First, get the two codes straight
Between July 23 and August 5, 2026 — just two weeks — two child deaths in China’s gene-editing field were disclosed. One took 16 months to come into public view; the other was disclosed about 12 months late. They belong to two companies, two hospitals, two editing modalities and two routes of administration — but they went through the same channel, fell to the same class of toxicity, and exposed the same structural problem.
These two codes are not the same kind of thing, not even at the same level, so separate them first.
The internal institutional number entered by Xinhua Hospital, affiliated with Shanghai Jiao Tong University School of Medicine, when registering NCT06860672 on ClinicalTrials.gov (the registry record gives only the number itself; XH = Xinhua, 25 = 2025, 002 = sequence number that year is the author’s guess based on common numbering conventions).
Official title: A study of the safety, tolerability and preliminary efficacy of a single intrathecal injection of dual-vector AAV-CHD3-R1025W base editor for developmental disorder caused by the CHD3 R1025W mutation.
Responsible party: Sponsor-Investigator Yu Yongguo; principal investigators Yu Yongguo and Qiu Zilong (Songjiang Research Institute, Shanghai Jiao Tong University School of Medicine).
Enrollment: planned for children aged 2–10, actual enrollment 1 patient.
HuidaGene’s CRISPR/hfCas12Max DNA editing therapy.
Mechanism: a single AAV vector delivers the editing system, targeting the splice donor site of exon 51 of the human DMD gene, disrupting this site to induce exon 51 skipping, so that exon 50 rejoins the first exon after the deleted region, restoring the correct open reading frame and expressing partially functional dystrophin. The joined exon depends on genotype: patients with an exon 52 deletion join to 53, those with a 52–61 deletion to 62, and those with a 52–63 deletion to 64 — “50–53” applies only to one of these genotypes.
First-in-human study: study number HG30201, registration NCT06594094, code-named M.U.S.C.L.E., conducted at Shanghai Children’s Medical Center, affiliated with Shanghai Jiao Tong University School of Medicine.
One is “where it was done”, the other “what was done”. They are looked at together precisely because — Xinhua Hospital and Shanghai Children’s Medical Center are both affiliated hospitals of Shanghai Jiao Tong University School of Medicine; both studies are investigator-initiated trials (IITs).
Case one: XH-25-002, the world’s first in-brain base editing
What was done technically. The child carried a heterozygous c.3073C>T (p.Arg1025Trp) mutation in the CHD3 gene, clinically diagnosed as Snijders Blok–Campeau syndrome — a neurodevelopmental disorder mainly presenting with delayed language, cognitive and motor development and autism-like behavior.
The research team used an adenine base editor (called TeABE in the team’s mouse paper), split and packaged into two AAV vectors, delivered into the central nervous system by intrathecal injection, aiming to change the mutant A–T base pair back to the normal G–C. This was reported as the world’s first in vivo base-editing treatment targeting the brain.
This disease is not fatal. The syndrome mainly causes intellectual and developmental disability and usually does not directly threaten life. On the risk–benefit scale, this is not in the same league as life-shortening diseases such as DMD, SMA or XLMTM. The core question raised by seven experts interviewed by Science was precisely this: for a non-fatal indication, taking on a high-risk first-in-human systemic AAV exposure — was risk disclosure adequate, and was the probability of success overestimated?
Dose and process details have still not been published. Media reports used the phrase “trillions of viral particles”, with no verifiable vg/kg value; neither the investigators nor the funder has issued a technical announcement.
Timeline (verifiable parts).
| Date | Event |
|---|---|
| 2025-01-02 | Hospital ethics committee approval (reported by media based on documents) |
| 2025-02-17 | Macaque toxicology report finalized, reportedly showing moderate to severe liver injury (reported by media based on documents) |
| 2025-02-19 | Family signs informed consent; study start date in the registry record (registry record + media) |
| 2025-03-06 | First posted on ClinicalTrials.gov (submitted February 24) (registry record) |
| 2025-03-24 | Dosing (media reports) |
| 2025-03-31 | Death (day 7 after dosing) (registry record + media) |
| 2025-09 | Yangpu District health authority in Shanghai fines Xinhua Hospital (reported as RMB 24,000–25,000, about US$3,600) (reported by media based on the penalty decision) |
| 2026-02-18 | The team’s mouse study paper published in Nature (journal record) |
| 2026-07-23 | Joint investigation by Science and Retraction Watch published (primary reporting) |
| 2026-07-26 | Shanghai Jiao Tong University School of Medicine issues a statement and sets up a special working group (official statement) |
| 2026-07-29 | Nature adds an editor’s note to the paper; registry record updated to “terminated” (journal + registry record) |
Ethics approval (January 2) came before the toxicology report was finalized (February 17) — the most glaring inversion of sequence on the whole timeline. When the ethics committee approved, it could not see the report that later showed liver injury in macaques.
That passage in the registry record. On July 29, 2026 — six days after the Science report — the status of NCT06860672 was updated to TERMINATED, with this sentence in the reason-for-termination field:
There are three key pieces of information in one sentence: the cause of death was thrombotic microangiopathy (TMA), causality was judged “definitely related”, and this sentence only appeared in the public record 16 months after the event, after media exposure.
Four structural problems.
Choice of indication
Non-fatal neurodevelopmental disorder + first-in-human + high-dose intrathecal AAV — the bar for a risk–benefit argument for this combination should have been much higher than for a fatal disease.
The money flowed the wrong way
Several media outlets reported that the child’s family raised more than US$800,000 (about US$860,000 / about RMB 5.82 million) themselves for the project (including preclinical research); reports also mentioned that a researcher had received US$130,000 wired into a personal account and later returned it. A trial subject paying to buy a first-in-human exposure is a red-line issue for subject protection. On timing: dosing took place in March 2025, while the Measures for the Administration of Investigator-Initiated Clinical Research Conducted by Medical and Health Institutions (Guo Wei Ke Jiao Fa [2024] No. 32) had already taken effect on October 1, 2024, explicitly stating that “medical and health institutions or investigators are strictly prohibited from illegally charging research participants research-related fees” — this ban is not a new rule from 2026.
Funder and disclosure
Based on business registration information, reports pointed out that the funder was a gene technology company founded in August 2024, whose only patent lists two of the project’s researchers as inventors, although the investigators themselves are not registered shareholders. One of the grounds for the September 2025 penalty was precisely failure to register the trial as a company-funded study in the relevant database.
The paper was decoupled from the human results
The team’s mouse study (Yang K et al., Nature, online 2026-02-18, 651:785–795) described the same strategy correcting behavioral abnormalities in mice, but did not mention that a human trial already existed, let alone a death. A Nature deputy editor responded that “no one ever told us a human trial was planned or under way”, and an editor’s note has been added to launch an investigation; the family has written requesting retraction.
Case two: HG302 — technically more “orthodox”, governance equally breached
How the trial was designed. No second-hand accounts; look directly at the raw fields of registry record NCT06594094:
| Field | Detail |
|---|---|
| Sponsor | HuidaGene (registry category: INDUSTRY) |
| Official title | An Investigator-initiated Clinical Study Evaluating the CRISPR-hfCas12Max Gene Editing Therapy in DMD |
| Phase | EARLY_PHASE1, single-arm, sequential |
| Dose groups | Low dose 2–3 patients; high dose 2–3 patients |
| Actual enrollment | 4 patients |
| Population | Boys aged 4–8 with DMD gene deletions of exon 52, 52–61 or 52–63, able to walk ≥10 m independently, willing to undergo muscle biopsy |
| Administration | Single intravenous infusion; about 110 weeks per patient (6-week screening + 104-week follow-up) |
| Primary endpoint | Incidence and severity of systemic AEs / SAEs / dose-limiting toxicities within 26 weeks |
| Data monitoring committee | Not declared — registry field oversightHasDmc: false |
| Site / start | Shanghai Children’s Medical Center / 2024-11-06 |
The sponsor is registered as a company, yet the title states it is investigator-initiated. This is precisely the most typical form of China’s IIT channel — a company’s asset, company-supplied drug, company-named sponsor, but not going through the path of a drug registration clinical trial. Precision is needed here: an IIT is not “just ethics approval”. Under the Measures for the Administration of Investigator-Initiated Clinical Research Conducted by Medical and Health Institutions, effective October 1, 2024, interventional clinical research requires project approval review by the medical and health institution, independent scientific review and ethics review, and study information must be uploaded to the national medical research registration and filing information system (which assigns a unified number) upon project approval; the medical and health institution is the entity responsible for implementation. The real divide between an IIT and a registration clinical trial is that it does not go through prior technical review of a drug IND by the national drug regulator.
The primary endpoint is safety, yet the data monitoring committee field in the registry record says “none”. The probative force of this field must be pinned down precisely: it can only show that the study did not register a formal independent data monitoring committee, and cannot further prove that dose escalation had no independent review of any kind — the company said in its announcement that the event was reported through the hospital ethics committee and project oversight procedures, and whether there were other safety review arrangements cannot be found in public information. But for a first-in-human dose-escalation study whose primary purpose is safety, not having a formal DMC is itself a design choice that needs explaining; and this is not anyone’s inference — it is written in the public registry record.
Dose: the heart of the problem. According to STAT (via a Chinese account in The Intellectual), the company’s former CEO Lu Yingming disclosed at the ASGCT annual meeting in May 2025 that only trace dystrophin was detected in the first low-dose patient; the dose was subsequently raised to the order of about 1×10¹⁴ viral particles per kilogram of body weight. The company’s August 5 announcement itself gave no dose value.
Put this number into the coordinate system of the literature and the problem is clear. A review by Duan and Herzog published in Molecular Therapy in January 2026 (a mechanistic, narrative review, without the search methodology of a systematic review) compiled reported deaths from systemically administered AAV gene therapy:
| Program | Dose (vg/kg) | Outcome |
|---|---|---|
| PF-06939926 (Pfizer, AAV9) | 2×10¹⁴ | 16-year-old patient died of cardiogenic shock on day 6 |
| CRD-TMH-001 (AAV9, CRISPR transcriptional activation) | 1×10¹⁴ | 27-year-old patient died of ARDS on day 8 |
| Elevidys (Sarepta, AAVrh74) | 1.33×10¹⁴ | Two adolescents died of liver failure after 11–12 weeks |
| GEN6050X (dual AAV9, cytosine base editing) | 5×10¹³ | 10-year-old child developed ARDS on days 4–5, recovered without sequelae |
HG302’s high dose sat in the literature’s danger zone
The review’s mechanistic conclusion: acute deaths 1–14 days after dosing are driven by an excessive innate immune response to the AAV vector, with endothelial injury as the core pathology; the clinical spectrum includes ARDS, capillary leak syndrome, hemophagocytic syndrome and TMA — covering exactly the two causes of death in the two cases in this article. First among the mitigation strategies the review offers is dose reduction: hoping to push clinical doses down to <5×10¹³ vg/kg through engineering such as muscle-tropic capsids.
In other words: HG302’s high-dose group fell within a range the literature had long flagged as dangerous; the range around 1×10¹⁴ had already seen fatal events with at least three products. This is not hindsight — the vast majority of these cases were public before August 2025.
There is another mechanism worth probing — it must be stated that this is a hypothesis, not a conclusion: the low-dose group detecting only trace dystrophin objectively created pressure to escalate. But the actual basis for dose escalation, internal discussions and safety assessment records have not been made public, so “insufficient efficacy drove unsafe escalation” is for now only a hypothesis worth investigating and cannot be written as the cause of the incident. It is worth noting because it points to a design problem that can be prevented in advance: when escalation decisions also carry pressure to produce data, independent review is most valuable.
Disclosure timeline.
| Date | Event |
|---|---|
| 2024-08 | Shanghai Children’s Medical Center publicly recruits subjects |
| 2024-11 | Study start (11-06); first dose completed in December |
| 2025-05 | Early data presented at the ASGCT annual meeting Presidential Symposium (low dose well tolerated, functional trends) |
| 2025-08 | Death of a high-dose subject (per company announcement) |
| ~15 months after | The company issued no further official press releases (per STAT) |
| 2026-01 | Full investigation results submitted for peer review (company announcement) |
| 2026-02 | Registry status at one point marked “completed” (per media reports) |
| 2026-07-31 | Registry update submitted, with status “active, not recruiting” |
| 2026-08-05 | Company issues an HG302-01 statement; STAT publishes an investigative report the same day |
The core wording of the company’s announcement: the subject, “after high-dose systemic administration of an adeno-associated viral vector, developed acute respiratory distress syndrome in the context of severe complement and cytokine activation”; the subject was the last patient enrolled under the established protocol, the other 3 did not develop similar serious clinical syndromes and remain in long-term follow-up; the event was reported under GCP through the hospital ethics committee and project oversight procedures; full scientific details will be made public after the paper is published. Follow-up reporting by Science added that the boy was the 4th patient enrolled and received the highest dose at the time.
As of the time of query (registry data version 2026-08-19), the public record of NCT06594094 still did not reflect this death — no reason for termination, no results, status “active, not recruiting”. By contrast, six days after being exposed, XH-25-002 wrote the cause of death, date of death and causality assessment in full into its reason-for-termination field. Same period, same type of event, same registry — two different treatments. This is not a technical issue; it is a matter of disclosure policy.
The January 2026 review by Duan and Herzog aimed to exhaustively compile deaths from systemic AAV, to give the whole industry a decision basis on dose and immunosuppression. Search the full text: there is no HG302, no hfCas12Max, no HuidaGene. The reason is simple — when the review was written, this death had not yet been disclosed. That is the real cost of concealment: every team designing a systemic AAV trial in DMD at the time saw one fewer data point.
Four things visible only when the two are viewed together
4.1 The same channel, the same system. Both studies went through the IIT route — to stress again, this does not equal “just ethics approval”: the measures in effect from October 2024 require institutional project approval review, scientific review, ethics review, upload to the national medical research registration and filing information system, and the institution’s whole-process quality and safety management; the divide from registration clinical trials is that there is no prior technical review of a drug IND by the national drug regulator. The two implementing institutions — Xinhua Hospital and Shanghai Children’s Medical Center — are both affiliated hospitals of Shanghai Jiao Tong University School of Medicine.
An analysis by Yang et al. published in Clinical and Translational Science in 2025 (data as of April 30, 2024) counted 1,033 cell and gene therapy IITs conducted by institutions in mainland China, with enrollment about twice that of the 329 IND trials in the same period, cumulatively more than 30,000 people; IITs in this field have grown rapidly since 2015 — that is “accelerating since 2015”, not “a thousand added since 2015”. This volume itself is not the problem — IITs are an important channel for early exploration in rare diseases worldwide. The problem is that under this volume, the density of oversight has not grown proportionally.
4.2 Immunotoxicity is not an accident but a known principal risk. One TMA and one ARDS both fall within the “acute-phase endothelial injury spectrum” described by the review; TMA in the AAV field already has a dedicated literature of case reviews and is known to be associated with anti-capsid antibodies and complement activation. These two deaths are not “black swans no one could have foreseen” — the type of risk, mechanism and mitigation measures are all written in the literature. As for which mitigation measures each study took and to what extent they were implemented, public information is insufficient to judge.
Notably, the XH-25-002 eligibility criteria explicitly excluded anti-AAV neutralizing antibody titers >1:50, required normal coagulation and platelets, and included protocol-specified prophylactic medication — the risk was identified. Identified and still an incident — what was missing was not knowledge but a braking mechanism.
4.3 A counterexample exists — and it’s in China. At the same time there was a third Chinese DMD base-editing IIT: GEN6050X (GenAssist), conducted at Peking Union Medical College Hospital, NCT06392724, principal investigator Dai Yi.
| Aspect | HG302 | GEN6050X |
|---|---|---|
| Dose | Reported at about 1×10¹⁴ vg/kg | 5×10¹³ vg/kg |
| Data monitoring committee | Not declared in registry record | Declared in registry record |
| Acute immune event | ARDS, death | ARDS (days 4–5), recovered without sequelae |
| Event disclosure | Made public only about 12 months after the event, after media questioning | Reported publicly at ASGCT 2025, now included in review literature |
| Regulatory extension | Holds FDA orphan drug and rare pediatric disease designations | FDA IND clearance in 2025-03, FDA orphan drug designation in 2025-06 |
| Subsequently | Study reported to have stopped | On 2026-07-24 the company disclosed one-year follow-up of 3 patients (aged 6.5–10): NSAA / PUL 2.0 stable or slightly improved (n=3, uncontrolled, self-disclosed by the company — insufficient to prove efficacy) |
This table overturns a conclusion that is easy to blurt out — “the IIT channel is inherently unsafe”: same channel, same indication, same class of modality, and another study followed a completely different process to a completely different outcome.
But it must also be said fully: this is not a comparison from which causality can be inferred. The two differ in vector, editor, dose, immunosuppression regimen and patients, with samples of only 4 and 3 patients respectively; a side-by-side comparison cannot prove “lower dose + a DMC, therefore a better outcome”. Its value lies not in causality but in providing a contemporaneous, same-country, same-indication example of practice: dose chosen within the range recommended by the literature, a formal DMC registered, acute events reported truthfully at an academic meeting — all actions that can be copied directly. Whether the outcome changed as a result is another question that can only be answered with individual-level data.
4.4 A mismatch of risks and benefits. Lay out where the benefits flow in the two cases: the family pays (XH-25-002), the child bears first-in-human risk, papers and conference data go to the team and the company, and financing and licensing value go to shareholders. And once an SAE occurs — Order No. 818 requires suspending the research, having the ethics committee assess it and reporting to the competent authority within 5 working days, but it has not established a mechanism for immediate public disclosure of individual SAEs. (Overseas situations cannot be generalized: listed companies may also be bound by continuous securities disclosure obligations, and requirements differ across jurisdictions and registries.)
Registry updates depend on self-discipline, reporting at academic meetings is voluntary, and media questioning depends on luck. These two events came to light, one because the family gave documents to reporters, the other because reporters kept asking for months.
What Order No. 818 changed, and what it didn’t
The Regulations on the Administration of Clinical Research and Clinical Translational Application of New Biomedical Technologies (State Council Order No. 818), adopted at a State Council executive meeting on September 12, 2025, signed on September 28 and published in October, took effect on May 1, 2026, with 7 chapters and 58 articles. Both deaths occurred before it took effect, but comparing against its provisions shows the direction of regulation clearly.
Article 10 The initiating institution must be “a legal person lawfully established within China”, and must ensure the technology has been shown safe and effective through non-clinical research.
Article 11 The implementing institution must be a Grade III-A medical institution, with both a clinical research academic committee and an ethics committee (dual committees).
Article 15 Filing with the State Council health authority within 5 working days of passing academic and ethics review.
Article 17 The competent authority shall publish filed studies and their initiating and implementing institutions; where there are major technical or ethical risks, termination shall be required.
Article 20 No research-related fees may be charged to subjects; violators are ordered to refund and fined up to 5 times the amount (Article 49).
Article 25 When a serious adverse reaction occurs, clinical research shall be suspended, assessed by the ethics committee, and reported to the competent authority within 5 working days.
Scale of penalties Up to RMB 10 million for prohibited research, with individuals barred from practice for 10 years to life; RMB 500,000–5 million for proceeding without non-clinical proof; RMB 100,000–500,000 for falsifying, altering or concealing research records.
There is no mechanism for immediate public disclosure of individual SAEs. Article 25 requires suspending research, ethics assessment and reporting to the competent authority — not public disclosure; synchronized updating of registries still depends on self-discipline. What these two events really have in common — one 16 months and one about 12 months before entering public view — the new regulations have not directly plugged.
Boundary rules have landed, but the power to classify has been given to the initiating institutions themselves. The supporting document authorized by Article 55 has already been issued: the Guiding Principles on the Classification of New Biomedical Technologies versus Drugs and Medical Devices (Interim) (Guo Wei Ke Jiao Fa [2026] No. 12, agreed with the NMPA), issued by the National Health Commission on April 28, 2026, published on April 30 and effective from May 1. The mechanism: the NHC organizes the formulation of a Guiding List for Filing Clinical Research on New Biomedical Technologies (the first edition covers 17 items in 5 major categories, including new gene therapy technologies, new cell and derivative therapy technologies and new brain–computer interface technologies), and initiators of clinical research “self-classify” with reference to the list — those classified as technologies file promptly under the list’s standard names and filing requirements; those classified as drugs or medical devices must, per law and regulation, conduct clinical trials after approval and then apply for marketing registration. The list also has a removal mechanism: technologies with major safety or efficacy problems or uncontrollable risks are removed directly, and projects filed before removal should be terminated promptly. So the accurate statement is not “the rules haven’t landed so the gray zone remains”, but: the rules have been established; what remains is execution and oversight of self-classification.
Which track HG302 should follow is itself an unresolved question. Article 57 requires studies already under way before the effective date to complete filing before June 1, 2026 — but this only applies if the project is classified as a “new biomedical technology”; if classified as “a clinical trial conducted to develop a drug”, it should follow the drug administration path and does not belong on the filing track of Order No. 818. The right question is two-part: if classified as a new biomedical technology, its status for existing-study filing has not been seen publicly; if treated as drug development, it needs to be explained why the first-in-human study did not go through the NMPA IND path.
Compared with the fine of just over RMB 20,000 actually imposed in September 2025 — this is a gap of two orders of magnitude. But one point of timing must be made clear: the new regulations took effect on May 1, 2026, while the two deaths occurred in March and August 2025 respectively, so using the new regulations to retroactively characterize these two events does not hold. Its significance is that the same practices today would trigger completely different consequences.
Practical implications for practitioners
If you work in gene therapy R&D
Three things go from “best practice” to “must be written into SOPs”: first, first-in-human systemic AAV dose escalation must have data monitoring arrangements independent of the research team, whether or not it is an IIT; second, dose ceilings need a written justification benchmarked against the literature, and a dose on the order of 1×10¹⁴ vg/kg today needs additional justification; third, SAE disclosure policies must define triggers and time limits in advance, rather than being decided afterwards by PR.
If you do BD or investment due diligence
From today, the credibility of Chinese IIT data is a due diligence item that must be listed separately. No longer just ask “are there data”, but: does the registry record match actual enrollment, was a DMC declared, who reviewed dose-escalation decisions, were historical SAEs disclosed simultaneously on registries and at academic meetings, and which track was the study classified and filed under. Discrepancies between the registry record and the company’s statements are themselves a quantifiable governance signal — the point in this article that NCT06594094 still does not reflect the death is something anyone can check themselves.
If you care about the direction of policy
What the industry really worries about is not these two events themselves but a swing back. Some practitioners have publicly expressed concern about a “regulatory backlash” — the efficiency advantage of China’s IIT channel in early exploration is real, and does indeed exert competitive pressure on Western peers. But preserving this channel depends not on defending it, but on making the GEN6050X way of doing things the industry default.
Five questions still without answers
The exact dose and immunosuppression regimen of HG302’s high-dose group, and the basis for the decision to escalate from low to high dose
The company says full results were sent for peer review in January 2026, pending publication.
The vector dose of XH-25-002, the full macaque toxicology data, and the specific decision process by which ethics approval came before the toxicology report was finalized
The conclusions of the Shanghai Jiao Tong University School of Medicine special working group have not been published.
The outcome for the Nature mouse paper
Currently an editor’s note; whether it proceeds to correction or retraction is undecided.
Determination of responsibility and follow-up in both cases
To date, public information shows no formal disciplinary decisions against the researchers involved.
How self-classification will be overseen
The classification guiding principles took effect on May 1, 2026, handing the power to classify to initiating institutions; what triggers review, who reviews, and the consequences of misclassification have no public precedents yet.
Conclusion
In these two events, the most easily misread takeaway is “gene editing is unsafe”.
It isn’t. In the same period, in the same country, for the same type of indication, GEN6050X chose its dose within the range recommended by the literature, registered an independent data monitoring committee, reported the case truthfully at an academic meeting after ARDS occurred, and the child recovered; in July 2026 the company disclosed one-year follow-up of 3 patients. These data do not constitute proof of efficacy (n=3, uncontrolled), but they prove something else: within the same channel, a different way of doing things is feasible — and someone was already doing it at the time.
Restraint is needed here: the direct cause of the two deaths was acute immunotoxicity triggered by high-dose AAV — one TMA, one ARDS, with clear mechanisms; the attribution of responsibility is still under investigation, and an article should not substitute for the investigation in drawing conclusions. XH-25-002 enrolled only 1 patient and had no dose escalation at all, so attributing both cases to “nobody holding back the hand that raised the dose” does not hold.
But one thing need not wait for the investigation: both events only came into public view more than a year after they happened. That XH-25-002’s registry record went 16 months without an update is established fact; for HG302, internal reporting and public disclosure must be distinguished — the company says it reported within the required time limits through the hospital ethics committee and project oversight procedures, and there is currently no public regulatory investigation conclusion finding it failed to meet statutory reporting obligations, but its external disclosure was indeed about a year late.
The cost is concrete: the January 2026 review that should have captured all deaths from systemic AAV did not include HG302. The whole industry therefore saw one fewer data point, and lost a year.
Data & Sources
Sources of fact: ClinicalTrials.gov registry records (NCT06860672, NCT06594094, NCT06392724, data version 2026-08-19); original text of State Council Order No. 818; Guiding Principles on the Classification of New Biomedical Technologies versus Drugs and Medical Devices (Interim) (Guo Wei Ke Jiao Fa [2026] No. 12); Measures for the Administration of Investigator-Initiated Clinical Research Conducted by Medical and Health Institutions (Guo Wei Ke Jiao Fa [2024] No. 32); HuidaGene announcement of 2026-08-05; Shanghai Jiao Tong University School of Medicine statement of 2026-07-26; Duan D & Herzog RW, Molecular Therapy, January 2026; Yang K et al., Nature 651:785–795; Yang et al., Clinical and Translational Science, 2025; and public reporting by Science/Retraction Watch, STAT, CNN, Bloomberg, SCMP, The Intellectual, Economic Observer, 21st Century Business Herald and others. For content based on media reports, such as doses, funding and penalty amounts, the nature of the source is flagged at each point in the text. Information as of August 19, 2026 (registry data version). This article is based on public information and represents only the author’s personal analysis and judgment. It does not represent any institution’s position and does not constitute investment or medical advice.