Skip to contents

Restricted mean survival time and cause-specific life-years lost

Source: R/getRMST.R
getRMST.Rd

getRMST() turns a fitted "ConcreteEst" object into restricted mean survival time (RMST) and cause-specific life-years-lost (LYL) estimands, which are collapsible, clinically interpretable summaries that regulators increasingly prefer to a hazard ratio. Both are linear functionals of the cumulative-incidence curves that concrete already targets, so their influence functions are time-integrals of the per-subject influence functions of the absolute risks. The integral is taken over the target times the model was fit on, so request a reasonably dense TargetTime grid in formatArguments() for an accurate RMST.

  • RMST (event-free): \(\int_0^\tau S(t)\,dt\), the mean amount of follow-up time spent free of all events up to the horizon \(\tau\). Only returned when every event type in the data was targeted.

  • Life-years lost to cause \(j\): \(\int_0^\tau F_j(t)\,dt\), the mean time lost to cause \(j\) by \(\tau\).

Usage

getRMST(
  ConcreteEst,
  Horizon = NULL,
  Intervention = seq_along(ConcreteEst),
  Contrasts = TRUE,
  Signif = 0.05,
  NIMargin = NULL,
  NIDirection = c("lower", "upper")
)

Arguments

ConcreteEst

a "ConcreteEst" object returned by doConcrete().

Horizon

numeric: the restriction horizon \(\tau\). Defaults to the largest target time. Snapped to the nearest target time at or below it.

Intervention

numeric (default seq_along(ConcreteEst)): which interventions to summarize. For contrasts the first two are treated as treatment and control.

Contrasts

logical: also return RMST / LYL differences between the first two interventions.

Signif

numeric (default 0.05): alpha for two-sided confidence intervals and two-sided Wald p-values.

NIMargin

numeric (optional): a non-inferiority margin for the contrast estimands. When supplied, a one-sided non-inferiority assessment is added.

NIDirection

one of "lower" or "upper": which side of the margin is "non-inferior". Use "lower" when a larger value is better (e.g. an RMST difference) and "upper" when a smaller value is better.

Value

a data.table of class "ConcreteOut" with point estimates, influence-function standard errors, confidence intervals, and p-values. If the fit was built with Strata (see formatArguments()), the standard errors are corrected for the stratified / covariate-adaptive randomization design.

See also

targetRMST() for the directly-targeted version that fluctuates the hazards for the RMST estimating equation instead of integrating pointwise-targeted risks; getOutput() for absolute risks, risk differences, and risk ratios.