rpact

Confirmatory Adaptive Clinical Trial Design

Guide users through confirmatory adaptive clinical trial design and analysis using the rpact R package. Use this skill when the user asks about: adaptive designs, sample size reassessment, conditional power, inverse normal combination test, Fisher combination test, multi-stage designs, or rpact design objects.

Adaptive Clinical Trial Design with rpact

Note: This skill targets rpact >= 4.4.0 (CRAN). Dev version 4.4.0.9305 is on GitHub.

API reference

Full function docs: references/llms.txt (source: https://docs.rpact.org/reference/)
Workflow patterns: references/code_patterns.md

Key functions

Trial design

getDesignGroupSequential() - Group sequential design (O’Brien-Fleming, Pocock, alpha-spending, etc.)
getDesignInverseNormal() - Inverse normal combination test for adaptive designs
getDesignFisher() - Fisher combination test for adaptive designs
getDesignConditionalDunnett() - Conditional Dunnett test for multi-arm
getDesignCharacteristics() - Design characteristics and properties
getDesignSet() - Compare multiple designs
getFutilityBounds() - Convert futility bounds between scales (z-value, p-value, conditional power, predictive power, reverse conditional power, effect estimate) (new in 4.3.0)

Sample size

getSampleSizeMeans() - Continuous endpoints
getSampleSizeRates() - Binary endpoints
getSampleSizeSurvival() - Time-to-event endpoints
getSampleSizeCounts() - Count data endpoints

Power

getPowerMeans() - Continuous endpoints
getPowerRates() - Binary endpoints
getPowerSurvival() - Time-to-event endpoints
getPowerCounts() - Count data endpoints
getPowerAndAverageSampleNumber() - Power and ASN

Simulation

getSimulationMeans() / getSimulationRates() / getSimulationSurvival() / getSimulationCounts() - Two-arm
getSimulationMultiArmMeans() / getSimulationMultiArmRates() / getSimulationMultiArmSurvival() - Multi-arm
getSimulationEnrichmentMeans() / getSimulationEnrichmentRates() / getSimulationEnrichmentSurvival() - Enrichment

Analysis

getDataset() - Enter stage-wise observed data
getStageResults() - Compute stage-wise test statistics
getAnalysisResults() - Final adaptive analysis results
getClosedCombinationTestResults() - Closed testing for multi-arm
getClosedConditionalDunnettTestResults() - Dunnett closed testing
getConditionalPower() - Conditional power at interim
getConditionalRejectionProbabilities() - CRP for adaptive recalculation

Inference

getFinalPValue() - Adjusted final p-value
getFinalConfidenceInterval() - Adjusted confidence interval
getRepeatedPValues() - Repeated p-values across stages
getRepeatedConfidenceIntervals() - Repeated CIs across stages

Utilities

getAccrualTime() - Define accrual period
getPiecewiseSurvivalTime() - Piecewise exponential survival
getEventProbabilities() - Event probability computation
getNumberOfSubjects() - Expected accrual over time
getObservedInformationRates() - Observed information rates
getTestActions() - Decision at each stage
getPerformanceScore() - Design performance evaluation

Plotting

plot() methods for all design, power, simulation, and analysis objects
getAvailablePlotTypes() - List available plot types for an object

I/O

readDataset() / readDatasets() - Read data from CSV
writeDataset() / writeDatasets() - Write data to CSV
getObjectRCode() - Generate reproducible R code for any rpact object

Workflow patterns

For detailed code templates, read references/code_patterns.md.

Topics covered: - Group sequential designs (OF, Pocock, alpha-spending, HSD, custom futility) - Adaptive designs (inverse normal combination, Fisher combination) - Sample size for survival (HR, medians, event probabilities, piecewise) - Sample size for means and rates (one-sample, two-sample, risk ratio) - Power computation (survival, means, rates) - Simulation for survival (two-arm, piecewise/NPH, multiple scenarios) - Multi-arm designs and simulation (Dunnett, arm selection, dose-response) - Enrichment designs (subgroup-specific effects) - Entering observed data with getDataset() (survival, means, rates) - Stage results and analysis (getStageResults(), getAnalysisResults(), getTestActions()) - Conditional power and sample size reassessment - Adjusted inference (final p-values, CIs, repeated p-values/CIs) - Piecewise survival and accrual specification - Comparing designs with getDesignSet() - Plotting (boundaries, power curves, simulation results) - Reproducible code with getObjectRCode()

Important design considerations

typeOfDesign choices: “OF” (O’Brien-Fleming), “P” (Pocock), “asOF”/“asHSD”/“asKD” (alpha-spending variants). Alpha-spending versions (“as*“) allow unequal information rate spacing
getDesignInverseNormal() vs getDesignGroupSequential(): Same signature, but inverse normal is for adaptive designs where sample size can be recalculated between stages
bindingFutility = FALSE: Non-binding futility is standard for confirmatory trials; efficacy bounds computed ignoring futility
Survival parameterization: Can specify treatment effect via hazardRatio, median1/median2, lambda1/lambda2, or pi1/pi2 (event probabilities) — use only one
plannedEvents in simulation: Vector of cumulative event counts at each analysis (not information fractions)
Multi-arm intersectionTest: “Dunnett” (parametric, most powerful), “Simes” (non-parametric), “Bonferroni” (conservative)
typeOfSelection: Controls arm selection at interim — “best” (highest effect), “rBest” (r best arms), “epsilon” (within epsilon of best), “all” (no dropping)
getObjectRCode(): Use to generate reproducible R code for any rpact object — invaluable for reporting and validation
futilityBoundsScale (v4.3.0+): In getDesignInverseNormal() and getDesignGroupSequential(), specify futility bounds on alternative scales (p-value, conditional power, etc.) that are internally converted
alpha0Scale (v4.3.0+): In getDesignFisher(), specify alpha0 bounds on alternative scales
efficacyStops / futilityStops (v4.2.1+): Control which stages have efficacy/futility stopping
Dose-response (v4.2.0+): Multi-arm simulations support doseLevels for linear or sigmoid Emax models
Unequal variances (v4.2.0+): getSampleSizeMeans(), getPowerMeans(), getSimulationMeans() support different variances per group
rpact defaults: alpha = 0.05, beta = 0.2, kMax = NA (inferred), sided = 1 — always set explicitly for clinical trial designs

Code Patterns

Code Patterns for rpact

Note: These patterns target rpact >= 4.x (CRAN). See https://docs.rpact.org for full documentation.

Group sequential designs
Adaptive designs (inverse normal and Fisher)
Sample size for survival endpoints
Sample size for means and rates
Power computation
Simulation for survival
Multi-arm designs and simulation
Enrichment designs
Entering observed data with getDataset
Stage results and analysis
Conditional power and sample size reassessment
Adjusted inference (p-values and CIs)
Piecewise survival and accrual specification
Comparing designs with getDesignSet
Plotting
Reproducible code with getObjectRCode

Group sequential designs

getDesignGroupSequential() creates classical group sequential designs with spending functions or fixed boundary types.

O’Brien-Fleming design (default)

library(rpact)

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  typeOfDesign = "OF"
)
design

Pocock design

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  typeOfDesign = "P"
)

Alpha-spending (Lan-DeMets O’Brien-Fleming)

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  informationRates = c(0.5, 0.75, 1),
  typeOfDesign = "asOF"   # Alpha-spending OF approximation
)

Alpha-spending with Hwang-Shih-DeCani (gamma)

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  sided = 1,
  informationRates = c(0.5, 0.75, 1),
  typeOfDesign = "asHSD",
  gammaA = -4              # Gamma parameter (negative = conservative early)
)

Non-binding futility with beta-spending

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  informationRates = c(0.5, 0.75, 1),
  typeOfDesign = "asOF",
  typeBetaSpending = "bsOF",   # Beta-spending for futility
  bindingFutility = FALSE
)

Custom futility bounds

design <- getDesignGroupSequential(
  kMax = 3,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  informationRates = c(0.5, 0.75, 1),
  typeOfDesign = "asOF",
  futilityBounds = c(0, 0.5),  # Z-scale futility bounds at IA1, IA2
  bindingFutility = FALSE
)

Design characteristics

# Examine properties of any design
chars <- getDesignCharacteristics(design)
chars  # Shows power, ASN, stopping probabilities

Adaptive designs (inverse normal and Fisher)

Inverse normal combination test

# For adaptive designs where sample size can be recalculated mid-trial
design_in <- getDesignInverseNormal(
  kMax = 2,
  alpha = 0.025,
  beta = 0.2,
  sided = 1,
  typeOfDesign = "asOF"
)

Fisher combination test

design_fisher <- getDesignFisher(
  kMax = 2,
  alpha = 0.025,
  method = "equalAlpha",    # Equal alpha at each stage
  alpha0Vec = 0.5           # Futility bound (p-value scale)
)

Fisher with no interaction method

design_fisher <- getDesignFisher(
  kMax = 3,
  alpha = 0.025,
  method = "noInteraction",
  alpha0Vec = c(0.5, 0.5)
)

Sample size for survival endpoints

Basic survival design (hazard ratio)

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

ss <- getSampleSizeSurvival(
  design = design,
  hazardRatio = 0.7,
  accrualTime = c(0, 12),
  accrualIntensity = 30,      # 30 patients/month
  followUpTime = 12,
  dropoutRate1 = 0.025,
  dropoutRate2 = 0.025
)
ss

Using median survival times

ss <- getSampleSizeSurvival(
  design = design,
  median1 = 18,               # Experimental arm median
  median2 = 12,               # Control arm median
  accrualTime = c(0, 6, 12),
  accrualIntensity = c(15, 30),
  followUpTime = 18,
  dropoutRate1 = 0.025,
  dropoutRate2 = 0.025
)

Using event probabilities

ss <- getSampleSizeSurvival(
  design = design,
  pi1 = 0.35,                 # 12-month event rate, experimental
  pi2 = 0.50,                 # 12-month event rate, control
  eventTime = 12,             # Reference time for pi1, pi2
  accrualTime = c(0, 12),
  accrualIntensity = 30,
  followUpTime = 12
)

Piecewise exponential survival

ss <- getSampleSizeSurvival(
  design = design,
  piecewiseSurvivalTime = c(0, 6),
  lambda2 = c(0.04, 0.03),   # Control hazard rates per period
  hazardRatio = 0.7,
  accrualTime = c(0, 12),
  accrualIntensity = 30,
  followUpTime = 18
)

Fixed design (no interim)

ss_fixed <- getSampleSizeSurvival(
  hazardRatio = 0.7,
  accrualTime = c(0, 12),
  accrualIntensity = 30,
  followUpTime = 12
)

Sample size for means and rates

Continuous endpoint (two-sample)

design <- getDesignGroupSequential(
  kMax = 2, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

ss_means <- getSampleSizeMeans(
  design = design,
  groups = 2,
  alternative = 0.5,          # Mean difference
  stDev = 1,
  normalApproximation = TRUE
)
ss_means

One-sample mean

ss_one <- getSampleSizeMeans(
  design = design,
  groups = 1,
  alternative = 0.3,
  stDev = 1
)

Binary endpoint (two-sample)

ss_rates <- getSampleSizeRates(
  design = design,
  groups = 2,
  pi1 = 0.40,                 # Experimental response rate
  pi2 = 0.20,                 # Control response rate
  normalApproximation = TRUE
)
ss_rates

Binary endpoint with risk ratio

ss_rr <- getSampleSizeRates(
  design = design,
  groups = 2,
  riskRatio = TRUE,
  thetaH0 = 1,                # H0: RR = 1
  pi1 = 0.40,
  pi2 = 0.20
)

Power computation

Power for survival endpoint

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

pwr <- getPowerSurvival(
  design = design,
  hazardRatio = c(0.6, 0.7, 0.8),  # Multiple HR values
  maxNumberOfEvents = 300,
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  followUpTime = 12
)
pwr

Power for means

pwr_means <- getPowerMeans(
  design = design,
  alternative = seq(0.2, 0.8, 0.1),
  stDev = 1,
  maxNumberOfSubjects = 200
)

Power for rates

pwr_rates <- getPowerRates(
  design = design,
  pi1 = seq(0.3, 0.6, 0.05),
  pi2 = 0.2,
  maxNumberOfSubjects = 200
)

Simulation for survival

Two-arm survival simulation

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF",
  typeBetaSpending = "bsOF",
  bindingFutility = FALSE
)

sim <- getSimulationSurvival(
  design = design,
  hazardRatio = 0.7,
  plannedEvents = c(100, 200, 300),
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  followUpTime = 18,
  maxNumberOfIterations = 10000,
  seed = 12345
)
sim
summary(sim)

Simulation under multiple scenarios

sim <- getSimulationSurvival(
  design = design,
  hazardRatio = c(0.6, 0.7, 0.8, 1.0),  # Multiple HR values
  plannedEvents = c(100, 200, 300),
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  followUpTime = 18,
  maxNumberOfIterations = 10000,
  seed = 12345
)

Simulation with piecewise survival (delayed effect)

sim_nph <- getSimulationSurvival(
  design = design,
  piecewiseSurvivalTime = c(0, 6),
  lambda2 = c(0.04, 0.03),
  lambda1 = c(0.04, 0.03 * 0.6),   # HR = 1 early, 0.6 after 6 months
  plannedEvents = c(100, 200, 300),
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  maxNumberOfIterations = 10000,
  seed = 12345
)

Multi-arm designs and simulation

Multi-arm survival simulation (Dunnett)

design_in <- getDesignInverseNormal(
  kMax = 2,
  alpha = 0.025,
  sided = 1,
  typeOfDesign = "asOF"
)

sim_ma <- getSimulationMultiArmSurvival(
  design = design_in,
  activeArms = 3,
  plannedEvents = c(50, 100),
  intersectionTest = "Dunnett",
  typeOfSelection = "best",        # Select best arm at interim
  effectMeasure = "effectEstimate",
  successCriterion = "atLeastOne",
  hazardRatio = c(0.6, 0.7, 0.8),  # HR for each active arm
  maxNumberOfSubjects = 400,
  accrualTime = c(0, 12),
  accrualIntensity = 40,
  maxNumberOfIterations = 10000,
  seed = 12345
)
sim_ma

Multi-arm with dose-response (means)

sim_dose <- getSimulationMultiArmMeans(
  design = design_in,
  activeArms = 4,
  typeOfShape = "sigmoidEmax",
  muMaxVector = seq(0.2, 1, 0.2),   # Maximum effect sizes
  gED50 = 2,
  intersectionTest = "Dunnett",
  typeOfSelection = "best",
  stDev = 1,
  maxNumberOfSubjects = 300,
  maxNumberOfIterations = 10000
)

Multi-arm with arm dropping

sim_drop <- getSimulationMultiArmSurvival(
  design = design_in,
  activeArms = 3,
  plannedEvents = c(50, 100),
  intersectionTest = "Simes",
  typeOfSelection = "epsilon",     # Drop if within epsilon of best
  epsilonValue = 0.1,
  hazardRatio = c(0.6, 0.7, 0.8),
  maxNumberOfSubjects = 400,
  accrualTime = c(0, 12),
  accrualIntensity = 40,
  maxNumberOfIterations = 10000
)

Enrichment designs

Enrichment survival simulation

design_in <- getDesignInverseNormal(
  kMax = 2,
  alpha = 0.025,
  sided = 1,
  typeOfDesign = "asOF"
)

# Effect sizes: rows = scenarios, columns = populations (S, F\S)
effectList <- list(
  subGroups = c("S", "R"),         # Subgroup and complement
  prevalences = c(0.4, 0.6),
  effects = matrix(c(
    0.7, 1.0,     # Scenario 1: effect only in subgroup
    0.7, 0.8,     # Scenario 2: larger effect in subgroup
    0.7, 0.7      # Scenario 3: equal effect
  ), ncol = 2, byrow = TRUE)
)

sim_enrich <- getSimulationEnrichmentSurvival(
  design = design_in,
  effectList = effectList,
  intersectionTest = "Simes",
  typeOfSelection = "best",
  successCriterion = "atLeastOne",
  plannedEvents = c(80, 160),
  maxNumberOfSubjects = 400,
  maxNumberOfIterations = 10000,
  seed = 12345
)

Entering observed data with getDataset

Survival data (cumulative events and logrank statistics)

# Stage-wise observed data
data_surv <- getDataset(
  cumulativeEvents = c(50, 100, 150),
  cumulativeLogRanks = c(1.2, 1.8, 2.3)
)

Means data (two-arm)

data_means <- getDataset(
  n1 = c(30, 30),             # Stage-wise sample sizes, arm 1
  n2 = c(30, 30),             # Stage-wise sample sizes, arm 2
  means1 = c(1.5, 1.8),      # Stage-wise means, arm 1
  means2 = c(1.0, 1.1),      # Stage-wise means, arm 2
  stDevs1 = c(2.0, 2.1),     # Stage-wise SDs, arm 1
  stDevs2 = c(1.9, 2.0)      # Stage-wise SDs, arm 2
)

Rates data (two-arm)

data_rates <- getDataset(
  n1 = c(50, 50),
  n2 = c(50, 50),
  events1 = c(20, 25),
  events2 = c(10, 15)
)

Reading from CSV

data_csv <- readDataset("trial_data.csv")

Stage results and analysis

Computing stage results

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

data <- getDataset(
  cumulativeEvents = c(50, 100),
  cumulativeLogRanks = c(1.5, 2.1)
)

stageResults <- getStageResults(design, data)
stageResults

Full analysis results

results <- getAnalysisResults(
  design = design,
  dataInput = data,
  directionUpper = FALSE        # Lower is better (HR < 1)
)
results

Checking test actions

# What decision at each stage?
getTestActions(results)
# Returns: "continue", "reject", or "accept" (futility)

Conditional power and sample size reassessment

Conditional power at interim

design <- getDesignInverseNormal(
  kMax = 2, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

data <- getDataset(
  n1 = c(50),
  n2 = c(50),
  means1 = c(1.5),
  means2 = c(1.0),
  stDevs1 = c(2.0),
  stDevs2 = c(1.9)
)

stageResults <- getStageResults(design, data)

# Conditional power for planned n
cp <- getConditionalPower(
  stageResults,
  nPlanned = c(100)           # Planned total n for stage 2
)
cp

Conditional rejection probabilities (for SSR)

crp <- getConditionalRejectionProbabilities(stageResults)
crp

# Sample size reassessment: increase stage 2 n to achieve target CP
# Use conditional power to determine new n

Adjusted inference (p-values and CIs)

Final p-value

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

data <- getDataset(
  n1 = c(30, 30, 30),
  n2 = c(30, 30, 30),
  means1 = c(1.5, 1.8, 2.0),
  means2 = c(1.0, 1.1, 1.0),
  stDevs1 = c(2.0, 2.1, 1.9),
  stDevs2 = c(1.9, 2.0, 2.0)
)

stageResults <- getStageResults(design, data)

# Stage-wise adjusted p-value
pval <- getFinalPValue(stageResults)
pval

Final confidence interval

ci <- getFinalConfidenceInterval(
  design = design,
  dataInput = data,
  directionUpper = TRUE
)
ci

Repeated p-values and confidence intervals

results <- getAnalysisResults(design, data, directionUpper = TRUE)

# Repeated p-values (available at each stage)
results$repeatedPValues

# Repeated CIs
results$repeatedConfidenceIntervalLowerBounds
results$repeatedConfidenceIntervalUpperBounds

Piecewise survival and accrual specification

Accrual time with ramp-up

# Vector specification
accrual <- getAccrualTime(
  accrualTime = c(0, 3, 6, 12),
  accrualIntensity = c(10, 20, 30),   # Patients/month per period
  maxNumberOfSubjects = 500
)

# List specification (more readable)
accrual <- getAccrualTime(
  list(
    "0 - <3"  = 10,
    "3 - <6"  = 20,
    "6 - 12"  = 30
  ),
  maxNumberOfSubjects = 500
)

Piecewise exponential survival

# Delayed treatment effect
pw_surv <- getPiecewiseSurvivalTime(
  piecewiseSurvivalTime = c(0, 6),
  lambda2 = c(0.04, 0.03),            # Control hazard
  lambda1 = c(0.04, 0.03 * 0.6)       # Experimental hazard (HR = 1, then 0.6)
)

# Using hazard ratio (applied to all periods)
pw_surv <- getPiecewiseSurvivalTime(
  piecewiseSurvivalTime = c(0, 6),
  lambda2 = c(0.04, 0.03),
  hazardRatio = 0.7
)

# List specification
pw_surv <- getPiecewiseSurvivalTime(
  list(
    "0 - <6"  = 0.04,
    ">=6"     = 0.03
  ),
  hazardRatio = 0.7
)

Using piecewise objects in sample size

ss <- getSampleSizeSurvival(
  design = getDesignGroupSequential(kMax = 3, alpha = 0.025, typeOfDesign = "asOF"),
  piecewiseSurvivalTime = c(0, 6),
  lambda2 = c(0.04, 0.03),
  hazardRatio = 0.7,
  accrualTime = c(0, 3, 6, 12),
  accrualIntensity = c(10, 20, 30),
  followUpTime = 18
)

Comparing designs with getDesignSet

Varying a single parameter

# Compare O'Brien-Fleming, Pocock, and HSD designs
ds <- getDesignSet(
  design = getDesignGroupSequential(kMax = 3, alpha = 0.025, sided = 1),
  typeOfDesign = c("OF", "P", "asHSD")
)
plot(ds)

Comparing gamma values

ds <- getDesignSet(
  design = getDesignGroupSequential(
    kMax = 3, alpha = 0.025, sided = 1,
    typeOfDesign = "asHSD"
  ),
  gammaA = c(-4, -2, 0, 2)
)
plot(ds)

From pre-built designs

d1 <- getDesignGroupSequential(kMax = 3, alpha = 0.025, typeOfDesign = "OF")
d2 <- getDesignGroupSequential(kMax = 3, alpha = 0.025, typeOfDesign = "P")
d3 <- getDesignGroupSequential(kMax = 3, alpha = 0.025, typeOfDesign = "asOF")
ds <- getDesignSet(designs = c(d1, d2, d3))
plot(ds)

Plotting

Design boundaries

design <- getDesignGroupSequential(
  kMax = 4, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF",
  typeBetaSpending = "bsOF",
  bindingFutility = FALSE
)

# Default: boundary plot
plot(design)

# Specific plot type
plot(design, type = 1)   # Boundaries (Z-scale)
plot(design, type = 3)   # Stage levels (adjusted significance levels)
plot(design, type = 4)   # Error spending
plot(design, type = 5)   # Power and early stopping

Sample size results

ss <- getSampleSizeSurvival(
  design = design,
  hazardRatio = seq(0.5, 0.9, 0.05),
  accrualTime = c(0, 12),
  accrualIntensity = 30,
  followUpTime = 12
)
plot(ss)

Power curves

pwr <- getPowerSurvival(
  design = design,
  hazardRatio = seq(0.5, 1, 0.05),
  maxNumberOfEvents = 300,
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  followUpTime = 12
)
plot(pwr)

Simulation results

sim <- getSimulationSurvival(
  design = design,
  hazardRatio = c(0.6, 0.7, 0.8, 1.0),
  plannedEvents = c(75, 150, 225, 300),
  maxNumberOfSubjects = 500,
  accrualTime = c(0, 12),
  accrualIntensity = 42,
  maxNumberOfIterations = 10000
)
plot(sim)

Available plot types

# Check what plot types are available for any object
getAvailablePlotTypes(design)
getAvailablePlotTypes(ss)

Reproducible code with getObjectRCode

design <- getDesignGroupSequential(
  kMax = 3, alpha = 0.025, sided = 1,
  typeOfDesign = "asOF"
)

# Print reproducible R code
getObjectRCode(design, output = "cat")

# Include default parameters too
getObjectRCode(design, includeDefaultParameters = TRUE, output = "cat")

# Get as character vector
code <- getObjectRCode(design, output = "vector")

# Markdown output (for reports)
getObjectRCode(design, output = "markdown")

Adaptive Clinical Trial Design with rpact

API reference

Key functions

Trial design

Sample size

Power

Simulation

Analysis

Inference

Utilities

Plotting

I/O

Workflow patterns

Important design considerations

Code Patterns

Code Patterns for rpact

Table of Contents

Group sequential designs

O’Brien-Fleming design (default)

Pocock design

Alpha-spending (Lan-DeMets O’Brien-Fleming)

Alpha-spending with Hwang-Shih-DeCani (gamma)

Non-binding futility with beta-spending

Custom futility bounds

Design characteristics

Adaptive designs (inverse normal and Fisher)

Inverse normal combination test

Fisher combination test

Fisher with no interaction method

Sample size for survival endpoints

Basic survival design (hazard ratio)

Using median survival times

Using event probabilities

Piecewise exponential survival

Fixed design (no interim)

Sample size for means and rates

Continuous endpoint (two-sample)

One-sample mean

Binary endpoint (two-sample)

Binary endpoint with risk ratio

Power computation

Power for survival endpoint

Power for means

Power for rates

Simulation for survival

Two-arm survival simulation

Simulation under multiple scenarios

Simulation with piecewise survival (delayed effect)

Multi-arm designs and simulation

Multi-arm survival simulation (Dunnett)

Multi-arm with dose-response (means)

Multi-arm with arm dropping

Enrichment designs

Enrichment survival simulation

Entering observed data with getDataset

Survival data (cumulative events and logrank statistics)

Means data (two-arm)

Rates data (two-arm)

Reading from CSV

Stage results and analysis

Computing stage results

Full analysis results

Checking test actions

Conditional power and sample size reassessment

Conditional power at interim

Conditional rejection probabilities (for SSR)

Adjusted inference (p-values and CIs)

Final p-value

Final confidence interval

Repeated p-values and confidence intervals

Piecewise survival and accrual specification

Accrual time with ramp-up

Piecewise exponential survival

Using piecewise objects in sample size

Comparing designs with getDesignSet

Varying a single parameter

Comparing gamma values

From pre-built designs

Plotting

Design boundaries