Simulate Non-Proportional Hazards

A toolkit for simulation studies concerning time-to-event endpoints with non-proportional hazards. 'SimNPH' encompasses functions for simulating time-to-event data in various scenarios, simulating different trial designs like fixed-followup, event-driven, and group sequential designs. The package provides functions to calculate the true values of common summary statistics for the implemented scenarios and offers common analysis methods for time-to-event data. Helper functions for running simulations with the 'SimDesign' package and for aggregating and presenting the results are also included. Results of the conducted simulation study are available in the paper: "A Comparison of Statistical Methods for Time-To-Event Analyses in Randomized Controlled Trials Under Non-Proportional Hazards", Klinglmüller et al. (2025) .

SimNPH

Simulate Non Proportional Hazards

This package provides several functions to simulate survival data with non proportional hazards using the general purpose simulation package SimDesign.

This package follows the structure of SimDesing and provides functions that can readily be used as generate, analyse and summarise arguments in SimDesign's runSimulation function.

Usage

Installation

The current development version can be installed with:

remotes::install_git("https://github.com/SimNPH/SimNPH.git")

Getting Started

Documentation for all functions can be found in the respective help topics in the package after installation or here

Some examples of data generation, testing and estimation can be found in this vignette.

Results of the Simulation Study

The results of the simulation study done by the CONFIRMS consortium in the course of the EMA tender are published in the paper A Comparison of Statistical Methods for Time-To-Event Analyses in Randomized Controlled Trials Under Non-Proportional Hazards, Klinglmüller et al. (2025) and are presented in a shinylive App.

Reproducing the Results of the Simulation study

The parameters for the simulations were set using the following scripts:

• scripts/set_parameters_delayed_effect.R
• scripts/set_parameters_crossing_hazards.R
• scripts/set_parameters_subgroup.R
• scripts/set_parameters_progression.R

The simulations were then run using the following scripts:

• scripts/run_simulations_delayed_effect.R
• scripts/run_simulations_crossing_hazards.R
• scripts/run_simulations_subgroup.R
• scripts/run_simulations_progression.R

Which in turn use scripts/run_simulations_common.R.

All scripts are executed with the base directory of this repository as working directory.

Changes since the conduct of the simulation study:

• The version number of SimNPH was increased, to reproduce the simulation study, remove the check from run_simulations_...
• Simulation parameters and results were previously stored in the folder data. This folder was renamed to data_sim_study because data is reserved for the use in R packages. Change the paths accordingly.
• The run_simulations_... scripts use the parameters from a specified data. If you want to use parameter-values you generated yourself, update the path accordingly.

Re-Using the Scenarios

If you want to re-use the scenarios from the simulation study to investigate additional methods, use the scripts and documentation provided in the SimTemplate github repository. Also refer to the documentation on contributing. If you do so please cite the simulation study, you can get a bibtex entry with citation(package="SimNPH").