Treatment-emergent adverse events: grading against a participant's baseline

A treatment-emergent adverse event (TEAE) is an adverse event that first appears on or after the first dose of study treatment, or a condition present before treatment that gets worse after it. ICH E9 defines treatment emergent as "an event that emerges during treatment having been absent pre-treatment, or worsens relative to the pre-treatment state." Treatment-emergent describes when an event happened relative to baseline; it says nothing about cause.

In oncology trials, participants often enroll with neuropathy left over from an earlier line of therapy, kidney impairment or diabetes, so every event raises two questions: what grade it is, and what caused it. Both answers depend on the record from before the first dose.

• Treatment-emergent is timing: new, or worse than baseline, on or after the first dose.
• A pre-existing condition unchanged from baseline is not an adverse event.
• The baseline grade is set under the CTCAE version the protocol names, and v5.0 and v6.0 can grade the same baseline differently.

What is a treatment-emergent adverse event?

A TEAE is new on or after the first dose, or a pre-existing condition that worsens after it, within a window the protocol defines.

ICH E9 (1998, section 6.3) offers it for trials with "a substantial background noise of signs and symptoms," recording adverse events "only if they emerge or worsen relative to pretreatment baseline."

The Center for Drug Evaluation and Research (CDER) Standard Safety Tables and Figures: Integrated Guide (version 2.0, April 2025) defines a TEAE as "the occurrence of an AE or worsening of an existing AE during or after the first treatment," and its adverse event tables carry the footnote "Treatment-emergent AE defined as [definition]."

For oncology applications, FDA's Oncology Center of Excellence and Office of Oncologic Diseases give a default in their pilot standard safety data requests (version 1.3): TEAEs are "new or worsening events occurring in the safety population at or after the first drug treatment up to and including 30 days after last dose of study drug or the day prior to start of subsequent therapy (whichever comes first)."

Events starting more than 30 days after the last dose that the sponsor determines are related also count, and a sponsor using another definition discusses it with the review division before submission. The ADaM analysis flag TRTEMFL (Treatment Emergent Analysis Flag) records the answer. The statistical analysis plan classifies the edge cases, such as onset on the first-dose date or a partial or missing onset date.

The definitions side by side:

Source	What counts as treatment-emergent	Window
ICH E9 (1998), glossary	New during treatment, or worse than the pre-treatment state	During treatment
FDA Integrated Guide (2025)	A new adverse event, or worsening of an existing one; each table footnote states the definition used	During or after the first treatment; no end point in the quoted definition
FDA OCE/OOD data requests (v1.3)	New or worsening events in the safety population	First dose to 30 days after the last dose or the day before subsequent therapy, whichever is first; later related events also count
NCI CTEP guidelines for DCTD INDs (2024), section 5.1	A baseline condition that worsens in severity, or resolves and then recurs	Re-assessed throughout the study

TEAE vs treatment-related adverse event: what is the difference?

A TEAE answers a question about time; a treatment-related adverse event answers a question about cause. Under 21 CFR 312.32, an adverse event is "any untoward medical occurrence associated with the use of a drug in humans, whether or not considered drug related." Relatedness is a separate call: the investigator's under 21 CFR 312.64(b), and the sponsor's when it decides whether a serious, unexpected event is a suspected adverse reaction for an IND safety report under 21 CFR 312.32(c).

The literature shows how hard that second call is:

• Hillman et al., Journal of Clinical Oncology, 2010. A recurring adverse event's attribution changes between visits for the same patient in 31 to 36% of cases, one of the findings in adverse event attribution and the drug label.
• George et al., Cancer Treatment Reviews, 2019. A workshop of regulators, sponsors, investigators and patients named "insufficient baseline assessments" among the problems with attribution and recommended "more complete baseline information on patients' clinical status at trial entry."

Baseline is the input both questions share.

How are pre-existing conditions handled at baseline?

A pre-existing condition becomes an adverse event when it worsens after the first dose, or resolves and then recurs; unchanged from baseline, it is not one. NCI's reporting guidelines for Division of Cancer Treatment and Diagnosis (DCTD) INDs say "A pre-existing medical condition identified on baseline assessment is not considered an AE," and it is re-assessed throughout the study (NCI CTEP, effective August 30, 2024, section 5.1):

1. It worsens in severity. The worsened event is reported as a routine adverse event, and the investigator reassesses whether an expedited report is required.
2. It resolves and then recurs. The recurrence is reported as a routine adverse event and reassessed the same way.

NCI's example: a participant enters with grade 1 dyspnea. If it increases in grade, it is a routine adverse event evaluated for expedited reporting. Unchanged, it needs no expedited report; hospitalization, or a grade increase that meets seriousness criteria, can change that.

Which value is the baseline?

The protocol and statistical analysis plan name the assessment, which ICH E9 places before treatment. In the datasets FDA's oncology offices request, that record carries the Baseline Record Flag (ABLFL) and its value is stored as BASE (OCE/OOD data requests).

How does CTCAE grade a change from baseline?

Some CTCAE criteria are written as a change from baseline; others grade the current state against the baseline grade. NCI names "grading AEs in patients with abnormal baseline laboratory values" as an area CTCAE v6.0 addresses.

Grade 1, quoted from NCI's CTCAE v5.0 and CTCAE v6.0:

Term	CTCAE v5.0 grade 1	CTCAE v6.0 grade 1
Diarrhea	Increase of <4 stools per day over baseline; mild increase in ostomy output compared to baseline	Change in consistency or frequency
Alanine aminotransferase increased	>ULN - 3.0 x ULN if baseline was normal; 1.5 - 3.0 x baseline if baseline was abnormal	>ULN - 3.0 x ULN if baseline was normal or less than normal; 1.0 - 1.5 x baseline if baseline was >ULN
Peripheral sensory neuropathy	Asymptomatic	Mild symptoms

ctcae.ai lists every CTCAE v6.0 change, term by term.

Diarrhea. Under v5.0, a participant who averaged three stools a day at baseline and now has four has an increase of one: grade 1. Read against zero, the same day falls in the grade 2 band ("Increase of 4 - 6 stools per day over baseline"). Under v6.0 the same day is grade 1 ("Change in consistency or frequency"), and grades 2 and 3 still read "over baseline," so the baseline stool count still decides the higher grades.

Liver enzymes. When baseline ALT was above the upper limit of normal (ULN), both versions grade against the participant's own baseline; v6.0 starts grade 1 at 1.0 times baseline, not 1.5. The full ALT and AST bands are in Grading ALT and AST from a high baseline on ctcae.ai.

Neuropathy. This term grades current symptoms, and its grade 1 change matters at enrollment. Mild residual symptoms from an earlier taxane are grade 1 under v6.0. Under v5.0, grade 1 reads "Asymptomatic" and grade 2 reads "Moderate symptoms; limiting instrumental ADL", where the semicolon means "or". Mild symptoms match neither text, so the baseline grade depends on the version the protocol names and on the study's grading convention for that gap.

Both versions grade symptoms limiting instrumental activities of daily living (ADL), such as preparing meals, as 2, and self-care ADL, such as bathing, as 3.

Why baseline comorbidities make adverse event grading and attribution harder

Worked example: kidney disease, diabetes and taxane neuropathy at week 11

Take a hypothetical participant on a combination arm who enrolls with stage 3 chronic kidney disease, controlled type 2 diabetes and grade 1 peripheral sensory neuropathy under CTCAE v6.0 (mild residual symptoms) from a taxane two lines earlier. Baseline creatinine is 1.9 mg/dL against a ULN of 1.2 mg/dL. At week 11, the neuropathy limits instrumental ADL and creatinine reads 2.1 mg/dL.

Grading the neuropathy. Under v6.0, grade 2 is one grade above baseline: treatment-emergent, and under NCI's rule for DCTD INDs a routine adverse event reassessed for expedited reporting. Under v5.0 the answer rests on the baseline grade the study recorded. If the mild symptoms were recorded as grade 2, the week-11 grade matches baseline, and the record has to show how the study decided whether the event worsened.

Grading the creatinine. CTCAE v5.0 reads Creatinine increased grade 2 as ">1.5 - 3.0 x baseline; >1.5 - 3.0 x ULN," and NCI defines the semicolon as "or." CTCAE v6.0 applies the baseline arm only "if baseline is below LLN" (the lower limit of normal), so here only the ULN multiples count.

Value	x ULN	x baseline	v5.0 grade	v6.0 grade
1.9 mg/dL, enrollment	1.58	1.00	2	2
2.1 mg/dL, week 11	1.75	1.11	2	2
3.9 mg/dL	3.25	2.05	2 by the baseline arm, 3 by the ULN arm	3

At week 11 the grade has not moved in either version, so the creatinine is no new event. At 3.9 mg/dL the v5.0 text fits two grades, and the study's grading convention has to say which the record carries.

Attribution. Diabetes and the earlier taxane are alternative explanations for the neuropathy; kidney disease is one for the creatinine. The investigational agent and the backbone are both candidates for each; attribution across a combination regimen walks the method step by step. Two careful investigators can disagree, because the competing explanation is in the chart.

The label. Under 21 CFR 201.57(c)(7), the Adverse Reactions section of labeling covers events for which "there is some basis to believe there is a causal relationship" with the drug. The same oncology data requests ask for "Investigator-assessed causality as noted on CRF" (AEREL, from the case report form), one variable per study drug where captured, so the site's call travels with the application.

For the laboratory abnormalities table, FDA's oncology offices count patients with a baseline and at least one post-baseline value, and call a lab worsening when its maximum grade exceeds the baseline grade (OCE/OOD data requests). A baseline recorded at enrollment lets that analysis separate an abnormality the participant brought from one that worsened on treatment; whether the drug caused it stays the attribution call.

What a defensible TEAE record shows

A central reviewer should be able to re-derive each call from:

1. The baseline state and grade for the term, with date and source, under the CTCAE version the protocol names.
2. The current value and grade, with the source line and the criterion text that produced it.
3. Onset relative to exposure: first dose, last dose and the treatment-emergent window.
4. The change: new, worsened by how many grades, recurrent or unchanged.
5. Attribution per suspect drug, with each comorbidity named as an alternative cause.
6. The investigator's signature on what was accepted or changed.

How Burna AI carries baseline comorbidities into grading and attribution

When a participant's comorbidities are on file, Burna AI passes them into grading, and an event that overlaps a documented comorbidity carries a note saying so for the investigator to weigh. How Burna AI separates pre-existing conditions from new events has the short version.

Each event is graded under the protocol's CTCAE version, coded to a MedDRA lowest level term (LLT), and scored against every drug in the regimen with WHO-UMC and Kramer: per-suspect attribution with the working shown.

Every grade traces to the line in the medical record that produced it, and a grade whose citation set is incomplete is flagged before it is signed. A grade becomes a record when a qualified investigator signs it with a 21 CFR Part 11 electronic signature, as listed on the Burna AI capability record. AI suggests, clinicians decide. Human-in-the-loop, always.

For a sponsor, attribution scored the same way at every site, with comorbidities flagged beside the events they may explain, helps elucidate the safety profile while the trial can still act on it.

Common questions about treatment-emergent adverse events

Is every TEAE caused by the study drug?

No. Treatment-emergent is a timing classification, and placebo arms record TEAEs too. For a serious event, 21 CFR 312.64 requires the investigator's assessment of whether there is a reasonable possibility the drug caused it.

Is a TEAE the same as a serious adverse event?

No. Treatment-emergent describes timing; serious describes outcome. 21 CFR 312.32(a) defines a serious adverse event by result: death, a life-threatening event, inpatient hospitalization or its prolongation, persistent or significant incapacity, a congenital anomaly, or an important medical event. A TEAE can be serious or not; serious is not the same as severe.

How long after the last dose is an event treatment-emergent?

FDA's oncology default runs from the first dose to 30 days after the last dose, or to the day before subsequent therapy, whichever comes first; another window is agreed with the review division before submission.

Is an adverse event that starts after consent but before the first dose treatment-emergent?

No. FDA's oncology default window opens at the first dose; the protocol's collection window decides whether it is recorded as a pre-treatment adverse event. If it is present at the baseline assessment, NCI's guidelines treat it as a pre-existing condition, and it becomes an adverse event only if it worsens, or resolves and recurs, during the study.

---

If the comorbid participants in your combination trial are the cases central review keeps sending back, book a 15-minute conversation.