New features implemented in rpact and RPACT Cloud

Friedrich Pahlke

October 6, 2025

New features implemented in rpact

rpact Package

rpact 4.2.0: New features

For the multi-arm simulation functions getSimulationMultiArmMeans(), getSimulationMultiArmRates(), and getSimulationMultiArmSurvival() it is now possible to specify a parameter doseLevels to define the dose levels for a linear or sigmoidEmax dose-response relationship (see feature request #63)

Example:

getDesignInverseNormal(informationRates = c(0.2, 0.6, 1)) |>
    getSimulationMultiArmRates(
        doseLevels = c(2.1, 3, 5.25, 9), # New!
        typeOfShape = "sigmoidEmax", # Define dose-response relationship
        gED50 = 2, # Dose with half of the maximum effect
        slope = 0.5, # Slope of the dose-response curve
        activeArms = 4,
        plannedSubjects = c(10, 30, 50),
        piControl = 0.1,
        piMaxVector = seq(0.35, 0.45, 0.1),
        intersectionTest = "Sidak",
        conditionalPower = 0.8,
        minNumberOfSubjectsPerStage = c(10, 4, 4),
        maxNumberOfSubjectsPerStage = c(10, 100, 100),
        maxNumberOfIterations = 10
    )

rpact 4.2.0: New features (cont’d)

Added support for unequal variances between two groups in getSampleSizeMeans(), getPowerMeans(), and getSimulationMeans() functions (see enhancement #70): You can specify unequal variances by providing the stDev argument as a vector of length 2, where each element represents the standard deviation of one group

Example:

getSampleSizeMeans(
    groups = 2,
    thetaH0 = 0,
    stDev = c(2, 2.2), # New: unequal standard deviations!
    normalApproximation = TRUE,
    alternative = 1.3,
    allocationRatioPlanned = 3
)

rpact 4.2.0: New features (cont’d)

Argument conservative added to getSampleSizeRates() function: Only relevant for one treatment group with normalApproximation = FALSE.
- If conservative = TRUE (default), the sample size is calculated so that any larger sample size guarantees a power greater than 1 - β.
- If conservative = FALSE, the minimum sample size for which
  power exceeds 1 - β is used (see enhancement #39)

rpact 4.2.0: New features (cont’d)

Example:

getSampleSizeRates(
    groups = 1,
    thetaH0 = 0.5,
    pi1 = 0.8,
    beta = 0.2,
    normalApproximation = FALSE,
    conservative = FALSE
)  |> fetch(maxNumberOfSubjects)

maxNumberOfSubjects 
                 20

getSampleSizeRates(
    groups = 1,
    thetaH0 = 0.5,
    pi1 = 0.8,
    beta = 0.2,
    normalApproximation = FALSE,
    conservative = TRUE
) |> fetch(maxNumberOfSubjects)

maxNumberOfSubjects 
                 23

rpact 4.2.0: New features (cont’d)

Save or reset all rpact options
- saveOptions() saves the current rpact options to a configuration file (located in the user’s configuration directory)
- resetOptions() resets the rpact options to their default values
- Note that the rappdirs package needs to be installed

rpact 4.2.0: New features (cont’d)

Enable futility boundaries in Boundaries p Values Scale plot plot (type = 3) using options("rpact.plot.show.futility.on.pvalue.scale" = TRUE) or argument showFutilityBounds = TRUE (see enhancement #79)

rpact 4.2.0: New features (cont’d)

Example:

options("rpact.plot.show.futility.on.pvalue.scale" = TRUE)
getDesignInverseNormal(
    informationRates = c(0.3333, 0.6667, 1),
    futilityBounds = c(-0.1, 0.2)
) |>
    getSampleSizeMeans(alternative = 0.2) |>
    plot(type = 3)

rpact 4.2.0: New features (cont’d)

Enable beta-spending in Error Spending plot (type = 4) using options("rpact.plot.show.beta.spent" = TRUE) or argument showBetaSpent = TRUE, see enhancement #80. Furthermore, options("rpact.plot.show.alpha.spent" = FALSE) or argument showAlphaSpent = FALSE can be used to show only beta-spending in the plot

rpact 4.2.0: New features (cont’d)

testPackage() produces a comprehensive installation qualification report in html and pdf format (see new vignette Installation Qualification of rpact or rpact.org/iq)

Examples:
- rpact_test_result_report.html
- rpact_test_result_report.pdf
setupPackageTests() downloads the test files and installs them into the rpact directory, providing access to the full unit test suite, which is now maintained in a private repository

rpact 4.2.1: New features

New parameters efficacyStops and futilityStops have been added to getDesignGroupSequential() and getDesignInverseNormal(), allowing users to flexibly enable or disable interim stops for efficacy and/or futility (see enhancement #88)

Example:

getDesignGroupSequential(
    kMax = 3,
    typeOfDesign = "asOF",
    typeBetaSpending = "bsP",
    futilityStops = c(TRUE, FALSE),
    efficacyStops = c(FALSE, TRUE)
)

rpact 4.2.1: New features

getAnalysisResults() and getFinalConfidenceInterval() now support the parameter stdErrorEstimate ("pooled" or "unpooled") to calculate final confidence intervals for two-sample rate comparisons

rpact 4.2.1: New features (cont’d)

Example:

design <- getDesignInverseNormal(kMax = 3, typeOfDesign = "OF")
dataset <- getDataset(
    events1 = c(23, 27, 14),
    events2 = c(14, 12, 8),
    n1 = c(40, 50, 33),
    n2 = c(41, 48, 26)
)
getFinalConfidenceInterval(
    design,
    dataset,
    thetaH0 = -0.1,
    stdErrorEstimate = "pooled"
)$finalConfidenceInterval

[1] 0.06340536 0.40022306

getFinalConfidenceInterval(
    design,
    dataset,
    thetaH0 = -0.1,
    stdErrorEstimate = "unpooled"
)$finalConfidenceInterval

[1] 0.05741957 0.38189911

rpact 4.2.1: New features (cont’d)

The new InstallationQualificationResult object returned by testPackage() introduces a default print output in a clear, detailed, and user-friendly format, making it easier to review qualification results

Optimal conditional error function design support

Developed by Morten Dreher

Pull Request Overview

This PR adds comprehensive support for optimal conditional error function designs to a statistical package. The optimal conditional error function approach allows for adaptive sample size re-estimation in clinical trials while maintaining strict type I error control.

Key changes include:

Implementation of a complete design framework for optimal conditional error functions with various likelihood ratio distributions
Addition of utility functions for calculating conditional errors, second-stage information, and overall power Comprehensive test suite covering different scenarios and edge cases

RPACT Cloud

RPACT Cloud: Highlights from Last Year

Over the past year, three major pharmaceutical companies have licensed RPACT Cloud for internal installation and use on their own servers. To support these enterprise deployments, we published extensive documentation covering setup on Posit Connect, configuration, and full qualification/validation:

How to Programmatically Clone and Deploy the RPACT Cloud Shiny App to Posit Connect
Deployment: RPACT Cloud with RStudio Workbench to Posit Connect
Installation: Running RPACT Cloud as an R Package
Guide to Environment Variables for RPACT Cloud
RPACT Cloud User Requirements Specification (URS)
RPACT Cloud Test Report \(\rightarrow\) The Test Report maps directly to the URS and documents the results of numerous shinytest2 unit tests, providing full transparency on validation outcomes

RPACT Cloud 2.6.1

Log Messages: New menu item to view and share server logs for debugging.
Custom Report Metadata: Specify user name and report name for generated reports.
Email Integration: Send reports as email attachments directly from the app.
CI/CD & Testing:
- Integrated shinytest2 unit tests.
- Added GitHub Actions workflows for package checks and tests.
Manual Table Updates: Tables now update only after clicking the Apply button.

RPACT Cloud 2.7.0

Installation Qualification & Quality Assurance
- Added User Requirements Specification (URS). Details: https://docs.cloud.rpact.com/articles/urs.html
- Perform installation qualification directly in the System Status section while the app is running.
- Download qualification reports as HTML or PDF.
- Enhanced UX with a progress spinner and real-time status updates.

RPACT Cloud 2.7.0 (cont’d)

Table Rendering Improvements
- New options in the project comparison UI to toggle rounding and set decimal precision via checkbox and select inputs.
Session Timeout Handler
- Introduced session timeout logic with an environment variable fallback.
- Display remaining session time in the footer.
Report Legend Position
- Added a drop-down menu to select the legend position of plots in generated reports.

RPACT Cloud 2.8.0

Comparison Tool
- Added a Comparison Tool Help page to the comparison tab
- Comparison tables are now generated with the flextable package
- Added filters for relevant parameters
Saving Designs
- Sessions (including all designs) can now be downloaded as a file and restored via an HTML link
- Sessions can be bookmarked in the browser, preserving all designs
- All designs are automatically restored after a session timeout (just reload the page, e.g. with the F5 key, when the app is greyed out)

RPACT Cloud 2.8.0 (cont’d)

Survival Design Simulation (Two Groups)
- Users can define the number of raw datasets per stage (maxNumberOfRawDatasetsPerStage)
- Raw data can be exported as an Excel file
Standard Deviation
- For two-armed trials, it is possible to specify the standard deviations separately, i.e., different standard deviations for group 1 and 2 can be defined.
Enhancements
- Improved performance by replacing all conditional panels with a server-side rendering solution
- Several design comparison settings have been moved to a central settings tab on the right for easier access and usability

Thank you!

Questions and Answers