sim_gs_nbinom: Simulate group sequential clinical trial for negative binomial outcomes

Description

Simulates multiple replicates of a group sequential clinical trial with negative binomial outcomes, performing interim analyses at specified calendar times.

Usage

sim_gs_nbinom(
  n_sims,
  enroll_rate,
  fail_rate,
  dropout_rate = NULL,
  max_followup,
  event_gap = 0,
  analysis_times = NULL,
  n_target = NULL,
  design = NULL,
  data_cut = cut_data_by_date,
  cuts = NULL
)

Value

A data frame containing simulation results for each analysis of each trial. Columns include:

sim: Simulation ID
analysis: Analysis index
analysis_time: Calendar time of analysis
n_enrolled: Number of subjects enrolled
n_ctrl: Number of subjects in control group
n_exp: Number of subjects in experimental group
events_total: Total events observed
events_ctrl: Events in control group
events_exp: Events in experimental group
exposure_at_risk_ctrl: Exposure at risk in control group (adjusted for event gaps)
exposure_at_risk_exp: Exposure at risk in experimental group (adjusted for event gaps)
exposure_total_ctrl: Total exposure in control group (calendar follow-up)
exposure_total_exp: Total exposure in experimental group (calendar follow-up)
z_stat: Z-statistic from the Wald test (positive favors experimental if rate ratio < 1)
estimate: Estimated log rate ratio from the model
se: Standard error of the estimate
method_used: Method used for inference ("nb" or "poisson")
dispersion: Estimated dispersion parameter from the model
blinded_info: Estimated blinded statistical information (ML)
unblinded_info: Observed unblinded statistical information (ML)
info_unblinded_ml: Observed unblinded statistical information (ML)
info_blinded_ml: Estimated blinded statistical information (ML)
info_unblinded_mom: Observed unblinded statistical information (Method of Moments)
info_blinded_mom: Estimated blinded statistical information (Method of Moments)

Arguments

n_sims: Number of simulations to run.
enroll_rate: Enrollment rates (data frame with rate and duration).
fail_rate: Failure rates (data frame with treatment, rate, dispersion).
dropout_rate: Dropout rates (data frame with treatment, rate, duration).
max_followup: Maximum follow-up time.
event_gap: Event gap duration.
analysis_times: Vector of calendar times for interim and final analyses. Optional if cuts is provided.
n_target: Total sample size to enroll (optional, if not defined by enroll_rate).
design: An object of class gsNB or sample_size_nbinom_result. Used to extract planning parameters (lambda1, lambda2, ratio) for blinded information estimation.
data_cut: Function to cut data for analysis. Defaults to cut_data_by_date(). The function must accept sim_data, cut_date, and event_gap as arguments.
cuts: A list of cutting criteria for each analysis. Each element of the list should be a list of arguments for get_cut_date() (e.g., planned_calendar, target_events, target_info). If provided, analysis_times is ignored (or used as a fallback if planned_calendar is missing in a cut).

Examples

Run this code

set.seed(123)
enroll_rate <- data.frame(rate = 10, duration = 3)
fail_rate <- data.frame(
  treatment = c("Control", "Experimental"),
  rate = c(0.6, 0.4),
  dispersion = 0.2
)
dropout_rate <- data.frame(
  treatment = c("Control", "Experimental"),
  rate = c(0.05, 0.05),
  duration = c(6, 6)
)
design <- sample_size_nbinom(
  lambda1 = 0.6, lambda2 = 0.4, dispersion = 0.2, power = 0.8,
  accrual_rate = enroll_rate$rate, accrual_duration = enroll_rate$duration,
  trial_duration = 6
)
cuts <- list(
  list(planned_calendar = 2),
  list(planned_calendar = 4)
)
sim_results <- sim_gs_nbinom(
  n_sims = 2,
  enroll_rate = enroll_rate,
  fail_rate = fail_rate,
  dropout_rate = dropout_rate,
  max_followup = 4,
  n_target = 30,
  design = design,
  cuts = cuts
)
head(sim_results)

Run the code above in your browser using DataLab