How adverse event attribution shapes the drug label

Adverse event attribution is the judgment of whether a drug caused an adverse event recorded in a clinical trial. Attribution shapes the drug label because FDA's labeling rule limits the Adverse Reactions section to events with "some basis to believe there is a causal relationship" (21 CFR 201.57). The same rule requires a warning about a clinically significant hazard as soon as there is "reasonable evidence of a causal association."

Much of the evidence behind those judgments (timing, the course after the drug stops, alternative causes) is recorded at the encounter. Settled during the trial, a causality question can still change what the sponsor reports under the IND; settled after approval, it can mean a labeling supplement. Complete attribution evidence helps elucidate the safety profile the label will carry.

What is adverse event attribution (causality assessment) in clinical trials?

Adverse event attribution, also called causality assessment or relatedness, is the judgment of whether the drug caused an adverse event. The event itself is defined broadly: "any untoward medical occurrence associated with the use of a drug in humans, whether or not considered drug related" (21 CFR 312.32(a)). In a US IND trial, attribution happens at two levels:

• The investigator reports every serious adverse event to the sponsor immediately, with "an assessment of whether there is a reasonable possibility that the drug caused the event" (21 CFR 312.64(b)).
• The sponsor makes its own determination for IND safety reporting, where a reasonable possibility means "there is evidence to suggest a causal relationship between the drug and the adverse event" (21 CFR 312.32(a)).

Common scales include NCI's five attribution categories (Unrelated, Unlikely, Possible, Probable, Definite), the WHO-UMC system and the Kramer algorithm. Each turns the same questions into a category: when the event started, what happened when the drug stopped or restarted, and what else explains it.

How does adverse event attribution reach the drug label?

Three safety sections of prescription drug labeling under 21 CFR 201.57 turn on a causal judgment, and trial attribution evidence feeds each.

Label section	What the regulation asks	Where attribution evidence feeds in
Boxed Warning, 201.57(c)(1)	Serious warnings "particularly those that may lead to death or serious injury"; the box "ordinarily must be based on clinical data"	Case evidence that a serious or fatal reaction belongs to this drug
Warnings and Precautions, 201.57(c)(6)	Revise "as soon as there is reasonable evidence of a causal association with a drug; a causal relationship need not have been definitely established"	Which reactions belong to this drug, how they presented and what resolved them
Adverse Reactions, 201.57(c)(7)	Only "those adverse events for which there is some basis to believe there is a causal relationship between the drug and the occurrence of the adverse event"	Whether an event seen in the trial counts as a reaction to this drug

Warnings and Precautions also describes "steps that should be taken if they occur (e.g., dosage modification)" and laboratory tests helpful "in identifying possible adverse reactions": both rest on how individual cases presented and resolved.

Each section is written for one product. In a combination program, an event assigned to the backbone may never reach the investigational agent's warnings; multi-drug attribution in combination regimens covers the reverse error and works a case drug by drug.

What does FDA weigh before listing an adverse reaction on the drug label?

FDA's 2006 guidance on the Adverse Reactions section calls the decision "a matter of judgment" based on factors such as:

1. The frequency of reporting
2. Whether the adverse event rate for the drug exceeds the placebo rate
3. The extent of dose-response
4. Consistency with the pharmacology of the drug
5. The timing of the event relative to drug exposure
6. Challenge and dechallenge experience
7. Whether related drugs are known to cause the event

The first three are aggregate measures; the fourth and seventh come from prior knowledge. Timing and dechallenge are facts about individual cases, and exist only if someone recorded them.

The same guidance discourages computing a reaction's rate from events "that individual investigators believe to be causally related," because doing so "introduces bias and inconsistency in rate determinations." Where a control arm exists, the comparison with it carries much of the causal weight for common events.

When does attribution at the encounter carry the most weight?

Attribution at the encounter decides the outcome where a comparator arm cannot answer the causal question on its own:

1. Rare, serious reactions. For serious events "unusual in the absence of drug therapy," such as liver failure or agranulocytosis, FDA's 2006 Adverse Reactions guidance says they are generally listed in the Adverse Reactions section "even if there are only one or two reported events, unless it is clear that a causal relationship can be excluded." With one or two cases, the case record is the evidence, in labeling as in IND safety reporting (21 CFR 312.32(c)(1)(i)(A)).
2. Single-arm trials. FDA's 2023 draft guidance on accelerated approval of oncology therapeutics notes that single-arm designs have most commonly been used for oncology accelerated approval. Their safety databases are "typically small," and "attribution of adverse events to the drug under study can be limited in the absence of a comparator arm."
3. Combination regimens. With a backbone plus an investigational agent, the question is which drug, answered for each.
4. First-in-human and early phase. With no prior human record, the sponsor's causality determination, with seriousness and unexpectedness, decides whether an event becomes an IND safety report (21 CFR 312.32(c)(1)(i)).

A worked example: one case in a single-arm trial

Take an illustrative single-arm phase 2 trial of one oral agent, Agent Y. In week 8, a patient whose baseline chest scan was clear develops cough and shortness of breath, and imaging shows new bilateral infiltrates.

• Timing. Onset eight weeks after the first dose; no new concomitant medication.
• Alternative causes. Cultures and a viral panel are negative, no chest radiation, and the radiology read does not suggest progression.
• Dechallenge. Agent Y is held per protocol, and the findings improve.

Recorded that way, the case meets the WHO-UMC criteria for Probable/Likely: a reasonable time relationship, unlikely to be attributed to disease or other drugs, a clinically reasonable response to withdrawal, rechallenge not required. Recorded only as "possible," it gives the sponsor a category without the evidence. With no comparator arm, cases like this are most of what the safety database holds when Warnings and Precautions is drafted. Treatment decisions belong to the protocol and the investigator.

What does the published evidence say about attribution?

Published studies show that individual attribution calls vary and over-estimate relatedness, and that a randomized comparison answers the causal question better where one exists:

• Hillman et al., Journal of Clinical Oncology, 2010. In the placebo arms of two phase III trials (398 patients, 7,736 adverse events), 47% and 50% of events were reported as related to study drug. When the same event was reported for the same patient on multiple visits, its attribution category changed at least once 36% and 31% of the time (Hillman et al.).
• Le-Rademacher et al., Annals of Oncology, 2017. Across nine placebo-controlled trials (2,155 patients), clinician-reported attribution over-estimated treatment-related events; the authors concluded it should be eliminated in randomized double-blind placebo-controlled trials (Le-Rademacher et al.).
• George et al., Cancer Treatment Reviews, 2019. Reporting on a workshop of regulators, sponsors, investigators and patients, the authors called attribution processes "often sub-optimal, unreliable, and inefficient," and recommended fewer tiers and more complete baseline information (George et al.).

Where no comparison exists, the call at the encounter remains the evidence, and the remedy is better inputs for capable investigators: a documented baseline, a named method and the reasoning attached.

What happens after launch: label changes and safety communications

After approval, a causality finding can lead to a labeling supplement or an FDA-ordered safety labeling change, and to a communication to prescribers and patients. During the trial, the same correction changes only the record and the reports resting on it.

• A required safety labeling change. Section 505(o)(4) of the FD&C Act lets FDA require safety labeling changes for certain drugs, and FDA "can also issue an SLC order if a labeling supplement or rebuttal statement is not submitted within 30 calendar days" (FDA, Safety Labeling Change Orders).
• A sponsor-initiated supplement. A sponsor can add or strengthen a warning or adverse reaction through a changes-being-effected supplement when "the evidence of a causal association satisfies the standard for inclusion in the labeling under § 201.57(c)" (21 CFR 314.70(c)(6)(iii)(A)): the same standard the trial data were measured against.
• Drug Safety Communications. FDA Drug Safety Communications are "intended to provide important information to patients and health care professionals about new safety issues with the medicines they are taking or prescribing."
• A visible change on the label. A substantive change to the Boxed Warning or Warnings and Precautions is listed under Recent Major Changes in Highlights, with its month and year (21 CFR 201.57(a)(5)), and FDA's Drug Safety-related Labeling Changes database records such changes since January 2016.

Of 222 novel therapeutics FDA approved from 2001 through 2010, 71 (32.0%) had a postmarket safety event (a withdrawal, an added boxed warning or a safety communication), the first arriving a median of 4.2 years after approval; accelerated approval carried an incidence rate ratio of 2.20 (Downing et al., JAMA, 2017). Not every event traces to attribution, and most arrived years after approval.

What should a sponsor ask of attribution support?

1. One score per drug, or one verdict per event? In a combination regimen, a single verdict discards the comparison between drugs the label depends on.
2. What does the record keep when the investigator disagrees? If an override replaces the suggestion, the record no longer shows what was proposed or why.
3. Where is the attribution formed? Support begins where the evidence is first recorded, at the encounter; a review of the case report form begins after the call is made.

How Burna AI supports attribution at the encounter

Burna AI scores each adverse event for attribution against every drug in the regimen, using WHO-UMC and Kramer, with the reasoning shown. Every grade traces to the line in the medical record that produced it, and each graded event carries a CTCAE v5 or v6 grade and a MedDRA LLT code.

The investigator accepts or overrides each score. A grade becomes a record when a qualified investigator signs it with a 21 CFR Part 11 electronic signature, and the signature carries what was accepted, what was overridden and the citation set as signed. The platform produces FDA MedWatch, IRB and serious adverse event reports from the graded record, all listed in the public capability record. AI suggests, clinicians decide. Human-in-the-loop, always.

A sponsor reads graded safety data for its own trials in a view separate from the site's. More on Burna AI for pharma sponsors and attribution in pharmacovigilance.

Common questions about adverse event attribution and the label

What is the difference between an adverse event and an adverse reaction?

An adverse event is any untoward medical occurrence during drug use, caused by the drug or not (21 CFR 312.32(a)). For labeling, an adverse reaction is "an undesirable effect, reasonably associated with use of a drug" (21 CFR 201.57(c)(7)). Attribution is the step that separates the two.

Who makes the causality assessment in a clinical trial?

The investigator assesses each serious adverse event under 21 CFR 312.64(b), and the sponsor makes its own determination for IND safety reporting under 21 CFR 312.32(a). For the label, the applicant proposes the wording and FDA reviews it.

Does an investigator's "related" call put an event on the label?

Not by itself. FDA weighs factors such as frequency of reporting, the rate against placebo, dose-response, timing and dechallenge across the whole safety database. The call and its evidence are inputs, and for rare serious events and single-arm trials they can be most of the evidence.

---

If you are responsible for the safety profile of a program whose label is still being written, book a 15-minute conversation.