Published on

The cancer vaccine that cleared phase 3 is not universal. It is personal.

Authors

Listen to this article · The cancer vaccine that cleared phase 3 is not universal. It is personal.

0:00 / 0:00

Watch this article · The cancer vaccine that cleared phase 3 is not universal. It is personal.

On 19 August 2026, Merck and Moderna announced that a cancer vaccine had met the endpoints of a phase 3 trial — the first positive phase 3 for an mRNA cancer therapy, and the first for a treatment designed around one patient's tumour.

Nine weeks later, a different vaccine, aimed at a mutation carried by roughly 90% of pancreatic cancers, published its phase 2 result in Nature Medicine. It missed its primary endpoint.

Read together, the two results answer the question the headlines keep getting wrong. There is no single shot that teaches the immune system to hunt every cancer. There is something narrower: a vaccine built from one person's tumour, which works only after the immune system has had months to learn from it.

82%
free of recurrence at 24 months, against 70.0% on Keytruda alone
42%
lower risk of recurrence or death with the vaccine added
1,137
patients, all with completely resected stage IIB–IV melanoma

What the melanoma trial showed

INTerpath-001 enrolled 1,137 people whose melanoma had been removed but who remained at high risk of it returning. They were randomised two to one: the larger group received intismeran autogene — 1 mg every three weeks, up to nine doses — alongside pembrolizumab, the standard adjuvant immunotherapy; the smaller group received pembrolizumab alone.

After a median follow-up of 24.6 months, adding the vaccine reduced the risk of recurrence or death by 42% (hazard ratio 0.58; 95% confidence interval 0.45 to 0.75; P < 0.001). A prespecified sensitivity analysis excluding patients who missed planned doses put the reduction at 49%.

Vaccine + Keytruda82.2%
Keytruda alone70%
Recurrence-free survival at 24 months. The gap is what a phase 3 trial exists to measure.

Safety matched the combination's earlier studies: grade 3 to 5 treatment-related events in 14.2% of the vaccine arm against 13.4% of the control, with fever (52.2%), injection-site pain (51.0%) and fatigue (38.5%) the common complaints.

What it has not shown is whether people live longer. Overall survival is still maturing. That is usually the first thing a headline drops.

The effect arrives late, and that is the mechanism

The most informative number here is not the 42%. It is the timing.

During the first six months, while patients were still being dosed, the two arms were indistinguishable: a hazard ratio of 0.99. After the immunisation course finished, the gap opened to 0.42 — a 58% reduction.

That is what a teaching vaccine should look like. The shots do not attack the tumour. They hand T cells a list of proteins to recognise, and the killing happens afterwards, by an immune system that had to be trained first.

How one dose is made

  1. Tumour and blood

    Sequencing compares the cancer genome with normal DNA to find the mutations only the tumour carries.

  2. Up to 34 targets

    Computational models pick the abnormal proteins most likely to be visible to T cells.

  3. One mRNA strand

    The chosen sequences are encoded into a single molecule carrying a different message.

  4. Nine doses, 56 weeks

    1 mg every three weeks alongside pembrolizumab. The response keeps building after the last shot.

The doubtful part was logistics, and it held: the therapy was manufactured for more than 99% of randomised patients.

That is the real engineering achievement — a personalised medicine surviving sampling, sequencing, selection, manufacture, identity checks and distribution, on schedule, for a thousand people.

The pancreatic trial that missed

The second readout is the honest counterweight.

AMPLIFY-7P tested camupepimut, an off-the-shelf vaccine from Elicio Therapeutics carrying seven mutated-KRAS peptides attached to a lipid that ferries them into lymph nodes, where immune responses are organised. 144 patients with resected, KRAS-mutated pancreatic cancer were randomised two to one against observation.

The primary endpoint failed: median disease-free survival of 23.8 months with the vaccine against 21.2 months without, a hazard ratio of 0.85 that did not reach significance.

Vaccine arm23.8 months
Observation21.2 months
The primary endpoint, missed. A real result, reported as one.

Underneath the miss, the immunology behaved exactly as designed.

99%
of evaluable patients mounted T cells against mutated KRAS
44×
median rise in those T cells over baseline
96%
still responding at 46 weeks or later

Patients whose T cells responded strongly did better: above a 9.17-fold rise, median disease-free survival has not been reached, against 6.08 months below it. Among those with no detectable tumour DNA in their blood at baseline, it was not reached on the vaccine against 21.2 months on observation, with 70% against 56% free of relapse at 18 months.

A trial can miss its endpoint and still tell you where to aim. This one did.

Why "universal" is the wrong word

A vaccine for all cancers would need a target nearly every cancer shares. No such target has been found, and no vaccine has been shown in humans to work across cancer types.

What exists is narrower and, for a patient, better. Each tumour accumulates its own mutations, and those mutations produce proteins the immune system can learn to see. The melanoma trial reads that fingerprint per person. The pancreatic trial bets on one gene family mutated in most pancreatic cancers and produces the strongest T-cell responses of its class, but not yet a survival benefit on its own.

Off-the-shelf for many, bespoke for the few: that is the field's live argument, and the phase 3 went to the bespoke side.

What would have to be true

The vaccine is investigational. Moderna has said approval could come in 2027; nine trials are running across melanoma, lung, bladder and kidney cancer, plus an early study in pancreatic cancer.

Three things decide whether this becomes standard care. Whether overall survival follows recurrence-free survival. Whether manufacture scales past a thousand patients without a weeks-long turnaround becoming the bottleneck. And whether the design works outside melanoma, a cancer the immune system already half-recognises.

Prevention is further out. Moderna lists an investigational candidate, mRNA-4194, being studied in Lynch syndrome, where people are born with very high inherited risk. That is the first real test of a vaccine given before a tumour exists, and it is early.

The honest sentence is this: for one cancer, in one setting, after surgery, a vaccine built from that patient's own mutations measurably reduced the chance the cancer came back. Not the end of cancer — the first time the approach survived a trial big enough to prove it.

Source: Merck and Moderna, INTerpath-001 topline (19 August 2026) and detailed results presented at ESMO Congress 2026; George et al., "Lymph node-targeted, mKRAS-specific amphiphile vaccine in pancreatic cancer: the randomized phase 2 AMPLIFY-7P trial", Nature Medicine, 23 October 2026.

Contact: [email protected]