The announcements answer different questions
Moderna and Merck reported on August 19 that intismeran autogene plus Keytruda met the primary endpoint of recurrence-free survival and the key secondary endpoint of distant metastasis-free survival in INTerpath-001. The Phase 3 trial enrolled 1,137 people with high-risk melanoma whose visible tumors had been removed. Its control arm received Keytruda plus placebo. The companies have not yet published the hazard ratios, absolute recurrence rates, survival curves, detailed safety results, or overall survival analysis.
Nine days later, BioNTech said it would terminate BNT122-01, a Phase 2 trial of autogene cevumeran in people with ctDNA-positive, resected high-risk stage II or stage III colorectal cancer. The vaccine was tested as adjuvant monotherapy against watchful waiting. The trial had crossed a prespecified futility boundary in October 2025, but the data were then too immature for a reliable efficacy conclusion. In the latest review, the independent monitoring board found a numerical imbalance in overall survival and concluded that continued treatment was unlikely to change the efficacy outcome. BioNTech said no new safety signal had been identified. It did not disclose the direction or size of the survival imbalance.
A simple reading would rank Moderna's vaccine above BioNTech's. The trial designs make that comparison unreliable. They used different products in different tumors, at different clinical phases, with different control arms and different companion therapies. A personalized mRNA sequence is one part of an anticancer regimen. Tumor biology and the treatment around the vaccine can decide whether the immune response becomes clinically useful.
The trials placed different burdens on the vaccine
Both programs begin with the same broad idea. A patient's tumor is sequenced, software selects mutations that may produce useful neoantigens, and an individualized mRNA product carries those targets back to the immune system. The resemblance ends before treatment starts.
| Feature | Moderna and Merck | BioNTech and Roche |
|---|---|---|
| Product | Intismeran autogene, also called V940 or mRNA-4157 | Autogene cevumeran, also called BNT122 or RO7198457 |
| Disease | Completely resected stage IIB through IV melanoma | ctDNA-positive, resected high-risk stage II or stage III colorectal cancer |
| Experimental regimen | Vaccine plus pembrolizumab | Vaccine monotherapy |
| Comparator | Placebo plus pembrolizumab | Watchful waiting |
| Trial | Randomized, blinded Phase 3 | Randomized, open-label Phase 2 |
| Reported outcome | RFS and DMFS endpoints met; effect size not disclosed | Trial terminated after futility review and a numerical OS imbalance; full results not disclosed |
INTerpath-001 asks whether adding a personalized vaccine improves an active PD-1 regimen. BNT122-01 asked whether the vaccine could prevent recurrence without a checkpoint inhibitor or another immune partner. The first design measures incremental benefit on top of established immunotherapy. The second puts much more of the biological burden on the vaccine itself.
Moderna and Merck entered Phase 3 after a randomized Phase 2b study in resected melanoma had already shown a recurrence-free survival signal. BioNTech and Roche were testing whether monotherapy could produce a sufficient clinical effect in a colorectal population selected for residual disease. That history changes what each result says. One confirms an earlier signal in a larger study. The other closes a development path before Phase 3.
Melanoma supplied targets and an immune partner
Melanoma was a favorable place to test personalized vaccination. Ultraviolet damage often leaves melanoma cells with many mutations, creating a larger pool of possible neoantigens. The disease can also respond to PD-1 blockade, evidence that T cells can control it when immune suppression is relieved. Neither property guarantees a vaccine response, but both improve the starting conditions.
The Moderna regimen divided the work between two medicines. Intismeran was designed to prime and expand T cells against patient-specific targets. Keytruda blocked PD-1, a checkpoint that can restrain activated T cells. The control arm also received Keytruda, so a positive result indicates that the personalized component added clinical value beyond checkpoint therapy alone, at least at the interim analysis.
The amount of disease was favorable too. Participants had undergone complete surgery and were being treated for microscopic residual disease. An immune response did not have to shrink a large tumor within weeks. It had to find and control malignant cells that remained after surgery, before they produced a detectable recurrence.
The Phase 3 announcement is incomplete. Without the effect size, clinicians cannot calculate how many patients must receive the combination to prevent one recurrence or distant metastasis. Overall survival remains under evaluation. The result gives strong support to an adjuvant combination in melanoma. It says much less about other tumors.
ctDNA found risk, not immune sensitivity
BNT122-01 used ctDNA to identify patients at high risk after colorectal cancer surgery. Tumor-derived DNA in the blood can reveal that malignant cells remain even when imaging shows no visible disease. That makes ctDNA useful for enriching a trial with people who are more likely to recur.
The same signal says nothing about whether those residual cells are vulnerable to a vaccine. A blood test for tumor DNA cannot measure whether useful T cells have entered the tumor, whether cancer cells present the selected antigens, or whether the surrounding tissue suppresses immune activity. It is a marker of residual disease and recurrence risk, not a direct marker of vaccine responsiveness.
The trial's eligibility criteria make the distinction sharper. Patients with diagnosed microsatellite instability-high tumors were excluded. That removed the colorectal subgroup with high mutation burden and established sensitivity to checkpoint inhibition. Most remaining colorectal cancers are mismatch repair-proficient or microsatellite stable. They tend to have fewer immunogenic mutations, less T-cell infiltration, and several routes of immune exclusion.
Selecting ctDNA-positive patients may have made clinical events easier to observe. The population remained immunologically cold. That helps explain how a well-targeted personalized product could fail to change disease-free survival, although the unpublished dataset is needed to identify the mechanism. Weak immune priming, poor T-cell trafficking, antigen escape, and patient selection are all possible contributors.
Monotherapy tested the hardest version of the platform
A vaccine has to do more than encode the correct mutations. Antigen-presenting cells must take up the product and activate T cells. Those T cells must expand, reach residual tumor deposits, recognize antigens displayed by cancer cells, and remain functional inside an immunosuppressive environment. Personalization addresses target selection. The downstream barriers remain.
Checkpoint inhibitors and chemotherapy can change that environment. PD-1 blockade can release a brake on activated T cells. Chemotherapy can reduce the number of malignant cells and, in some settings, alter antigen release or immune suppression. These effects vary by drug and tumor, which is why combination evidence must come from randomized trials rather than mechanism alone.
BioNTech's own development decisions reflect this constraint. The company said the colorectal result did not affect IMcode003, a continuing Phase 2 trial in resected pancreatic ductal adenocarcinoma. That study combines autogene cevumeran with the PD-L1 inhibitor atezolizumab and mFOLFIRINOX chemotherapy. The same vaccine platform is therefore being tested under a different biological and therapeutic load.
The colorectal termination narrows the credible claim. It argues against autogene cevumeran monotherapy in this specific ctDNA-positive, non-MSI-high adjuvant population. It does not establish that another mRNA sequence, another combination, or another colorectal subgroup would fail. It also gives BioNTech and Roche a reason to require stronger combination logic before moving other immune-cold tumors into larger trials.
Company value rests on different evidence
Moderna gets the larger platform read-through from melanoma. The result supports a real oncology asset and shows that shared sequencing, prediction, and manufacturing can work in a Phase 3 program. Applying the melanoma probability of success to every tumor in the pipeline would still be a mistake. Each indication carries its own biological and trial risk.
Merck's exposure is more asymmetric. Keytruda is the active backbone in INTerpath-001, and the trial tests whether intismeran adds enough benefit to expand that backbone. A successful combination gives Merck a place in a personalized regimen. Vaccine failures elsewhere have less effect on a franchise that already spans many tumors and treatment partners.
BNT122-01 removes one monotherapy path for BioNTech and Roche and raises the evidentiary bar for immune-cold tumors. BioNTech is already moving toward combinations involving checkpoint agents, bispecific antibodies, antibody-drug conjugates, and chemotherapy. That strategy may improve the biological odds, but it creates more variables and more toxicity to measure. It also makes partners more important. Development becomes slower and more expensive when a vaccine needs an immune backbone.
The production system deserves some platform value because it can be reused. Clinical efficacy remains local to a tumor and regimen, so each indication needs its own probability of success.
The next data must explain the divergence
The Moderna and Merck disclosure needs the Phase 3 hazard ratios and absolute RFS and DMFS curves first. Safety, treatment discontinuations, subgroup consistency, quality of life, and overall survival will determine whether intismeran is a modest addition to Keytruda or a new standard after melanoma surgery.
BioNTech and Roche need to publish the BNT122-01 dataset even though the trial stopped. The direction and magnitude of the overall survival imbalance are essential. Baseline balance and causes of death come next. Data on ctDNA, induced T-cell responses, and recurrence patterns could then help distinguish a product problem from a population or regimen problem. The monitoring board's decision is clinically decisive, but the scientific explanation remains incomplete.
Results from other tumors will show how broad the platform can become. The most informative trials will compare the same vaccine approach across immune environments and test whether adding checkpoint blockade or chemotherapy changes outcomes. Repeated success in several randomized settings would support platform economics. Mixed outcomes would still leave useful medicines, but with narrower markets and fewer shared assumptions.
The evidence available on August 31 supports an mRNA vaccine combination in resected melanoma. It did not support continuing vaccine monotherapy in one ctDNA-positive colorectal population. Clinical development should respond to that difference. A universal oncology platform still needs randomized success across several tumor types and treatment backbones.
Sources
The Moderna and Merck result remains a company-reported Phase 3 interim analysis. BioNTech had not released the complete BNT122-01 dataset by August 31, 2026.
- Moderna and Merck, INTerpath-001 Phase 3 topline results
- BioNTech, termination of the BNT122-01 colorectal cancer trial
- ClinicalTrials.gov, BNT122-01 study record
- ClinicalTrials.gov, IMcode003 pancreatic cancer study record
- Weber et al., randomized Phase 2b intismeran study in The Lancet
- Ganesh et al., colorectal cancer immunotherapy biology in Nature Reviews Gastroenterology and Hepatology