-rheumatic atrial fibrillation, Statistics in Medicine 28:1861-1881. doi:10.1002/sim.3594

Examples

###Blocker data example
blocker
network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})
network

Make a deviance plot

Description

This makes a deviance plot which plots residual deviance (dev_arm) vs. all the arms for each study.

Usage

network.deviance.plot(result)

Arguments

Value

None

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.deviance.plot(result)

Draws forest plot

Description

Draws forest plot of pooled treatment effect. Reports odds ratio for binomial and multinomial outcomes and continuous scale for normal outcomes.

Usage

network.forest.plot(
  result,
  level = 0.95,
  ticks.position = NULL,
  label.multiplier = 0.2,
  label.margin = 10,
  title = "Network Meta-analysis Forest plot",
  only.reference.treatment = FALSE
)

Arguments

Value

None

References

W. Viechtbauer (2010), Conducting meta-analyses in R with the metafor package, Journal of Statistical Software, 36(3):1-48. doi:10.18637/jss.v036.i03

Examples

network <- with(certolizumab, {
 network.data(Outcomes, Study, Treat, N=N, response="binomial", Treat.order,
 covariate = covariate, covariate.type = "continuous", hy.prior = list("dhnorm", 0, 9.77))
})

result <- network.run(network)
network.forest.plot(result)

Use coda package to find Gelman-Rubin diagnostics

Description

This function uses coda package to find Gelman-Rubin diagnostics.

Usage

network.gelman.diag(result, extra.pars = NULL, only.pars = NULL)

Arguments

Value

Returns gelman-rubin diagnostics

Examples

network <- with(statins, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial",
 Treat.order = c("Placebo", "Statin"), covariate = covariate, covariate.type = "discrete")
})

result <- network.run(network)
network.gelman.diag(result, extra.pars = c("Eta"))

Use coda package to plot Gelman-Rubin diagnostic plot

Description

This function plots Gelman-Rubin diagnostic using coda package.

Usage

network.gelman.plot(result, extra.pars = NULL, only.pars = NULL)

Arguments

Value

None

Examples

network <- with(statins, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial",
 Treat.order = c("Placebo", "Statin"), covariate = covariate, covariate.type = "discrete")
})

result <- network.run(network)
network.gelman.plot(result)

Plotting comparison of posterior mean deviance in the consistency model and inconsistency model

Description

This function compares posterior mean deviance of inconsistency model and consistency model. Such comparison provides information that can help identify the loops in which inconsistency is present.

Usage

network.inconsistency.plot(result1, result2, with.label = T)

Arguments

Details

This function draws network graph using igraph package

Value

None

References

S. Dias, N.J. Welton, A.J. Sutton, D.M. Caldwell, G. Lu, and A.E. Ades (2013), Evidence synthesis for decision making 4: inconsistency in networks of evidence based on randomized controlled trials, Medical Decision Making 33(5):641-656. doi:10.1177/0272989X12455847

Examples

network1 <- with(smoking, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial", type = "random")
})

network2 <- with(smoking, {
 ume.network.data(Outcomes, Study, Treat, N = N, response = "binomial", type = "random")
})

result1 <- network.run(network1)
result2 <- ume.network.run(network2)
network.inconsistency.plot(result1, result2)

Make a leverage plot

Description

This function makes a leverage vs. square root of residual deviance plot (mean for each study)

Usage

network.leverage.plot(result, per.study = FALSE)

Arguments

Value

None

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.leverage.plot(result)

Create a treatment rank plot

Description

This plot displays how each treatment is ranked. For each rank, we show how likely each treatment will be at that rank.

Usage

network.rank.tx.plot(
  result,
  txnames = NULL,
  catnames = NULL,
  legend.position = c(1, 1)
)

Arguments

Value

None

See Also

rank.tx

Examples

network <-with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
network.rank.tx.plot(result, txnames = c("a", "b"))

Run the model using the network object

Description

This is the core function that runs the model in our program. Before running this function, we need to specify data, prior, JAGS code, etc. using network.data.

Usage

network.run(
  network,
  inits = NULL,
  RNG.inits = NULL,
  n.chains = 3,
  max.run = 1e+05,
  setsize = 10000,
  n.run = 50000,
  conv.limit = 1.05,
  extra.pars.save = NULL
)

Arguments

Value

Examples

#parkinson's example (normal)
network <- with(parkinsons,{
  network.data(Outcomes, Study, Treat, SE = SE, response = "normal")
})

result <- network.run(network)

Make a network object containing data, priors, and a JAGS model file

Description

This function makes a network object that can be used to run network meta-analysis using nodesplit.network.run. User needs to specify Outcomes, Study, Treat, N or SE, and response. Prior parameters are filled in automatically based on the data type if not specified. The input data should be arm-level so that we have observations for each treatment in each study. The input data is preprocessed to fit the format necessary to run model in JAGS.

Usage

nodesplit.network.data(
  Outcomes = NULL,
  Study = NULL,
  Treat = NULL,
  N = NULL,
  SE = NULL,
  response = NULL,
  Treat.order = NULL,
  pair = NULL,
  type = "random",
  dic = FALSE
)

Arguments

Value

Creates list of variables that are used to run the model using nodesplit.network.run

References

S. Dias, N.J. Welton, D.M. Caldwellb, A.E. Ades (2010), Checking consistency in mixed treatment, Statistics in Medicine 29(7-8, Sp. Iss. SI): 932-944. doi:10.1002/sim.3767

Examples

###Thrombolytic data example
network <- with(thrombolytic,{
 nodesplit.network.data(Outcomes, Study, Treat, N, response = "binomial", pair = c(3,9))
})
network

Run the model using the nodesplit network object

Description

This is similar to the function network.run, except this is used for the nodesplitting model.

Usage

nodesplit.network.run(
  network,
  inits = NULL,
  n.chains = 3,
  max.run = 1e+05,
  setsize = 10000,
  n.run = 50000,
  conv.limit = 1.05,
  extra.pars.save = NULL
)

Arguments

Value

Examples

###Thrombolytic data example
network <- with(thrombolytic,{
 nodesplit.network.data(Outcomes, Study, Treat, N, response = "binomial", pair = c(3,9))
})

result <- nodesplit.network.run(network)

Dopamine agonists as adjunct therapy in Parkinson's disease

Description

A dataset of 7 studies investigating the mean lost work-time reduction in patients given 4 dopamine agonists and placebo as adjunct therapy for Parkinson's disease. There is placebo and four active drugs coded 2 to 5.

Usage

parkinsons

Format

A list of Outcomes, Treat, Study, N, covariate, and Treat.order

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Dopamine agonists as adjunct therapy in Parkinson's disease

Description

A contrast level (i.e. treatment difference) dataset of 7 studies investigating the mean lost work-time reduction in patients given 4 dopamine agonists and placebo as adjunct therapy for Parkinson's disease. Placebo is coded as 1, and four active drugs are coded 2 to 5. There is placebo, coded as 1, and four active drugs coded 2 to 5.

Usage

parkinsons_contrast

Format

A list of Outcomes, Treat, SE, na, and V

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Plot traceplot and posterior density of the result using contrast data

Description

This function uses plotting function in coda package to plot mcmc.list object

Usage

## S3 method for class 'contrast.network.result'
plot(x, ...)

Arguments

Value

None

Examples

network <- with(parkinsons_contrast, {
 contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)
plot(result)

Plot traceplot and posterior density of the result

Description

This function uses plotting function in coda package to plot mcmc.list object

Usage

## S3 method for class 'network.result'
plot(x, ...)

Arguments

Value

None

Examples

network <- with(statins, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial",
 Treat.order = c("Placebo", "Statin"), covariate = covariate, covariate.type = "discrete")
})

result <- network.run(network)
plot(result, only.pars = "sd")

Plot traceplot and posterior density of the result using contrast data

Description

This function uses plotting function in coda package to plot mcmc.list object

Usage

## S3 method for class 'ume.network.result'
plot(x, ...)

Arguments

Value

None

Examples

network <- with(smoking, {
 ume.network.data(Outcomes, Study, Treat, N = N, response = "binomial", type = "random")
})

result <- ume.network.run(network)
plot(result)

Create a treatment rank table

Description

This function makes a table of ranking for each treament. Each number in the cell represents a probability certain treatment was in such rank. This table is also stored as an output from network.run.

Usage

rank.tx(result)

Arguments

Value

Returns a table of ranking

See Also

network.rank.tx.plot

Examples

network <- with(blocker, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- network.run(network)
rank.tx(result)

Find relative effects for base treatment and comparison treatments

Description

This function calculates relative effects for base treatment and comparison treatments.

Usage

relative.effects(
  result,
  base.treatment = NULL,
  comparison.treatments = NULL,
  base.category = NULL,
  comparison.categories = NULL,
  covariate = NULL
)

Arguments

Value

This returns a mcmc.list sample of relative effects for the base treatment specified. This allows user to obtain relative effects of different base.treatment after the sampling has been done. For a simple summary, use relative.effects.table.

See Also

relative.effects.table

Examples

network <- with(parkinsons, {
 network.data(Outcomes, Study, Treat, SE = SE, response = "normal")
})

result <- network.run(network)
summary(relative.effects(result, base.treatment = "Placebo"))

Make a summary table for relative effects

Description

This function creates a summary table of relative effects. Relative effects are in units of log odds ratio for binomial and multinomial data and real number scale for normal data.

Usage

relative.effects.table(
  result,
  summary_stat = "mean",
  probs = NULL,
  base.category = NULL
)

Arguments

Value

Returns relative effects table

See Also

relative.effects

Examples

#cardiovascular
network <- with(cardiovascular,{
 network.data(Outcomes, Study, Treat, N, response = "multinomial")
})

result <- network.run(network)
exp(relative.effects.table(result)) #look at odds ratio instead of log odds ratio

Smoking cessation counseling programs

Description

Twenty-four studies, including 2 three-arm trials, compared 4 smoking cessation counseling programs and recorded the number of individuals with successful smoking cessation at 6 to 12 month. Counseling programs include 1 = no intervention, 2 = self-help, 3 = individual counseling, and 4 = group counseling.

Usage

smoking

Format

A list of Outcomes, Treat, Study, and N

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Trials of statins for cholesterol lowering vs. placebo or usual care

Description

A dataset of 19 trials of statins for cholesterol lowering vs. placebo. Each trial has a subgroup indicator for primary prevention (patients included had no previous heart disease) or secondary prevention (patients had previous heart disease). Dummy variable is coded such that covariate is equal to 1 if a study is a secondary prevention study and 0 if a study is a primary prevention study.

Usage

statins

Format

A list of Outcomes, Treat, Study, N, covariate, and Treat.order

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013b), Heterogeneity-Subgroups, Meta-Regression, Bias, and Bias-Adjustment, Medical Decision Making 33(5):618-640. doi:10.1177/0272989X13485157

Calculate SUCRA

Description

SUCRA is the surface under the cumulative ranking distribution defined in Salanti et al. (2011)

Usage

sucra(result, txnames = NULL, catnames = NULL)

Arguments

Value

Returns SUCRA for each treatment

References

G. Salanti, A.E. Ades, J.P.A. Ioannidisa (2011), Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial, Journal of Clinical Epidemiology 64(2):163-71. doi:10.1016/j.jclinepi.2010.03.016

See Also

rank.tx

Examples

########### certolizumab (with baseline risk)
network <- with(certolizumab, {
 network.data(Outcomes, Study, Treat, N=N, response = "binomial", Treat.order,
 baseline = "common", hy.prior = list("dhnorm", 0, 9.77))
})

result <- network.run(network)
sucra(result)

Summarize result run by contrast.network.run

Description

This function uses summary function in coda package to summarize mcmc.list object. Monte carlo error (Time-series SE) is also obtained using the coda package and is printed in the summary as a default.

Usage

## S3 method for class 'contrast.network.result'
summary(object, ...)

Arguments

Value

Returns summary of the contrast network model result

Examples

network <- with(parkinsons_contrast, {
 contrast.network.data(Outcomes, Treat, SE, na, V)
})

result <- contrast.network.run(network)
summary(result)

Summarize result run by network.run

Description

This function uses summary function in coda package to summarize mcmc.list object. Monte carlo error (Time-series SE) is also obtained using the coda package and is printed in the summary as a default.

Usage

## S3 method for class 'network.result'
summary(object, ...)

Arguments

Value

Returns summary of the network model result

Examples

network <- with(statins, {
 network.data(Outcomes, Study, Treat, N = N, response = "binomial",
 Treat.order = c("Placebo", "Statin"), covariate = covariate, covariate.type = "discrete")
})

result <- network.run(network)
summary(result)

Summarize result run by nodesplit.network.run

Description

This function uses summary function in coda package to summarize mcmc.list object. Monte carlo error (Time-series SE) is also obtained using the coda package and is printed in the summary as a default.

Usage

## S3 method for class 'nodesplit.network.result'
summary(object, ...)

Arguments

Value

Returns summary of the nodesplit network model result

Examples

###Parkinsons data example
network <- with(parkinsons, {
 nodesplit.network.data(Outcomes, Study, Treat, SE = SE, response = "normal", 
 Treat.order = Treat.order, pair = c("Placebo", "Ropinirole"))
})

result <- nodesplit.network.run(network)
summary(result)

Summarize result run by ume.network.run

Description

This function uses summary function in coda package to summarize mcmc.list object. Monte carlo error (Time-series SE) is also obtained using the coda package and is printed in the summary as a default.

Usage

## S3 method for class 'ume.network.result'
summary(object, ...)

Arguments

Value

Returns summary of the ume network model result

Examples

network <- with(smoking, {
 ume.network.data(Outcomes, Study, Treat, N = N, response = "binomial", type = "random")
})

result <- ume.network.run(network) 
summary(result)

Thrombolytic drugs and percutaneous transluminal coronary angioplasty

Description

A dataset consisting of 50 trials comparing 8 thrombolytic drugs and percutaneous transluminal coronary angioplasty, following acute myocardial infarction. Data consist of the number of deaths in 30 or 35 days and the number of patients in each treatment arm. There are 9 treatments in total: streptokinase (1), alteplase (2), accelerated alteplase (3), streptokinase + alteplase (4), reteplase (5), tenecteplase (6), percutaneous transluminal coronary angioplasty (7), urokinase (8), anistreptilase (9)

Usage

thrombolytic

Format

A list of Outcomes, Treat, Study, and N

References

S. Dias, A.J. Sutton, A.E. Ades, and N.J. Welton (2013a), A Generalized Linear Modeling Framework for Pairwise and Network Meta-analysis of Randomized Controlled Trials, Medical Decision Making 33(5):607-617. doi:10.1177/0272989X12458724

Make a network object for the unrelated mean effects model (inconsistency model) containing data, priors, and a JAGS model file

Description

This is similar to the function network.data, except this is used for the unrelated mean effects model.

Usage

ume.network.data(
  Outcomes,
  Study,
  Treat,
  N = NULL,
  SE = NULL,
  response = NULL,
  Treat.order = NULL,
  type = "random",
  mean.mu = NULL,
  prec.mu = NULL,
  mean.d = NULL,
  prec.d = NULL,
  hy.prior = list("dunif", 0, 5),
  dic = TRUE
)

Arguments

Value

Creates list of variables that are used to run the model using ume.network.run

References

S. Dias, N.J. Welton, A.J. Sutton, D.M. Caldwell, G. Lu, and A.E. Ades (2013), Evidence synthesis for decision making 4: inconsistency in networks of evidence based on randomized controlled trials, Medical Decision Making 33(5):641-656. doi:10.1177/0272989X12455847

Examples

network <- with(thrombolytic, {
 ume.network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})
network

Run the model using the network object

Description

This is similar to the function network.run, except this is used for the unrelated mean effects model.

Usage

ume.network.run(
  network,
  inits = NULL,
  n.chains = 3,
  max.run = 1e+05,
  setsize = 10000,
  n.run = 50000,
  conv.limit = 1.05,
  extra.pars.save = NULL
)

Arguments

Value

Examples

network <- with(thrombolytic, {
 ume.network.data(Outcomes, Study, Treat, N = N, response = "binomial")
})

result <- ume.network.run(network, n.run = 10000)

Calculate correlation matrix for multinomial heterogeneity parameter.

Description

This function calculates correlation matrix from the variance matrix for heterogeneity parameter. Only used for multinomial.

Usage

variance.tx.effects(result)

Arguments

Value

Returns correlation matrix

Examples

#cardiovascular
network <- with(cardiovascular, {
 network.data(Outcomes, Study, Treat, N, response = "multinomial")
})

result <- network.run(network)
variance.tx.effects(result)