Skip to content

Clinical Trial Simulation

NeoPKPD provides comprehensive clinical trial simulation capabilities for study design, power analysis, and decision-making.

Overview

Clinical trial simulation enables:

  • Virtual trial execution before enrollment
  • Power and sample size estimation
  • Study design optimization
  • Risk assessment and mitigation
graph TB
    A[Define Design] --> B[Generate Population]
    B --> C[Apply Dosing]
    C --> D[Simulate PK/PD]
    D --> E[Evaluate Endpoints]
    E --> F[Statistical Analysis]
    F --> G[Decision Metrics]

Trial Types

  • Parallel Design

    Independent treatment groups

    Parallel

  • Crossover Design

    Within-subject comparisons

    Crossover

  • Dose Escalation

    3+3, mTPI, CRM designs

    Escalation

  • Power Analysis

    Sample size and power estimation

    Power

Quick Start

Parallel Design

using NeoPKPD

# Define trial design
design = ParallelDesign(
    n_arms = 3,
    arm_names = ["Placebo", "Low Dose", "High Dose"],
    randomization_ratio = [1, 1, 1]
)

# Define dosing regimens
regimens = [
    DosingRegimen(0.0, 28, 24.0),      # Placebo
    DosingRegimen(50.0, 28, 24.0),     # 50 mg QD
    DosingRegimen(100.0, 28, 24.0)     # 100 mg QD
]

# Generate virtual population
population = generate_population(
    n = 150,                           # Total subjects
    spec = HealthyVolunteerSpec(),
    seed = 42
)

# Define trial specification
trial = TrialSpec(
    name = "Phase 2 Dose Finding",
    design = design,
    regimens = regimens,
    population = population,
    pk_model = OneCompIVBolus(),
    pk_params = OneCompIVBolusParams(5.0, 50.0),
    omega = OmegaMatrix([0.09 0.0; 0.0 0.04])
)

# Run simulation
result = simulate_trial(trial, seed = 12345)

# Analyze results
for arm in result.arms
    println("$(arm.name): n=$(arm.n_completed), mean_auc=$(arm.mean_auc)")
end

Crossover Design

# 2×2 crossover
design = CrossoverDesign(
    n_periods = 2,
    n_sequences = 2,
    sequences = [["A", "B"], ["B", "A"]],
    washout_days = 14
)

# Generate balanced population
population = generate_population(
    n = 24,                            # 12 per sequence
    spec = HealthyVolunteerSpec(),
    seed = 42
)

trial = TrialSpec(
    name = "BE Crossover",
    design = design,
    population = population,
    ...
)

result = simulate_trial(trial, seed = 12345)

Population Generation

Demographics

# Healthy volunteer specification
spec = DemographicSpec(
    age_mean = 35.0,
    age_sd = 10.0,
    age_min = 18.0,
    age_max = 55.0,
    weight_mean = 75.0,
    weight_sd = 12.0,
    weight_min = 50.0,
    weight_max = 100.0,
    female_fraction = 0.5,
    race_distribution = Dict(
        "white" => 0.7,
        "black" => 0.15,
        "asian" => 0.10,
        "other" => 0.05
    )
)

pop = generate_population(n = 100, spec = spec, seed = 42)

Patient Population

# Renal impairment study
spec = DemographicSpec(
    age_mean = 65.0,
    age_sd = 12.0,
    age_min = 18.0,
    age_max = 85.0,
    egfr_distribution = Dict(
        "normal" => 0.0,           # eGFR ≥ 90
        "mild" => 0.33,            # eGFR 60-89
        "moderate" => 0.34,        # eGFR 30-59
        "severe" => 0.33           # eGFR 15-29
    )
)

Dosing Regimens

# Once daily
regimen = dosing_qd(dose = 100.0, days = 28)

# Twice daily
regimen = dosing_bid(dose = 50.0, days = 14)

# Custom schedule
regimen = DosingRegimen(
    doses = [100.0, 100.0, 100.0],
    times = [0.0, 8.0, 16.0],
    repeat_days = 7
)

# Titration
regimen = TitrationRegimen(
    start_dose = 25.0,
    target_dose = 100.0,
    steps = [25, 50, 75, 100],
    days_per_step = 7
)

Power Analysis

Analytical Power

# Calculate power for given sample size
power = estimate_power(
    n_per_arm = 50,
    effect_size = 0.5,          # Cohen's d
    sd = 1.0,
    alpha = 0.05,
    test = :two_sample_t
)

println("Power: ", power.power)
println("Effect size: ", power.effect_size)

Sample Size Estimation

# Find n for target power
result = estimate_sample_size(
    target_power = 0.80,
    effect_size = 0.5,
    sd = 1.0,
    alpha = 0.05
)

println("Required n per arm: ", result.n_per_arm)
println("Achieved power: ", result.achieved_power)

Simulation-Based Power

# Power via simulation (more flexible)
power = estimate_power_simulation(
    trial_spec = trial,
    n_simulations = 1000,
    endpoint = :auc_comparison,
    success_criterion = auc_diff -> auc_diff > 0 && pvalue < 0.05,
    seed = 42
)

println("Simulated power: ", power.power)
println("95% CI: ", power.ci)

Statistical Analysis

Arm Comparison

# Compare treatment to control
comparison = compare_arms(
    treatment = result.arms["High Dose"].auc_values,
    control = result.arms["Placebo"].auc_values,
    test = :ttest
)

println("Difference: ", comparison.difference)
println("95% CI: ", comparison.ci)
println("p-value: ", comparison.pvalue)

Bioequivalence Analysis

# 90% CI for geometric mean ratio
be_result = bioequivalence_analysis(
    test = result.arms["Test"].auc_values,
    reference = result.arms["Reference"].auc_values
)

println("GMR: ", be_result.gmr)
println("90% CI: ", be_result.ci_90)
println("BE conclusion: ", be_result.is_bioequivalent)

Trial Result Structure

struct TrialResult
    # Design info
    name::String
    design::TrialDesign

    # Arm results
    arms::Dict{String, ArmResult}

    # Population summary
    n_enrolled::Int
    n_completed::Int
    dropout_rate::Float64

    # Timing
    enrollment_duration::Float64
    study_duration::Float64
end

struct ArmResult
    name::String
    n_enrolled::Int
    n_completed::Int

    # PK endpoints
    auc_values::Vector{Float64}
    cmax_values::Vector{Float64}

    # Summary statistics
    mean_auc::Float64
    sd_auc::Float64
    cv_auc::Float64
end

Next Steps