Adverse Event Reporting

Overview

VIGIA’s Adverse Event module provides comprehensive tools for coding, classifying, and assessing the causal relationship between drugs and adverse events using internationally recognized algorithms and standardized medical terminology.

MedDRA Integration

AI-powered MedDRA coding with LLT→PT mapping and SOC hierarchy

Causality Algorithms

Karch-Lasagna and French (FT-42) assessment algorithms

Multi-Event Cases

Track multiple concurrent adverse events per ICSR

Automated Extraction

AI extraction of events from case narratives

MedDRA Coding

What is MedDRA?

MedDRA (Medical Dictionary for Regulatory Activities) is the international medical terminology used in pharmacovigilance. VIGIA supports the complete MedDRA hierarchy:

SOC (System Organ Class)
  ↓
HLGT (High Level Group Term)
  ↓
HLT (High Level Term)
  ↓
PT (Preferred Term) ← Primary coding level
  ↓
LLT (Lowest Level Term) ← User verbatim terms

AI-Powered MedDRA Suggestion

VIGIA uses OpenAI and local NLP to suggest MedDRA codes:

GET /api/v1/meddra/suggest?q=náuseas&limit=10&scope=pt

Query Parameters:

q: Search term (Spanish or English)
limit: Maximum results (default: 15)
scope: Search scope - llt, pt, hlt, hlgt, soc, or all
version_id: MedDRA version (optional)

Response:

{
  "suggestions": [
    {
      "code": "10028813",
      "term": "Nausea",
      "level": "PT",
      "soc": "Gastrointestinal disorders",
      "score": 0.95,
      "match_type": "exact"
    },
    {
      "code": "10047699",
      "term": "Vomiting",
      "level": "PT",
      "soc": "Gastrointestinal disorders",
      "score": 0.78,
      "match_type": "related"
    }
  ]
}

Smart Matching: The system uses multiple strategies:

Exact match (normalized, accent-insensitive)
Whole word match (prevents “epicardio” matching “picar”)
Prefix match on tokens
Semantic similarity via OpenAI embeddings

Event Extraction from Narrative

Automatically extract adverse events from case narratives:

POST /api/v1/meddra/extract
Content-Type: application/json

{
  "text": "Paciente presenta náuseas, vómitos y dolor abdominal intenso desde hace 3 días.",
  "auto_code": true,
  "include_literal": true
}

Response:

{
  "events": [
    {
      "verbatim": "náuseas",
      "llt_code": "10028813",
      "llt_term": "Náusea",
      "pt_code": "10028813",
      "pt_term": "Nausea",
      "confidence": 0.92,
      "found_literal": true
    },
    {
      "verbatim": "vómitos",
      "llt_code": "10047700",
      "llt_term": "Vómito",
      "pt_code": "10047699",
      "pt_term": "Vomiting",
      "confidence": 0.89,
      "found_literal": true
    },
    {
      "verbatim": "dolor abdominal",
      "llt_code": "10000081",
      "llt_term": "Dolor abdominal",
      "pt_code": "10000081",
      "pt_term": "Abdominal pain",
      "confidence": 0.95,
      "found_literal": true
    }
  ]
}

Coding Workflow in ICSR

Create or edit ICSR

Navigate to the ICSR detail page

Add adverse events

Events can be added manually or extracted automatically from the narrative

PUT /api/v1/icsr/{icsr_id}/eventos
{
  "eventos": [
    {
      "texto": "Náuseas severas",
      "fecha_inicio": "2024-01-15",
      "gravedad": "Moderada",
      "tipo_notificacion": "Espontáneo"
    }
  ]
}

Apply MedDRA codes

Use the search interface or automated coding:

PUT /api/v1/icsr/{icsr_id}/meddra
{
  "eventos": [
    {
      "texto": "Náuseas severas",
      "meddra_llt_code": "10028813",
      "meddra_llt_term": "Náusea",
      "meddra_pt_code": "10028813",
      "meddra_pt_term": "Nausea"
    }
  ]
}

Review and validate

Pharmacovigilance specialists review auto-coded events and make adjustments

Causality Assessment

Karch-Lasagna Algorithm

The Karch-Lasagna algorithm evaluates causality using 8 criteria:

Criterion	Score Range	Description
Temporalidad	0-2	Temporal relationship between drug and event
Conocimiento previo	0-1	Known ADR in literature/labeling
Dechallenge	0-1	Event resolution upon drug discontinuation
Rechallenge	0-2	Event recurrence upon re-exposure
Alternativas	0-2	Other possible causes excluded
Factores contribuyentes	-1-0	Contributing factors present
Exploraciones	0-1	Objective evidence (labs, imaging)

Run Karch-Lasagna Assessment

GET /api/v1/causality/karch-lasagna/suggest/{icsr_id}

AI-Powered Evaluation: The system uses LLM (Gemini 1.5 Pro) to automatically evaluate each criterion based on case data:

{
  "ok": true,
  "suggestion": {
    "criterios": {
      "temporalidad": {
        "resp": "si",
        "score_sys": 2,
        "score_qf": 2,
        "rationale_sys": "El evento inició 2 días después del medicamento",
        "rationale_qf": "Relación temporal clara"
      },
      "conocimiento_previo": {
        "resp": "si",
        "score_sys": 1,
        "rationale_sys": "Náuseas es RAM conocida según ficha técnica"
      },
      "dechallenge": {
        "resp": "si",
        "score_sys": 1,
        "rationale_sys": "Síntomas cesaron al suspender el fármaco"
      }
    },
    "score_total_sys": 7,
    "score_total_qf": 7,
    "clasificacion_sys": "Probable",
    "clasificacion_qf": "Probable"
  }
}

Classification Scale:

Definitiva (8-10 points): Rechallenge positive, no alternatives
Probable (5-7 points): Good temporal relationship, known ADR
Posible (3-4 points): Temporal relationship present, alternatives exist
Improbable (1-2 points): Poor temporal relationship, likely alternatives
Inclasificable (0 points): Insufficient information

Dual Scoring System:

score_sys: AI-generated score
score_qf: Pharmacist-validated score

This allows AI suggestions while maintaining human oversight.

Confirm Assessment

After review, save the causality assessment:

PUT /api/v1/causality/karch-lasagna/confirm/{icsr_id}
Content-Type: application/json

{
  "criterios": {
    "temporalidad": {
      "resp": "si",
      "score_qf": 2,
      "rationale_qf": "Confirmed by QF: Clear temporal relationship"
    },
    "conocimiento_previo": {"resp": "si", "score_qf": 1},
    "dechallenge": {"resp": "si", "score_qf": 1},
    "rechallenge": {"resp": "no", "score_qf": 0},
    "alternativas": {"resp": "no_sabe", "score_qf": 0},
    "factores_contribuyentes": {"resp": "no", "score_qf": 0},
    "exploraciones": {"resp": "no", "score_qf": 0}
  },
  "score_total_qf": 4,
  "clasificacion_qf": "Posible",
  "rationale_global_qf": "Buena relación temporal pero sin dechallenge documentado"
}

French Algorithm (FT-42)

Alternative causality method focusing on pharmaceutical data sheet analysis:

POST /api/v1/causality/ft/analyze
Content-Type: multipart/form-data

file: [PDF of pharmaceutical data sheet]
mode: "ft42"

Features:

OCR extraction from pharmaceutical data sheets (Ficha Técnica)
Smart header/footer removal
Multi-page PDF support
Automatic re-challenge/de-challenge detection
Expected vs. Unexpected ADR classification

Event Severity Classification

VIGIA follows ICH E2A guidelines for severity:

Severity	Criteria
Leve	Minor discomfort, no intervention required
Moderada	Interferes with daily activities, may require intervention
Grave	Life-threatening, requires hospitalization, causes disability/death

Serious Event Criteria (Grave):

Death
Life-threatening
Hospitalization (initial or prolonged)
Persistent/significant disability
Congenital anomaly
Important medical event

API Endpoints Summary

MedDRA

# Search MedDRA terms
GET /api/v1/meddra/suggest?q={term}&scope=pt&limit=15

# Extract events from narrative
POST /api/v1/meddra/extract

# Get hierarchy (PT → SOC)
GET /api/v1/meddra/hierarchy/{code}

ICSR Events

# List events for ICSR
GET /api/v1/icsr/{icsr_id}/eventos

# Replace all events
PUT /api/v1/icsr/{icsr_id}/eventos

# Update MedDRA coding
PUT /api/v1/icsr/{icsr_id}/meddra

# Bulk operations
GET /api/v1/icsr/_bulk/eventos?ids=1&ids=2&ids=3

Causality

# Get last assessment
GET /api/v1/causality/karch-lasagna/last/{icsr_id}

# AI suggestion
GET /api/v1/causality/karch-lasagna/suggest/{icsr_id}

# Confirm assessment
PUT /api/v1/causality/karch-lasagna/confirm/{icsr_id}

# French algorithm
POST /api/v1/causality/ft/analyze

Code References

Feature	Implementation
Event Model	`backend/app/models/icsr_evento.py:16-76`
MedDRA Search	`backend/app/routers/meddra_unified.py:186-200`
Event Extraction	`backend/app/routers/meddra_unified.py`
Karch-Lasagna	`backend/app/routers/causality_karch_lasagna.py:119-199`
French Algorithm	`backend/app/routers/causality_ft.py:1-200`
ICSR Events CRUD	`backend/app/routers/icsr.py:1077-1209`

Best Practices

Coding Quality

Use PT (Preferred Term) level for primary coding
Include LLT when verbatim differs from PT
Verify AI suggestions before confirming
Document manual coding rationale

Causality Assessment

Review all available medical records
Document re-challenge/de-challenge data
Consider temporal plausibility
Involve clinical pharmacist in complex cases

ICSR Management - Full case management
Surveillance - Signal detection across events
IPS Reports - Periodic safety reporting

Get Started

Core Features

User Guides

Modules

Regulatory

Administration

Adverse Event Reporting

Overview

MedDRA Integration

Causality Algorithms

Multi-Event Cases

Automated Extraction

MedDRA Coding

What is MedDRA?

AI-Powered MedDRA Suggestion

Event Extraction from Narrative

Coding Workflow in ICSR

Causality Assessment

Karch-Lasagna Algorithm

Run Karch-Lasagna Assessment

Confirm Assessment

French Algorithm (FT-42)

Event Severity Classification

API Endpoints Summary

MedDRA

ICSR Events

Causality

Code References

Best Practices

Coding Quality

Causality Assessment

Build docs developers (and LLMs) love

Get Started

Core Features

User Guides

Modules

Regulatory

Administration

​Overview

MedDRA Integration

Causality Algorithms

Multi-Event Cases

Automated Extraction

​MedDRA Coding

​What is MedDRA?

​AI-Powered MedDRA Suggestion

​Event Extraction from Narrative

​Coding Workflow in ICSR

​Causality Assessment

​Karch-Lasagna Algorithm

​Run Karch-Lasagna Assessment

​Confirm Assessment

​French Algorithm (FT-42)

​Event Severity Classification

​API Endpoints Summary

​MedDRA

​ICSR Events

​Causality

​Code References

​Best Practices

Coding Quality

Causality Assessment

​Related Features

Build docs developers (and LLMs) love

Overview

MedDRA Coding

What is MedDRA?

AI-Powered MedDRA Suggestion

Event Extraction from Narrative

Coding Workflow in ICSR

Causality Assessment

Karch-Lasagna Algorithm

Run Karch-Lasagna Assessment

Confirm Assessment

French Algorithm (FT-42)

Event Severity Classification

API Endpoints Summary

MedDRA

ICSR Events

Causality

Code References

Best Practices

Related Features