Clinical Scenario Evaluation Workflow

Overview

Comprehensive Clinical Scenario Evaluation (CSE) using the Mediana package to evaluate trial designs across multiple scenarios, analysis strategies, and success criteria.

Workflow Phases

Phase 1: Data Model Construction

<Task> subagent_type: cse-specialist prompt: | Build the Data Model: 1. Select appropriate outcome distribution: - Continuous: NormalDist - Binary: BinomDist - Survival: ExpoDist, WeibullDist - Count: PoissonDist, NegBinomDist - Multivariate: MVNormalDist, MVExpoDist, etc. 2. Define sample sizes or event counts to evaluate 3. Specify treatment effect scenarios: - Conservative (pessimistic) - Expected (primary assumption) - Optimistic 4. Configure design parameters (if event-driven): - Enrollment period and distribution - Study duration - Dropout rates 5. Build complete DataModel object </Task>

Phase 2: Analysis Model Construction

<Task> subagent_type: cse-specialist prompt: | Build the Analysis Model: 1. Define all statistical tests: - Match test method to endpoint type - Specify sample order correctly 2. Add descriptive statistics for monitoring 3. Configure multiplicity adjustments (if needed) </Task> <Task> subagent_type: multiplicity-expert condition: multiple_hypotheses == TRUE prompt: | Add multiplicity adjustment to Analysis Model: 1. Select appropriate procedure based on structure 2. Configure procedure parameters 3. Specify test ordering/families 4. Validate FWER control </Task>

Phase 3: Evaluation Model

<Task> subagent_type: cse-specialist prompt: | Build the Evaluation Model: 1. Define power criteria: - MarginalPower for individual tests - DisjunctivePower for "at least one" - ConjunctivePower for "all" - WeightedPower for prioritized 2. Add summary statistics (means, counts) 3. Ensure labels are clear for reporting </Task>

Phase 4: Simulation Execution

<Task> subagent_type: cse-specialist prompt: | Execute CSE and generate results: 1. Configure SimParameters: - n.sims (10000+ for stable estimates) - proc.load (parallel computation) - seed (reproducibility) 2. Run CSE() 3. Extract and summarize results 4. Generate Word report (optional) </Task>

Success Criteria

• All three models (Data, Analysis, Evaluation) complete
• Simulations run successfully
• Results interpretable and documented
• Report generated (if requested)

Final Deliverables

1. Complete R code for CSE
2. Results summary table
3. Word report (optional)
4. Scenario comparison visualization
5. Recommendations based on results

Configuration Options

• n_sims: Number of simulations (default: 10000)
• proc_load: "full", "high", "med", "low", or integer
• generate_report: TRUE/FALSE
• report_sections: Section organization for report
• scenarios: Named list of treatment effect scenarios
• sample_sizes: Vector of sample sizes to evaluate

Example Structure

# Data Model
data.model <- DataModel() +
  OutcomeDist(outcome.dist = "NormalDist") +
  SampleSize(seq(80, 120, 10)) +
  Sample(id = "Control", outcome.par = parameters(...)) +
  Sample(id = "Treatment", outcome.par = parameters(...))

# Analysis Model
analysis.model <- AnalysisModel() +
  Test(id = "Primary", samples = samples("Control", "Treatment"), method = "TTest") +
  MultAdjProc(proc = "HolmAdj")

# Evaluation Model
evaluation.model <- EvaluationModel() +
  Criterion(id = "Power", method = "MarginalPower",
            tests = tests("Primary"), par = parameters(alpha = 0.025))

# Run CSE
results <- CSE(data.model, analysis.model, evaluation.model,
               SimParameters(n.sims = 10000, proc.load = "full", seed = 12345))