ATX-GD-59

The entire evidence base is one open-label, uncontrolled, single-arm study in twelve people ^[1]. There is no control group, no randomisation, no blinding and no efficacy hypothesis test; the registered primary endpoint was safety, and no sample size calculation was performed.

Of ten completers, five had free T3 back in the normal range at week 18 and two more had lower values than at baseline, which is where the frequently quoted seven-of-ten figure comes from. ⚠️ Three of ten got worse, with progressive rises in thyroid hormone.

⚠️ Several things make that response rate hard to attribute to the drug. Six of the ten completers were taking propranolol, which on its own is associated with spontaneous remission in 10 to 25 percent of patients over three to eight months. One participant's raised hormone levels had already normalised after a single dose, before any tolerance could plausibly have developed. One of the responders was unintentionally underdosed for the whole study. Response was defined after the fact as any reduction from baseline. And with ten people, the confidence interval around a 50 percent normalisation rate runs from roughly 19 to 81 percent.

⚠️ In the sponsor's posted results every standard deviation exceeds its own mean, for every measure at every timepoint. The registered secondary endpoints that would actually show tolerance being induced, covering T-cell activity and interleukin-10 expression, have no data posted against them at all.

The successor phase 2, testing three doses against placebo in 176 people, is registered as suspended because recruiting suitable participants was not feasible. Eight years after the phase 1 finished there is still no controlled efficacy data.

Over 22 weeks in twelve people the tolerability was good, and injection-site reactions dominated. Eleven of twelve reported adverse events, 180 of which were considered treatment-related, and 153 of those were injection-site reactions, all mild ^[1]. Nothing was severe, nobody died, and no adverse event caused a dose interruption or withdrawal.

Three serious events occurred in one participant: nausea judged possibly related, and vomiting and an episode of atrial fibrillation judged unrelated.

⚠️ Two findings deserve more weight than their frequency suggests. One participant developed antibodies against the drug at week 14 and still had them at week 22. For a therapy whose entire purpose is to induce tolerance, an immune response against the treatment is the specific failure mode the approach was designed to avoid. And one participant developed first-degree heart block at week 4, which improved when a concurrent beta blocker was stopped.

Twelve people cannot establish the safety of an immunotherapy, and the study was not designed to.

The apitope approach came out of work on how the immune system decides between attacking and tolerating: peptides that can load onto presenting molecules without being processed first are displayed by resting cells, and a T cell meeting its target that way is switched off. The platform was developed at Bristol and taken forward by Apitope, which applied it first to multiple sclerosis and then to Graves' disease, choosing thyrotropin receptor fragments already known to be the dominant T-cell targets in animal models. The phase 1 ran across eight UK hospitals and finished in 2018 with results that the company presented as encouraging and its own authors described more cautiously. The announced phase 2 never started under Apitope; the assets passed to another company, and the successor trial was registered and then suspended.