How? I mean the way every drug for example is tested for safety, they give it to people (healthy volunteers) in escalating doses, and watch them for adverse effects.
Rarely, it happens that a drug that was expected to be safe based on animal testing turns out to be dangerous in humans, and healthy volunteers die. It’s a tragedy, it happens rarely, but it happens, and safety trials are still the best way to start the process of clinical trials for new drugs. For some treatments, such as gene therapy, it’s not ethical to ask a volunteer to just try it to see if it’s safe. So they look for sick people, who are consenting to risking a new treatment with unproven safety, because their other choice is to not be treated.
This is how it’s always done. I wonder what you mean when you say the treatment should be tested for safety before giving it to patients.

Edit: this was a personalized experimental treatment developed specifically for this girl, on the parents’ initiative. The clinical team did test for antibodies against the virus before injecting the AAV into the csf. Lots of shady and wrong things on the research side, I’m not protecting them, just disagreeing with the previous comment that this tragedy would have been prevented by a simple test.

Turns out they did on both mice and primates, and the side effect is also evaluated:

Another potential concern, however, emerged about 1 month before Mei was treated: the final conclusions of a primate toxicology study conducted at PriMed. The company’s report, dated 17 February 2025, concluded all four monkeys that received the therapy, whether at low or high doses, developed moderate to severe liver damage.

However, the ethics board did not review the safety report:

An ethics panel at Xinhua Hospital in China approved a gene-editing trial, but had not reviewed primate data suggesting safety concerns. Another hospital panel later concluded the experimental treatment was “definitely related” to Mei’s death.

Even worse, the brain development data from the primate study is entirely unconvincing:

But four of the experts Science asked to assess this figure, including one of the original reviewers of the Nature manuscript, expressed concerns about how well the primate proof-of-concept study was conducted. The submitted paper had no control samples to show the stain was selectively highlighting the gene editor. “It’s completely unconvincing,” says David Sanders, a biochemist at Purdue University who has studied gene therapy. “One can’t have the confidence that one isn’t mostly looking at background staining.”