Pharmacovigilance Safety Analyzer

Systematic drug safety analysis using FAERS adverse event data, FDA labeling, PharmGKB pharmacogenomics, and clinical trial safety signals.

KEY PRINCIPLES:

• Report-first approach - Create report file FIRST, update progressively

• Signal quantification - Use disproportionality measures (PRR, ROR)

• Severity stratification - Prioritize serious/fatal events

• Multi-source triangulation - FAERS, labels, trials, literature

• Pharmacogenomic context - Include genetic risk factors

• Actionable output - Risk-benefit summary with recommendations

• English-first queries - Always use English drug names and search terms in tool calls, even if the user writes in another language. Only try original-language terms as a fallback. Respond in the user's language

When to Use

Apply when user asks:

• "What are the safety concerns for [drug]?"

• "What adverse events are associated with [drug]?"

• "Is [drug] safe? What are the risks?"

• "Should I be concerned about [specific adverse event] with [drug]?"

• "Compare safety profiles of [drug A] vs [drug B]"

• "Pharmacovigilance analysis for [drug]"

Critical Workflow Requirements

1. Report-First Approach (MANDATORY)

Create the report file FIRST:

• File name: [DRUG]_safety_report.md

• Initialize with all section headers

• Add placeholder text: [Researching...]

Progressively update as you gather data

Output separate data files:

• [DRUG]_adverse_events.csv

• Ranked AEs with counts/signals

• [DRUG]_pharmacogenomics.csv

• PGx variants and recommendations

2. Citation Requirements (MANDATORY)

Every safety signal MUST include source:

Signal: Hepatotoxicity

• PRR: 3.2 (95% CI: 2.8-3.7)
• Cases: 1,247 reports
• Serious: 892 (71.5%)
• Fatal: 23

Source: FAERS via FAERS_count_reactions_by_drug_event (Q1 2020 - Q4 2025)

Phase 0: Tool Verification

CRITICAL: Verify tool parameters before calling.

Known Parameter Corrections

Tool WRONG Parameter CORRECT Parameter

FAERS_count_reactions_by_drug_event

drug

drug_name

DailyMed_search_spls

name

drug_name

PharmGKB_search_drug

drug

query

OpenFDA_get_drug_events

drug_name

search

Workflow Overview

Phase 1: Drug Disambiguation ├── Resolve drug name (brand → generic) ├── Get identifiers (RxCUI, ChEMBL, DrugBank) └── Identify drug class and mechanism ↓ Phase 2: Adverse Event Profiling (FAERS) ├── Query FAERS for drug-event pairs ├── Calculate disproportionality (PRR, ROR) ├── Stratify by seriousness └── OUTPUT: Ranked AE table ↓ Phase 3: Label Warning Extraction ├── DailyMed boxed warnings ├── Contraindications ├── Warnings and precautions └── OUTPUT: Label safety summary ↓ Phase 4: Pharmacogenomic Risk ├── PharmGKB clinical annotations ├── High-risk genotypes ├── Dosing recommendations └── OUTPUT: PGx risk table ↓ Phase 5: Clinical Trial Safety ├── ClinicalTrials.gov safety data ├── Phase 3/4 discontinuation rates ├── Serious AEs in trials └── OUTPUT: Trial safety summary ↓ Phase 5.5: Pathway & Mechanism Context (NEW) ├── KEGG: Drug metabolism pathways ├── Reactome: Mechanism-linked pathways ├── Target pathway analysis └── OUTPUT: Mechanistic safety context ↓ Phase 5.6: Literature Intelligence (ENHANCED) ├── PubMed: Published safety studies ├── BioRxiv/MedRxiv: Recent preprints ├── OpenAlex: Citation analysis └── OUTPUT: Literature evidence ↓ Phase 6: Signal Prioritization ├── Rank by PRR × severity × frequency ├── Identify actionable signals ├── Risk-benefit assessment └── OUTPUT: Prioritized signal list ↓ Phase 7: Report Synthesis

Phase 1: Drug Disambiguation

1.1 Resolve Drug Identity

def resolve_drug(tu, drug_query): """Resolve drug name to standardized identifiers.""" identifiers = {}

# DailyMed for NDC and SPL
dailymed = tu.tools.DailyMed_search_spls(drug_name=drug_query)
if dailymed:
    identifiers['ndc'] = dailymed[0].get('ndc')
    identifiers['setid'] = dailymed[0].get('setid')
    identifiers['generic_name'] = dailymed[0].get('generic_name')

# ChEMBL for molecule data
chembl = tu.tools.ChEMBL_search_drugs(query=drug_query)
if chembl:
    identifiers['chembl_id'] = chembl[0].get('molecule_chembl_id')
    identifiers['max_phase'] = chembl[0].get('max_phase')

return identifiers

1.2 Output for Report

1. Drug Identification

Source: DailyMed via DailyMed_search_spls, ChEMBL

Phase 2: Adverse Event Profiling

2.1 FAERS Query Strategy

def get_faers_events(tu, drug_name, top_n=50): """Query FAERS for adverse events."""

# Get event counts
events = tu.tools.FAERS_count_reactions_by_drug_event(
    drug_name=drug_name,
    limit=top_n
)

# For each event, get detailed breakdown
detailed_events = []
for event in events:
    detail = tu.tools.FAERS_get_event_details(
        drug_name=drug_name,
        reaction=event['reaction']
    )
    detailed_events.append({
        'reaction': event['reaction'],
        'count': event['count'],
        'serious': detail.get('serious_count', 0),
        'fatal': detail.get('death_count', 0),
        'hospitalization': detail.get('hospitalization_count', 0)
    })

return detailed_events

2.2 Disproportionality Analysis

Proportional Reporting Ratio (PRR):

PRR = (A/B) / (C/D)

Where: A = Reports of drug X with event Y B = Reports of drug X with any event C = Reports of event Y with any drug (excluding X) D = Total reports (excluding drug X)

Signal Thresholds:

Measure Signal Threshold Strong Signal

PRR

2.0 3.0

Chi-squared

4.0 10.0

N (case count) ≥3 ≥10

2.3 Severity Classification

Category Definition Priority

Fatal Death outcome Highest

Life-threatening Immediate death risk Very High

Hospitalization Required/prolonged hospitalization High

Disability Persistent impairment High

Congenital anomaly Birth defect High

Other serious Medical intervention required Medium

Non-serious No serious criteria Low

2.4 Output for Report

2. Adverse Event Profile (FAERS)

Data Period: Q1 2020 - Q4 2025 Total Reports for Drug: 45,234

2.1 Top Adverse Events by Frequency

2.2 Serious Adverse Events Only

2.3 Signal Interpretation

Strong Signal: Lactic Acidosis ⚠️

• PRR of 15.2 indicates 15x higher reporting rate than expected
• 89% classified as serious
• 156 fatalities (12.5% case fatality)
• Known class effect of biguanides
• Risk factors: renal impairment, hypoxia, contrast agents

Source: FAERS via FAERS_count_reactions_by_drug_event

Phase 3: Label Warning Extraction

3.1 DailyMed Query

def extract_label_warnings(tu, setid): """Extract safety sections from FDA label."""

label = tu.tools.DailyMed_get_spl_by_set_id(setid=setid)

warnings = {
    'boxed_warning': label.get('boxed_warning'),
    'contraindications': label.get('contraindications'),
    'warnings_precautions': label.get('warnings_and_precautions'),
    'adverse_reactions': label.get('adverse_reactions'),
    'drug_interactions': label.get('drug_interactions')
}

return warnings

3.2 Warning Severity Categories

Category Symbol Description

Boxed Warning ⬛ Most serious, life-threatening

Contraindication 🔴 Must not use

Warning 🟠 Significant risk

Precaution 🟡 Use caution

3.3 Output for Report

3. FDA Label Safety Information

3.1 Boxed Warning ⬛

LACTIC ACIDOSIS

Metformin can cause lactic acidosis, a rare but serious complication. Risk increases with renal impairment, sepsis, dehydration, excessive alcohol intake, hepatic impairment, and acute heart failure.

Contraindicated in patients with eGFR <30 mL/min/1.73m²

3.2 Contraindications 🔴

3.3 Warnings and Precautions 🟠

Source: DailyMed via DailyMed_get_spl_by_set_id

Phase 4: Pharmacogenomic Risk

4.1 PharmGKB Query

def get_pharmacogenomics(tu, drug_name): """Get pharmacogenomic annotations."""

# Search PharmGKB
pgx = tu.tools.PharmGKB_search_drug(query=drug_name)

annotations = []
for result in pgx:
    if result.get('clinical_annotation'):
        annotations.append({
            'gene': result['gene'],
            'variant': result['variant'],
            'phenotype': result['phenotype'],
            'recommendation': result['recommendation'],
            'level': result['level_of_evidence']
        })

return annotations

4.2 PGx Evidence Levels

Level Description Clinical Action

1A CPIC/DPWG guideline, implementable Follow guideline

1B CPIC/DPWG guideline, annotation Consider testing

2A VIP annotation, moderate evidence May inform

2B VIP annotation, weaker evidence Research

3 Low-level annotation Not actionable

4.3 Output for Report

4. Pharmacogenomic Risk Factors

4.1 Clinically Actionable Variants

4.2 Clinical Implications

OCT1 (SLC22A1) Poor Metabolizers:

• ~9% of Caucasians carry two loss-of-function alleles
• Reduced hepatic uptake of metformin
• May have decreased efficacy
• Consider higher doses or alternative agent

No CPIC/DPWG guidelines currently exist for metformin

Source: PharmGKB via PharmGKB_search_drug

Phase 5: Clinical Trial Safety

5.1 ClinicalTrials.gov Query

def get_trial_safety(tu, drug_name): """Get safety data from clinical trials."""

# Search completed phase 3/4 trials
trials = tu.tools.search_clinical_trials(
    intervention=drug_name,
    phase="Phase 3",
    status="Completed",
    pageSize=20
)

safety_data = []
for trial in trials:
    if trial.get('results_posted'):
        results = tu.tools.get_clinical_trial_results(
            nct_id=trial['nct_id']
        )
        safety_data.append(results.get('adverse_events'))

return safety_data

5.2 Output for Report

5. Clinical Trial Safety Data

5.1 Phase 3 Trial Summary

5.2 Common Adverse Events in Trials

Source: ClinicalTrials.gov via search_clinical_trials

Phase 5.5: Pathway & Mechanism Context (NEW)

5.5.1 Drug Metabolism Pathways (KEGG)

def get_drug_pathway_context(tu, drug_name, drug_targets): """Get pathway context for mechanistic safety understanding."""

# KEGG drug metabolism
metabolism = tu.tools.kegg_search_pathway(
    query=f"{drug_name} metabolism"
)

# Target pathways
target_pathways = {}
for target in drug_targets:
    pathways = tu.tools.kegg_get_gene_info(gene_id=f"hsa:{target}")
    target_pathways[target] = pathways.get('pathways', [])

return {
    'metabolism_pathways': metabolism,
    'target_pathways': target_pathways
}

5.5.2 Output for Report

5.5 Pathway & Mechanism Context

Drug Metabolism Pathways (KEGG)

Target Pathway Analysis

Primary Target: AMPK

• Pathway: AMPK signaling (hsa04152)
• Downstream: mTOR inhibition, autophagy
• Safety relevance: Explains metabolic effects

Mechanistic Basis for Key AEs:

Source: KEGG, Reactome

Phase 5.6: Literature Intelligence (ENHANCED)

5.6.1 Published Safety Studies

def comprehensive_safety_literature(tu, drug_name, key_aes): """Search all literature sources for safety evidence."""

# PubMed: Peer-reviewed
pubmed = tu.tools.PubMed_search_articles(
    query=f'"{drug_name}" AND (safety OR adverse OR toxicity)',
    limit=30
)

# BioRxiv: Preprints
biorxiv = tu.tools.BioRxiv_search_preprints(
    query=f"{drug_name} mechanism toxicity",
    limit=10
)

# MedRxiv: Clinical preprints
medrxiv = tu.tools.MedRxiv_search_preprints(
    query=f"{drug_name} safety",
    limit=10
)

# Citation analysis for key papers
key_papers = pubmed[:10]
for paper in key_papers:
    citation = tu.tools.openalex_search_works(
        query=paper['title'],
        limit=1
    )
    paper['citations'] = citation[0].get('cited_by_count', 0) if citation else 0

return {
    'pubmed': pubmed,
    'preprints': biorxiv + medrxiv,
    'key_papers': key_papers
}

5.6.2 Output for Report

5.6 Literature Evidence

Key Safety Studies

Recent Preprints (Not Peer-Reviewed)

⚠️ Note: Preprints have NOT undergone peer review.

Evidence Summary

Source: PubMed, BioRxiv, MedRxiv, OpenAlex

Phase 6: Signal Prioritization

6.1 Signal Scoring Formula

Signal Score = PRR × Severity_Weight × log10(Case_Count + 1)

Severity Weights:

• Fatal: 10
• Life-threatening: 8
• Hospitalization: 5
• Disability: 5
• Other serious: 3
• Non-serious: 1

6.2 Output for Report

6. Prioritized Safety Signals

6.1 Critical Signals (Immediate Attention)

6.2 Moderate Signals (Monitor)

6.3 Known/Expected (Manage Clinically)

Report Template

File: [DRUG]_safety_report.md

Pharmacovigilance Safety Report: [DRUG]

Generated: [Date] | Query: [Original query] | Status: In Progress

Executive Summary

[Researching...]

1. Drug Identification

1.1 Drug Information

[Researching...]

2. Adverse Event Profile (FAERS)

2.1 Top Adverse Events

[Researching...]

2.2 Serious Adverse Events

[Researching...]

2.3 Signal Analysis

[Researching...]

3. FDA Label Safety Information

3.1 Boxed Warnings

[Researching...]

3.2 Contraindications

[Researching...]

3.3 Warnings and Precautions

[Researching...]

4. Pharmacogenomic Risk Factors

4.1 Actionable Variants

[Researching...]

4.2 Testing Recommendations

[Researching...]

5. Clinical Trial Safety

5.1 Trial Summary

[Researching...]

5.2 Adverse Events in Trials

[Researching...]

6. Prioritized Safety Signals

6.1 Critical Signals

[Researching...]

6.2 Moderate Signals

[Researching...]

7. Risk-Benefit Assessment

[Researching...]

8. Clinical Recommendations

8.1 Monitoring Recommendations

[Researching...]

8.2 Patient Counseling Points

[Researching...]

8.3 Contraindication Checklist

[Researching...]

9. Data Gaps & Limitations

[Researching...]

10. Data Sources

[Will be populated as research progresses...]

Evidence Grading

Tier Symbol Criteria Example

T1 ⚠️⚠️⚠️ PRR >10, fatal outcomes, boxed warning Lactic acidosis

T2 ⚠️⚠️ PRR 3-10, serious outcomes Hepatotoxicity

T3 ⚠️ PRR 2-3, moderate concern Hypoglycemia

T4 ℹ️ PRR <2, known/expected GI side effects

Completeness Checklist

Phase 1: Drug Identification

• Generic name resolved

• Brand names listed

• Drug class identified

• ChEMBL/DrugBank ID obtained

• Mechanism of action stated

Phase 2: FAERS Analysis

• ≥20 adverse events queried

• PRR calculated for top events

• Serious/fatal counts included

• Signal thresholds applied

• Time period stated

Phase 3: Label Warnings

• Boxed warnings extracted (or "None")

• Contraindications listed

• Key warnings summarized

• Drug interactions noted

Phase 4: Pharmacogenomics

• PharmGKB queried

• Actionable variants listed (or "None")

• Evidence levels provided

• Testing recommendations stated

Phase 5: Clinical Trials

• Phase 3/4 trials searched

• Serious AE rates compared

• Discontinuation rates noted

Phase 6: Signal Prioritization

• Signals ranked by score

• Critical signals flagged

• Actions recommended

Phase 7-8: Synthesis

• Risk-benefit assessment provided

• Monitoring recommendations listed

• Patient counseling points included

Fallback Chains

Primary Tool Fallback 1 Fallback 2

FAERS_count_reactions_by_drug_event

OpenFDA_get_drug_events

Literature search

DailyMed_get_spl_by_set_id

FDA_drug_label_search

DailyMed website

PharmGKB_search_drug

CPIC_get_guidelines

Literature search

search_clinical_trials

ClinicalTrials.gov API PubMed for trial results

Tool Reference

See TOOLS_REFERENCE.md for complete tool documentation.