QPPV.COM

How to Write an ICSR Case Narrative

Explains the regulatory purpose of the ICSR case narrative, what information belongs in it, how to organise chronology, distinguish reported facts from assessment, manage follow-up and contradictions, and produce a concise stand-alone medical story suitable for regulatory review.

Take test

How to Write an ICSR Case Narrative

An ICSR case narrative is the clinical story of an individual safety case. Its purpose is not to repeat every structured database field in prose. It is to present the medically relevant information in a form that allows a reviewer to understand what happened to the patient, when it happened, how the medicinal product exposure relates in time to the event, what investigations and treatment followed, how the patient evolved, and what information supports or argues against a causal relationship.

ICH E2D(R1) describes the narrative as a comprehensive, stand-alone medical story. It should summarise relevant patient characteristics, therapy details, medical history, concurrent conditions, the clinical course of the event, diagnosis, outcome, laboratory evidence and other information important to causal assessment.

The defining principle is chronology.

The narrative should normally follow the patient's clinical experience, not the order in which the company received pieces of information.

A useful conceptual sequence is:

patient context → relevant treatment exposure → event onset → clinical course → investigations → interventions → outcome → follow-up → remaining uncertainty

A strong narrative therefore allows a medical reviewer to reconstruct the case without repeatedly moving between unrelated fields.

Purpose and Scope

This article addresses post-authorisation ICSR case narratives in the EU pharmacovigilance context, using current ICH E2D(R1), EU GVP Module VI and E2B(R3) principles.

It focuses on:

It does not attempt to replace structured ICSR fields.

The narrative is one component of the ICSR and should remain consistent with the structured data.

Regulatory Framework

Article 28(3)(m) of Commission Implementing Regulation (EU) No 520/2012 requires, where possible, a case narrative providing all relevant information for individual cases with the exception of non-serious adverse reactions.

This is the legally binding EU requirement.

The same Regulation requires ICSRs to be as complete as possible and communicated accurately and reliably.

The legal rule should not be expanded into a claim that every non-serious EU ICSR must contain a narrative. A company may choose to create narratives more broadly as an operational practice, but that is different from the statutory requirement.

GVP Module VI

Current GVP Module VI operationalises the legal requirement.

For case narratives, GVP states that information should be presented:

GVP also states that the narrative should be a comprehensive, stand-alone medical report containing relevant patient, therapy, history, diagnostic, clinical-course and laboratory information.

ICH E2D(R1)

ICH E2D(R1), adopted at Step 4 in September 2025, modernises the post-approval case-management framework.

Its narrative section emphasises that:

E2D(R1) is harmonised guidance. Regional and local reporting requirements remain applicable.

E2B(R3)

In E2B(R3), the main narrative field is:

H.1 — Case Narrative Including Clinical Course, Therapeutic Measures, Outcome and Additional Relevant Information

EU GVP directs serious reports to use this field for the case narrative.

Other related E2B(R3) fields serve different functions, including:

A case narrative should not absorb every comment or assessment merely because free text is available.

Narrative Design Principles

What the Narrative Is — and Is Not

The narrative is a clinical synthesis of case facts.

It is not:

A useful test is:

Could a medically trained reviewer understand the patient's course from this narrative alone, while still distinguishing what was reported from what was assessed?

If yes, the narrative is performing its intended function.

Stand-Alone Does Not Mean Exhaustive

A stand-alone narrative should contain enough information to support independent understanding of the case.

It does not need every available data point.

The writer should include information that is relevant to:

Irrelevant administrative detail reduces clarity.

For example, the reviewer usually needs to know:

The reviewer usually does not need a prose description of:

Those may still belong elsewhere in the audit trail.

The Narrative and Structured Data Must Agree

EU GVP explicitly states that narrative information should be consistent with corresponding E2B data elements.

This does not mean the narrative must repeat every structured value.

It means contradictions should not arise without explanation.

Examples of inconsistency include:

A QC review should compare the narrative with structured case data before submission.

Chronology Is the Organising Principle

A good narrative follows clinical time.

A weak narrative often follows processing time:

The case was first received on 1 May. Follow-up was received on 4 May. Additional follow-up was received on 8 May. The reporter then stated...

A stronger narrative reorganises those fragments into the patient's course:

The patient started Product X on 20 April. On 27 April, the patient developed progressive dyspnoea. On 28 April, the patient presented to hospital...

The dates on which the information arrived remain important for Day Zero and follow-up processing, but they do not need to dominate the medical story.

This distinction is fundamental.

A Useful Narrative Skeleton

A practical narrative can often be organised around seven elements:

  1. patient context;
  2. relevant medical history and concomitant conditions;
  3. suspect and relevant concomitant treatment;
  4. event onset and clinical course;
  5. investigations and diagnosis;
  6. therapeutic measures, dechallenge/rechallenge and outcome;
  7. follow-up and unresolved information.

The exact order can be adapted where the case requires it.

For example, a fatal case may need the outcome and autopsy findings developed more prominently. A medication-error case may need the use error described before the clinical consequence.

Building the Opening Clinical Context

Start With the Patient, Not the Company

The opening sentence should orient the reader clinically.

A common pattern is:

A 67-year-old woman with a history of hypertension and chronic kidney disease received Product X for Condition Y.

This establishes:

The sentence can be adapted when age, sex or medical history are unknown.

Do not invent demographic information to create a polished opening.

For example:

An adult patient received Product X for Condition Y.

can be appropriate if that is all the source supports.

Patient Characteristics

Include clinically relevant patient information such as:

The purpose is not to reproduce the entire patient profile.

Select characteristics that help the reviewer interpret:

Medical History

Medical history should be included when it is relevant to case interpretation.

Examples include:

A remote, unrelated condition should not be included merely because it exists in the database.

The narrative should remain medically focused.

Concurrent Conditions

Current illnesses can be as important as past history.

For example:

At the time of the event, the patient had pneumonia and was receiving systemic corticosteroids.

This may materially affect interpretation of:

Concurrent conditions should be integrated into the chronology where they matter, not placed in a detached list.

Suspect Product Information

Relevant suspect-product information can include:

For biological medicines, exact product identity and batch information can be relevant where available.

The narrative should not copy every product field mechanically.

Include the details that help explain the event chronology and assessment.

Concomitant Medicines

Concomitant medicines should be included when they are clinically relevant.

Examples include:

A long stable medication list can overwhelm the narrative.

Where many concomitants exist, focus on those that materially affect interpretation and ensure full structured medication data remain available elsewhere in the ICSR.

Dose and Exposure Changes

Dose changes can be clinically important.

For example:

Product X was increased from 10 mg once daily to 20 mg once daily on 5 June. Three days later, the patient developed...

This chronology may support a dose-response assessment.

If the source does not provide exact dose information, do not reconstruct it from package strength or other assumptions.

The narrative should reflect the evidence, not complete the story by inference.

Writing the Clinical Course

The clinical course is the centre of the narrative.

It should answer, in sequence:

The reader should not have to reconstruct the timeline from scattered facts.

Event Onset

State the event onset as precisely as the source supports.

Examples include:

If the date is unknown, say so.

Do not fabricate precision.

For example:

Approximately two weeks after starting Product X, the patient developed progressive abdominal pain.

is better than assigning a specific calendar date that was never reported.

Multiple Events

When several events occur, the narrative should preserve their relationship.

A useful sequence is:

first event → second event → diagnostic work-up → treatment → outcome.

Avoid listing multiple MedDRA terms without showing whether they occurred:

For example:

The patient first developed fever and cough. Two days later, dyspnoea worsened and pneumonia was diagnosed.

is more informative than:

The patient experienced pyrexia, cough, dyspnoea and pneumonia.

Signs, Symptoms and Diagnoses

Narratives should preserve the distinction between:

If the source says:

“possible pulmonary embolism”

the narrative should not silently rewrite this as:

“the patient developed pulmonary embolism”.

If later imaging confirms the diagnosis, the narrative can show that evolution:

Pulmonary embolism was initially suspected. CT pulmonary angiography subsequently confirmed bilateral pulmonary emboli.

This preserves both chronology and diagnostic certainty.

Laboratory and Investigation Results

Include investigations that materially support or refute:

Relevant details can include:

ICH E2D(R1) specifically identifies laboratory evidence, including normal ranges, as relevant narrative information.

Abnormal Results

State clinically meaningful abnormal results clearly.

For example:

Alanine aminotransferase was 425 U/L (reference range 7–56 U/L).

This is usually more informative than:

Liver tests were abnormal.

Normal or Negative Results

Normal or negative findings can be highly relevant where they help evaluate alternative explanations.

Examples include:

Do not include every normal test.

Include those that contribute materially to interpretation.

Test Timing Matters

A test result has meaning in relation to the clinical course.

For example:

On the day of admission, creatinine was 3.2 mg/dL compared with a baseline of 1.0 mg/dL.

This conveys both:

The same result presented without context is less useful.

Therapeutic Measures

Describe clinically important treatment given for the event.

This can include:

The purpose is to show what happened medically.

Do not convert the narrative into a full medication administration record.

Dechallenge

Dechallenge information can be important to causal assessment.

Where relevant, state:

For example:

Product X was discontinued on 12 August. The rash improved over the following four days and had resolved by 18 August.

Avoid the conclusion:

Positive dechallenge occurred.

unless that is a clearly attributed assessment.

The narrative should first present the facts.

Rechallenge

Rechallenge should be described precisely.

Relevant information includes:

For example:

Product X was restarted at 10 mg daily on 2 September. Within 24 hours, the rash recurred. Treatment was discontinued again on 3 September.

This is more useful than simply stating “positive rechallenge”.

Outcome

State the outcome at the latest point known.

Possible outcomes include:

The narrative wording should be consistent with the structured outcome field.

If follow-up changes the outcome, update the narrative accordingly.

Hospitalisation

Hospitalisation is both a clinical event in the chronology and, where applicable, a seriousness criterion.

The narrative should explain:

For example:

The patient was admitted for management of severe dehydration and acute kidney injury.

This is more useful than:

The event was serious due to hospitalisation.

The latter is a regulatory classification rather than the clinical story.

Life-Threatening Events

If an event was life-threatening, the narrative should describe the clinical facts supporting that assessment where available.

Examples can include:

Do not use “life-threatening” merely as an adjective for a severe event if the source does not support the regulatory seriousness criterion.

Disability or Incapacity

When disability is relevant, explain the nature and persistence of the impairment where known.

For example:

Six months after the event, the patient continued to require assistance with activities of daily living because of persistent left-sided weakness.

This provides more useful context than a standalone seriousness label.

Fatal Cases

Fatal cases require particularly careful chronology.

Include, where available:

EU law and GVP specifically identify date and cause of death, including autopsy-determined cause where available, as relevant ICSR information.

Death Is Not Automatically the Event

The narrative should distinguish:

For example:

The patient developed intracranial haemorrhage and died two days later.

is clearer than:

The patient experienced death.

Autopsy and Post-Mortem Findings

Relevant autopsy or post-mortem findings should be summarised.

If documents are held separately, the narrative can state that the report is available.

Do not copy an entire autopsy report into H.1 when a focused summary communicates the medically important findings.

Supplementary Records

The case may include:

ICH E2D(R1) states that key information from supplementary records should be included in the report and their availability noted.

The narrative should therefore extract clinically meaningful findings rather than say only:

“Hospital records received.”

A better approach is:

Hospital records confirmed admission from 4–9 March for acute pancreatitis. Lipase was 1,240 U/L and CT showed pancreatic inflammation. The discharge summary is held by the sender.

Source Attribution and Assessment Boundaries

Source Attribution

Source attribution is essential when information represents an opinion rather than an objective fact.

For example:

The treating physician considered the reaction probably related to Product X.

is different from:

The reaction was probably related to Product X.

The first identifies the source of the opinion.

The second sounds like the narrative writer's conclusion.

Similarly:

The patient reported that symptoms began within one hour of dosing.

preserves the source.

Reporter Comments Versus Narrative Facts

E2B(R3) provides dedicated fields for reporter comments.

The narrative can include clinically important reporter observations, but should preserve attribution.

Examples include:

Do not blend these into objective narrative statements.

Sender Comments Versus Narrative

The MAH or authority may have its own interpretation.

E2B(R3) H.4 is designed for sender comments.

Examples include:

A sender assessment should not be inserted into the narrative in a way that obscures the original case facts.

A useful separation is:

H.1 → medical story

H.2 → reporter comments

H.4 → sender comments

structured causality fields → source-specific drug-event assessments.

Do Not Rewrite the Source to Fit the Company Assessment

The narrative should remain source-faithful even if the company disagrees with the reporter.

For example, if the reporter states:

“The doctor said the medicine caused hepatitis.”

and the MAH considers viral hepatitis more likely, the narrative can say:

The reporter stated that the treating physician considered the hepatitis related to Product X.

The company's alternative assessment belongs in sender comments or the structured causality assessment.

The narrative should not erase the reporter's statement.

Contradictory Information

Cases often contain contradictions.

Examples include:

Do not silently select one version unless the later information clearly corrects the earlier report.

Instead, present the conflict transparently.

For example:

The initial consumer report stated that symptoms began on 3 June. In follow-up, the treating physician documented onset on 5 June.

If the reporter later explicitly corrects the earlier information, state that:

The reporter subsequently corrected the onset date from 3 June to 5 June.

Unknown Information

A good narrative can state uncertainty directly.

Useful phrases include:

This is preferable to writing around the missing information in a way that implies completeness.

Narrative Style

Avoid Unnecessary Abbreviations

ICH E2D(R1) advises avoiding abbreviations and acronyms, except where they are appropriate for laboratory parameters and units.

For example:

computed tomography

is clearer than:

CT

if the abbreviation is not obvious in context.

Common laboratory abbreviations can be acceptable where medically unambiguous.

If an abbreviation could mean several things, spell it out.

Avoid Promotional or Defensive Language

Narratives should not contain language such as:

unless this is a clearly attributed source or sender assessment in the appropriate field.

The narrative's role is to present evidence, not argue the case.

Follow-Up Narratives

Follow-up reports should improve the medical story rather than append disconnected text.

ICH E2D(R1) states that new information in follow-up reports should be clearly identified.

There are two practical objectives:

  1. preserve the complete current narrative;
  2. make it possible to identify what changed.

An organisation may accomplish this through:

The important point is that the latest narrative should remain medically coherent.

Do Not Append Follow-Up Mechanically

A weak follow-up narrative often looks like:

Follow-up 1: patient admitted. Follow-up 2: laboratory results received. Follow-up 3: patient recovered.

This preserves receipt history but forces the reviewer to reconstruct the patient course.

A stronger integrated narrative is:

The patient developed fever on 1 March and was admitted on 3 March. Blood cultures obtained on admission were negative. Product X was discontinued on 4 March. The patient recovered and was discharged on 8 March.

New follow-up information can then be identified separately or in the appropriate system metadata.

Preserve What Changed

Follow-up may change:

The narrative should reflect the current best source information while preserving relevant uncertainty and corrections.

For example:

Initial:

“Possible seizure.”

Follow-up:

“Neurology review concluded that the episode represented syncope; EEG was normal.”

The updated narrative should show that diagnostic evolution rather than leave the earlier provisional diagnosis as if it remained confirmed.

Corrections Versus Additional Information

A follow-up can:

These are not the same.

Additional Information

“The patient was later found to have hypokalaemia.”

Correction

“The reporter corrected the age from 74 to 47 years.”

Clarification

“The term ‘fainted’ referred to a brief loss of consciousness without seizure activity.”

Narrative wording should make the type of change understandable where it affects interpretation.

Case Type and Source-Specific Narratives

Serious Cases

Serious cases usually require the most developed narratives because the regulator needs to understand:

The narrative should explain the facts that support seriousness rather than merely repeat the seriousness category.

For example:

The patient was admitted for four days because of severe vomiting, dehydration and acute kidney injury.

is more informative than:

The event was serious due to hospitalisation.

Non-Serious Cases

EU law does not require a case narrative for non-serious adverse reactions under Article 28(3)(m).

However, organisations may choose to generate narratives for:

When a non-serious narrative is created, the same principles of chronology, source fidelity and consistency apply.

Literature Cases

Literature cases require careful separation of:

A narrative should summarise the individual patient's course, not reproduce the entire article.

Relevant elements can include:

If the article contains several patients, the narrative should ensure that the information belongs to the specific identifiable patient represented by the ICSR.

The literature citation and article itself belong in the appropriate structured/supplementary fields.

Digital-Platform Cases

Digital reports can be brief, informal and incomplete.

The narrative should preserve:

For example:

The patient posted that Product X “made me pass out” approximately one hour after dosing. No further clinical details were available.

Do not convert “passed out” into a specific diagnosis unless the source supports it.

If screenshots or platform exports are held, clinically relevant information can be summarised and their availability documented.

Consumer Reports

Consumer descriptions should retain their clinical meaning without being rewritten into unsupported medical terminology.

For example:

The patient reported “heart racing” for approximately 20 minutes.

is preferable to:

The patient developed supraventricular tachycardia.

unless a diagnosis was actually established.

MedDRA coding may standardise the reported event in structured fields, but the narrative should remain faithful to the source.

Healthcare-Professional Reports

Healthcare-professional reports may contain:

The narrative should preserve these distinctions.

For example:

The physician diagnosed drug-induced liver injury and considered Product X the likely cause.

This represents the physician's assessment.

It should not be rewritten as an unqualified company conclusion.

Special Situations

Medication-Error Cases

Medication-error narratives should explain both:

  1. the use error;
  2. the clinical consequence, if any.

A useful chronology is:

intended use → what actually happened → why/how the error occurred if known → patient exposure → clinical consequence → intervention → outcome.

For example:

The patient was prescribed 10 mg once daily but mistakenly took 20 mg once daily for three days. On the third day, the patient developed symptomatic hypotension and attended the emergency department.

This is more useful than listing:

Overdose

An overdose narrative should include, where known:

Intent should not be inferred.

If the source does not state whether the overdose was intentional, the narrative should say that intentionality was not reported.

Misuse and Abuse

For misuse or abuse, relevant context can include:

Again, distinguish reported facts from interpretation.

Off-Label Use

An off-label-use narrative should describe:

Do not make a regulatory conclusion about off-label status unless the case and regional product information support it.

The narrative can state:

The physician reported that Product X was being used off label for Condition Y.

This is different from the case processor independently declaring a use off label without sufficient regulatory context.

Pregnancy Cases

Pregnancy narratives often require a longer time horizon.

Relevant information can include:

The chronology should distinguish:

maternal exposure → pregnancy course → outcome → infant course.

Do not compress these into a single sentence merely because the structured fields contain the details.

Breastfeeding Exposure

Relevant elements can include:

If no infant event occurred, the narrative should not create one.

Parent–Child/Foetus Cases

E2B(R3) can represent parent-child relationships.

The narrative should make clear:

For example:

The mother received Product X during the first trimester. The infant was born with...

This avoids ambiguity about whether the child directly received the medicine.

Lack of Efficacy or Effect

Where lack of efficacy is reportable or clinically relevant, the narrative should describe:

Do not equate every worsening disease with lack of efficacy unless the source supports that interpretation.

Product-Quality Complaints With Adverse Events

A combined quality/PV case should explain:

For example:

The patient reported that the prefilled syringe leaked during administration and that only part of the dose appeared to be delivered. The patient's symptoms worsened over the following two days.

The narrative should not attempt to conclude the technical root cause unless the quality investigation established it.

Counterfeit or Falsified Products

If product authenticity is uncertain, the narrative should preserve that uncertainty.

Relevant information can include:

Do not describe the authentic product as the exposure if the actual contents are unknown.

Biologicals and Vaccines

For biologicals, product-specific details can be important.

Where available and relevant, include:

For vaccines, include:

Narrative detail should support product traceability without replacing the structured fields.

Devices and Combination Products

Where a medicinal product is delivered by a device, distinguish:

For example:

The autoinjector failed to deploy fully. The patient believed that only part of the dose was administered. No immediate adverse event occurred.

Do not collapse the device problem and clinical outcome into one ambiguous sentence.

Multi-Source Cases and Disagreement

Multi-Source Cases

One ICSR may contain information from:

The narrative should synthesise the sources without erasing attribution.

If sources disagree, say so.

If one source provides a later, better-supported diagnosis, the chronology can show that evolution while preserving the earlier report.

The Narrative Is Not the Place to Resolve Every Disagreement

Some disagreements require sender comments or medical assessment rather than narrative editing.

For example:

Reporter:

“The reaction was caused by Product X.”

MAH:

“Alternative cause more likely.”

The narrative should preserve the reporter's view with attribution.

The MAH's assessment belongs in:

This separation improves transparency.

Privacy and Personal Data

A case narrative is a clinical document, not a source-document archive.

EU GVP states that narratives should not include information that could lead to identification of the patient, including references to healthcare professionals or treatment centres.

This principle is now supplemented by GVP Module VI Addendum II on masking of personal data in ICSRs submitted to EudraVigilance, adopted in 2025.

The practical objective is to preserve information needed for pharmacovigilance while avoiding unnecessary personal identifiers.

Clinical Relevance Versus Identifiability

The narrative may need:

It generally does not need:

Data minimisation should not remove clinically important information.

For example, replacing an exact birth date with age can preserve medical meaning while reducing identifiability.

Treatment Centres and Healthcare Professionals

Avoid unnecessary statements such as:

The patient was admitted to St Example Hospital under Dr Smith.

The medically relevant information is usually:

The patient was admitted to hospital.

If the type of centre matters clinically, a generic description may be enough:

The patient was transferred to a tertiary transplant centre.

Any retention of identifying detail should follow the applicable EU masking guidance and the organisation's controlled privacy process.

Geographic Information

Country can be important for:

Exact town or facility can be unnecessary and potentially identifying.

The narrative should preserve the minimum geographic detail needed for the clinical/regulatory purpose.

Rare Diseases and Re-Identification Risk

A combination of apparently harmless details can identify a patient in a rare case.

For example:

Narrative QC should therefore consider combined identifiability, not only obvious names and addresses.

Free-Text Copying Is a Privacy Risk

Copying text directly from:

can import identifiers inadvertently.

Examples include:

The narrative should summarise clinically relevant content rather than paste source text uncritically.

E2B Technical and Supporting-Document Boundaries

H.1 Is Not an Attachment Field

The EU ICSR Implementation Guide states that case narratives should not be sent as file attachments.

E2B(R3) H.1 is the designated narrative field.

The current EU implementation guide allows up to 100,000 characters in H.1.

That technical capacity is not a target length.

A good narrative should still be focused, factual and clear.

If a narrative approaches the technical limit, the writer should first ask whether the text contains:

that could be synthesised more effectively.

Supplementary Documents

Supporting documents can be important where the case cannot be interpreted fully without them.

Examples include:

The EU implementation guide does not recommend routine attachment of every document.

The narrative should summarise key information.

Documents can then be provided according to the applicable technical and regional requirements.

Narrative Language

In EU electronic reporting, the applicable E2B language requirements should be followed.

A global safety system may receive source information in many languages.

The writer should ensure that translation preserves:

The original-language source should remain traceable according to the organisation's records process.

Translation Should Not Improve the Source

A translator or narrative writer should not transform an imprecise phrase into a more specific medical diagnosis.

For example:

source:

“pressure in the chest”

should not become:

“angina”

unless the source establishes that diagnosis.

The translation should preserve meaning, not optimise clinical terminology.

Interfaces With Structured Data and Other Standards

Consistency With MedDRA Coding

The narrative and MedDRA-coded event fields serve different purposes.

The narrative preserves source meaning and clinical chronology.

MedDRA standardises medical concepts for transmission and analysis.

If the source says:

“heart racing”

and the coded term is a suitable standardised concept, the narrative can still retain the reporter's wording while the structured field contains the MedDRA representation.

The companion article MedDRA Coding in Pharmacovigilance: A Practical Guide explains that coding should not manufacture diagnoses unsupported by the source.

The same principle applies to narrative writing.

Consistency With WHODrug Coding

The narrative can contain the reported product name and clinically relevant treatment detail.

WHODrug can standardise the medicine in structured data.

A narrative should not replace the reporter's uncertain product description with a specific trade product merely because the coding system later found a probable match.

The companion article WHODrug Coding in Pharmacovigilance explains how product identity should be supported by the source and context.

Consistency With Day Zero

Narrative chronology and regulatory receipt chronology are different.

The narrative may say:

The patient developed the event on 4 May.

Day Zero may be:

10 May, when sufficient reportable information reached the MAH.

Both can be correct.

Do not rewrite the patient timeline to match the regulatory awareness date.

The distinction is developed in Day Zero in Pharmacovigilance: When Does the Reporting Clock Start?.

Practical Writing Method

Narrative Template Versus Narrative Thinking

Templates can improve consistency.

A useful template may prompt for:

But a template should not force every case into identical prose.

A medication-error case and a fatal hepatotoxicity case need different emphasis.

The writer should use the template as a completeness aid, not as a substitute for medical reasoning.

Six-Pass Writing Method

A high-quality narrative can be built in six passes.

Pass 1 — Extract the Timeline

Before writing prose, list:

Do this in chronological order.

Pass 2 — Identify Clinically Relevant Context

Add:

Exclude unrelated background information.

Pass 3 — Separate Facts From Opinions

Mark each statement as:

This prevents attribution errors.

Pass 4 — Draft the Medical Story

Write the story in the patient's chronology.

Use direct, factual sentences.

Avoid:

Pass 5 — Compare With Structured Data

Check:

Resolve contradictions before submission.

Pass 6 — Privacy and Readability Review

Remove unnecessary identifiers.

Then read the narrative as a clinician who has never seen the case.

If the clinician cannot answer:

the narrative needs revision.

Sentence Construction

Narratives benefit from short factual sentences connected by chronology.

For example:

Product X was started on 2 January at 50 mg once daily for rheumatoid arthritis. On 14 January, the patient developed fever and productive cough. On 16 January, the patient was admitted with hypoxia. Chest radiography showed right lower-lobe consolidation.

This is clearer than one long sentence containing every fact.

Use Dates Strategically

Exact dates are useful when they establish:

Do not overload every sentence with repeated dates.

Once a temporal anchor is established, relative wording can improve readability.

For example:

Two days later...

is acceptable if the reference point is unambiguous.

Use Temporal Relationships Carefully

A narrative can state:

Three days after starting Product X, the patient developed rash.

That describes timing.

It should not automatically state:

Because Product X had been started three days earlier, it caused the rash.

Temporal association is evidence, not proof of causality.

Use Medical Terminology Without Over-Interpreting

A professional narrative should be medically precise.

But precision must come from the source.

Use:

serum creatinine increased

rather than:

renal failure

if renal failure was not reported or established.

Use:

the patient reported palpitations

rather than:

arrhythmia

if no rhythm diagnosis exists.

Active Voice Helps Attribution

Compare:

It was considered related.

with:

The treating physician considered the event related to Product X.

The second is clearer because the source of assessment is explicit.

Avoid Redundant Coding Language

Do not turn structured codes into awkward prose.

Weak:

The outcome of the reaction was recovered/resolved and seriousness criterion was hospitalisation.

Better:

The patient recovered and was discharged after four days.

The structured outcome and seriousness fields can carry the regulatory labels.

The narrative should carry the clinical meaning.

Quality Control

Narrative QC should assess more than grammar.

A useful QC framework has five dimensions.

Completeness

Does the narrative include all clinically relevant information?

Accuracy

Does every statement reflect the source or a clearly attributed assessment?

Chronology

Can the sequence of exposure, event, investigation, treatment and outcome be followed?

Consistency

Does the narrative agree with structured ICSR data?

Privacy

Has unnecessary identifying information been removed or handled according to current EU guidance?

Narrative-to-Field Reconciliation

A focused QC comparison can use:

Narrative element Structured fields to compare
Patient age/sex D section
Medical history D.7 / related history elements
Reaction/event E section
Seriousness E.i seriousness criteria
Outcome E.i outcome
Test results F section
Suspect/concomitant products G section
Drug dates/dose/route G dosing elements
Causality G.k.9.i
Reporter opinion H.2 / relevant structured fields
Sender assessment H.4
Narrative H.1

This is not a requirement to repeat each field in H.1.

It is a consistency check.

Quality Review Should Be Risk Based

Not every case requires identical review intensity.

Higher scrutiny may be appropriate for:

The organisation should define its QC model.

The principle is that narrative quality should be sufficient for the medical and regulatory importance of the case.

Automation and Version Control

Automated Narrative Generation

Safety systems increasingly use templates, rules or language models to draft narratives.

Automation can help with:

It can also introduce serious errors.

Common risks include:

Automated narrative generation therefore requires human oversight and validation appropriate to its intended use.

Source-Grounded Automation

A safer automated approach constrains generation to:

The system should be able to distinguish:

reported from derived from assessed.

A fluent narrative is not evidence of an accurate narrative.

Automated Consistency Checks

Automation can be especially useful for detecting discrepancies such as:

These checks support human review rather than replace it.

Version Control of Narratives

Each submitted follow-up may change the narrative.

The safety system should preserve:

This enables reconstruction during:

The latest narrative should not erase the existence of earlier information.

Worked Examples

The following examples are illustrative. They are designed to demonstrate narrative reasoning, not to prescribe fixed wording.

Example 1 — Serious Spontaneous Case

Source information

Weak narrative

A 62-year-old male patient experienced pneumonia, fever, cough and dyspnoea. The event was serious due to hospitalisation. Product X was stopped. The patient recovered. The physician considered the event unrelated.

The facts are present, but the medical course is fragmented and the timing is lost.

Stronger narrative

A 62-year-old man with a history of hypertension started Product X 100 mg once daily on 1 April for Condition Y. On 12 April, he developed fever and productive cough. His dyspnoea subsequently worsened, and he was admitted to hospital on 14 April with an oxygen saturation of 86%. Chest radiography showed bilateral pulmonary infiltrates, and pneumonia was diagnosed. Product X was discontinued on admission. The patient was treated with intravenous antibiotics and supplemental oxygen and improved clinically. He was discharged on 21 April. The treating physician considered the pneumonia unrelated to Product X.

Why this works

The narrative:

Example 2 — Provisional Diagnosis Becomes Confirmed

Source information

Initial report: - severe abdominal pain; - nausea; - lipase 880 U/L; - possible pancreatitis.

Follow-up: - CT showed pancreatic inflammation; - gastroenterologist confirmed acute pancreatitis; - Product X stopped; - patient recovered.

Stronger integrated narrative

The patient developed severe upper abdominal pain and nausea during treatment with Product X. Initial laboratory testing showed lipase 880 U/L, and pancreatitis was suspected. Follow-up abdominal CT demonstrated pancreatic inflammation, and the gastroenterologist subsequently confirmed acute pancreatitis. Product X was discontinued. The patient's abdominal pain and nausea resolved, and the patient recovered.

Why this works

The narrative preserves the evolution from:

provisional diagnosis → investigation → confirmed diagnosis.

It does not retrospectively make the diagnosis appear certain from the beginning.

Example 3 — Conflicting Onset Dates

Source information

Consumer: - onset 3 June.

Physician: - medical record onset 5 June.

No explicit correction from either source.

Stronger narrative

The consumer reported that symptoms began on 3 June. In follow-up, the treating physician documented symptom onset on 5 June. No additional information was available to reconcile the discrepancy.

Why this works

The conflict is transparent.

The writer does not choose whichever date makes the temporal relationship look stronger or weaker.

Example 4 — Medication Error With Clinical Consequence

Source information

Stronger narrative

The patient was prescribed Product X 5 mg once daily but accidentally took 10 mg once daily for three consecutive days. On the third day, the patient developed dizziness and symptomatic hypotension and attended the emergency department. Product X was withheld. The symptoms resolved after supportive treatment. No further clinical consequences were reported.

Why this works

The narrative explains:

It is more useful than simply listing “medication error, overdose, hypotension”.

Example 5 — Fatal Case With Competing Cause

Source information

Stronger narrative

A patient with metastatic malignancy received Product X for Condition Y. During treatment, the patient developed pulmonary embolism, which was treated with anticoagulation and subsequently considered resolved by the treating physician. The patient's underlying malignancy continued to progress, and the patient died on 18 September. The physician reported disease progression as the cause of death. No autopsy was performed.

Why this works

The narrative separates:

It avoids implying that the pulmonary embolism was fatal merely because the patient eventually died.

Example 6 — Pregnancy Exposure

Source information

Stronger narrative

The patient received Product X from approximately gestational weeks 4 through 8, when pregnancy was recognised and treatment was discontinued. No maternal adverse event was reported. Prenatal assessments were reported as normal. The pregnancy progressed to term, and a healthy infant was delivered. No congenital anomaly or neonatal adverse event was reported at the time of follow-up.

Why this works

The narrative follows:

exposure → pregnancy course → outcome → infant status.

It does not invent reassurance beyond the follow-up available.

Common Narrative Failure Modes

The following are hypothetical failure modes, not actual inspection findings.

Receipt Chronology Replaces Clinical Chronology

Narrative:

Initial report received Monday. Follow-up received Wednesday. Further follow-up received Friday...

Why it fails: the patient's clinical course is buried inside administrative chronology.

Structured Fields Copied Into Sentences

Narrative:

The outcome was recovered/resolved. The seriousness criterion was hospitalisation. Action taken with drug was withdrawn.

Why it fails: regulatory labels are repeated without explaining the medical story.

Unsupported Diagnosis Added

Source:

“chest tightness and shortness of breath.”

Narrative:

“The patient developed anaphylaxis.”

Why it fails: the narrative writer has created a diagnosis.

Reporter Opinion Becomes Company Fact

Source:

“The physician considered Product X related.”

Narrative:

“Product X caused the event.”

Why it fails: attribution and uncertainty have been removed.

Company Assessment Rewrites the Source

Reporter states:

“drug-induced hepatitis.”

MAH considers:

“viral hepatitis more likely.”

Narrative:

“The patient developed viral hepatitis.”

Why it fails: the narrative has silently replaced the primary-source assessment.

Contradiction Hidden

Two sources give different onset dates, but the narrative uses one without explanation.

Why it fails: traceability is lost and the selected date may affect causality assessment.

Every Laboratory Value Included

The narrative reproduces an entire laboratory table.

Why it fails: clinically important findings become difficult to identify.

Important Negative Finding Omitted

The case concerns suspected pulmonary embolism, but negative CT angiography is left out because the result was “normal”.

Why it fails: evidence that argues against the diagnosis is clinically relevant.

Long Concomitant List Dominates the Narrative

Twenty stable medications are copied into prose.

Why it fails: important temporal and causal information becomes obscured.

Follow-Up Appended Instead of Integrated

Each follow-up becomes a new paragraph beginning “Follow-up received...”.

Why it fails: the narrative stops functioning as a stand-alone clinical story.

Personal Identifiers Copied From Hospital Records

Names, medical-record numbers and treating-centre details appear in H.1.

Why it fails: unnecessary identifying information has entered the transmitted narrative.

Automation Creates a Fluent Hallucination

An automated system converts “possible seizure” into “confirmed seizure” and inserts an unreported date.

Why it fails: readability has replaced source fidelity.

Inspection and Governance

Inspection Perspective

An inspector reviewing narrative quality is likely to assess whether the organisation can consistently convert source information into medically useful, traceable ICSRs.

Potential questions include:

The evidence can include:

The objective is effectiveness, not stylistic uniformity.

Narrative Quality Metrics

Narrative quality is difficult to reduce to one percentage.

Useful indicators can include:

Metrics should be interpreted by defect type and clinical significance.

A low number of grammatical corrections does not prove medical accuracy.

Vendor-Prepared Narratives

Where vendors prepare case narratives, the MAH should ensure alignment on:

Vendor quality should be assessed on medical accuracy and traceability, not only turnaround time.

Medical Review

A medical reviewer should focus on whether the narrative supports correct clinical interpretation.

Questions can include:

Medical review should not make the narrative more polished at the cost of source fidelity.

Practical Narrative Checklist

Before finalising H.1, verify:

Relationship With Other QPPV.com Articles

Narrative quality depends on several upstream processes.

GVP Module VI: ICSR Validity, Minimum Information and Case Assessment explains whether the report constitutes a valid ICSR.

GVP Module VI: Causality Assessment and Clinical Evaluation of ICSRs addresses medical assessment that should remain distinguishable from source facts.

GVP Module VI: ICSR Data Quality, Follow-Up and Reconciliation covers the broader quality and follow-up system.

Day Zero in Pharmacovigilance: When Does the Reporting Clock Start? explains regulatory awareness dates, which should not be confused with patient-event chronology.

MedDRA Coding in Pharmacovigilance: A Practical Guide explains standardisation of clinical concepts.

WHODrug Coding in Pharmacovigilance explains standardisation of medicinal-product information.

Together, these processes create the structured and narrative representation of the same clinical case.

Key Takeaways

References

  1. European Union. Commission Implementing Regulation (EU) No 520/2012, consolidated version current from 12 February 2026. Article 28 specifies ICSR content, including the case narrative requirement in Article 28(3)(m).
    https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02012R0520-20260212

  2. International Council for Harmonisation. ICH E2D(R1): Post-Approval Safety Data: Definitions and Standards for Management and Reporting of Individual Case Safety Reports. Final Step 4 guideline, adopted 15 September 2025. Section 6.2 addresses the role of narratives.
    https://database.ich.org/sites/default/files/ICH_E2D%28R1%29_Step4_FinalGuideline_2025_0819.pdf

  3. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module VI — Collection, management and submission of reports of suspected adverse reactions to medicinal products, Rev. 2. Section VI.C.6.2.2.4 addresses case narrative, comments and causality assessment.
    https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/guideline-good-pharmacovigilance-practices-gvp-module-vi-collection-management-submission-reports-suspected-adverse-reactions-medicinal-products-rev-2_en.pdf

  4. European Medicines Agency. GVP Module VI Addendum II — Masking of personal data in individual case safety reports submitted to EudraVigilance. EMA/178902/2025, adopted 22 July 2025.
    https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/guideline-good-pharmacovigilance-practices-gvp-module-vi-addendum-ii-masking-personal-data-individual-case-safety-reports-submitted-eudravigilance_en.pdf

  5. International Council for Harmonisation. E2B(R3) Individual Case Safety Report Specification and Related Files. Current implementation package and Q&A material; package updated January 2026.
    https://admin.ich.org/node/348

  6. European Medicines Agency. European Union Individual Case Safety Report (ICSR) Implementation Guide. EMA/51938/2013 Rev. 2. Includes EU technical implementation guidance for H.1 and supplementary documents.
    https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/european-union-individual-case-safety-report-icsr-implementation-guide_en.pdf

  7. Council for International Organizations of Medical Sciences. Current Challenges in Pharmacovigilance: Pragmatic Approaches — Report of CIOMS Working Group V. Geneva: CIOMS; 2001. GVP Module VI cites this historical report as an example source for a standard narrative template.

Regulatory Note

This article describes post-authorisation ICSR narrative practice as reviewed on 2 October 2026.

In the EU, the legal narrative requirement derives from Commission Implementing Regulation (EU) No 520/2012, Article 28(3)(m), which requires a case narrative where possible for individual cases other than non-serious adverse reactions.

GVP Module VI Rev. 2 remains the adopted operational guidance for ICSR collection, management and submission. ICH E2D(R1) provides the current harmonised good-case-management framework and should be read together with applicable EU requirements. E2D(R1) does not replace regional legal requirements.

GVP Module VI Addendum II provides current supplementary EU guidance on masking of personal data in ICSRs submitted to EudraVigilance and should be applied when preparing free-text narrative content.

The worked examples and failure modes in this article are illustrative. They are not actual cases or inspection findings.

Revision History

Related Articles

Day Zero in Pharmacovigilance: When Does the Reporting Clock Start?

Explains when the ICSR reporting clock starts, why Day Zero is not simply the database-entry date, how source-specific rules operate under ICH E2D(R1) and EU GVP, and how to control awareness dates across vendors, partners, literature, digital channels and follow-up.

ICH E2D(R1): What Changed in Post-Approval Pharmacovigilance

Explains why ICH E2D was revised, what changed in E2D(R1), how the new source and report classifications affect ICSR management, and what MAHs and QPPVs should have implemented in the EU after the 2026 transition.

GVP Module VI Inspection Findings: Common ICSR Case-Processing Deficiencies

A practical inspection-focused guide to recurring weaknesses in individual case safety report processing and the evidence that demonstrates effective control.

Last reviewed: 2026-10-02

QPPV.com