Somewhere in a lab right now, a piece of tumour that used to belong to one particular person is being read letter by letter, and what comes out the other end is a medicine nobody else on earth can take.

That is the odd, quietly enormous thing about the result Merck and Moderna announced on 19 August. A cancer vaccine built from a single patient’s own mutations passed a Phase 3 trial. No individualised neoantigen therapy had ever done that, and neither had an mRNA cancer treatment of any description.

What the trial measured

The study, INTerpath-001, enrolled 1,137 people whose melanoma had been surgically removed but who remained at high risk of it coming back. Two thirds received the personalised therapy, intismeran autogene, alongside Merck’s immunotherapy Keytruda. The rest got Keytruda on its own, already the standard of care in this setting.

At a pre-planned interim look at the data, independent monitors found the combination beat Keytruda alone on recurrence-free survival, meaning patients went longer before the cancer returned. It also won on distant metastasis-free survival, which tracks whether the disease reaches other organs. Merck reported no new safety signals, and the study continues so investigators can establish whether people actually live longer.

That last point matters most.

How you build a drug for one person

How do you make a medicine that only one human being can use? Start with the reason cancer is hard to fight at all. Tumour cells are corrupted copies of your own tissue, which is precisely why the immune system so often waves them through. Mutations change the calculation. A mutated gene can produce a protein that exists nowhere else in your body, and those oddities are called neoantigens. They are, in effect, flags. The immune system just needs someone to point at them.

So surgeons remove the tumour, a lab sequences it to find that patient’s particular mutational fingerprint, and software selects up to 34 neoantigens most likely to be visible to that person’s immune system. Those get encoded into a synthetic mRNA strand, wrapped in a lipid nanoparticle and injected. The body reads the instructions, builds the flagged proteins itself and presents them to the T cells. Keytruda handles the other half of the job by releasing the brakes those T cells normally operate under.

Every batch is bespoke. As BioPharma Dive noted in its coverage, this is a manufacturing model with no real precedent at commercial scale, which is a different sort of problem to the scientific one and arguably the harder of the two.

The numbers nobody has seen yet

Nobody outside the two companies has laid eyes on the actual data. What was released is topline: the endpoints were met, the effect was statistically significant. No hazard ratios, no confidence intervals, no survival curves. As NBC News pointed out, none of it has been peer reviewed, and the full dataset is being saved for a medical meeting.

The public evidence so far comes from the earlier Phase 2b study, KEYNOTE-942, whose five-year follow-up was presented at the American Society of Clinical Oncology meeting in June. There the combination cut the risk of recurrence or death by 49 per cent, and the risk of the disease spreading by 59 per cent. Encouraging figures, drawn from 157 patients in a single mid-stage trial. One trial is one trial, however good the numbers look.

Why melanoma was first

Australians get melanoma at a rate few other populations can match, which is part of why a Sydney oncologist ran this study. Professor Georgina Long, medical director of Melanoma Institute Australia and the trial’s principal investigator, called the result a landmark moment for adjuvant melanoma treatment. Adjuvant here means treatment given after surgery to stop the thing returning.

Biology also made it the obvious first target. Sun-damaged skin cells pile up mutations by the thousand, which hands the vaccine designers a generous menu to choose from, and this particular cancer already responds unusually well to immunotherapy. The T cells being recruited arrive somewhere they can plausibly do some good.

What the result does not prove

A researcher quoted by CBC News made the sensible point that everyone in this trial had early-stage disease which surgeons had completely removed. Those are the easiest conditions the field can offer: minimal tumour left, an immune system that has not yet been ground down, time to mount a response. Success there does not guarantee success in later-stage melanoma, where previous attempts have produced negative results.

The approach also depends entirely on having a tumour to sample in the first place. No surgery, no sequencing, no shot.

Where this goes next

The wider program now runs to nine mid and late-stage trials spanning melanoma, lung, bladder and kidney cancer, with earlier work under way in pancreatic and stomach tumours. Those are less forgiving targets. Kidney cancers carry fewer mutations, so there is less for a vaccine to aim at, and pancreatic tumours are famously good at smothering immune attacks before they start. Roche and BioNTech are testing a rival personalised vaccine, with colon cancer results expected in 2027, as CNN reported.

Investors, not a group known for restraint, sent Moderna’s share price up by roughly 177 per cent in a day.

I have spent years reading predictions that mRNA would eventually do more than fight viruses, and mostly filing them under wishful. This is the first time that filing looks like a mistake. The question now is not whether the science holds, because on current evidence it probably does, but whether any health system can be talked into building a separate factory run for every patient who walks through the door.