Gene therapy long-term follow-up: keeping adverse event capture comparable for years

Gene therapy long-term follow-up (LTFU) is the period after a trial's active phase in which participants who received the product are observed for delayed adverse events. FDA's 2020 guidance recommends 15 years of follow-up for integrating vectors such as lentiviral and gammaretroviral vectors, up to 15 years for genome editing products, and up to 5 years for AAV vectors.

Across that span the staff and the systems change, and so can the CTCAE version where the protocol grades with CTCAE. Each adverse event record therefore has to carry its own source text, criterion, version and reasoning to stay comparable with year one.

What is at stake is the product label. In January 2024 FDA initiated class safety labeling changes so that the boxed warning on six approved BCMA-directed and CD19-directed autologous CAR T cell products highlights the serious risk of T cell malignancies. Its April 2024 safety communication states: "Patients and clinical trial participants receiving treatment with these products should be monitored life-long for secondary malignancies." The 2020 guidance counts such ex vivo genetically modified cells as gene therapy products.

How long is long-term follow-up for gene therapy?

CBER's Long Term Follow-Up After Administration of Human Gene Therapy Products (January 2020, section V.C) sets durations by product type:

Product type	Recommended LTFU duration
Integrating vectors: gammaretroviral, lentiviral, transposon elements	15 years
Genome editing products	Up to 15 years
Herpes virus vectors or oncolytics capable of latency	Up to 15 years
Microbial vectors known to establish persistent infection	Up to 15 years
AAV vectors	Up to 5 years

Three points matter in practice:

1. It is guidance. Not every gene therapy product needs LTFU; the sponsor starts from a risk assessment, and can propose a different duration from persistence and clinical data through an IND amendment discussed with FDA.
2. Fifteen years is not new. The November 2006 guidance it supersedes already advised up to 15 years: at least 5 years of annual examinations, then 10 years of annual queries (section III.B).
3. It continues after approval. Licensed biologics carry adverse event reporting under 21 CFR 600.80, and FDA may recommend a patient registry. See postmarket pharmacovigilance on Burna AI.

What delayed adverse events does gene therapy LTFU look for?

FDA recommends a dedicated LTFU protocol setting the visit schedule, sampling plan, monitoring tests and clinical events of interest (section V.D). For the first 5 years or more, investigators need a method to record new clinical conditions:

• New malignancies
• New incidence or exacerbation of a pre-existing neurologic disorder
• New incidence or exacerbation of a prior rheumatologic or other autoimmune disorder
• New incidence of a hematologic disorder
• New incidence of infection, potentially product-related

The guidance also asks the LTFU protocol to include:

1. A baseline history before exposure, as part of the case history required by 21 CFR 312.62(b).
2. A record of exposures to mutagenic agents and other medicinal products.
3. An annual physical examination with laboratory tests such as a hematology profile.
4. Testing for persistent vector sequences at least annually until they become undetectable.

Period	What the 2020 guidance recommends
Before exposure	Baseline history under 21 CFR 312.62(b)
Years 1 to 5 or more	Annual physical examination with laboratory tests, recording of new clinical conditions, vector persistence testing
After year 5	Contact at least once a year, by telephone or questionnaire if the sponsor chooses

Integrating vectors get more. Because their integration carries a risk of leukemias and clonal expansion, section V.F recommends integration site analysis in surrogate cells such as peripheral blood, with PCR testing for vector sequences at least every six months for 5 years, then yearly for 10 or until none are detectable. Persistent monoclonality, clonal expansion or integration near an oncogenic locus calls for close monitoring for malignancy. FDA's companion replication competent retrovirus guidance (January 2020, section IV.A) adds RCR testing before treatment, at 3, 6 and 12 months, and yearly for up to 15 years. A malignancy in year nine is read against exactly these records.

Reporting continues throughout. IND safety reporting under 21 CFR 312.32 still applies, and FDA recommends a Long Term Follow-Up section in the IND annual report under 21 CFR 312.33, or in a DSUR (section V.D.3). That report summarizes the most frequent and most serious adverse experiences by body system, year after year, which is where comparability is tested.

Severity and seriousness stay separate: a CTCAE grade describes severity, while seriousness follows the outcome criteria in 21 CFR 312.32(a).

Why does a 15-year adverse event record drift?

Three things change, and each moves what an old record means.

The vocabulary. Where the protocol uses CTCAE, the version moves. NCI published CTCAE v4.03 in 2010, v5.0 in 2017 and v6.0 in 2025, the last two aligned to MedDRA 20.1 and 28.0 (NCI CTEP). A participant treated in 2010 reached year fifteen in 2025, two major versions later.

CTCAE v6.0 has 850 terms: 45 new, 7 renamed, and 376 reworded at one grade or more since v5.0. Those counts come from comparing the NCI publications of CTCAE v5.0 and CTCAE v6.0 term by term (ctcae.ai/v6-changes). A name can match while the cutoff moves: an absolute neutrophil count of 700/mm3 is Grade 3 under v5.0 and Grade 2 under v6.0 (Neutrophil count decreased; on why, A paradigm shift in neutrophil adverse event grading).

The people. Coordinators, investigators and the participant's own physician change. FDA recommends naming suitable health care providers outside the trial, such as the primary healthcare provider, with a template for reporting observations to the investigator (section V.D).

Participants change too. FDA asks sponsors to make every effort to prevent loss to follow-up (section V.C). Its September 2025 draft guidance on postapproval data for cell and gene therapy products notes that pediatric patients will often be followed into adulthood, which means obtaining their own consent as adults, and that remote and telehealth visits improve convenience and retention.

The systems. A study that outlives its data capture platform has its safety data exported, mapped and loaded again, and each migration keeps only the fields somebody mapped. The 2025 draft asks for a follow-up plan, including funding, in case the sponsor ceases to operate the study before LTFU ends. It also states that software producing postapproval data is subject to 21 CFR Part 11.

What should each long-term follow-up adverse event record carry?

Each record should carry the quoted source text, the CTCAE version and criterion, the proposed and signed grades, seriousness apart from the grade, the inputs to attribution, the MedDRA LLT and version, and the signature, inside the record rather than beside it.

Field on the event	Why year twelve needs it
The source text, quoted	The grade can be re-derived without the person who assigned it
The CTCAE version and criterion applied	A grade means little across a version boundary without its publication
The proposed grade and the signed grade	A reviewer sees where a clinician changed the assessment
Seriousness criteria, apart from the grade	Severity and seriousness answer different questions
Time since administration and other exposures	A causality assessment at year nine needs both
Attribution per suspect, with algorithm and inputs	A reviewer can follow the score to its evidence
The MedDRA LLT and MedDRA version	Body-system tabulations stay consistent across years
Reviewer, signature and timestamp	Who decided, held as a Part 11 electronic record

With the version on the event, a numeric criterion such as a neutrophil count can be re-derived against any version from the recorded value, and a descriptive one reread in the quoted source text. For the operational side, see how sites and sponsors run CTCAE v5.0 and v6.0 at the same time.

How do you capture delayed adverse events between annual contacts?

FDA asks sponsors to encourage participants to report adverse events proactively, through diaries and wallet cards with investigator contacts (section V.D.1). After year 5 a participant may hear from the study once a year, while the events the protocol watches for develop over months.

A structured channel turns a recollection at an annual call into a dated entry that can be graded. NCI built PRO-CTCAE to capture symptomatic adverse events from patients on cancer trials.

Healthnotes, the participant app sponsors deploy through their sites, is that kind of channel between visits. The site issues the participant a PRO token tied to the study and site. The participant records what happened in their own words, by voice or text, and reviews it before sending. When the site's encounter is graded, the engine reads the report, proposes a CTCAE grade, shows the criterion behind it and anchors it to the participant's words; the grade stays a suggestion until a clinician confirms it. PRO-CTCAE is supported, and a sponsor schedules instruments without an app release. More in between-visit symptom capture.

What should you ask of an LTFU safety system?

1. Where is the CTCAE version for an event stored? If on the study rather than the event, every comparison across versions becomes a recoding project.
2. What survives a migration to another platform? If the reasoning lives in an interface rather than the record, the export leaves the basis behind.
3. What can a reviewer without the original system recover from one event? If only a coded term and a grade, year twelve cannot be read against year one.

How Burna AI keeps a long-horizon safety record comparable

What the platform records on each graded event today:

• Every grade traces to the line in the medical record that produced it, with the CTCAE criterion attached, and an incomplete citation set is flagged before signing (capabilities).
• Grading runs against the CTCAE version the protocol names, with v5.0 and v6.0 on one engine, and each graded event records its version and whether the clinician, the trial or the platform default chose it.
• The 21 CFR Part 11 electronic signature carries what was accepted, what was overridden and the citation set as signed.
• Each event is scored against every drug in the regimen with WHO-UMC and Kramer attribution, with the reasoning shown.
• Each event carries its MedDRA LLT code, and the graded record produces FDA MedWatch, IRB and serious adverse event reports and exports for a sponsor or monitor, with every access to patient data recorded.

AI suggests, clinicians decide. Human-in-the-loop, always.

Common questions about gene therapy long-term follow-up

How does EMA's gene therapy follow-up guideline compare with FDA's?

EMA sets no fixed 15-year figure. Its Guideline on follow-up of patients administered with gene therapy medicinal products (EMEA/CHMP/GTWP/60436/2007, in effect since May 2010) ties follow-up duration to a risk stratification consistent with the risk management plan. For vectors that can integrate or reactivate it recommends at least 5 years, then yearly until data show no remaining risk; for non-integrating viral vectors, at least 5 years. FDA's 2020 guidance sets nonbinding durations by product type.

Does long-term follow-up continue after approval?

When the 15-year period has not elapsed for every participant by licensure, FDA recommends a pharmacovigilance plan with the BLA, which the ongoing LTFU study can be part of, and LTFU data collected after licensure go in the BLA annual report (section VI).

Which CTCAE version should a long-term follow-up study use?

The version the protocol names. When a program spans versions, keep the version on every grade rather than relabeling earlier ones, and re-derive from the recorded value wherever the criterion is numeric.

If your program carries a follow-up obligation that will outlast the people and systems that start it, book a 15-minute conversation.