Post-delay log hazard ratio that gives a target Cox estimand — time_to_event_delayed_log

Under a delayed treatment effect the hazard ratio is 1 for the first delay years and exp(beta) afterwards. The Cox model fitted to such a trial converges to an average of the two, weighted by when the events fall under the trial's censoring. This returns the beta for which that average equals log_hr, so that the delayed-effect scenario keeps the treatment effect - and hence the drift, the bias and the null hypothesis - of the proportional-hazards scenario it replaces. Under no effect beta is 0 and the two arms coincide.

Usage

time_to_event_delayed_log_hr(
  log_hr,
  control_parameter,
  event_time_distribution,
  weibull_shape,
  delay,
  accrual_period,
  final_follow_up,
  max_follow_up_time,
  dropout_rate
)

Arguments

log_hr: The Cox estimand, the scenario's target treatment effect.
control_parameter: Control rate (exponential) or scale (Weibull).
event_time_distribution: Either "exponential" or "weibull".
weibull_shape: The Weibull shape, unused for exponential times.
delay: Time before the treatment effect starts, in years.
accrual_period, final_follow_up, max_follow_up_time: The calendar design.
dropout_rate: Rate of loss to follow-up.

Value

The post-delay log hazard ratio, beta.