Skip to contents

Simulate time-to-event (survival) outcomes

Source: R/simulate_survival.R
simulate_survival.Rd

This function simulates a survival time (e.g., time to event or censoring) for each patient, with an option to specify treatment effect on hazard. It assumes an exponential survival distribution for simplicity.

Usage

simulate_survival(data, median_time_control, hr = 1, study_duration = NULL)

Arguments

data

A data frame of patients that includes a treatment group column (typically Arm from simulate_allocation). If no Arm column is present or only one level, all patients are treated as a single group.

median_time_control

Numeric, median event time for the control group (e.g., median survival in control arm).

hr

Numeric or named numeric vector specifying hazard ratios for treatment arms relative to control. If a single numeric is given (default = 1), and exactly two arms exist, it is interpreted as the hazard ratio for the treatment arm vs control. If multiple arms or a named vector is provided, each arm's hazard is adjusted accordingly (the control arm should have hazard ratio 1).

study_duration

Numeric, maximum follow-up time (e.g., study end in time units). If provided, events beyond this time are censored at study_duration. (Default = NULL, meaning no administrative censoring apart from any later dropout handling.)

Value

The input data frame with two new columns:

  • EventTime: Simulated event time or censoring time.

  • EventStatus: Event occurrence indicator (1 = event occurred, 0 = censored by end of follow-up).

Details

Exponential survival times are generated. The control arm event rate is derived from median_time_control (rate = log(2)/median). Hazard ratios < 1 indicate longer survival in treatment arms (reduced hazard). For multiple arms, provide hazard ratios as a named vector (names matching Arm levels; the control arm can be omitted or set to 1). If no Arm column is present in data, all patients are assumed to share the same baseline hazard.

Examples

df <- simulate_patients(100)
df <- simulate_allocation(df)
# Simulate survival with median 12 months in control, and hazard ratio 0.75 for treatment
df <- simulate_survival(df, median_time_control = 12, hr = 0.75, study_duration = 24)
head(df[c("Arm","EventTime","EventStatus")])
#>         Arm EventTime EventStatus
#> 1   Control  11.67065           1
#> 2   Control  24.00000           0
#> 3   Control  24.00000           0
#> 4   Control  10.11397           1
#> 5 Treatment  24.00000           0
#> 6 Treatment  24.00000           0