On 13 November 2026 a task force set up by the Shanghai Party committee and municipal government published its findings on the death of a six-year-old girl in a gene-editing study at Xinhua Hospital, which is affiliated with Shanghai Jiao Tong University School of Medicine [1][2]. The girl had Snijders Blok-Campeau syndrome, a rare neurodevelopmental disorder caused by variants in the CHD3 gene. She received a single injection of an experimental base-editing treatment on 24 March 2025 and died on 31 March 2025. The hospital attributed her death to thrombotic microangiopathy (TMA), a known complication of gene therapy [1].
The task force found failures at almost every point where safety information should have moved: from the primate toxicology study to the clinical team, from the research team to the ethics committee, from the ethics committee to the consent form, and from the study to the published record [1]. Two researchers were dismissed, others were demoted or disciplined, and the hospitals involved received administrative penalties [1]. Shanghai's health commission has started special inspections of clinical research on new biomedical technologies [1].
This article sets out what is established, where the safety chain broke, and how the relevant rules in China, the EU and the US compare. It is analysis, not legal or medical advice.
Key takeaways
- Safety findings from a primate study completed in February 2025 did not reach the hospital's clinical study lead in time, so the warning signals in that report "did not receive due attention" [1].
- The ethics committee approved the study despite insufficient toxicology and safety evidence, and without the extra materials it had asked for [1].
- The consent form mentioned serious complications but did not highlight the risk of death [1].
- The February 2026 Nature paper disclosed neither the primate safety findings, the death, nor the money and gifts some authors had received from the girl's family [1]. Investigators found no data fabrication but did find academic irregularities [1].
- China's regulation on biomedical new technologies, in force since 1 May 2026, now requires nonclinical proof of safety, national filing and a pause after a serious adverse reaction. It did not apply when the girl was treated [10].
Sources and what they cover
The primary source is the task force's notice. We could not find it on a government website, so we used the full text published by The Paper (澎湃新闻) and Xinhua's summary, which agree [1][2]. Both are state-owned media, as are Global Times and China Daily [3][4]. Caixin, an independent outlet, confirms the main findings, though mostly behind a paywall [5][6]. Further detail comes from a Science and Retraction Watch investigation of 23 July 2026, based on family documents, recordings and a district health commission report [7]. Where we rely on Science alone, we say so. Translations are our own.

What happened
According to the notice, the girl's family contacted the lead researcher in July 2023, asking for a customised gene-editing therapy for the central nervous system [1]. His team at the Songjiang Research Institute did the preclinical work. A Guangzhou biotechnology company made the viral vectors, and a Sichuan company ran the large-animal safety studies [1]. The team submitted its paper to Nature on 21 December 2024 [1][8]. The hospital's ethics committee approved the clinical study on 2 January 2025, and the child and her family signed the consent form on 20 February [1]. She was dosed on 24 March and died on 31 March. On 2 April the ethics committee held an emergency serious adverse event (SAE) review, attributed the death to TMA, and required the study to be stopped [1].
The ClinicalTrials.gov record (NCT06860672) describes an early phase 1, single-participant study of one intrathecal injection of a dual-vector AAV base editor, with the hospital's clinical lead as sponsor-investigator [9]. It was first submitted on 24 February 2025 and was updated to record the death and termination only on 30 July 2026 [9].
The Nature paper, published on 18 February 2026, reports mouse studies and delivery of the editor in non-human primates, and declares no competing interests [8]. On 29 July 2026 Nature added an editor's note that concerns had been raised, including about the data [8]. On 16 November 2026 the paper had not been retracted.
| Area | Task force finding [1] | Other reporting |
|---|---|---|
| Nonclinical safety | Primate safety results not passed promptly to the clinical lead; warning signals not given due attention | Report dated 17 Feb 2025 found moderate to severe liver damage in all four treated monkeys, and kidney damage in one (Science) [7] |
| Ethics review | Approved despite insufficient toxicology and safety evidence, and without the requested extra materials; scientific review not independent or thorough | Committee did not review the final primate report before approval (Science, citing a district health commission document) [7] |
| Consent | Serious complications disclosed; death risk not highlighted | Form named TMA as possible; death not mentioned in the form or in conversations, the parents say (Science) [7] |
| Publication | Primate safety report and death not disclosed; conflicts of interest not disclosed; no fabrication found, but actual methods differed from those described | Editor's note added 29 Jul 2026 [8] |
| Money | The girl's family funded 5.337m yuan; about 1.06m yuan in cash and gifts to four people, since returned | Science reports US$130,000 paid to one researcher's personal account [7] |

Where the safety chain broke
Nonclinical-to-clinical handover. The notice says the preclinical team did not tell the clinical study lead about the results of the primate safety report, so its warning signals were not acted on [1]. Science reports that the contract lab's report, dated 17 February 2025, five weeks before dosing, found moderate to severe liver damage in all four monkeys that received the treatment, at both low and high doses. One high-dose monkey also had kidney damage. That could have reflected TMA, but because data were not collected in the first month after dosing, the cause could not be confirmed [7]. These details come from Science, not the notice. Either way, the people who held the toxicology data were not the people deciding on dose, immunosuppression and monitoring.
Ethics review. The committee asked for more toxicology and safety material, did not receive it, and approved the study anyway; the hospital's scientific review did not independently assess the protocol's safety [1].
Consent. The consent form did mention serious complications and adverse events, but it did not single out the risk of death [1]. For a first-in-human, brain-directed gene-editing treatment given to a child whose condition, Science reports, was not life-threatening, that omission matters [7].
Publication. The paper reported primate delivery data but not the primate safety report or the death of the only person treated [1][8]. Four people had received money or gifts from the family, which was not disclosed as a conflict of interest [1].


Which rules applied, and which apply now
The notice does not say under which regulatory pathway the study was approved. Science reports that it ran as an investigator-initiated study that did not need national regulatory approval, and that a district health commission fined the hospital about US$3,600 in September 2025, partly for not registering it as commercially sponsored research [7]. We could not confirm that fine officially; the notice says the Yangpu District Health Commission has now penalised the hospital [1]. China has three relevant regimes (Table 2).
- The 2024 rules on investigator-initiated trials (IITs), issued by the National Health Commission (NHC) and in force from 1 October 2024, covered the period of treatment. They require "a solid preliminary research base" and bar approval where preparation is inadequate, risk exceeds what the institution can control, or improper financial relationships may exist [11].
- The regulation on biomedical new technologies (State Council Decree No. 818), in force since 1 May 2026 and cited by the task force [1][10], requires nonclinical proof of safety and efficacy before clinical research, national filing within five working days of institutional approval, and, after a serious adverse reaction, a pause, ethics committee assessment and a report within five working days [10].
- The revised Good Clinical Practice (GCP), in force since 1 September 2026, applies only to drug trials approved by or filed with the National Medical Products Administration (NMPA). Its Center for Drug Evaluation (CDE) receives expedited reports of suspected unexpected serious adverse reactions (SUSARs) [12].
Whether a product like this would now be treated as a new technology or a drug is not settled; Decree 818 leaves the boundary to future guidance [10].
| Element | IIT rules (2024) | New-technology regulation (2026) | GCP (2026) |
|---|---|---|---|
| In force | 1 Oct 2024 [11] | 1 May 2026 [10] | 1 Sep 2026 [12] |
| Scope | Investigator-initiated, not for product registration (Art. 2) | Cell- or molecule-level technologies not yet used clinically in China (Art. 3) | Drug trials approved by or filed with NMPA (Art. 2) |
| Before first dose | Scientific and ethics review, with outside experts generally included; registration (Arts. 22–26) | Nonclinical proof of safety and efficacy; filing with nonclinical reports and consent form (Arts. 9, 15–16) | NMPA approval; ethics committee reviews the investigator's brochure (IB) and safety data (Arts. 14, 38) |
| Serious reaction | Institution decides to continue, pause or stop (Arts. 36, 38) | Pause; ethics committee assesses; report within 5 working days (Art. 25) | PI informs sponsor and ethics committee immediately; SUSAR expedited to CDE (Arts. 26, 44) |
| Fees | None charged to participants (Art. 31) | None charged to participants (Art. 20) | Not addressed in the articles we reviewed |
How China compares with the EU and the US
In the EU, a gene therapy is an advanced therapy medicinal product (ATMP), and a trial needs Member State authorisation after scientific and ethical review, with nonclinical data from GLP-compliant studies [13][14]. EMA's investigational ATMP guideline, in force since 1 July 2025, sets minimum data before first human use: proof of concept, support for the route and starting dose, and "appropriate safety data" [15]. The hospital exemption covers custom-made ATMPs outside clinical trials; it does not replace trial authorisation [14].
In the US, a trial cannot start until an IND is in effect, and the IND must contain toxicology data showing the trial is "reasonably safe" [16]. FDA's 2024 genome-editing guidance says first-in-human trials "generally should enroll only subjects for whom no other treatment options are available or justified" [18]. Consent must describe "any reasonably foreseeable risks", and children may face more than minimal risk only on specific IRB findings [17].
| Element | China | EU | US |
|---|---|---|---|
| Regulator before first dose | NMPA for drug trials; national filing for new technologies [10][12] | Member State authorisation [13] | IND in effect [16] |
| Nonclinical package | Proof of safety and efficacy before clinical research [10] | GLP studies; EMA minimum data list [13][15] | Toxicology adequate to show "reasonably safe" [16][19] |
| Who must see the toxicology | Ethics committee (IB and safety data) [12] | Assessed in authorisation [13] | Investigators, through the IB [16] |
| Consent | Disclose possible risks; no deception or inducement [10][12] | Risks explained; no incentives for minors [13] | Reasonably foreseeable risks [17] |
| Fatal SUSAR | CDE within 7 days (GCP trials); new technologies: pause, report within 5 working days [10][23] | Within 7 days [13] | Within 7 calendar days [16] |
| Public results | Results reported to NHC (new technologies) [10] | Summary within one year [13] | Registry results not required for phase 1 [20] |
For the routine trial-safety clocks, see our earlier analysis of CDE Notice 52/2026.
What sponsors, investigators and ethics committees can check
These points follow from the findings and texts cited, and apply most to academic programmes where developer, sponsor and treating physician differ.
Sponsors and developers
- Name one owner for nonclinical safety signals, responsible for getting every draft and final toxicology report to the principal investigator (PI) before dosing decisions, with a record of receipt.
- Write findings into the IB and protocol. Target-organ toxicity in primates should lead to a written assessment of dose, immunosuppression and monitoring.
- Treat a death in a single-patient study as a full safety case: expedited reports on the applicable clocks, the development safety update report (DSUR) where one applies, and a prompt registry update [9][13][16].
Investigators
- Do not dose before reading the final toxicology report, and confirm the IB reflects it.
- Route family funding through institutional accounts, never individuals, and disclose it to the ethics committee and in publications [1][11].
Ethics committees
- Do not approve while requested toxicology is missing; make the final nonclinical report a condition of approval [1].
- Check for plain wording on death where nonclinical data or similar products suggest a fatal risk.
- Ask who is paying, and check the study against the no-fee rules [10][11].
Authors and journals
- Disclose related clinical outcomes and conflicts of interest when a preclinical paper supports a study already under way.
- Register before enrolment and report harms fully. ICMJE asks editors to require registration at or before first enrolment [21]. CONSORT 2025, written for randomised trials, builds harms into its core checklist (items 7, 15, 21a, 23a and 27) [22].
What is established and what is not
Established by the official notice [1]: the dates of approval, consent, dosing, death and the ethics committee's review; the failures in safety handover, ethics review, consent and disclosure; the funding and gifts; the absence of data fabrication; and the sanctions.
Reported by Science but not confirmed officially [7]: the date and detailed findings of the primate toxicology report; the September 2025 fine; the payments to individuals in US dollars; and the claim that the ethics committee did not see the final report.
Not known: the regulatory pathway; whether the study was filed nationally or any expedited safety report was made; the dose and immunosuppression (Science mentions prednisone) [7]; and whether Nature will act further [8]. The notice is available only through state media [1][2].
A toxicology report protects patients only if it reaches the people who set the dose, approve the study and write the consent form.
References
- Shanghai Municipal Investigation and Handling Task Force: notice on the "SJTU School of Medicine gene-editing research incident", 13 November 2026, full text as published by The Paper (澎湃新闻)
- Xinhua: Shanghai publishes investigation results on the SJTU School of Medicine gene-editing research incident, 13 November 2026
- Global Times: Shanghai probes gene-editing study after child's death, finds research irregularities and oversight failures, 13 November 2026
- China Daily: Shanghai releases investigation results on gene-editing research incident, 13 November 2026
- Caixin (财新网): 女童基因编辑死亡事件终局:传递监管趋严信号, 13 November 2026
- Caixin Global: Shanghai Fires Gene-Editing Researcher After Child's Death, 14 November 2026
- Borrell B (Retraction Watch). A fatal reaction. Science, 23 July 2026 (corrected 24 July)
- In vivo base editing of Chd3 rescues behavioural abnormalities in mice. Nature 2026;651:785–795, with editor's note of 29 July 2026
- ClinicalTrials.gov: NCT06860672, Clinical Trial of the Dual Vector Base Editor for the Treatment of the CHD3-R1025W Mutation
- State Council Decree No. 818: Regulation on the Administration of Clinical Research and Clinical Translational Application of Biomedical New Technologies, signed 28 September 2025, effective 1 May 2026
- NHC, NATCM and NDCPA: Measures for the Administration of Investigator-Initiated Clinical Research by Medical and Health Institutions (国卫科教发〔2024〕32号), effective 1 October 2024
- NMPA, NHC, NATCM and NDCPA: Good Clinical Practice for Drug Clinical Trials (Announcement No. 50 of 2026), effective 1 September 2026
- Regulation (EU) No 536/2014 on clinical trials on medicinal products for human use, Articles 4, 25, 29, 32, 37, 41 and 42
- Regulation (EC) No 1394/2007 on advanced therapy medicinal products, including Article 28 (hospital exemption)
- EMA: Guideline on quality, non-clinical and clinical requirements for investigational advanced therapy medicinal products in clinical trials (EMA/CAT/22473/2025), effective 1 July 2025
- 21 CFR Part 312: sections 312.20, 312.23, 312.32 and 312.55 (eCFR)
- 21 CFR 50.25, 50.52–50.53 and 56.111: informed consent, children and IRB approval criteria (eCFR)
- FDA: Human Gene Therapy Products Incorporating Human Genome Editing, Guidance for Industry, January 2024
- FDA: Preclinical Assessment of Investigational Cellular and Gene Therapy Products, Guidance for Industry, November 2013
- 42 CFR 11.22: applicable clinical trials (eCFR)
- ICMJE Recommendations: Clinical Trials (registration) and Disclosure of Financial and Non-Financial Relationships and Activities
- Hopewell S et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ 2025;389:e081123
- CDE: Standards and Procedures for Expedited Reporting of Safety Data during Drug Clinical Trials, 27 April 2018
Safety Strategy · Pharmacovigilance and safety risk management · Vaud, Switzerland
This article is for information only and is not medical or regulatory advice.
Related services: regulatory filings including DSURs and signal management and benefit-risk assessment.
Follow Safety Strategy on LinkedIn (opens in a new tab) for new articles and updates. Back to all news