RMSTpowerBoost: Power and Sample Size Calculations for RMST-Based Trials

Overview

RMSTpowerBoost provides power and sample size tools for study designs that use restricted mean survival time (RMST) as a summary metric of time-to-event outcomes. The package supports covariate adjustment with analytical and simulation-based procedures for settings that include nonproportional hazards, stratification or multi-center effects, and dependent censoring.

The package includes both an R interface and a Shiny application for interactive use.

Key Features

• Linear IPCW-based RMST power and sample size calculations.
• Additive and multiplicative stratified models for multi-center or highly stratified studies.
• Bootstrap-based semiparametric GAM procedures for nonlinear covariate effects.
• Analytical and simulation-based methods for covariate-dependent censoring under a single censoring mechanism.

Unified Interface

The main package surface is organized around three top-level helpers:

• rmst.sim() generates pilot or reference survival data across the supported AFT and PH model families.
• rmst.power() computes a power curve from a Surv(time, status) ~ ... formula.
• rmst.ss() searches for the minimum sample size that reaches a target power.

library(RMSTpowerBoost)
library(survival)

set.seed(42)
pilot <- rmst.sim(
  n = 150,
  model = "aft_lognormal",
  baseline = list(mu = 2.2, sigma = 0.5),
  treat_effect = -0.3,
  covariates = list(covar_continuous("age"), covar_binary("female")),
  L = 12,
  seed = 42
)

power_fit <- rmst.power(
  Surv(time, status) ~ age + female,
  data = pilot,
  arm = "arm",
  sample_sizes = c(75, 100, 125),
  L = 12
)

rmst.power() and rmst.ss() route automatically based on a few key arguments:

• type = "analytical" or "boot" selects the analytical or bootstrap path.
• strata and strata_type activate additive or multiplicative stratified models.
• dep_cens = TRUE selects the dependent-censoring workflow.
• Smooth terms such as s(age) in the formula select the GAM bootstrap workflow.

seed in rmst.sim() is stored in the recipe metadata for provenance, but reproducible simulation still comes from calling set.seed() before rmst.sim().

Installation

Install the development version from GitHub:

install.packages("remotes")
remotes::install_github("UTHSC-Zhang/RMSTpowerBoost-Package")

Shiny App

Interactive web application:

Launch the Shiny App