Undisclosed Deaths in Gene Therapy Trials: What Clinical Authors Must Report to Journals in 2026

On July 23, 2026, Science and Retraction Watch published an investigation describing how a six-year-old girl in China died after receiving an experimental gene therapy treatment her parents had paid more than $800,000 to access. The hospital's internal ethics board concluded her death was "definitely related" to the therapy. That conclusion was never made public. In the months that followed, the lead researcher published findings from related animal studies in Nature, with no mention that a human participant had died. Nature added an editor's note on July 29 and announced a formal investigation the following day. Within a week, university-level scrutiny had been launched as well.

The case is now a reference point in a much older question, one that medical authors working anywhere on clinical research need to understand before their next submission: what are you required to tell a journal about serious adverse events in any trial your paper touches, and what happens if you do not?

What the Investigation Actually Revealed

The gene therapy case is not primarily a fabrication story. The animal study paper published in Nature is not alleged to contain invented data. What the Science and Retraction Watch investigation found was a specific kind of omission: a researcher published results from animal experiments while a concurrent human trial of the same or closely related therapeutic approach had produced a participant death, and neither the death nor the existence of the human trial was mentioned in the journal submission.

Nature's response was instructive. The journal stated that no information about a human trial was ever shared with the editorial team during submission or peer review. The journal was not deceived by false statements in the manuscript. It was left without information it would have needed to evaluate the paper properly. That is a different kind of integrity problem, and it is one that the COPE framework and most major publishers' policies are equipped to address, even if the response looks different from a straightforward data fabrication case.

The investigation also revealed that the company's own animal safety data was not reassuring: all four monkeys given the therapy, whether at low or high doses, developed moderate to severe liver damage. That information existed before the human participant was treated. Whether that context should have appeared in the Nature paper is a question the journal's investigation will presumably address. From the outside, the case is a readable example of how concurrent lines of work can create undisclosed material context for a published paper.

What ICMJE Actually Requires in Published Trial Reports

The ICMJE Recommendations for clinical trial reporting are explicit on adverse events. For any published report of a clinical trial, authors must report serious adverse events and deaths in the results section, with enough detail to allow a reader to evaluate the causal relationship. This means naming the adverse events that occurred, identifying which treatment arm they appeared in, reporting the time of onset, and describing the investigators' or ethics board's assessment of relatedness to the treatment.

The January 2026 update to the ICMJE Recommendations reinforced the clinical trial registration obligations and added language about data access in industry-sponsored research. Authors must have meaningful access to the data underlying the paper they are signing. That principle applies equally to adverse event data. If you do not know what happened to participants in a trial your paper describes, you should not be listed as an author.

There is also a specific ICMJE obligation around registry correspondence. Journals adhering to ICMJE guidelines will not publish a clinical trial report unless the trial was registered before patient enrollment began. The registered record includes a section on adverse events. If the registry record shows adverse events, including deaths, and the manuscript does not, authors are required to explain the discrepancy. This is not a technicality. It is an enforceable submission requirement at ICMJE-member journals including JAMA, the New England Journal of Medicine, The Lancet, and BMJ, among others.

The gene therapy case involved a publication about animal studies, not the human trial itself. Whether ICMJE obligations attach to the animal paper is a more contested question. But the broader principle is clear: authors are responsible for ensuring the published record is not materially misleading, and omitting the existence of a concurrent human trial with a related fatality can render an animal study misleading in practice, even if it is technically accurate in isolation.

The COPE Framework: What Journals Do When Adverse Events Surface After Publication

The Committee on Publication Ethics released Version 3 of its Retraction Guidelines in August 2025, and the graduated response it recommends is now the standard approach at most major biomedical publishers. Understanding that framework matters because it tells you exactly what will happen to your paper if a similar situation arises.

The first step, when concerns are raised but not yet resolved, is an editor's note: a brief, neutrally worded statement attached to the article flagging that the journal is investigating concerns. This is not a verdict. It does not claim the authors did something wrong. It signals to readers that findings in the paper should be held provisionally while the journal gathers information. Nature used this mechanism when it added a note to the gene therapy paper on July 29. The note will stay until the investigation concludes, however long that takes.

If the investigation reveals that the paper is substantively misleading but the authors are not found to have acted in bad faith, the outcome is typically a correction: a formal erratum that supplements the original paper with the omitted information. The correction becomes part of the permanent record and is indexed alongside the original paper. If the investigation reveals that a paper cannot be trusted in a material way, and especially if the authors are found to have known the relevant information at the time of submission, the outcome is retraction.

For authors whose papers are under investigation, the practical message is this: if you become aware of information that was missing from a submission and that information is relevant to the paper's safety claims, contact the journal proactively before an investigation reaches you. A voluntary author-initiated correction is handled under a different track than one that emerges from external reporting. The 2025 COPE guidelines explicitly distinguish between authors who come forward and those who are found out, and journal editors know the difference in practice as well.

Regulatory Reports Are Not Journal Disclosure

The most persistent misconception among clinical researchers on this topic is that filing a serious adverse event with a regulatory body constitutes scientific disclosure. It does not, and the distinction is fundamental.

In most jurisdictions, SAE reports submitted to regulatory agencies are confidential. They are not publicly accessible. They are not indexed. They do not appear in ClinicalTrials.gov unless the trial sponsor specifically includes adverse event summary data there (which FDAAA 801 requires for applicable clinical trials, but which is regularly violated, as the FDA's 2026 enforcement campaign made clear). A researcher can maintain a technically complete regulatory safety file, with every death properly reported to every relevant authority, while that death remains entirely invisible to the scientific community.

The China gene therapy case illustrates this exactly. Whatever regulatory and ethics board processes took place around the girl's death, the outcome of those processes did not enter the scientific literature. The internal ethics board conclusion that the death was "definitely related" to the treatment was a document in a file, not a published finding. It had no path into the scientific record unless the researchers chose to put it there. The journal that published the animal study knew nothing about it.

This gap is structural, and journal submissions are where individual authors can bridge it. If your research team has run or sponsored a clinical study, and that study produced adverse safety findings, the submission of any related paper without disclosure of those findings creates integrity risk. The defense that "we reported it to the regulators" does not satisfy the separate obligation to the scientific record that journal authorship carries.

Co-Author Responsibilities When You Did Not Run the Trial

Large clinical and translational research teams create a specific accountability problem. In multi-investigator studies, co-authors often have deep expertise in a specific part of the work, the animal pharmacology, the biomarker analysis, the statistical modeling, and limited visibility into the trial's overall safety profile. The lead investigator and the corresponding author are the ones who typically know about participant adverse events. In some team structures, co-authors are not informed about safety events unless they ask.

The ICMJE authorship criteria require that every author agrees to be accountable for all aspects of the work, not just for the piece they contributed. That is a high standard, and in practice it means that accepting authorship on a clinical study carries an implicit obligation to ask whether the full picture of what happened to participants is represented in the submission you are approving. It is not enough to verify your own contribution is accurately described.

The practical way to operationalize this: before approving any clinical manuscript for submission, ask the corresponding author specifically whether any participants in this study or in any concurrent study of the same agent experienced a serious adverse event, and whether those events are reflected in the manuscript. Ask this as a direct question, in writing, so there is a record of the exchange. If the answer is yes, follow up to confirm how the event appears in the methods and results sections. If the answer is that no concurrent human trials are ongoing, make a note of that assurance as well.

This kind of documentation feels like administrative overhead until the day it matters. If your paper is investigated after publication, the question of whether co-authors knew or should have known about omitted safety data will come up. Having asked directly, and having received a written answer, is meaningful protection. It also changes the incentive structure for the corresponding author: if they know the co-authors are asking, they are less likely to assume everyone has implicitly agreed to omit something.

Concurrent Studies and the Question of Cross-Reference

The specific mechanism in the gene therapy case, submitting an animal study without disclosing a concurrent human trial, raises a question that has not been formally resolved in most journal policies. Current ICMJE guidance addresses adverse event disclosure within the reported trial. It does not have explicit requirements about disclosing concurrent trials of the same agent in separately submitted papers. The obligation to cross-reference is implied rather than stated.

That implied obligation is becoming less implied over time. In the past two years, several journals have begun revising their instructions for authors to ask whether any concurrent studies of the same investigational product or approach are underway, and whether any participants in those studies have experienced unexpected serious adverse events. These questions are not yet universal. They are, however, appearing at journals that have handled post-publication integrity investigations and have drawn lessons from them.

For authors currently preparing papers: if you are publishing animal data for a compound that is concurrently being tested in humans, or if you are publishing a first-in-human safety study while a larger efficacy trial is already generating safety signals, the question of what the animal or early phase paper needs to say about the concurrent work is worth thinking through carefully before submission. The answer will depend on the nature of the concurrent data, but the default of assuming that concurrent work is simply irrelevant to the submitted paper is harder to defend now than it was two years ago.

A Pre-Submission Checklist for Clinical Authors

Before submitting any paper that reports on clinical trial data or that is part of a research program involving human participants, work through the following questions. These are not bureaucratic formalities. They are the questions that will be asked if an integrity review opens after publication, and the best time to answer them is before submission rather than under the pressure of an editorial investigation.

None of these questions require you to append a regulatory filing to your methods section. They require an informed decision about what a reader needs to know to interpret your paper accurately. In most cases the answers will confirm that the manuscript is already complete. In some cases they will reveal a gap that is better addressed before submission than discovered afterward.

What Journals Are Beginning to Ask

The gene therapy investigation will accelerate a shift in journal submission systems that was already underway. The kinds of questions that appeared on submission checklists after the CONSORT 2025 update (about pre-registration, open science reporting, and protocol adherence) are going to be joined by questions about concurrent trial safety in fields where the gap between animal and human work is narrow and fast.

Some journals already ask a version of this. Submission systems for gene therapy, gene editing, and cell and gene therapy journals frequently include a field asking whether any human administration of the described agent has occurred, and whether any adverse events were observed. These fields exist precisely because the field has learned from previous disclosure failures, including the Jesse Gelsinger case in 1999, where the death of an early gene therapy participant was not fully disclosed in related publications.

The 2026 case suggests the question needs to be broader, covering not just direct human administration but any concurrent human work involving related therapeutic approaches. That expansion is likely to reach submission checklists in gene therapy and similar fields within the next year or two. Authors in high-stakes translational fields should begin building this information into their submission documentation now, rather than adapting to a new requirement at a moment when a trial is already generating safety questions.

Echoes of an Earlier Failure

The 2026 case echoes the CRISPR baby controversy that surfaced in late 2018, when He Jiankui announced the birth of gene-edited children without prior peer-reviewed publication and without the kind of regulatory scrutiny that the field expected. China subsequently tightened its regulations on gene editing in clinical contexts. The 2026 investigation suggests that those tightened rules have not been uniformly effective, and that the gap between regulatory compliance in a domestic context and transparency in international scientific publishing remains significant.

For international research collaborators and co-authors, this gap has direct practical meaning. If you are a researcher at an institution outside China who is listed as a co-author on a paper arising from a Chinese clinical trial, the same questions apply to you. You have the same ICMJE accountability. The fact that the trial happened elsewhere, under a different regulatory jurisdiction, does not change what authorship means. The questions in the pre-submission checklist above are the same questions you should be asking, regardless of where the research was conducted.

The Practical Implication for Every Clinical Author

The girl in the China investigation died in early 2025. As of August 2026, the Nature paper carries an editor's note flagging ongoing concerns. The university and journal investigations are open. Whatever the final outcome, the researchers involved will spend years managing the consequences of a disclosure that could have been made at the moment of journal submission.

The lesson is straightforward, even if the underlying ethics are not simple. Regulatory systems exist to protect participants during a trial. Journal publications exist to inform the field after a trial. These are related but distinct obligations. Meeting one does not satisfy the other.

If your work involves human participants, or if it exists in a pipeline where animal or early phase results flow into human trials, the full picture of what happened to those participants belongs in the scientific record. The appropriate place for that information is the journal submission. What you put in a regulatory filing and what you put in your methods section need to tell the same story. When they diverge, and an investigation eventually finds the gap, the scientific record absorbs the damage long after the paper is published and long after the underlying trial has ended.

This is why the pre-submission checklist matters, and why building adverse event verification into your team's standard submission workflow is a better approach than relying on the corresponding author to remember what the regulatory file says. It is the same diligence that applies to conflicts of interest, CRediT authorship statements, and data availability. The standards around clinical trial transparency are converging on the position that anything material to a reader's interpretation of your findings needs to appear in print.

Further Reading