The question the trial was built to answer
The study published in Nature Medicine is an open-label 3+3 dose-escalation trial delivering TX103, an autologous CAR-T product targeting B7-H3, directly into the skull in recurrent glioblastoma. Its primary endpoints were safety, the maximum tolerated dose and the recommended phase 2 dose. In the trial 15 patients received a total of 72 infusions, 13 underwent repeated infusions, and three dose levels of 20 million, 60 million and 150 million cells per infusion were tested.[1]
When safety is the primary endpoint, survival and response stay secondary and descriptive, and the paper presents them that way. Who the figures describe is set by the eligibility criteria: patients aged 18 to 75 whose tumours are at least 30 per cent positive for B7-H3. Being well enough to receive repeated infusions into the skull is itself a selection threshold.[1]
What the survival numbers can carry
The reported results run as follows: 12-month overall survival of 66.7 per cent, median overall survival of 19.1 months from the first infusion, and a 95 per cent confidence interval for that estimate that starts at 8.93 months and never reaches its upper bound. Disease control was achieved in 8 of 14 patients with measurable disease, and one complete response was sustained through the latest follow-up.[1]
A wide interval that never reaches its upper bound makes 19.1 months hard to read as a single number; the honest description is the interval itself. The same median could also arise from selecting patients able to tolerate repeated intracranial infusions rather than from any effect of the cells. With no comparator arm, the two explanations cannot be separated from inside the trial.[1]
The harm side and the recommended dose
Treatment-related adverse events included low-grade cytokine release syndrome at 86.7 per cent, sinus tachycardia at 53.3 per cent, vomiting at 53.3 per cent, hypertension at 53.3 per cent and raised intracranial pressure at 46.7 per cent. Three grade 3 events were counted as serious, and two of them occurred at the top dose. The recommended phase 2 dose was set at the middle level of 60 million cells per infusion rather than the highest. If a phase 2 trial with a comparator arm is run, the survival estimate will for the first time become readable against a control group.[1]
Meanwhile the picture facing the patient does not change: options in recurrent glioblastoma are few, and this therapy was designed for patients whose tumours are at least 30 per cent positive for B7-H3, in whom a surgical access route can be established, and who can return for repeated infusions. What a phase 1 delivers is a tolerability picture and a dose; it promises no cure.[1]