Analysis limits. The New England Journal of Medicine paper has been accepted but was not online when this was written, so the Kaplan–Meier curves, supplement and full subgroup tables were not available. Numbers here come from the companies' October 22 release, the ESMO abstract (LBA1) and trade-press reports of the October 24 presentation. Nothing here has been checked against the full paper.
This follows the August piece written when the companies announced the trial had met its endpoints without releasing numbers: The Vial That Fits One Person.
INTerpath-001 randomized 1,137 patients with completely resected stage IIB to IV cutaneous melanoma to intismeran autogene plus pembrolizumab (Keytruda) or placebo plus pembrolizumab. Intismeran is an mRNA product made for each patient from sequencing of their own tumor, encoding up to 34 neoantigens.
At a median follow-up of 24.6 months, adding intismeran reduced the risk of recurrence or death (hazard ratio 0.58) and of distant metastasis or death (hazard ratio 0.62). Overall survival is immature and not statistically significant. The data were scheduled for ESMO on October 24 but became public on October 22 after an embargo breach.
| Feature | Detail |
|---|---|
| Population | Resected stage IIB/IIC, III or IV cutaneous melanoma; no prior systemic therapy; ECOG 0 to 1; tumor sample adequate for sequencing |
| Randomization | 2:1, double-blind, placebo-controlled; stratified by risk-based stage and age |
| Intismeran arm (n = 757) | Intismeran 1 mg IM every 3 weeks, up to 9 doses, plus pembrolizumab 400 mg IV every 6 weeks, up to 9 cycles |
| Control arm (n = 380) | Placebo injections plus the same pembrolizumab |
| Primary endpoint | Recurrence-free survival (RFS): any recurrence, investigator-assessed, or death from any cause |
| Testing order | RFS, then distant metastasis-free survival (DMFS), then overall survival (OS), each tested only if the prior one was significant |
| Median age / sex | 61 years; about 60% male |
| Data cutoff | June 30, 2026 (prespecified interim analysis) |
Pembrolizumab started at randomization. Intismeran or placebo started once the individualized product was manufactured; median time to first dose was 64 days.
| Measure | Intismeran + pembro | Pembro alone | Hazard ratio (95% CI) |
|---|---|---|---|
| RFS events | 136 (18.0%) | 111 (29.2%) | 0.58 (0.45 to 0.75); p < 0.001 |
| 2-year RFS | 82.2% | 70.0% | |
| DMFS events | 89 (11.8%) | 71 (18.7%) | 0.62 (0.45 to 0.85); p = 0.002 |
| 2-year DMFS | 88.5% | 80.2% | |
| Overall survival | Immature | 0.77 (0.45 to 1.31); p = 0.33 | |
| RFS, months 0 to 6 | Piecewise, prespecified sensitivity analysis | 0.99 (0.66 to 1.49) | |
| RFS, after 6 months | Piecewise, prespecified sensitivity analysis | 0.42 (0.30 to 0.58) | |
| RFS, modified ITT | 84.2% at 2 yr | 70.4% at 2 yr | 0.51 (0.39 to 0.67) |
Medians were not reached in either arm for RFS or DMFS. Modified ITT excludes patients who did not receive at least one planned dose of intismeran or placebo (95.9% and 96.1% did). Absolute differences at two years are about 12 percentage points for RFS and 8 for DMFS, which corresponds to roughly 8 patients treated to prevent one recurrence or death and roughly 12 to prevent one distant metastasis or death at two years (calculated here, not reported by the trial).
Subgroups. RFS benefit was similar in the lower-risk stages (IIB to IIIB; HR 0.60) and the higher-risk stages (IIIC to IV; HR 0.57), and in patients under 65 (HR 0.65) and 65 or older (HR 0.53). Several small subgroups had confidence intervals crossing 1, including the US subgroup for DMFS (HR 1.77), which rested on 10 events.
Manufacturing. Of 1,749 patients screened, 302 (17.3%) were excluded because the tumor sample was inadequate. Intismeran was manufactured for more than 99% of patients assigned to it, and 91% of randomized patients had the full 34 neoantigens identified.
| Measure | Intismeran + pembro | Pembro alone |
|---|---|---|
| Treatment-related adverse events, any grade | 95.8% | 90.0% |
| Grade 3 to 5 treatment-related | 14.2% | 13.4% |
| Serious treatment-related | 9.8% | 11.1% |
| Grade 3 to 5, any cause | 23.6% | 20.3% |
| Discontinued any study drug for treatment-related AE | 14.0% | 8.7% |
| Fever | 52.2% | 8.2% |
| Injection site pain | 51.0% | 9.5% |
| Fatigue | 38.5% | 29.7% |
| Influenza-like illness | 25.4% | 9.2% |
| Pembrolizumab immune-mediated AEs | 36.0% | 34.7% |
Two treatment-related deaths (0.3%) occurred in the intismeran arm, both attributed to pembrolizumab. BioSpace reported eight deaths in that arm overall, five before any intismeran dose and three not attributed to it. Investigators described the vaccine-type reactions as low grade and transient.
Accurate as a restatement of the hazard ratio (0.58) in the intention-to-treat population. This was the headline in the company release and most trade and general press.
What gets lostA hazard ratio is a relative measure across the whole follow-up. The absolute difference at two years is about 12 percentage points.
The two-year event rates (17.8% vs 30.0%) give a relative reduction of about 41%, so the number is close.
What gets lostThe sentence treats a hazard ratio as a two-year probability, and the endpoint includes death as well as recurrence. The article does state that overall survival has not been shown.
The piecewise analysis was prespecified. The near-zero effect in the first six months (HR 0.99) fits the design, since patients waited a median of 64 days for their first dose and the immunization course runs about six months.
What gets lostIt is a sensitivity analysis, and the release placed it second in its headline bullets. The primary result is 0.58, and the later-segment estimate excludes the early months by design.
82.2% is the two-year recurrence-free survival rate.
What gets lostBeing alive and being recurrence-free are different endpoints. The overall survival difference is not statistically significant (HR 0.77, confidence interval 0.45 to 1.31).
Intismeran is a therapeutic product made after diagnosis from an individual tumor and given with a checkpoint inhibitor. It would be the first individualized and first mRNA cancer therapy of this type to reach approval.
What gets lostMany readers hear "vaccine" as prevention in healthy people. A therapeutic cancer vaccine is already approved (sipuleucel-T, for prostate cancer, since 2010). The New York Times phrase "of its kind" carries the qualification.
Serious and grade 3 to 5 treatment-related events were similar between arms, and intismeran did not increase pembrolizumab's immune-mediated adverse events.
What gets lostDiscontinuation of any study drug for a treatment-related adverse event was higher with intismeran (14.0% vs 8.7%), and about half of patients had fever and injection site pain.
The Phase 3 DMFS hazard ratio was 0.62. The 157-patient Phase 2b (KEYNOTE-942) reported 0.35 at its primary analysis and 0.41 at five years. Phase 2b RFS was 0.51 at five years.
What gets lostSmaller trials often overestimate effects, so some shrinkage from Phase 2b to Phase 3 is common.
Both outlets profiled trial participants who are doing well. The CNN participant said she still does not know her assignment and guesses she received the vaccine because she felt mildly ill after each dose. The New York Times participant attributes his course to the vaccine.
What gets lostThe trial remains blinded, and the Times does not say whether its participant was unblinded. Seventy percent of the pembrolizumab-alone arm was also recurrence-free at two years, so a good outcome does not identify the arm.
- Overall survival. Follow-up continues. The companies report a favorable trend; the hazard ratio's confidence interval still includes no effect.
- Comparison with neoadjuvant immunotherapy. Checkpoint inhibitors given before surgery are an alternative standard for some resectable stage III patients. RBC analysts argued it achieves similar results at much lower cost. This trial did not compare against it.
- Who is eligible. About one in six screened patients lacked a usable tumor sample, so the trial population was already selected for adequate tissue.
- Price and turnaround. No price has been announced. Median time from randomization to first dose was 64 days.
- Other cancers. Merck and Moderna list nine Phase 2 and Phase 3 trials, including three in non-small cell lung cancer. The renal cell carcinoma readout is expected late 2026 or early 2027.
- The full paper. Accepted by the New England Journal of Medicine; publication date not announced.
On August 19, the companies announced the trial had met both endpoints without giving a hazard ratio, p value or confidence interval. Moderna's stock rose sharply that day (reports range from nearly doubling to 177 percent).
On October 22, an oncology news site posted embargoed data early. ESMO lifted the embargo on the abstract and the full dataset, and the companies issued their release that afternoon. Analysts had hoped for a hazard ratio near the Phase 2b figure of 0.51. Moderna fell about 21 percent that day and Merck about 1 percent. RBC and BMO both described the result as clinically meaningful and likely approvable.
Georgina Long of Melanoma Institute Australia presented the results on October 24 in the ESMO Presidential Symposium. The discussant, Inge Marie Svane of the University of Copenhagen, called the results compelling. The companies say they are engaging with regulators; analysts expect a possible US approval in 2027.
The New York Times placed the result in the wider history of cancer vaccines, a field with decades of failed trials. Points from that piece, checked where possible:
- Recent randomized vaccine trials in colon cancer and in head and neck cancer missed their goals.
- Melanoma has many mutations and responds well to immune therapy, so other tumors may be harder.
- In a 16-patient Phase 1 study of BioNTech and Genentech's personalized mRNA vaccine for pancreatic cancer, 8 patients mounted an immune response, and 7 of those 8 were alive 4 to 6 years after surgery (confirmed against the AACR 2026 update). The study had no control group. A 260-patient randomized trial is underway.
- Early prevention efforts include a MUC1 vaccine for ductal carcinoma in situ (Pittsburgh), an off-the-shelf vaccine for BRCA carriers (Penn) and a vaccine of more than 200 proteins for Lynch syndrome (MD Anderson).
- The $200 million federal cancer vaccine partnership is an NCI and Foundation for the NIH initiative, split $100 million federal and $100 million from companies, philanthropy and other partners.
A large, blinded Phase 3 trial shows that an individualized mRNA product added to pembrolizumab reduces melanoma recurrence. At two years that is about one recurrence or death prevented for every eight patients treated. Whether it extends life, how it compares with immunotherapy before surgery, and what it will cost are still open.
Sources
Company release: Merck and Moderna, Results from INTerpath-001 to be presented at ESMO 2026 Presidential Symposium, October 22, 2026. merck.com
ESMO presentation: Long GV. INTerpath-001, ESMO Congress 2026, abstract LBA1, as reported by CancerNetwork, October 24, 2026. cancernetwork.com
Embargo breach and safety detail: Targeted Oncology, October 23, 2026. targetedonc.com
Trial registry: NCT05933577. clinicaltrials.gov
Analyst expectations and Phase 2 comparison: Fierce Biotech, October 22, 2026. fiercebiotech.com
Market reaction and deaths: BioSpace, October 22, 2026. biospace.com
Presidential Symposium: STAT, October 24, 2026. statnews.com
August stock move: Biotech Blueprint, October 2026. biotechblueprint.com
New York Times: Agrawal N. First-of-its-kind vaccine ushers in new hope for cancer patients, October 23, 2026. nytimes.com
CNN: Experimental mRNA vaccine cuts risk that melanoma returns by 42%, October 22, 2026. cnn.com
Time: Park A. A personalized cancer vaccine lowers melanoma recurrence by 42%, October 22, 2026. time.com
Phase 2b five-year data: Practical Dermatology summary of KEYNOTE-942. practicaldermatology.com
Pancreatic vaccine follow-up: MDedge, AACR 2026. mdedge.com
Federal partnership: Becker's Hospital Review, citing the Wall Street Journal. beckershospitalreview.com