INTerpath-001 at the ESMO 2026 Presidential Symposium — the first phase 3 win for an individualized mRNA cancer vaccine, and the questions I still have. Presented by Dr Georgina Long.
Why I have been waiting a long time for this one
I have been a principal investigator on cancer vaccine trials since before most people had heard the letters mRNA strung together in that order. I have joked for years that if you are an anti-vaxxer, you are probably not going to like me very much. The joke lands less well than it used to, so let me put the serious version plainly: this technology was never only about infectious disease. The same platform that can teach an immune system to recognize a virus can teach it to recognize a patient’s own tumor — not a generic tumor, not a tumor from a cell line, but the specific mutational fingerprint of the cancer that was cut out of that one person three months earlier.
Today at the ESMO Presidential Symposium, it became a positive phase 3 trial.
So yes, I am excited. I also have a list of questions I cannot answer yet, and I think they are the right questions, so I am putting both in the same post.
Adjuvant melanoma is where good ideas go to die
To appreciate why this result matters, you have to appreciate how many reasonable strategies have failed in exactly this setting.
Adjuvant pembrolizumab and nivolumab are standard of care after resection of high-risk melanoma. They work. They also leave roughly 40–50% of stage III patients, and 30–40% of stage II patients, destined to recur. So for a decade the field has tried to beat single-agent PD-1 blockade by adding something.
| Trial | Added to PD-1 | Result |
|---|---|---|
| CheckMate 915 | Ipilimumab (anti-CTLA-4) | Missed primary endpoint |
| KEYVIBE-010 | Vibostolimab (anti-TIGIT) | Discontinued — more toxicity, no benefit |
| Relativity 098 | LAG-3 blockade | Disappointing in this setting |
Every one of those tried the same thing: release more immune brakes. None of them worked.
INTerpath-001 did something different. It did not add another checkpoint inhibitor or “release the brakes”. It added antigen specificity or simply “unique cancer targets”— telling the immune system what to look for rather than simply disinhibiting it further. That is the conceptual story of this trial, and it is why a result that could be dismissed as incremental is, I think, genuinely a turn in the road.
What is actually in the vial
Intismeran autogene is not one drug. It is a manufacturing process that produces a different drug for every patient.
- Sequencing. The resected tumor and the patient’s normal tissue are both sequenced by NGS. Comparing the two identifies the somatic mutations unique to the cancer. Roughly 1–2 months.
- Selection. Algorithms rank those mutations by how likely the resulting mutant peptides are to be presented on that patient’s own HLA molecules and recognized by T cells. Roughly 2 weeks.
- Immunogenicity testing. The selected candidates are screened for whether they are likely to provoke a real response. Roughly 2 weeks.
- GMP synthesis and release. The winning neoantigens — up to 34 of them — are encoded into a single mRNA construct, manufactured to clinical standard, and released for shipment. 90% of the time, patients received a vaccine that targeted 34 unique antigens!
That encoding number is a deliberate bet. Neoantigen prediction is imperfect; by most estimates only a small fraction of computationally predicted neoantigens turn out to be genuinely immunogenic. Rather than betting everything on picking the one perfect target, the design casts wide and lets the immune system sort it out.
The mechanism claim — and the trial showed supporting data for this — is that the vaccine does two things at once: amplifies T cell clones the patient already had, and creates entirely new ones that were undetectable before vaccination. The second part is the interesting one. That is the immune system learning something it did not previously know.
The data
Design. 1,137 patients with completely resected stage IIB–IV cutaneous melanoma, randomized 2:1 to intismeran plus pembrolizumab or placebo plus pembrolizumab. Nine vaccine doses over roughly one year alongside standard pembrolizumab. Stratified by stage and age. Formalin-fixed tissue was sufficient — no fresh specimen required. Median follow-up 24.6 months.
| Endpoint | Vaccine + pembro | Pembro alone |
|---|---|---|
| Recurrence-free survival | 82% | 70% |
| Distant metastasis-free survival | 88% | 80% |
| RFS hazard ratio | 0.58 — 42% risk reduction | — |
| DMFS hazard ratio | 0.62 — 38% risk reduction | — |
| Overall survival | Immature; early trend favorable | — |
Benefit was seen across both lower-risk and higher-risk stages. Stage III was the most common population enrolled; stage IV the least.
On safety, the honest summary is: there almost isn’t one. The added toxicity was fever and injection-site discomfort. Two deaths occurred on study, and both happened before the patient had received any vaccine at all — attributed to pembrolizumab.
I want to dwell on that for a moment, because in oncology we are conditioned to expect that efficacy is purchased with toxicity. Here it largely wasn’t. An eight-point absolute improvement in freedom from distant metastasis, for a sore arm and a low-grade fever, is not a trade most of us are used to being offered.
The bottleneck was not where anyone expected
This is the part of the presentation I found most operationally important, and it got the least attention.
1,789 patients were screened. 1,137 were randomized. The single most common reason for failure was inadequate tissue — roughly 17% of screened patients simply did not have enough tumor to build a vaccine from.
Now look at what happened to the patients who cleared that hurdle:
- 99% had a vaccine successfully manufactured
- 91% received the full complement of 34 neoantigens
The biotech worked. Pathology was the rate-limiting step.
I find this genuinely clarifying. Everyone worries about whether individualized manufacturing can scale, and the answer appears to be yes — it ran at 99%. What does not currently scale is making sure there is enough tumor left in the block after the diagnostic workup is done.
The questions I am still asking
I submitted several of these during the session. They are not objections — they are the things that determine how this gets used well rather than used indiscriminately.
1. Does ctDNA identify who benefits?
This is my biggest question. If post-operative circulating tumor DNA were collected, is the benefit concentrated in patients with detectable ctDNA — those with proven molecular residual disease — while ctDNA-undetectable patients derive little?
The caution is that melanoma sheds ctDNA poorly compared with, say, colorectal cancer. Detection is highly specific but relatively insensitive, which creates an asymmetry worth stating plainly: a positive ctDNA is a reason to escalate; a negative ctDNA is not yet a reason to withhold. But looking at those subgroup comparisons may shed some light on benefit.
2. Will overall survival ever improve?
The discussant made this point and it is a fair one. No adjuvant melanoma therapy has ever demonstrated an overall survival benefit — not pembrolizumab, not nivolumab. The entire adjuvant paradigm rests on recurrence-free and distant metastasis-free survival. So the honest claim today is fewer recurrences, not longer life, and we should say it that way until the data say otherwise.
3. Who were the 17% without adequate tissue?
I suspect they skewed toward stage IIB/IIC and early IIIA, where resected tumor volume is smallest. If that is true, the patients hardest to manufacture for are also the ones with the highest NNT — which compounds rather than offsets.
4. Does the vaccine work where the immune system already failed?
There is a lot of chatter among melanoma physicians about giving this to patients with no pathologic response after neoadjuvant immunotherapy, because that is where the unmet need is greatest. I understand the instinct. I am not sure the biology cooperates.
Pathologic non-response may be enriched for exactly the defects a vaccine depends on — B2M loss, HLA class I loss or loss of heterozygosity, impaired IFN-γ/JAK-STAT signaling. A vaccine does not bypass that machinery; it requires it. You can encode 34 beautifully selected neoantigens and present precisely none of them. A question for future research protocols I suspect.
The awkward question: where does this fit in the neoadjuvant era?
The discussant raised the sharpest critique of the day, and it deserves a straight answer rather than a defensive one.
Roughly 40% of patients enrolled were high-risk stage III/IV — the population that, in 2026, many of us would treat with neoadjuvant ipilimumab/nivolumab before surgery. NADINA has shown two-year outcomes in that range of 80–90%. So for a meaningful slice of this trial, the control arm was not what we would actually do today.
Here is the comparison as I see it:
| Adjuvant pembro + vaccine | Neoadjuvant ipi/nivo | |
|---|---|---|
| Toxicity | Minimal added | Substantial |
| Duration | ~1 year | 2 doses if major path response |
| Drug access | Individualized manufacture, high cost | Available today |
| Tissue requirement | Adequate specimen mandatory | None |
| Time to start | ~3 months | Immediate |
| Surgical delay | None | Yes |
| Patient selection | No biomarker | IFN-γ signature, PD-L1, TMB all predict benefit |
That last row is the one I keep returning to. NADINA has a biomarker signal and this trial does not. For the high-risk stage III patient, neoadjuvant therapy lets you select; the vaccine asks you to treat everyone and hope. That is not a reason to reject it — it is a reason to build the correlatives now.
Where I think this lands, for now:
- Stage IIB/IIC and resected stage III without macroscopic nodal disease — neoadjuvant is not an option here, and this is a genuine advance. Likely practice-changing once approved.
- Macroscopic stage III — the discussant is right that this trial does not settle it. My personal take is for a poor man’s approach of using upfront immunotherapy using patient’s own immune cells to recognize cancer targets while the tumor is intact. Then future trials can focus on the high risk non responders and perhaps in partial responders with adjuvant mRNA vaccine.
The second of those is the trial I would most like to see, with antigen-presentation status built in as a stratification factor rather than an afterthought.
Why I am excited anyway
Every question above is a question about how best to use this, not whether it works. That distinction matters.
What happened in Madrid is that an individualized mRNA cancer vaccine combination beat standard-of-care immunotherapy in a randomized phase 3 trial, with essentially no added toxicity, in a disease where every previous attempt to improve on adjuvant PD-1 blockade had failed. That is the first time any of that has been true.
It is also worth sitting with what the technology actually did. A tumor was removed from a patient. Its mutations were read, compared against that person’s own genome, ranked for immune visibility, encoded into a molecule, manufactured, and infused back into the same patient as a set of instructions. And it reduced the chance that their cancer came back.
That is not an incremental drug approval. That is a different category of medicine, and melanoma — as it has for checkpoint inhibitors, for targeted therapy, for cell therapy — got there first.
What I will be watching: overall survival as it matures, the stage IIB/IIC subgroup in detail, whatever ctDNA correlatives exist, and how quickly this moves through regulatory review. And on the operational side, whether the field takes the tissue problem seriously enough to fix it before approval rather than after.
I expect this to become available for patients with high-risk resected melanoma hopefully next year. When it does, the question in clinic will not be does it work — it will be is this the patient it works best for. We should start building the answer now.
Dr. Sajeve Thomas is a distinguished medical professional and a compassionate guide in the field of oncology. With over a decade of dedicated experience as a board-certified medical oncologist/internal medicine specialist, Dr. Thomas has become a trusted expert in the treatment of melanoma, sarcoma, and gastrointestinal conditions. He brings a wealth of expertise to the complex and challenging world of oncology.
Disclosures:
Dr. Thomas serves as a speaker for BMS, Merck, Ipsen, Natera, Immunocore, Pfizer, Sun Pharma, SpringWorks. He also receives industry grants in support of numerous clinical trials.
Patient cases described in this post are composites and have been altered — including but not limited to diagnosis specifics, demographics, timeline, and clinical details — to protect patient privacy and confidentiality. No case represents an actual, identifiable patient, and any resemblance to a specific individual is coincidental. Clinical scenarios involving other providers have similarly been altered and are presented for educational purposes only, not as a representation of any specific person’s practice. This post is for general educational purposes and does not constitute medical advice; it should not be used to guide individual treatment decisions. This article is for education and general information only. It is written about an anonymized case, shared with permission, and it is not personal medical advice. The individual described is not a patient of the author, is not treated in the United States or at the author’s institution, and no doctor–patient relationship is created by reading this. Statistics and treatment options are summarized from the current oncology literature and evolve quickly; specific percentages are ap
Leave a Reply