A function to simulate Survival data

This function simulates survival data with correlated time to progression and overall survival times. Optionally crossover from treatment arms can be simulated.

sim_adtte(
  rc_siminfo = FALSE,
  rc_origos = FALSE,
  id = 1,
  seed = 1234,
  rho = 0,
  pswitch = 0,
  proppd = 0,
  beta_1a = log(0.7),
  beta_1b = log(0.7),
  beta_2a = log(0.7),
  beta_2b = log(0.7),
  beta_pd = log(0.4),
  arm_n = 250,
  enroll_start = 0,
  enroll_end = 1,
  admin_censor = 2,
  os_gamma = 1.2,
  os_lambda = 0.3,
  ttp_gamma = 1.5,
  ttp_lambda = 2
)

Arguments

rc_siminfo

Should simulation params be included in simulated dataframe (logical). Defaults to FALSE.

rc_origos

Should OS without switching be included in simulated dataframe (logical). Defaults to FALSE.

id

Identifer added to simulated dataframe. Defaults to 1.

seed

Seed used for random number generator. Defaults to 1234.

rho

correlation coefficient between TTP and OS. Defaults to 0.

pswitch

proportion of patients who switch. Defaults to 0.

proppd

proportion of patients with PFS before switch allowed. Defaults to 0.

beta_1a

treatment effect (as log(Hazard Ratio)) for OS pre PFS. Defaults to log(0.7).

beta_1b

treatment effect (as log(Hazard Ratio)) for OS post PFS. Defaults to log(0.7).

beta_2a

treatment effect (as log(Hazard Ratio)) for OS pre PFS (switch). Defaults to log(0.7).

beta_2b

treatment effect (as log(Hazard Ratio)) for OS post PFS (switch). Defaults to log(0.7).

beta_pd

treatment effect on progression (as log(HR)). Defaults to log(0.4).

arm_n

patients per arm. Defaults to 250.

enroll_start

start of enrollment. Defaults to 0.

enroll_end

end of enrollment. Defaults to 1.

admin_censor

end of trial. Defaults to 2.

os_gamma

weibull shape - for OS. Defaults to 1.2.

os_lambda

weibull scale - for OS. Defaults to 0.3.

ttp_gamma

weibull shape - for TTP. Defaults to 1.5.

ttp_lambda

weibull scale - for TTP. Defaults to 2.

Details

The simulation times are derived from formulas in Austin 2012 adapted to enable correlations between endpoints. Austin, P.C. (2012), Generating survival times to simulate Cox proportional hazards models with time‐varying covariates. Statist. Med., 31: 3946-3958. https://doi.org/10.1002/sim.5452

Examples

require(survival)
require(dplyr)
#> Loading required package: dplyr
#> 
#> Attaching package: ‘dplyr’
#> The following objects are masked from ‘package:stats’:
#> 
#>     filter, lag
#> The following objects are masked from ‘package:base’:
#> 
#>     intersect, setdiff, setequal, union

ADTTE <- sim_adtte()
survfit(Surv(AVAL, event = CNSR ==  0) ~ ARM, data = filter(ADTTE, PARAMCD == "PFS")) %>%
  plot()


survfit(Surv(AVAL, event = CNSR ==  0) ~ ARM, data = filter(ADTTE, PARAMCD == "OS")) %>%
  plot()