Two children died in gene-editing trials run out of Shanghai hospitals within five months of each other, one on March 31, 2025, and one in August 2025. Neither death reached the public for roughly a year. The first surfaced on July 23, 2026, through an investigation by Science, and the second on August 5, 2026, through a press release from the trial's sponsor [1][4]. The registries built to carry exactly this information carried it late in one case and, so far, not at all in the other.
These are two distinct cases, and as the story circulates they are beginning to blur together. They should not. The dates, hospitals, diseases, and sponsors are different, and the record is precise about each.
The first case is the one Science broke. A 6-year-old girl carrying a CHD3-R1025W mutation, a genetic change associated with a rare neurodevelopmental disorder, received an experimental gene-editing treatment at Xinhua Hospital, which is affiliated with Shanghai Jiao Tong University. She died on March 31, 2025, of thrombotic microangiopathy, a condition in which clots form in small blood vessels and shut down organs [3][4]. Her trial is registered as NCT06860672, sponsored by Yongguo Yu of Xinhua Hospital. The registry entry lists exactly one participant. The study is marked Terminated, and the entry states the cause in the registry's own vocabulary: the death was "definitely related to the study drug" [3].
The registry did not say so until sixteen months after she died. The acknowledgment of the death in NCT06860672 dates to a July 30, 2026 update, one week after Science's investigation ran [3][4]. Her family had paid more than $800,000 for the treatment, according to Retraction Watch's account of the Science reporting [4].
The second case is a different trial at a different hospital. HuidaGene Therapeutics' HG302-01, called the MUSCLE study, is testing a gene-editing therapy for Duchenne muscular dystrophy, delivered by an AAV vector, at Shanghai Children's Medical Center. It is registered as NCT06594094 with four participants enrolled [2]. One of them died in August 2025 of acute respiratory distress syndrome following "severe complement and cytokine activation" after receiving a high dose of the vector, according to the company [1].
HuidaGene disclosed the death itself, in a press release dated August 5, 2026, roughly a year after it happened [1]. The company says the findings were submitted for peer review in January 2026 and that the three remaining participants continue in follow-up [1]. As of today, the trial's registry entry lists the study as active and not recruiting, gives an estimated completion date of June 2, 2027, and posts no results [2].
Set the two timelines side by side and the pattern is the same. A terminated one-child trial recorded a treatment-related death only after Science asked questions. An active four-child trial's registry still shows nothing a year after a participant died, with the death disclosed instead through corporate communications [1][2][3]. Trial registries exist so that this information does not depend on a press release or an investigative reporter. They are the document a family reads before signing a consent form, and in at least one of these cases, before paying more than $800,000.
For sixteen months, on the record that counted, the girl's death did not exist. For twelve, the Duchenne patient's did not either. The corrections arrived in the same two-week window this summer, one prompted by a journalist, one written by the sponsor. Whatever else Chinese regulators or the registries' operators do next, the baseline fact is now on the record twice over: the disclosure system worked backward, with the public filings trailing the press instead of leading it.
