In borealexander/tslrt: Function for treatment switching logrank test

knitr::opts_chunk$set(echo = TRUE)

library(data.table)
library(survival)
library(ggplot2)
library(dplyr)
library(tslrt)
library(knitr)
library(kableExtra)

Introduction

This tutorial will show how to use the 'tslrt' package for Monte Carlo simulations. For this we have 2 different models of treatment switching; the first where treatment switching occurs after a specified delay time, and the second where treatment switching occurs after progression.

Switching after delay time

We start by setting the parameters used for the simulation

# number of patients in each arm
n_c <- 100
n_e <- 100

# median in control before switch
median_c_1 <- 10
# median in control after switch
median_c_2 <- 15
# median in experimental
median_e <- 15

# delay time where switching occurs
delay <- 3

# number of events to stop (event-driven study)
end_event <- 150

# recruitment period
rec_period <- 12
# recruitment parameter (assuming recruitment follows a powermodel)
rec_power <- 1

# different proportion of patients that switches
p <- c(0, 0.25, 0.5, 0.75, 1)

# significance level (one-sided)
alpha <- 0.025

# number of simulations
M <- 100

Now to simulate the results we use the function

res_delay <- MC_delay_crossover(n_c = n_c,
                                n_e = n_e,
                                median_c_1 = median_c_1,
                                median_c_2 = median_c_2,
                                median_e = median_e,
                                delay = delay,
                                end_event = end_event,
                                rec_period = rec_period,
                                rec_power = rec_power,
                                p = p, 
                                alpha = alpha,
                                M = M)

The result can be viewed both as a table

res_delay$result

and as a graph for the power

res_delay$plot

and the efficiency

res_delay$plot.eff

Switching after progression

For the model where we have treatment switching after progression we start by setting up the parameters

# number of patients in each arm
n_c <- 100
n_e <- 100

# median in control before switch
median_c <- 10
# median in experimental
median_e <- 15
# median for progressions
median_prog <- 3

# number of events to stop (event-driven study)
end_event <- 150

# recruitment period
rec_period <- 12
# recruitment parameter (assuming recruitment follows a powermodel)
rec_power <- 1

# different proportion of patients that switches
p <- c(0, 0.25, 0.5, 0.75, 1)

# significance level (one-sided)
alpha <- 0.025

# number of simulations
M <- 100

To run the simulations we use the function

res_prog <- MC_exp_prog_crossover(n_c = n_c,
                                  n_e = n_e,
                                  median_c = median_c,
                                  median_e = median_e,
                                  median_prog = median_prog,
                                  end_event = end_event,
                                  rec_period = rec_period,
                                  rec_power = rec_power,
                                  p = p, 
                                  alpha = alpha,
                                  M = M)

We can see the results in a table

res_prog$result

and as a graph for the power

res_prog$plot

and the efficiency

res_prog$plot.eff

Notes

Note that the results are only based on r sprintf("%.0f", M) simulations and for a real evalution the number of simulations would need to be much higher. This is just meant as a tutorial on showing how to use the package, not to give clear results.

borealexander/tslrt documentation built on March 26, 2020, 4:11 p.m.