- Analysis tools for enrichment design testing means, rates, and hazard ratios: function getAnalysisResults() generalized for enrichment designs; function getDataset() generalized for entering stratified data; manual extended for enrichment designs
- Automatic boundary recalculations during the trial for analysis with alpha spending approach, including under- and over-running: setup via the optional parameters ‘maxInformation’ and ‘informationEpsilon’ in function getAnalysisResults()
- The new function getObjectRCode (short: rcmd) returns the original R command which produced any rpact result object, including all dependencies
- getWideFormat() and getLongFormat() return a dataset object in wide format (unstacked) or long format (narrow, stacked)
- Generic function kable() returns the output of an rpact result object formatted in Markdown.
- Generic function t() returns the transpose of an rpact result object

- New argument ‘plotSettings’ added to all plot functions
- Summary for design, simulation, and analysis unified and extended
- Issue in getDesignFisher fixed: getDesignFisher(method = “noInteraction”, kMax = 3) and getDesignFisher(method = “noInteraction”) produced different results
- ‘normalApproximation’ default value changed to TRUE for multi-arm analysis of rates
- Repeated p-values: in search algorithm, upper bound of significance level corrected when considering binding futility bounds
- testPackage: the default call is now running only a small subset of all available unit tests; with the new argument ‘connection’ the owners of the rpact validation documentation can enter a and a to get full access to all unit tests
- Scaling of grid plots improved
- Minor improvements

- Beta-spending function approach with binding futility bounds
- Pampallona & Tsiatis design with binding and non-binding futility bounds
- Argument ‘accrualIntensityType’ added to getSampleSizeSurvival, getSimulationSurvival, getNumberOfSubjects, and getEventProbabilities
- Specification of Weibull survival times possible through definition of hazard rates or medians in simulation tool
- Minor improvements

- New utility functions getParameterCaption() and getParameterName() implemented
- Design parameters added to simulation print output
- Generic function as.matrix improved for several result objects
- Issue in getAvailablePlotTypes for sample size and power results fixed
- Issue for getDesignFisher(kMax = 1) in getSimulationMultiArm…() fixed
- getSimulationMultiArmSurvival: correlation of log-rank statistics revised and improved
- getSimulationMultiArmMeans: name of the first effectMeasure option “effectDifference” changed to “effectEstimate”
- getSimulation[MultiArm][Means/Rates/Survival]: argument ‘showStatistics’ now works correctly and is consistently FALSE by default for multi-arm and non-multi-arm
- getSimulation[MultiArm]Survival: generic function summary() improved
- getAnalysisResults: generic function summary() improved
- getAccrualTime: improved and new argument ‘accrualIntensityType’ added
- Header text added to design summaries
- getSampleSizeSurvival: field ‘studyDurationH1’ in result object was replaced by ‘studyDuration’, i.e., ‘studyDurationH1’ is deprecated and will be removed in future versions
- Minor changes in the inline help and manual
- Minor improvements

- getSimulationMultiArmSurvival: plannedEvents redefined as overall events over treatment arms
- getStageResults: element overallPooledStDevs added; print output improved
- Unit tests improved: test coverage and references to the functional specification optimized
- plot type 13 of getSampleSizeSurvival with user defined lambdas with different lengths: issue fixed
- Minor improvements

- Vignette “rpact: Getting Started” included into the package
- New summary output option “rpact.summary.width” added
- Generic function summary() improved for several result objects
- Result output of function testPackage() improved
- getSimulationMultiArm[Means/Rates/Survival]: stage index corrected for user defined calcSubjectsFunction or calcEventsFunction
- getSimulationMultiArmRates: adjustment for identical simulated rates to account for ties
- getSimulationMultiArmSurvival: corrected correlation of test statistics
- Output formatting improved
- Minor improvements

- Simulation tools for multi-arm design testing means, rates, and hazard ratios
- Analysis tools for multi-arm design testing means, rates, and hazard ratios
- getSimulationRates: exact versions for testing a rate (one-sample case) and equality of rates (two-sample case)
- getDataset: multi-arm datasets for means, rates, and survival data
- Analysis of fixed designs
- Summary for analysis and simulation result objects newly implemented
- Summary for most rpact result objects substantially improved and enhanced
- getEventProbabilities: plot of result object
- getNumberOfSubjects: plot of result object
- Visual comparison of two designs: plot(design1, design2)
- Functions setOutputFormat and getOutputFormat implemented: definition of user defined output formats
- getSimulationMeans: thetaH1 and stDevH1 can be specified for assessment of sample size recalculation (replaces thetaStandardized)
- getSimulationSurvival: separate p-values added to the aggregated simulation data for Fisher designs
- getSimulationMeans, getSimulationRates: Cumulated number of subjects integrated in getData object
- getSimulation[MultiArm][Means/Rates/Survival]: new logical argument ‘showStatistics’ added
- Example datasets (csv files) added to the package
- plot type “all”: plot all available plots of an object in one step using plot(x, type = “all”)
- plot type improved: ‘type’ now can be a vector, e.g., plot(x, type = c(1, 3))
- plot(x, grid = 1): new plot argument ‘grid’ enables the plotting of 2 or more plots in one graphic

- getAnalysisResults: list output implemented analogous to the output of all other rpact objects
- getAnalysisResults: the following stage result arguments were removed from result object because they were redundant: effectSizes, testStatistics, and pValues. Please use the ‘.stageResults’ object to access them, e.g., results$.stageResults$effectSizes
- getAnalysisResults: the following design arguments were removed from result object because they were redundant: stages, informationRates, criticalValues, futilityBounds, alphaSpent, and stageLevels. Please use the ‘.design’ object to access them, e.g., results$.design$informationRates
- Optional argument ‘stage’ removed from functions getConditionalPower, getConditionalRejectionProbabilities, getFinalPValue, getRepeatedPValues, and getTestActions
- Function testPackage improved, e.g., results will be displayed now on screen
- Help system renewed and approved, e.g., help for corresponding generic functions (e.g., plot) linked where applicable
- Function getPiecewiseSurvivalTime improved: pi1 and pi2 will not be calculated any longer for lambda- or median-based definitions; eventTime only required for pi-based definitions
- plot(x, showSource = TRUE) improved for all rpact result objects x
- Performance of plotting analysis results of Fisher designs improved
- getSimulationRates: issue for futility stopping for Fisher’s combination test fixed
- getSimulationSurvival: issue for expected number of events fixed
- getSimulationSurvival: if eventsNotAchieved > 0, rejection/futility rate and analysis time is estimated for valid simulation runs
- getSimulationSurvival: output improved for lambda1/median1/hazardRatio with length > 1
- getSampleSizeSurvival: calculation of the maximum number of subjects given the provided argument ‘followUpTime’ improved
- getPiecewiseSurvivalTime: delayed response via list-based piecewiseSurvivalTime definition enabled
- getAccrualTime/getSimulationSurvival: issue with the calculation of absolute accrual intensity by given relative accrual intensity fixed
- getRawData: issue for multiple pi1 solved
- Implementation of the generic function ‘names’ improved
- Test coverage improved: lots of new unit tests added
- License information in the DESCRIPTION file corrected: changed from GPL-3 to LGPL-3
- Minor improvements

- Boundaries on effect scale for testing means now accounts for the unknown variance case
- getAnalysisSurvival: calculation of stage wise results not more in getStageResults
- getStageResults: the calculation of ‘effectSizes’ for survival data and thetaH0 != 1 was corrected
- getDataset of survival data: issue with the internal storage of log ranks fixed
- Sample size plot: issue for kMax = 1 fixed
- getSampleSizeSurvival with piecewise survival time: issue with calculation of ‘maxNumberOfSubjects’ for given ‘followUpTime’ fixed
- Internal Shiny app interface improved
- Minor improvements

- Assumed median survival time: get[SampleSize/Power/Simulation]Survival now support direct input of arguments ‘median1’ and ‘median2’
- Output of generic function ‘summary’ improved
- Plot type 5 of getPower[…] and getSimulation[…] objects improved
- Output of getSampleSizeSurvival with given maxNumberOfSubjects improved
- Output of get[SampleSize/Power]Survival for Kappa != 1 improved
- Assert function for minNumberOfSubjectsPerStage corrected for undefined conditionalPower
- Two-sided boundaries on effect scale in survival design improved
- Error in ‘summary’ for getDesign[…] fixed
- Other minor improvements

- Incorrect output of function ‘summary’ fixed for getSampleSize[…] and getPower[…]
- as.data.frame: default value of argument ‘niceColumnNamesEnabled’ changed from TRUE to FALSE

- Plot function for Fisher design implemented
- Generic function ‘summary’ implemented for getDesign[…], getSampleSize[…], getPower[…], and getSimulation[…] results: a simple boundary summary will be displayed

- Generic function as.data.frame improved for getDesign[…], getSampleSize[…], getPower[…], and getSimulation[…] results
- Ouput of getStageResults() improved
- Improvements for Shiny app compatibility and better Shiny app performance
- Repeated p-values are no longer calculated for typeOfDesign = “WToptimum”
- Piecewise survival time improved for numeric definition: median and pi will not be calculated and displayed any longer
- Plot: legend title and tick mark positioning improved; optional arguments xlim and ylim implemented
- Sample size/power: usage of argument ‘twoSidedPower’ optimized
- Performance of function rpwexp/getPiecewiseExponentialRandomNumbers improved (special thanks to Marcel Wolbers for his example code)
- For group sequential designs a warning will be displayed if information rates from design not according to data information
- Format for output of standard deviation optimized

- Minor corrections in the inline help
- Labeling of lower and upper critical values (effect scale) reverted
- Simulation for Fisher’s combination test corrected
- Parameter minNumberOfAdditionalEventsPerStage renamed to minNumberOfEventsPerStage
- Parameter maxNumberOfAdditionalEventsPerStage renamed to maxNumberOfEventsPerStage
- Parameter minNumberOfAdditionalSubjectsPerStage renamed to minNumberOfSubjectsPerStage
- Parameter maxNumberOfAdditionalSubjectsPerStage renamed to maxNumberOfSubjectsPerStage
- Output of function getAccrualTime() improved
- Validation of arguments maxNumberOfIterations, allocation1, and allocation2 added: check for positive integer
- Function getSampleSizeSurvival improved: numeric search for accrualTime if followUpTime is given
- Default value improved for analysis tools: if no effect was specified for conditional power calculation, the observed effect is selected
- Fixed: function getDataset produced an error if only one log-rank value and one event was defined
- Number of subjects per treatment arm are provided in output of simulation survival if allocation ratio != 1
- Function getSimulationSurvival improved: first value of minNumberOfEventsPerStage and maxNumberOfEventsPerStage must be NA or equal to first value of plannedSubjects

- Function base::isFALSE replaced to guarantee R 3.4.x compatibility
- C++ compiler warning on r-devel-linux-x86_64-debian-clang system removed
- C++ compiler error on r-patched-solaris-x86 system fixed

- Power calculation at given or adapted sample size for means, rates and survival data
- Sample size and power calculation for survival trials with piecewise accrual time and intensity
- Sample size and power calculation for survival trials with exponential survival time, piecewise exponential survival time and survival times that follow a Weibull distribution
- Simulation tool for survival trials; our simulator is very fast because it was implemented with C++. Adaptive event number recalculations based on conditional power can be assessed
- Simulation tool for designs with continuous and binary endpoints. Adaptive sample size recalculations based on conditional power can be assessed
- Comprehensive and unified tool for performing sample size calculation for fixed sample size design
- Enhanced plot functionalities

- Fisher design, analysis of means or rates, conditional rejection probabilities (CRP): calculation issue fixed for stage > 2
- Call of getSampleSize[Means/Rates/Survival] without design argument implemented
- For all ‘set.seed’ calls ‘kind’ and ‘normal.kind’ were specified as follows: kind = “Mersenne-Twister”, normal.kind = “Inversion”
- Minor code optimizations, e.g. ‘return()’ replaced by ‘return(invisible())’ if reasonable
- Bug in ‘readDatasets’ fixed: variable names ‘group’ and ‘groups’ are now accepted
- “Overall reject per stage” and “Overall futility per stage” renamed to “Overall reject” and “Overall futility”, respectively (also variable names)
- Labels “events..” and “..patients..” consistently changed to “# events..” and “# patients…”, respectively
- Output format for ‘allocationRatioPlanned’ specified
- Method ‘show’ of class ‘ParameterSet’ expanded: R Markdown output features implemented
- getSampleSizeSurvival(): argument ‘maxNumberOfPatients’ was renamed in ‘maxNumberOfSubjects’
- Result output, inline help and documentation: the word ‘patient’ was replaced by ‘subject’
- Variables ‘numberOfSubjectsGroup1’ and ‘numberOfSubjectsGroup2’ were renamed to ‘numberOfSubjects1’ and ‘numberOfSubjects1’
- Final p-values for two-sided test (group sequential, inverse normal, and Fisher combination test) available
- Upper and lower boundaries on effect scale for testing rates in two samples

- First release of rpact