Trialsim

TrialSim

Status: Active Development

TrialSim generates realistic synthetic clinical trial data for testing, training, and development purposes.

For Claude

Use this skill when the user requests clinical trial data, CDISC-compliant datasets, or regulatory submission test data. This is the primary skill for generating realistic synthetic clinical trial data.

When to apply this skill:

• User mentions clinical trials, studies, or protocols
• User requests CDISC, SDTM, or ADaM datasets
• User specifies trial phases (Phase I, II, III, IV)
• User mentions FDA/EMA submission data or regulatory requirements
• User asks for adverse events, safety data, or efficacy endpoints
• User mentions specific therapeutic areas (oncology, cardiovascular, CNS)
• User requests SDTM domains (DM, AE, VS, LB, CM, EX, DS, MH)

Key capabilities:

• Generate complete study definitions with protocol parameters
• Create multi-site, multi-country trial configurations
• Produce subject-level longitudinal data with realistic patterns
• Generate safety data (adverse events, labs, vitals) with MedDRA/LOINC coding
• Create efficacy endpoints for various therapeutic areas
• Output CDISC-compliant formats (SDTM, ADaM)

For specific trial phases, therapeutic areas, or SDTM domains, load the appropriate skill from the tables below.

Safety Guardrails

All data generated by TrialSim is synthetic and fictional. It must never be used for real clinical decisions, patient care, or regulatory submissions without explicit disclaimers.

• Synthetic data only: All subjects, sites, adverse events, and results are generated/simulated. Always confirm this when presenting output.
• No clinical advice: Never recommend treatments, prescribe medications, or interpret safety signals as if data were real. If asked "should this patient receive X?", remind the user this is synthetic test data.
• Real codes, synthetic entities: Use real standard code systems (ICD-10, CPT, LOINC, SNOMED CT, MedDRA, RxNorm, NDC, NPI, ATC, HCPCS) for coding accuracy, but all people, sites, and events are fictional.
• Do NOT generate: Real patient identifiers, actual investigator names, real site addresses, or any data that could be confused with actual clinical trial records.

Edge cases: If a user requests invalid visit windows, missing baseline assessments, or partial SDTM domains, flag the issue and suggest corrections. For unknown MedDRA terms, use the closest valid preferred term.

Trigger Phrases

Activate TrialSim when user mentions:

• "clinical trial" or "clinical study"
• "Phase I/II/III/IV" or "pivotal trial"
• "CDISC", "SDTM", "ADaM"
• "FDA submission data" or "regulatory data"
• "adverse events" or "safety data"
• "efficacy endpoints"
• Trial therapeutic areas (oncology, cardiology, etc.)
• SDTM domains (DM, AE, VS, LB, CM, EX, DS)

Quick Links

Core Skills

Trial Phase Skills

SDTM Domain Skills

Therapeutic Areas

Real World Evidence

Output Formats

Data Models & References

Core Entities

TrialSim uses 15 canonical entity schemas. See Data Models Reference for complete JSON schemas.

Entity Overview

Key Entity Examples

Study:

{
  "study_id": "ABC-123-001",
  "protocol_title": "A Phase 3, Randomized, Double-Blind Study...",
  "phase": "Phase 3",
  "therapeutic_area": "Oncology",
  "indication": "Non-Small Cell Lung Cancer",
  "sponsor": "Example Pharma Inc.",
  "status": "Ongoing"
}

Subject (with cross-product linking):

{
  "subject_id": "0001",
  "usubjid": "ABC-123-001-001-0001",
  "site_id": "001",
  "patient_ref": "MRN-12345",
  "screening_date": "2024-01-15",
  "randomization_date": "2024-01-22",
  "treatment_arm": "TRT",
  "status": "Active"
}

Integration with Other Products

TrialSim integrates with other HealthSim products for complete clinical trial data:

Cross-Product: PatientSim

Trial subjects are patients with additional trial-specific data:

• ../patientsim/oncology/ - Oncology trial subjects
• ../patientsim/heart-failure.md - CV outcomes trial subjects
• ../patientsim/behavioral-health.md - CNS trial subjects
• ../patientsim/diabetes-management.md - Metabolic trial subjects

Integration Pattern: Use PatientSim for baseline clinical characteristics. TrialSim adds protocol-specific assessments (RECIST, NYHA class changes), randomization, and SDTM-formatted data.

Cross-Product: PopulationSim Integration

PopulationSim provides real-world reference data (CDC PLACES, SVI, ADI) for evidence-based trial planning, site selection, and FDA diversity compliance. When geographies are specified, TrialSim grounds feasibility estimates and enrollment projections in actual prevalence and demographic data.

See populationsim-integration.md for detailed data-driven planning patterns, reference data sources, and site selection examples.

Development Status

All skill files complete: core skills, Phase 1-3, SDTM domains (DM, AE, VS, LB, CM, EX, DS, MH), therapeutic areas, and RWE.

Related Skills

• PatientSim - Clinical patient data
• MemberSim - Claims integration
• Code Systems - Standard terminologies (ICD-10, CPT, LOINC, SNOMED, RxNorm, NDC, HCPCS)

Usage Examples

Example 1: Basic Phase 3 Trial

Prompt: "Generate a Phase 3 oncology trial with 200 subjects"

Output:

{
  "study": {
    "study_id": "ONCO-2025-001",
    "protocol_title": "A Phase 3, Randomized, Double-Blind, Placebo-Controlled Study of ABC-123 in Advanced NSCLC",
    "phase": "Phase 3",
    "therapeutic_area": "Oncology",
    "indication": "Non-Small Cell Lung Cancer",
    "design": "Randomized, Double-Blind, Placebo-Controlled",
    "primary_endpoint": "Overall Survival",
    "target_enrollment": 200,
    "arms": [
      { "arm_id": "A", "name": "ABC-123 + SOC", "allocation_ratio": 1 },
      { "arm_id": "B", "name": "Placebo + SOC", "allocation_ratio": 1 }
    ]
  },
  "sites": [
    { "site_id": "001", "name": "Memorial Cancer Center", "country": "USA", "enrolled": 25 },
    { "site_id": "002", "name": "University Oncology", "country": "USA", "enrolled": 22 }
  ],
  "subjects": [
    {
      "subject_id": "001-0001",
      "site_id": "001",
      "arm": "A",
      "status": "Active",
      "randomization_date": "2025-01-15"
    }
  ]
}

Example 2: Adverse Events with MedDRA

Prompt: "Generate adverse events for a 50-subject immunotherapy trial"

Output:

{
  "domain": "AE",
  "adverse_events": [
    {
      "USUBJID": "IO-001-0023",
      "AESEQ": 1,
      "AETERM": "Fatigue",
      "AEDECOD": "Fatigue",
      "AEBODSYS": "General disorders and administration site conditions",
      "AESEV": "MILD",
      "AESER": "N",
      "AEREL": "POSSIBLY RELATED",
      "AESTDTC": "2025-02-10",
      "AEENDTC": "2025-02-18",
      "AEOUT": "RECOVERED/RESOLVED"
    },
    {
      "USUBJID": "IO-001-0007",
      "AESEQ": 1,
      "AETERM": "Immune-mediated colitis",
      "AEDECOD": "Colitis",
      "AEBODSYS": "Gastrointestinal disorders",
      "AESEV": "SEVERE",
      "AESER": "Y",
      "AESHOSP": "Y",
      "AEREL": "RELATED",
      "AEACN": "DRUG INTERRUPTED",
      "AESTDTC": "2025-03-05",
      "AEOUT": "NOT RECOVERED/NOT RESOLVED"
    }
  ]
}

Example 3: Screening and Enrollment

Prompt: "Generate screening data with 30% screen failure rate for 100 subjects"

Output:

{
  "screening_summary": {
    "total_screened": 143,
    "screen_passed": 100,
    "screen_failed": 43,
    "screen_failure_rate": 0.30
  },
  "screen_failures": [
    {
      "screening_id": "SCR-0015",
      "failure_reason": "Inclusion criteria not met",
      "failure_detail": "EGFR mutation negative",
      "screening_date": "2025-01-20"
    },
    {
      "screening_id": "SCR-0028",
      "failure_reason": "Exclusion criteria met",
      "failure_detail": "Prior immunotherapy within 6 months",
      "screening_date": "2025-01-25"
    }
  ]
}

Example 4: DM Domain SDTM Output

Prompt: "Generate DM domain for 10 subjects as SDTM"

Output:

{
  "domain": "DM",
  "records": [
    {
      "STUDYID": "ONCO-2025-001",
      "DOMAIN": "DM",
      "USUBJID": "ONCO-2025-001-001-0001",
      "SUBJID": "0001",
      "SITEID": "001",
      "RFSTDTC": "2025-01-22",
      "RFENDTC": null,
      "BRTHDTC": "1958-05-15",
      "AGE": 66,
      "AGEU": "YEARS",
      "SEX": "M",
      "RACE": "WHITE",
      "ETHNIC": "NOT HISPANIC OR LATINO",
      "ARMCD": "TRT",
      "ARM": "ABC-123 + SOC",
      "COUNTRY": "USA"
    }
  ]
}

Example 5: Laboratory Results with LOINC

Prompt: "Generate LB domain with liver function tests for safety monitoring"

Output:

{
  "domain": "LB",
  "records": [
    {
      "STUDYID": "SAFE-001",
      "DOMAIN": "LB",
      "USUBJID": "SAFE-001-001-0042",
      "LBSEQ": 1,
      "LBTESTCD": "ALT",
      "LBTEST": "Alanine Aminotransferase",
      "LBCAT": "CHEMISTRY",
      "LBORRES": "32",
      "LBORRESU": "U/L",
      "LBSTRESN": 32,
      "LBSTRESU": "U/L",
      "LBSTNRLO": 7,
      "LBSTNRHI": 56,
      "LBNRIND": "NORMAL",
      "LBLOINC": "1742-6",
      "LBBLFL": "Y",
      "VISITNUM": 2,
      "VISIT": "BASELINE"
    }
  ]
}

Example 6: ADaM Datasets (ADSL + ADTTE)

Prompt: "Derive ADSL and ADTTE datasets from the SDTM data for this trial"

ADSL (subject-level analysis dataset with baseline characteristics):

{
  "dataset": "ADSL",
  "records": [
    {
      "STUDYID": "ONCO-2025-001", "USUBJID": "ONCO-2025-001-001-0001",
      "SUBJID": "0001", "SITEID": "001", "TRT01P": "ABC-123 + SOC",
      "AGE": 66, "SEX": "M", "RACE": "WHITE",
      "RFSTDTC": "2025-01-22", "RFENDTC": "2025-07-15",
      "EOSSTT": "COMPLETED", "DCSREAS": null,
      "BMIBL": 24.3, "ECOGBL": 1
    }
  ]
}

ADTTE (time-to-event analysis dataset):

{
  "dataset": "ADTTE",
  "records": [
    {
      "STUDYID": "ONCO-2025-001", "USUBJID": "ONCO-2025-001-001-0001",
      "PARAMCD": "PFS", "PARAM": "Progression-Free Survival",
      "AVAL": 182, "AVALU": "DAYS",
      "CNSR": 0, "EVNTDESC": "Disease Progression (RECIST)",
      "STARTDT": "2025-01-22", "ADT": "2025-07-23"
    },
    {
      "STUDYID": "ONCO-2025-001", "USUBJID": "ONCO-2025-001-001-0002",
      "PARAMCD": "OS", "PARAM": "Overall Survival",
      "AVAL": 365, "AVALU": "DAYS",
      "CNSR": 1, "EVNTDESC": "Censored (alive at cutoff)",
      "STARTDT": "2025-01-29", "ADT": "2026-01-29"
    }
  ]
}

Generative Framework Integration

TrialSim integrates with the Generative Framework for specification-driven generation at scale.

Profile-Driven Generation

Use profile specifications to generate trial subject populations. The Profile Executor samples demographics meeting I/E criteria, generates baseline disease characteristics, applies randomization, and creates screening assessments.

Journey-Driven Generation

Attach protocol journey specifications to create visit sequences. The Journey Executor generates protocol visits at specified windows, creates assessments per schedule, applies visit variance, and handles protocol deviations and early termination.

Cross-Domain Sync

When generating across products, TrialSim entities are automatically linked:

See: ../generation/executors/cross-domain-sync.md