msb: Simulation of the Indirect Comparison Methods MAIC, STC, and the Bucher Method

Overview

The msb package provides tools to simulate and evaluate the performance of three indirect comparison methods:

• MAIC (Matching-Adjusted Indirect Comparison)
• STC (Simulated Treatment Comparison)
• Bucher Method

This package was developed as part of a university research project in the third year of my undergraduate degree, whilst also working full-time as a programmer at a pharmaceutical company.

Features

• Data Generation: Generate synthetic trial data.
• Simulation Implementation: Apply MAIC, STC, and Bucher methods to your data with the desired number of simulation runs.
• Analysis: Calculate performance metrics such as bias, standard errors, and coverage.
• Visualisation: Create plots and tables.

Adjustable Parameters:

• Sample Size - Examine how methods perform with varying trial sizes
• Population Overlap - Assess impact of different degrees of patient population similarity between studies
• Effect Modifier Strength - Evaluate sensitivity to the magnitude of treatment effect modification
• Adjustment Strategy - Compare full versus partial adjustment for effect modifiers to understand the trade-offs in bias reduction

The package enables exploration of how these four parameters influence the comparative performance of indirect comparison methods, allowing users to understand when each approach is most appropriate for their specific research context.

Installation

You can install msb from GitHub:

# Install from GitHub
devtools::install_github("lilyn09/msb")

Example Workflow

Select Parameters and Run Simulation

library(msb)

# Run Simulations with N = 100, overlap = 0.5 and 0.75
sim_1 <- do_simulation(N_sim = 100, 
                       data_overlap_param = 0.5)

sim_2 <- do_simulation(N_sim = 100, 
                       data_overlap_param = 0.75)

Analyse Simulation Data

analysis_1 <- analyse_simulations(sim_1)

analysis_2 <- analyse_simulations(sim_2) 

Generate Tables

format_table(analysis_1$shared_results)
format_table(analysis_2$shared_results)

Plot Bias

datasets <- list(analysis_1$shared_results, analysis_2$shared_results)
facet_bias_plot(datasets = datasets, 
                facet_names = c("Overlap = 0.5", "Overlap = 0.75"))

Plot Standard Errors

facet_se_plot(datasets = datasets, 
              facet_names = c("Overlap = 0.5", "Overlap = 0.75"))

Plot Coverage

datasets <- list(sim_1$results_df, sim_2$results_df)
facet_coverage_plot(datasets = datasets, 
                facet_names = c("Overlap = 0.5", "Overlap = 0.75"))

References

• Phillippo, D., S. Dias, A. Ades, and N. Welton (2020). Assessing the performance of population adjustment methods for anchored indirect comparisons: A simulation study. Statistics in Medicine 39, 4885–4911.
• Phillippo, D., A. Ades, S. Dias, S. Palmer, K. Abrams, and N. Welton (2016). NICE DSU Technical Support Document 18: Methods for population-adjusted indirect comparisons in submission to NICE. Technical support document, NICE Decision Support Unit.