Seriousness, Severity, and Expectedness: The Triad that Drives Reporting

Once an event is identified, three determinations control routing and timelines: seriousness, severity, and expectedness. Getting any of these wrong can lead to late or missing expedited reports, inappropriate unblinding, or misleading aggregate conclusions. Agencies such as the FDA, EMA, PMDA, and TGA expect these assessments to be traceable in the case file and consistent with ICH E2A/E2D principles under the ICH framework; the WHO offers public-health guidance in resource-limited and vaccine settings.

Severity grading. Use clinically appropriate grading systems (e.g., CTCAE for oncology/toxicities) or a standardized mild/moderate/severe scale for other indications. Consistency matters—do not mix scales within a study without governance. Severity informs clinical interpretation and dose-modification rules but does not decide reporting clocks.

Seriousness determination. Apply the outcomes-based criteria objectively, independent of causality. One or more criteria make the event “serious.” Document the specific criterion chosen and explain the clinical reasoning in the narrative (e.g., “hospitalization for IV antibiotics due to community-acquired pneumonia; unplanned, length 3 days”). For IMEs, justify why the event qualifies (e.g., “required urgent bronchodilator and systemic steroids to prevent respiratory failure”).

Expectedness against RSI. Expectedness refers to whether the nature or severity of an adverse reaction is consistent with the Reference Safety Information (RSI) in the IB (development) or the product label/SmPC (post-marketing). Practical rules:

• Compare to the most current, approved RSI for the study at the time of onset.
• Match at the appropriate specificity (preferred term or concept level), not just the SOC; “rash” does not automatically cover “Stevens-Johnson syndrome.”
• Intensity matters: if RSI lists “transaminase increase (mild to moderate),” a case meeting Hy’s Law (ALT/AST ≥3× ULN with total bilirubin ≥2× ULN and no cholestasis) is unexpected in severity.

From determinations to routing. The canonical logic chain is:

1. Is it an AE? If no, document rationale (e.g., prespecified clinical outcome not representing harm) and close.
2. Is it serious? If yes, proceed regardless of severity.
3. For serious events, is there a reasonable possibility that the product caused the event (causality positive or cannot be ruled out)? If yes, it is a serious adverse reaction (SSAR/SAR).
4. Is the reaction unexpected vs RSI? If yes, classify as SUSAR and follow expedited timelines and distribution lists per region.

Edge cases.

• Planned hospitalization (e.g., elective surgery) is not an SAE unless prolonged due to an AE or associated with a complication that meets criteria.
• Progression of underlying disease may be an AE or endpoint depending on protocol; clarity in the protocol and SAP prevents duplicate adjudication.
• Lack of efficacy in life-threatening disease: deterioration requiring rescue ICU admission may be an SAE and, if related, can be a serious adverse reaction.
• Vaccines: IMEs include anaphylaxis, ADEM, myocarditis/pericarditis patterns; vaccine PV leans heavily on expectedness tables and background rates for causality context, aligned with WHO guidance.

Device and combination products. Device-related incidents (malfunctions that could lead to death or serious injury) intersect drug PV. Record device model/lot, software version, and use a harmonized terminology (device problem codes) alongside MedDRA for patient impact. Route device vigilance in parallel to medicinal product safety as required by regional rules.

Attribution That Stands Up to Review: Causality Frameworks, Bias Guards, and Special Scenarios

Regulators expect a transparent and reproducible approach to attribution—was there a reasonable possibility that the investigational product caused the event? The answer determines whether an AE becomes an AR (and possibly a SUSAR) and appears in aggregate analyses and labeling decisions. Attribution is fundamentally clinical but should be disciplined by structured frameworks accepted by the FDA, EMA, PMDA, TGA, and coherent with ICH E2D/E2A and the WHO UMC model.

Minimum workable scale. Many sponsors adopt a 4- or 5-level scale mapped to “related” vs “not related” for reporting purposes:

• Not related / Unrelated
• Unlikely related
• Possible
• Probable / Likely
• Certain

For expedited-reporting logic, levels ≥“possible” generally equal “reasonable possibility.” Clearly document the study’s mapping in the SMP and protocol.

Structured tools. Two families of tools help standardize judgement:

• WHO-UMC causality categories (certain, probable/likely, possible, unlikely, conditional/unclassified, unassessable/unclassifiable) emphasize clinical plausibility, time relationship, dechallenge/rechallenge, and alternative explanations.
• Naranjo algorithm assigns points for temporal sequence, dechallenge/rechallenge, alternative causes, dose–response, and prior conclusive reports to yield a category. Useful when training new investigators; always overlay medical judgment.

Guarding against bias. Implement practices that reduce variability and hindsight bias:

• Collect first, judge later: ensure high-quality clinical facts (onset, meds, comorbidities, labs) before forcing a causality choice.
• Use structured prompts in the eCRF/SAE form (temporal plausibility, alternative causes including disease, infection, concomitant meds or procedures; dechallenge response; rechallenge info if ethical).
• Require rationale text for “related,” “not related,” and “uncertain” selections; narratives should echo this logic.
• Harmonize sponsor and investigator causality: capture both; for reporting, if either indicates “reasonable possibility,” treat as suspected for expedited purposes unless regional rules dictate otherwise.
• Segregate unblinded reviews to avoid operational unblinding; the independent unblinded physician may review patterns across arms when necessary.

Case-type nuances.

• Immune-mediated events (e.g., myocarditis with certain platforms): latency, biomarkers, imaging, and response to steroids inform plausibility; background rates by age/sex matter in risk assessment.
• Drug-induced liver injury (DILI): use RUCAM or similar structured causality tools; Hy’s Law signals deserve heightened scrutiny and expedited escalation pathways.
• Infusion reactions: premedication, timing relative to infusion, and recurrence upon rechallenge guide causality; distinguish from hypersensitivity mediated by IgE vs cytokine-release syndromes.
• Medication errors: causality relates to harm, not the error itself; if harm is due to dosing confusion from labeling, attribution to product may be reasonable.
• Lack of efficacy in life-threatening disease: may be causally related if pharmacology predicts attenuation or if product–disease interactions plausibly worsen outcomes; most are confounded by disease course and remain “not related.”
• Device malfunctions (combination products): attribute patient harm to the product system; document the device problem code and investigate lot/serial trends.

Expectedness and causality interplay. Expectedness is assessed only for reactions. If causality is “not related,” expectedness is moot. If related or cannot be ruled out, compare to RSI. Borderline examples (e.g., “rash” vs “SJS/TEN”) require conservative judgement and clear justification in the file.

Vaccines and background rates. For vaccines and high-incidence AEs in the general population (e.g., headaches, fevers), attribute using time-to-onset, biological plausibility, and background rates from reliable sources. The WHO and national programs provide frameworks for causality in immunization safety; document sources and reasoning.

From Definitions to Daily Practice: Forms, Workflows, Quality Controls, and Inspection Readiness

Definitions and attribution frameworks only help if they are embedded in daily operations and documented for inspectors. Build the safety system so that any reviewer from the FDA, EMA, PMDA, TGA, or public-health programs aligned with the WHO can reconstruct what happened, why you classified it the way you did, and whether timelines and distribution were met.

Design the SAE form to drive correct classification. Include mandatory fields for seriousness criteria (with explicit IME pick-list), severity grading scale selection, temporality (start/stop), outcome, hospitalization details (admission, prolongation, reason), and expectedness vs RSI with version/date. Provide structured causality questions (temporal plausibility, alternative causes, dechallenge/rechallenge). For device components, capture model/lot/software version and device problem codes.

Embed RSI governance. Maintain a controlled list of RSI versions per protocol and subject enrollment period. The safety database should time-stamp which RSI version applied at onset and preserve the text used for expectedness decisions. When the IB/label changes, update site training and informed-consent materials as needed and document communications.

MedDRA/WHO-DD discipline. Enforce version control, change-control logs, and coding QC for terms that affect seriousness and expectedness (e.g., distinguishing “Stevens-Johnson syndrome” from “rash”). Map commonly mis-coded terms and train on preferred term selection. Ensure concomitant medications are coded with WHO-DD to support alternative-cause analysis.

Case processing SOPs that reflect definitions. SOPs should specify: minimum criteria for a valid case; who sets day 0; how seriousness, severity, and expectedness are determined; how investigator and sponsor causality are captured; and how SUSAR logic triggers expedited reporting (including distribution lists to regulators, investigators, and IRBs/IECs). Integrate escalation rules for potential Hy’s Law, anaphylaxis, or other IMEs where rapid action is needed.

Quality controls and metrics.

• Classification accuracy: periodic audits comparing narratives to seriousness/severity/expectedness selections; error rate trends by site and vendor.
• Timeliness: receipt-to-submission cycle times; % SUSARs submitted within regulatory clocks (region-stratified).
• RSI alignment: proportion of expectedness decisions supported by the correct RSI version; mismatches escalated to CAPA.
• Causality rationale completeness: % of “possible/probable/certain” with explicit rationale fields populated.
• Coding agreement: dual coding concordance for high-impact terms.

Training and competency. Build role-based curricula for investigators, coordinators, case processors, coders, and safety physicians. Include case vignettes illustrating severity vs seriousness, IMEs, RSI expectedness, and borderline causality. Test competency, not just completion, and gate system access on passing scores.

Blinding and independence. Keep operational dashboards arm-agnostic. Unblinded safety review (when required) should be done by independent personnel with segregated systems and access logs. Emergency unblinding procedures should be scripted, time-stamped (with local time and UTC offset), and impact-assessed for analysis plans.

Reconciliation with EDC. Reconcile SAE forms in EDC with the safety database on a fixed cadence; resolve discrepancies in onset dates, seriousness, causality, and outcome. Track open mismatches and age; perform root-cause analysis for recurring issues at specific sites or for specific terms.

Documentation and inspection readiness. Maintain a rapid-pull index of: SOPs/WIs; RSI version history; MIE list; causality scale definitions and mapping to “reasonable possibility”; training logs; coding QC plans; expedited reporting logs with acknowledgments; and examples of narratives that justify seriousness and expectedness. Inspectors should be able to read the story of any case from first awareness to submission and communication to stakeholders.

Governance and continuous improvement. Safety committees should review metrics and case audits, update MIE lists, refine causality guidance, and adjust training. Feed learnings into DSURs/PBRERs and—where risk signals emerge—into labeling and Risk Management Plans (RMPs) or U.S. REMS, coordinating with regulatory authorities such as the FDA, EMA, PMDA, and TGA.

Quick reference—practical checklist.

• AE/SAE definitions and seriousness criteria are embedded in forms and SOPs; “SSAE” usage (if adopted) is defined and cross-walked to SAR/SSAR/SUSAR.
• Severity scale standardized and documented; seriousness decisions independent of severity but consistently justified.
• Expectedness assessed against the correct RSI version with text/section references captured in the case.
• Causality captured from both investigator and sponsor with rationale; mapping to “reasonable possibility” for expedited rules documented.
• MedDRA/WHO-DD versions controlled; high-impact terms dual-coded or QC-checked.
• Expedited logic (SUSAR) automated where possible and verified by medical review; distribution lists current by region.
• Training includes case vignettes on IMEs, DILI/Hy’s Law, anaphylaxis, infusion reactions, device incidents, and vaccine-specific patterns.
• Reconciliation processes with EDC are active; discrepancies aged and trended with CAPA when thresholds breach.
• Blinding safeguards in place; unblinded roles segregated; emergency unblinding workflow documented.
• Inspection pack ready: SOPs, RSI history, MIE list, coding QC, expedited logs, narratives, metrics, and CAPA records.

Bottom line. Clear terminology and disciplined attribution turn safety from a compliance burden into a decision system. When definitions are embedded in forms and SOPs, expectedness ties to a controlled RSI, causality is reasoned and documented, and operations are measured and improved, organizations can meet the expectations of the FDA, EMA, PMDA, TGA, aligned with ICH standards and the WHO mission to safeguard public health.