AE vs. SAE vs. SUSAR: Clinical Safety Terminology Explained

It is 6pm on a Friday and an intake email lands: a participant on your Phase II study was admitted to hospital overnight. The site called it "an AE, probably nothing." But the word you write on that case in the next few hours decides whether a reporting clock starts ticking, who has to be told, and how fast. Misjudge it and you either breach a regulatory timeline or bury a genuine safety signal under a pile of non-actionable reports. Few decisions in clinical trials are this small on paper and this consequential in practice.

Adverse event classification is the quiet backbone of pharmacovigilance. Get AE, SAE, and SUSAR right, consistently, and the safety system works: signals surface, regulators stay informed, participants stay protected. Get them wrong and the failure is rarely dramatic in the moment — it shows up later, as an inspection finding, a missed signal, or a reporting breach that erodes trust in the whole trial.

AE — Adverse Event

An adverse event is any untoward medical occurrence in a participant administered a pharmaceutical product, whether or not it is considered related to that product. The operative word is any. A headache, a mildly elevated liver enzyme, a twisted ankle in the car park — all are AEs if they occur during the trial. Causality is deliberately irrelevant at this stage; you capture first and assess later.

AEs are recorded on the CRF and coded, typically to MedDRA, so that "heart attack" and "myocardial infarction" roll up to the same preferred term in analysis. The discipline here is completeness: an event that never gets recorded can never be escalated, so the AE layer is the widest part of the funnel by design.

SAE — Serious Adverse Event

An adverse event becomes a serious adverse event when it meets at least one seriousness criterion. Under ICH definitions, an event is serious if it:

• Results in death,
• Is life-threatening,
• Requires inpatient hospitalisation or prolongs an existing hospitalisation,
• Results in persistent or significant disability or incapacity,
• Is a congenital anomaly or birth defect, or
• Is another important medical event that may jeopardise the participant or require intervention to prevent one of the outcomes above.

The single most common classification error lives here: confusing serious with severe. Serious is about outcome, not intensity. A severe migraine that resolves with paracetamol is not serious. A brief hospitalisation for observation after a fainting episode is serious. Severity is a three-point intensity scale (mild, moderate, severe); seriousness is a regulatory gate. Conflating them is how a reportable SAE gets logged as a routine AE and quietly misses its clock.

SAEs trigger expedited internal reporting — commonly the site must notify the sponsor within 24 hours of becoming aware — and the event enters the safety database for assessment of causality and expectedness.

SUSAR — Suspected Unexpected Serious Adverse Reaction

A SUSAR is the most tightly regulated category, and it is defined by three conditions that must all hold at once:

• Suspected — there is a reasonable possibility the event is related to the investigational product. Note the shift in language from event to reaction: relatedness is now in scope.
• Unexpected — the nature, severity, or frequency of the event is not consistent with the applicable reference safety information (typically the Investigator's Brochure or its equivalent).
• Serious — it meets at least one SAE seriousness criterion.

Miss any one condition and it is not a SUSAR. An expected, related, serious reaction (already listed in the Investigator's Brochure) is not a SUSAR because it is not unexpected. A serious, unexpected event with no plausible relationship to the product is not a SUSAR because it is not suspected. All three gates, every time.

Read the funnel from top to bottom and the logic is clean: every SUSAR is an SAE, and every SAE is an AE, but the reverse is never guaranteed. Each layer adds a criterion and narrows the population.

The clocks: 7 days, 15 days, and why they differ

SUSARs carry the strictest expedited reporting obligations to regulators and ethics committees, and the timelines split by outcome:

• Fatal or life-threatening SUSARs — reported as soon as possible and no later than 7 calendar days after the sponsor becomes aware, with a complete follow-up report within a further 8 days.
• All other SUSARs — reported within 15 calendar days of the sponsor becoming aware.

The clock starts on sponsor awareness, not on the event date — which is exactly why intake, triage, and reconciliation speed matter so much. A case that sits unread in an inbox over a weekend is already burning days off a 7-day clock.

Why misclassification is dangerous — in both directions

Under-classifying is the failure everyone fears: calling an SAE a routine AE, or failing to recognise a SUSAR, breaches reporting timelines and, far worse, delays a signal that might protect other participants still on the study. But over-classifying is a real harm too. Flooding regulators and ethics committees with non-actionable reports buries the meaningful signals in noise and burns out the people who have to read them. The goal is not to escalate everything; it is to classify accurately, every time, so the reports that go out are the ones that matter.

Two operational habits protect against both failures: clear, accessible reference safety information so expectedness calls are consistent, and disciplined SAE reconciliation between the clinical (EDC) and safety (pharmacovigilance) databases. A SUSAR living in the safety system with no matching event in the clinical data — or vice versa — is a classic inspection finding and a sign the two systems have drifted.

Where AI assists — and where it must not

The repetitive, high-volume parts of this work are well suited to machine support. AI can read source documents and flag potential adverse events and symptoms a busy coordinator might miss, surface events that appear to meet a seriousness criterion for human confirmation, help reconcile clinical and safety databases continuously rather than in a pre-lock scramble, and support safety signal detection across the accumulating dataset.

What AI must not do is make the calls that require medical judgment. Causality and expectedness are clinical determinations with regulatory weight; the appropriate pattern is machine-surfaces-candidates, clinician-decides. AI widens the net and speeds the triage; a qualified human still owns the classification and the report.

The bottom line

AE is any event. SAE adds a seriousness outcome. SUSAR adds relatedness and unexpectedness on top of seriousness. These are not academic distinctions — each layer starts or withholds a reporting clock that exists to protect the people in your trial. Precision here is not paperwork discipline; it is patient safety.