BioNTech and Genentech have stopped a Phase 2 trial of autogene cevumeran in colorectal cancer. An independent data safety monitoring board recommended the termination. The companies confirmed the decision on 28 August 2026. It arrives nine days after a landmark result in melanoma. Trade coverage described that readout as the first Phase 3 success for an individualised neoantigen therapy. Both products are therapeutic cancer vaccines, given to treat existing disease, not preventive vaccines. Taken together, the two results begin to define what works in this field.
What the monitoring board found
The study, BNT122-01, was registered as NCT04486378. It compared autogene cevumeran against watchful waiting after surgery. Eligible patients had resected stage II (high risk) or stage III colorectal cancer. All were positive for circulating tumour DNA following resection. Planned enrolment was 327 participants. BioNTech sponsored the trial, with Genentech as its development collaborator.
The trial crossed its futility boundary in October 2025. At that point the monitoring board judged the data too immature for firm conclusions. On later review the board identified a numerical imbalance in overall survival between the arms. It concluded that continuing was unlikely to change the efficacy outcome. Fierce Biotech reported that a Genentech spokesperson confirmed more deaths in the vaccine arm in this patient population. Neither company has released death counts or survival curves. No new safety signals were reported.
Why autogene cevumeran struggled in colorectal cancer
The design asked a great deal of a single agent. Autogene cevumeran was given alone, without a checkpoint inhibitor and without chemotherapy. Colorectal cancer is widely described as immunologically cold. Most tumours carry a low mutational burden and a suppressive microenvironment. That combination limits how far a vaccine-primed T cell response can travel. Adjuvant patients also have no measurable lesion to attack, only microscopic residual disease.
Professor Özlem Türeci, co-founder and outgoing chief medical officer at BioNTech, framed the result in those terms. BioNTech announced on 3 August 2026 that she will leave the role to lead a new company. She said the outcome was not what the company had envisioned for mRNA as a monotherapy in colorectal cancer. She added that it offers scientific insight into the challenges of treating immunotherapy-insensitive tumour types.
Analysts reached a similar conclusion. In a note reported by BioSpace, BMO Capital Markets said the termination would reduce confidence in autogene cevumeran. It expected the result to raise questions about whether the vaccine alone could prevent recurrence without additional therapies. That is now the central design question for the whole category.
The contrast with the melanoma readout
Nine days earlier, Merck and Moderna reported a very different result. Their Phase 3 INTerpath-001 trial enrolled 1,137 patients with completely resected stage IIB to IV melanoma. Intismeran autogene plus pembrolizumab met its primary endpoint of recurrence-free survival. It also met the key secondary endpoint of distant metastasis-free survival. The comparator was pembrolizumab alone. Overall survival data remain immature and the study continues.
The differences between the two programmes are instructive. Melanoma is highly mutated and responsive to checkpoint blockade. The Moderna candidate was added on top of an active immunotherapy backbone. Autogene cevumeran was tested without one, in a tumour type that resists immunotherapy.
Setting does not explain everything, however. Autogene cevumeran has already been tested in melanoma. The randomised Phase 2 IMcode001 study paired it with pembrolizumab in first-line advanced disease. Results presented at ESMO in October 2025 showed no improvement in progression-free survival over pembrolizumab alone. Overall survival favoured the combination numerically but the difference was not statistically significant.
That readout matters for how this week’s news is read. The candidate has now missed in a hot tumour, with a checkpoint backbone. That is the setting the melanoma comparison credits. Construct differences between the two platforms are therefore back in play alongside trial design. IMcode001 did report one encouraging signal: patients who mounted responses against more neoepitopes did better. Breadth of induced immunity, rather than the presence of a vaccine, may be the variable worth measuring. LSDN examined the manufacturing and logistics consequences of that Phase 3 win in melanoma last week.
What the halt means for UK cancer vaccine trials
BNT122-01 has a direct British footprint. The trial recruited through the NHS Cancer Vaccine Launch Pad, run by NHS England with Genomics England. The platform matches patients to trials using NHS records, surplus tissue and a blood sample. BioNTech studies were the first to recruit through it. The wider agreement aims to give up to 10,000 UK patients access to personalised immunotherapies by 2030.
UK research ethics approval for BNT122-01 came on 25 October 2022, under reference 22/LO/0621. The colorectal study then closed to new Launch Pad participants in March 2025. That was well over a year before this week’s announcement. The platform has since redirected its capacity. It now supports the BNT113-01 head and neck cancer study and the ROSETTA Lung-02 trial in non-small cell lung cancer. Launch Pad figures list 609 patients registered across 75 open research sites.
That sequencing limits the operational damage. No UK patients were left waiting on a study that had already stopped recruiting. The reputational question is harder to settle. The Launch Pad was built partly on the promise of adjuvant vaccination after curative surgery. Its first flagship trial has now failed on precisely that question. Recruitment messaging for the successor studies will need to reflect the change.
What survives in the autogene cevumeran programme
The candidate itself has not been abandoned, but what remains is narrow. The Phase 2 IMcode003 study continues in resected pancreatic ductal adenocarcinoma. That trial pairs autogene cevumeran with atezolizumab and mFOLFIRINOX chemotherapy. It is, in other words, the combination design that the colorectal study lacked. Genentech has said both partners remain committed to investigating the candidate’s efficacy and safety.
The adjuvant bladder study has also wound down. Roche’s patient trial listing for the muscle-invasive urothelial carcinoma study records it as active but not recruiting. The listing attributes that to a sponsor decision to phase the study out. Melanoma has missed, bladder is phasing out and colorectal is terminated. IMcode003 now looks like the last randomised test of this candidate still enrolling.
The pancreatic evidence base
Pancreatic work has produced the strongest evidence behind the candidate. Research published in Nature in 2023 showed the vaccine primed neoantigen-specific T cells after surgery. Follow-up work published in Nature in 2025 tracked those clones considerably further. It estimated an average lifespan of 7.7 years for vaccine-induced CD8+ T cell clones. At a median follow-up of 3.2 years, vaccine responders had not reached median recurrence-free survival. Non-responders reached it at 13.4 months. That separation was statistically significant.
Those numbers come from a small, non-randomised population. They are hypothesis-generating rather than confirmatory. The randomised IMcode003 study exists to test whether the signal holds. Its readout now carries considerably more weight for the platform than it did a week ago.
Commercial and regulatory read-across
Investors marked the news quickly. BioNTech shares fell by roughly 7 per cent during trading on 28 August, moving to around 103 dollars. Some outlets recorded intraday declines closer to 8 per cent. Moderna shares fell by around 6 per cent despite its own recent success. The market read the result as a verdict on the category, not on one sponsor.
BioNTech has the balance sheet to absorb it. The company reported 16.6 billion euros in cash and securities at the end of the second quarter of 2026. Its head and neck candidate BNT113 and its bispecific antibody BNT327 both continue. The wider portfolio is not dependent on this single indication.
Regulators will note the outcome for a different reason. Adjuvant oncology trials increasingly rely on recurrence-free survival as a primary endpoint. Overall survival typically follows years later, if at all. Here the survival signal moved the wrong way while the efficacy question was still open. That sequence strengthens the case for keeping overall survival under active monitoring in adjuvant designs.
What sponsors should take from this
Three practical lessons stand out for trial designers. First, neoantigen vaccination now carries a heavy evidential burden outside the adjuvant melanoma setting. Sponsors should expect ethics committees and monitoring boards to ask harder questions earlier. Second, ctDNA positivity identifies recurrence risk well but does not predict vaccine benefit. Patient selection may need an immunological marker alongside the molecular one.
Third, watchful waiting comparators leave nowhere to hide. A vaccine arm that underperforms observation produces the bluntest possible result. Active-control designs soften the optics but not the verdict, as IMcode001 showed. Sponsors weighing that trade-off now have two concrete cautionary cases to cite.
The field is not in retreat. It is being narrowed to the settings, and the constructs, where it demonstrably works. For now that means adjuvant combination regimens in immunologically responsive disease. Both companies have said they will analyse the BNT122-01 data in detail, with a focus on patient selection. Those findings should reach the literature before the next generation of adjuvant studies is designed.
