FAERS

FAERS operates as a passive surveillance system that accepts voluntary adverse event and medication error reports from:

• Healthcare professionals (physicians, pharmacists, nurses)
• Consumers (patients, caregivers)
• Manufacturers (mandatory reporting for marketed products)

Reports are submitted via the MedWatch program using Form FDA 3500 (voluntary) or Form FDA 3500A (mandatory for manufacturers). The database contains over 2 million individual case safety reports as of 2024, growing by approximately 1.5-2 million reports annually.

Each FAERS report captures standardized safety information using the ICH E2B (R3) electronic transmission format, including:

• Patient demographics: age, sex, weight (when available)
• Drug information: suspect and concomitant medications with dosage, route, and therapy dates
• Adverse events: coded using MedDRA preferred terms at multiple seriousness levels
• Outcomes: death, life-threatening, hospitalization, disability, congenital anomaly, other serious
• Reporter information: occupation, country, report source

The structured nature enables systematic disproportionality analysis across drug-event combinations.

FAERS data is publicly released as quarterly ASCII or XML extract files containing:

• DEMO: patient demographic and administrative information
• DRUG: medication details including drug characterization (suspect, concomitant, interacting)
• REAC: adverse event reactions coded to MedDRA preferred terms
• OUTC: patient outcomes for each event
• RPSR: report source information
• THER: drug therapy start and end dates
• INDI: indication for drug use coded to MedDRA

These flat files require significant data normalization and deduplication before analysis due to follow-up reports and multiple submission pathways.

FAERS serves as the primary data source for quantitative signal detection using disproportionality algorithms:

• Empirical Bayes Geometric Mean (EBGM): Bayesian shrinkage method using the Multi-item Gamma Poisson Shrinker (MGPS) algorithm
• Proportional Reporting Ratio (PRR): Frequentist measure comparing observed-to-expected reporting rates
• Reporting Odds Ratio (ROR): Case-non-case analysis for drug-event associations

FDA safety reviewers use these statistical scores alongside clinical case review to identify potential safety signals that warrant further investigation through the Signal Management Process.

Critical limitations must be considered when analyzing FAERS data:

• Underreporting: Estimated that only 1-10% of adverse events are reported
• No denominator data: Cannot calculate true incidence rates without exposure data
• Duplicate reports: Follow-up submissions create duplicate records requiring deduplication
• Confounding by indication: The underlying disease may cause the event, not the drug
• Stimulated reporting: Media attention or litigation can artificially inflate reporting rates
• No causality verification: Reports do not establish causation; they represent suspected associations

These limitations necessitate complementary data sources like Sentinel and claims databases for comprehensive safety assessment.

The core statistical engine for signal detection within FAERS. This methodology quantifies whether a specific drug-event pair is reported more frequently than expected by chance.

• Proportional Reporting Ratio (PRR): A frequentist measure comparing observed vs. expected reporting rates.
• Reporting Odds Ratio (ROR): Calculates the odds of an event being reported with a specific drug versus all other drugs.
• Empirical Bayes Geometric Mean (EBGM): A Bayesian metric that applies shrinkage to correct for instability in low-count pairs, reducing false positives.

The Medical Dictionary for Regulatory Activities is the standardized international terminology used to code adverse event reports in FAERS. It provides a hierarchical structure for precise medical encoding.

• Structure: Five levels from System Organ Class (SOC) down to Lowest Level Terms (LLT).
• Purpose: Enables consistent aggregation and retrieval of safety data across global regulatory submissions.
• Standardised MedDRA Queries (SMQs): Pre-defined groupings of terms designed to identify cases related to specific syndromes, such as anaphylactic reaction or acute pancreatitis.

The ICH E2B (R3) standard defines the electronic format for transmitting Individual Case Safety Reports (ICSRs) to databases like FAERS. It ensures structured, interoperable data exchange.

• Format: XML-based messaging using HL7 ICSR standards.
• Content: Encodes patient demographics, suspect drugs, adverse events, and medical history in a machine-readable schema.
• Automation: Enables pharmaceutical companies to submit thousands of reports programmatically, eliminating manual data entry errors.

A statistical association in FAERS does not equal a causal link. Signal validation is the rigorous process of confirming a new safety finding.

• Causality Assessment: Evaluates the likelihood that a drug caused an event using factors like temporal relationship and dechallenge/rechallenge data.
• Confounding by Indication: A critical bias where the treated disease, not the drug, causes the event. Disentangling this requires clinical review.
• Seriousness Criteria: Events are classified as 'serious' if they result in death, hospitalization, disability, or congenital anomaly.

FAERS is part of a global ecosystem of spontaneous reporting databases. Cross-referencing these systems strengthens signal detection.

• EudraVigilance: The European Medicines Agency's centralized database for adverse reactions within the European Economic Area.
• VigiBase: The WHO's global ICSR repository, maintained by the Uppsala Monitoring Centre, containing over 30 million reports.
• Data Mining: Modern pharmacovigilance applies NLP to simultaneously mine these databases alongside unstructured literature and social media for emerging safety signals.