In Nature this February, a team of Chinese scientists reported that they had used a base editor to rewrite a single faulty letter of DNA in the brains of mice and reverse the animals’ behavioral abnormalities. Near the end, on pages 785 to 795 of volume 651, the paper offered one sentence of restraint: bridging the gap between preclinical research and clinical translation, it said, remains a significant challenge. Its authors had already crossed that gap. They had put the same class of therapy into a six-year-old girl, and she was dead before the manuscript went to press.

Her death is not in the paper. It is not in the trial record. It was not public at all until 23 July 2026, when Retraction Watch and Science reported what her family had decided to tell them.

Start with the record the team did keep. The trial carries a ClinicalTrials.gov registration, NCT06860672, an entry that by the investigation’s account had not been updated in more than a year. The patient, identified by the pseudonym Mei, was six years old. She had a variant in the CHD3 gene, the cause of Snijders Blok-Campeau syndrome, one of the rarest neurodevelopmental disorders on record, described in only dozens of people worldwide. It is not a fatal disease. It presents as intellectual disability, autistic-like behavior, and motor delay. She was, by every account, a living child with a rare but survivable condition when she entered the trial.

The intervention was not modest. According to the investigation, the team infused trillions of viruses carrying a base-editor payload into her spinal fluid, aiming to correct a single mispaired DNA base, a mistaken T that needed to read C, in her neurons. The work was led by the neuroscientist Zilong Qiu, of Shanghai Jiao Tong University’s Songjiang Research Institute, and carried out at Xinhua Hospital in Shanghai, affiliated with the same university’s medical school. The infusion was given in late March 2025. Within 7 days she developed a fever and signs of kidney failure and died. A hospital report, as described to reporters, concluded that the most likely cause was an immune reaction to the viruses used to deliver the editor.

Now follow the money, because the financing is where a tragedy hardens into a structure. The parents did not receive this therapy. They paid for it. More than $860,000, drawn from their savings and from relatives, went to help develop the treatment that was then given to their daughter. That figure appears in the investigation. It does not appear in the Nature paper, which omitted any reference to the family or its contribution, and it does not appear in the trial records. A family bankrolled the development of an experimental therapy for their own child’s non-lethal condition, and the published scientific account of that same platform was scrubbed clean of both the money and the child.

Hold the line between what the record proves and what others allege. The omissions are documented: the paper names no patient, no death, and no private funding, and Nature itself says it was not aware of the issues surrounding the clinical trial before it published. The judgment that the trial should never have happened belongs to the outside experts the reporters brought in. Seven of them reviewed the details. They concluded that the team downplayed the risks in describing them to the parents, overlooked safety signals in the animal studies, and proceeded even though success was unlikely. “This shouldn’t have gone to trial,” Steven Gray of UT Southwestern Medical Center told the investigation. Several said the underlying data could warrant a retraction. That is their assessment of the science. What is not in dispute is that the human outcome was removed from the human record.

The regulatory pathway does not stay in Shanghai. The trial, by the investigation’s account, proceeded under a provision that does not require approval from national regulators, an investigator-initiated route that lets a hospital and its ethics committee green-light first-in-human gene editing without the national medical-products authority signing off. Joy Zhang, a sociologist at the University of Kent who studies Chinese science governance, called it evidence of “the gap between what is intended and what has been put in place.” That gap is not unique to one country. It is the same permission structure that lets an ambitious lab, chasing a landmark paper and a first-in-human claim, run ahead of the oversight meant to catch it, and it is how the animal-model foundation of a fatal experiment could appear in a prestige journal as a clean proof-of-concept while the fatality it was built toward lived only in a hospital file.

So account for who has not moved. Qiu did not respond to repeated requests for comment; neither the university nor Xinhua Hospital answered. Nature has acknowledged only that it was unaware of the trial when it published, and has not said whether it will correct or withdraw the paper. The girl’s parents, who say the missing safeguards changed how they view the entire project, have asked the authors to retract, and that request sits unanswered. As of the investigation, no regulator had announced a review of a trial that killed a child under a rule that required no regulator to look. The records, for now, are held by the people with the most reason to keep them closed.

Sources

  1. Retraction Watch / Science – investigation into the never-disclosed gene-editing death (2026-07-23)
  2. Science – “Exclusive: Death of girl in Chinese gene-editing trial was never made public”
  3. Nature – Qiu et al., “In vivo base editing of Chd3 rescues behavioural abnormalities in mice” (vol. 651, pp. 785–795, 2026)
  4. Live Science – coverage of the investigation and China’s gene-therapy oversight
  5. Medical Xpress – “Death of girl in Chinese gene-editing trial kept secret: report”
  6. RNZ – “Death of girl in Chinese gene-editing trial kept secret: report”