Individual Case Safety Report (ICSR)

The minimum criteria for a valid ICSR are the presence of four identifiable elements: an identifiable patient, an identifiable reporter, a suspected medicinal product, and an adverse event. Without all four, the case is considered incomplete and may not be reportable to regulatory authorities. The patient is typically identified by age, sex, or initials rather than full name to maintain pseudonymity.

The header of every ICSR contains globally unique identifiers for traceability. This includes the sender's safety report ID, the receipt date, the report type (spontaneous, clinical trial, literature), and the seriousness criteria classification. The ICH E2B(R3) standard mandates specific data elements such as the worldwide unique case identification number and the sender organization identifier to prevent duplicate reporting across databases like EudraVigilance and FAERS.

This section captures demographic and medical history data essential for causality assessment. Key fields include:

• Age at onset and age unit (decades, years, months)
• Sex and reproductive status (pregnant, breastfeeding)
• Relevant medical history including concurrent conditions
• Concomitant medications with dosage and duration
• Drug history including prior reactions

The patient's weight in kilograms is critical for weight-based dosing evaluation.

Each suspected medicinal product requires structured characterization. The medicinal product name must be coded using a standardized dictionary, with substance name mapped to the WHO Drug Dictionary or equivalent. Critical fields include:

• Dosage regimen and route of administration
• Therapy dates (start and stop)
• Indication for which the drug was prescribed
• Action taken with the drug (withdrawn, dose reduced, continued)
• Batch/lot number for biologics and vaccines

The dechallenge and rechallenge outcomes are documented here to support causality assessment.

The core clinical narrative of the ICSR. Each event must be coded using MedDRA terminology at the Lowest Level Term (LLT) level, which automatically maps to Preferred Terms and System Organ Classes. The section captures:

• Verbatim term as reported by the primary source
• Event onset date and end date for duration calculation
• Outcome (recovered, recovering, not recovered, fatal, unknown)
• Seriousness criteria (death, life-threatening, hospitalization, disability, congenital anomaly, other medically important)
• Causality assessment result from the reporter

The case narrative is a structured free-text summary that chronologically describes the patient's clinical course. It must include:

• Temporal sequence of drug administration and event onset
• Relevant laboratory data and diagnostic findings
• Treatment of the adverse event and response
• Reporter's causality assessment and rationale
• Autopsy or death details if applicable

This narrative is the primary target for Natural Language Processing models performing automated adverse event extraction from unstructured clinical documents.

The ICH E2B (R3) standard defines the structured XML schema for the electronic exchange of ICSRs between pharmaceutical companies and regulators. It specifies over 300 data elements, including patient demographics, drug identification, and reaction details, enabling automated acknowledgment, routing, and loading into safety databases like EudraVigilance and FAERS without manual re-keying.

The systematic evaluation of the likelihood that a suspect drug caused the reported adverse event. Key frameworks include the WHO-UMC system and the Naranjo algorithm, which weigh factors such as:

• Temporal plausibility (did the event occur after dosing?)
• Dechallenge/Rechallenge (did the event resolve on withdrawal and recur on re-administration?)
• Alternative etiologies (could concurrent disease explain the event?) This assessment is a critical narrative component within each ICSR.

A regulatory classification that determines the expedited reporting requirements for an ICSR. An adverse event is defined as serious if it results in:

• Death
• Life-threatening condition
• Inpatient hospitalization or prolongation
• Persistent or significant disability/incapacity
• Congenital anomaly/birth defect This binary flag triggers a 15-day reporting clock for manufacturers.

The computational process of mining aggregated ICSR databases to identify drug-event combinations reported more frequently than expected. Disproportionality analysis uses statistical metrics like the Proportional Reporting Ratio (PRR) and Empirical Bayes Geometric Mean (EBGM) to flag potential safety signals from the background noise of spontaneous reporting, initiating the signal validation workflow.