study_perc_contrib {rnmamod}R Documentation

Calculate study percentage contributions to summary treatment effects or regression coefficients

Description

A data-frame on the percentage contributions of each study to every possible pairwise comparison in the investigated network. Study percentage contributions are based on the proposed methodology of Donegan and colleagues (2018).

Usage

study_perc_contrib(
  study_name,
  base_t,
  exp_t,
  ref_t,
  obs_se,
  obs_cov = NULL,
  covar,
  covar_assum,
  model,
  tau = NULL,
  tau_beta = NULL
)

Arguments

study_name

A vector of labels with the names of the studies included in the investigated network. For multi-arm studies, the study name should appear as many times as the number of possible comparisons among the compared treatments.

base_t

A vector of numbers referring to the treatment identifier for the baseline arm (comparator) of each study.

exp_t

A vector of numbers referring to the treatment identifier for the experimental arm of each study.

ref_t

A scalar for the selected reference treatment in the network.

obs_se

A vector of numbers referring to the estimated standard error of the treatment effect of each study. For multi-arm studies, the standard error of the treatment effect of each possible comparison among the compared treatments should be included.

obs_cov

A vector of numbers referring to the covariance in the block variance-covariance matrix of the estimated treatments effects for the multi-arm studies only. This argument should be left unspecified if there are no multi-arm studies in the network.

covar

A vector of numbers referring to a continuous covariate that indicates a study characteristic or summary patient characteristic.

covar_assum

Character string indicating the structure of the treatment-by-covariate interaction, as described in Cooper et al. (2009) if interest also lies on the study percentage contributions to the estimated regression coefficients. Set covar_assumption equal to "no", "exchangeable", "independent", or "common". When covar_assum = "no", only the study percentage contributions to the summary treatment effects will be calculated. There is no default argument.

model

Character string indicating the analysis model with values "RE", or "FE" for the random-effects and fixed-effect model, respectively. There is no default argument.

tau

A scalar referring to the estimated between-study standard deviation obtained from network meta-analysis, if covar_assum = "no", or network meta-regression for a specific treatment-by-covariate interaction assumption. This argument should be left unspecified when model = "FE".

tau_beta

A scalar referring to the estimated standard deviation of the exchangeable regression coefficients obtained from network meta-regression with exchangeable treatment-by-covariate interaction. This argument should be left unspecified when covar_assum is not "exchangeable". There is no default argument.

Details

Note that the columns referring to the study percentage contributions to summary treatment effects are indicated by the letter 'd' with two numbers in decreasing order for the comparison: the first number refers to the comparator and the second number refers to the experimental treatment of the comparison. If interest lies also on the regression coefficients, the correspoding columns are indicated by 'beta'.

The function centers the covariate to the mean but presents the original version of the covariate.

Value

A list of the following two elements:

perc_contribute

A data-frame with four columns referring to the study name, baseline and experimental treatment arm, and the covariate and as many columns as the number of possible comparisons with the study percentage contributions to summary treatment effects referring to the basic parameters, and followed by the functional parameters. If interest lies also on the regression coefficients, extra columns appear referring to the study percentage contributions to the regression coefficients specified from the argument covar_assum.

covar_assumption

The estimated summary odd ratio in the logarithmic scale when measure = "RR" or measure = "RD".

Author(s)

Loukia M. Spineli

References

Cooper NJ, Sutton AJ, Morris D, Ades AE, Welton NJ. Addressing between-study heterogeneity and inconsistency in mixed treatment comparisons: Application to stroke prevention treatments in individuals with non-rheumatic atrial fibrillation. Stat Med 2009;28(14):1861–81. doi: 10.1002/sim.3594

Donegan S, Dias S, Tudur-Smith C, Marinho V, Welton NJ. Graphs of study contributions and covariate distributions for network meta-regression. Res Synth Methods 2018;9(2):243–60. doi: 10.1002/jrsm.1292

Examples

data("nma.malaria.donegan2018")

# Get study contributions to fixed-effect network meta-regression
# results under the assumption of independent treatment-by-covariate
# interaction
study_perc_contrib(study_name = nma.malaria.donegan2018$s,
                   base_t = nma.malaria.donegan2018$t1,
                   exp_t = nma.malaria.donegan2018$t2,
                   ref_t = 1,
                   obs_se = nma.malaria.donegan2018$se,
                   covar = nma.malaria.donegan2018$x,
                   covar_assum = "independent",
                   model = "FE")


[Package rnmamod version 0.5.0 Index]