Mei's medical team infused trillions of adeno-associated virus (AAV) base editors directly into her spinal fluid. Seven days later, she died from thrombotic microangiopathy — a severe immune and vascular reaction triggered by the viral vector .

The death was never publicly disclosed by the researchers. An investigation by Science magazine and Retraction Watch, published in July 2026, brought the case to light more than a year after it occurred . Shanghai Jiao Tong University School of Medicine subsequently launched a comprehensive investigation .

A second death in a Duchenne muscular dystrophy trial

In a separate trial in China, a child with Duchenne muscular dystrophy died after receiving a gene-editing treatment developed by Shanghai-based HuidaGene Therapeutics . The company attributed the August 2025 death to acute respiratory distress syndrome as part of a severe immune reaction to the therapy .

Both deaths share a common mechanism: the viral vector delivery system (AAV) provoked catastrophic systemic immune responses that animal studies had not adequately predicted.

Stanford expert: Human testing is 3–5 years away

Karoly Nikolich, an adjunct professor at the Stanford University School of Medicine and a leading figure in the commercialization of brain science, has stated that human testing for CNS gene-editing therapies could take up to five years to ramp up safely .

Nikolich identifies three core problems that must be resolved :

• Limited understanding of systemic genetic effects. Editing genes in the nervous system can trigger unpredictable off-target edits and immune reactions that are not well-characterized from animal models alone.
• Techniques not mature enough for human translation. Current delivery methods — AAV viral vectors — can cause severe inflammatory and thrombotic responses at the doses needed to reach the brain, as the China deaths tragically demonstrated.
• Better preclinical models and AI-based prediction tools are needed first. Nikolich argues that the field must invest in more sophisticated models — including AI-driven risk prediction — before moving to human subjects.

"The techniques are reasonably well established for animal experiments, but they really haven't been as mature for human studies," Nikolich said in a detailed interview .

The broader picture: An industry losing momentum

Nikolich's 5-year estimate aligns with broader analyses. A 2025 review in Frontiers in Genome Editing noted that CRISPR-based clinical applications for the CNS are projected at 5–10 years away . Industry tracking published in Nature Medicine showed that the total number of active gene therapy development programs actually declined in 2025, reversing a prior upward trend . Several trials were placed on temporary clinical hold in 2025 .

The key takeaway

The China fatalities demonstrate that current gene-editing approaches for the nervous system carry life-threatening risks — especially from vector-driven immune responses — that animal models failed to flag. The field needs deeper biological understanding of systemic off-target effects and safer delivery technologies before human trials can responsibly proceed. Credible expert estimates point to at least several more years of preclinical work.

Until then, gene editing for the brain remains a promise — not a treatment.

Mei's medical team infused trillions of adeno-associated virus (AAV) base editors directly into her spinal fluid. Seven days later, she died from thrombotic microangiopathy — a severe immune and vascular reaction triggered by the viral vector .

The death was never publicly disclosed by the researchers. An investigation by Science magazine and Retraction Watch, published in July 2026, brought the case to light more than a year after it occurred . Shanghai Jiao Tong University School of Medicine subsequently launched a comprehensive investigation .

A second death in a Duchenne muscular dystrophy trial

In a separate trial in China, a child with Duchenne muscular dystrophy died after receiving a gene-editing treatment developed by Shanghai-based HuidaGene Therapeutics . The company attributed the August 2025 death to acute respiratory distress syndrome as part of a severe immune reaction to the therapy .

Both deaths share a common mechanism: the viral vector delivery system (AAV) provoked catastrophic systemic immune responses that animal studies had not adequately predicted.

Stanford expert: Human testing is 3–5 years away

Karoly Nikolich, an adjunct professor at the Stanford University School of Medicine and a leading figure in the commercialization of brain science, has stated that human testing for CNS gene-editing therapies could take up to five years to ramp up safely .

Nikolich identifies three core problems that must be resolved :

• Limited understanding of systemic genetic effects. Editing genes in the nervous system can trigger unpredictable off-target edits and immune reactions that are not well-characterized from animal models alone.
• Techniques not mature enough for human translation. Current delivery methods — AAV viral vectors — can cause severe inflammatory and thrombotic responses at the doses needed to reach the brain, as the China deaths tragically demonstrated.
• Better preclinical models and AI-based prediction tools are needed first. Nikolich argues that the field must invest in more sophisticated models — including AI-driven risk prediction — before moving to human subjects.

"The techniques are reasonably well established for animal experiments, but they really haven't been as mature for human studies," Nikolich said in a detailed interview .

The broader picture: An industry losing momentum

Nikolich's 5-year estimate aligns with broader analyses. A 2025 review in Frontiers in Genome Editing noted that CRISPR-based clinical applications for the CNS are projected at 5–10 years away . Industry tracking published in Nature Medicine showed that the total number of active gene therapy development programs actually declined in 2025, reversing a prior upward trend . Several trials were placed on temporary clinical hold in 2025 .

The key takeaway

The China fatalities demonstrate that current gene-editing approaches for the nervous system carry life-threatening risks — especially from vector-driven immune responses — that animal models failed to flag. The field needs deeper biological understanding of systemic off-target effects and safer delivery technologies before human trials can responsibly proceed. Credible expert estimates point to at least several more years of preclinical work.

Until then, gene editing for the brain remains a promise — not a treatment.