metaRMST (version 1.0.0)

metaRMSTD: Meta-analysis of RMSTD at multiple time horizons

Description

Perform a meta-analysis with RMSTD using individual patient data. Methods include:

1. "mvma" a multivariate meta-analysis borrowing strength across time-points with within-trial covariance matrix derived analytically

2. "mvma_boot" a multivariate meta-analysis borrowing strength across time-points with within-trial covariance matrix derived by bootstrap

3. "uni" a univariate meta-analysis for combined effect at each time-point using only available data

4. "uni_flex" a univariate meta-analysis for combined effect at each time-point using estimates based on flexible parametric models as described by Wei et al (Stat Med 2015).

Usage

metaRMSTD(trialdata, time_horizons, MA_method, nboot = 500)

Arguments

trialdata

IPD trial data, see details for specifications

time_horizons

specified vector of time horizons for the meta-analysis

MA_method

the desired meta-analysis method; options are: "mvma", "mvma_boot", "uni", "uni_flex"

nboot

the number of bootstrap iterations, if using the MVMA with bootstrap covariance matrix; default=500

Value

The metaRMSTD function returns a list object containing the random-effects model results, the RMSTD and SE values for each trial at each available time horizon, and the estimated within-trial covariance matrix for each RCT.

Details

Specify the time horizons at which to calculate the meta-analytic results. The trialdata must be formatted as a dataframe containing the IPD for each single trial. Variable names must include Trial ID ("trialID"), Time ("Time"), Event status ("Event"), and randomization group ("Arm").

References

Wei, Y, Royston, P, Tierney, JF and Parmar, MKB. (2015). Meta-analysis of time-to-event outcomes from randomized trials using restricted mean survival time: application to individual participant data. Stat Med 34(21), 2881-2898.

Chen, Han, Alisa K. Manning, and Jos<U+00E9>e Dupuis. "A method of moments estimator for random effect multivariate meta-analysis." Biometrics 68.4 (2012): 1278-1284.

Jackson, Dan, Ian R. White, and Richard D. Riley. "A matrix-based method of moments for fitting the multivariate random effects model for meta-analysis and meta-regression." Biometrical Journal 55.2 (2013): 231-245.

Examples

# NOT RUN {
# read in built-in dataset 
data(AorticStenosisTrials)

# }
# NOT RUN {
# meta-analysis to obtain combined effect by multivariate model (method="mvma")
result <- metaRMSTD(AorticStenosisTrials, time_horizons=c(12,24,36), MA_method="mvma")

# generate figure: 
obj <- RMSTcurves(AorticStenosisTrials, time_horizons=c(12,24,36), tmax=40, nboot=500)
RMSTplot(obj, xlim=c(0,40), ylim=c(-0.25,2.75), yby=0.5, ylab="RMSTD (mos)", xlab="Time (mos)")
# }
# NOT RUN {
# }
# NOT RUN {
# }