Two small trials taught the immune system to see pancreatic cancer. Whether that buys patients time is still unsettled.
The team, from the Gillanders laboratory at Washington University School of Medicine in St. Louis, reported the result on September 4 in Science Advances ↗. They ran two phase 1 trials, testing two different ways to deliver a vaccine built from each patient's own tumor mutations, and both raised a real, specific T-cell response against the cancer. The survival number moved in the right direction but did not reach statistical significance, which means it could still be chance.
Pancreatic ductal adenocarcinoma is the cancer where immunotherapy keeps failing. The checkpoint drugs that transformed melanoma and lung cancer barely touch it, even though pancreatic tumors carry the same kind of raw material those drugs rely on. Every tumor accumulates mutations, and some of those mutations produce altered proteins, called neoantigens, that look foreign to the immune system because no healthy cell makes them. The premise of a personalized cancer vaccine is to read a patient's tumor, find those altered proteins, and hand the immune system a wanted poster.
What the trials actually did
Both trials enrolled patients after the standard sequence: surgery to remove the tumor, then adjuvant chemotherapy. That timing matters. The vaccine was not asked to shrink a visible mass. It was given when the scans were clean and the job was to hunt whatever microscopic disease surgery and chemo left behind.
To build each vaccine, the team sequenced the whole exome (the protein-coding stretch of the genome) of the tumor and of healthy tissue, sequenced the tumor's RNA to see which mutations were actually switched on, and ranked the candidate neoantigens with pVACtools, their own open-source neoantigen-prediction software. One trial (registered as NCT03956056) delivered the chosen targets as synthetic long peptides, short lab-made protein fragments. The other (NCT03122106) delivered them as DNA, letting the patient's own cells manufacture the fragments. Two formats, same idea.
The safety read was clean. Across both trials there were no grade 3 or higher adverse events, the threshold clinicians use for serious toxicity. The immune read was the point, and it held up under scrutiny. The vaccines produced neoantigen-specific T cells measured two ways, by interferon-gamma ELISPOT (a plate assay that counts immune cells firing at a target) and by intracellular cytokine staining. Then the team went further than most vaccine papers: they sequenced the expanded T-cell receptors, inserted those receptors into the patients' own blood cells, and confirmed the engineered cells recognized the intended neoantigens. That last step is a specificity check. It rules out the possibility that the counted T cells were reacting to something other than the tumor mutation.
Where the honesty lives
The immunology is the solid part. A vaccine made from a patient's private mutations, in the cancer most resistant to immunotherapy, raised a polyclonal T-cell response (many different T-cell lines, not one lucky clone) that provably targeted the tumor. Both a peptide and a DNA vaccine did it.
The clinical part is a hypothesis, not a finding. Vaccinated patients had a median overall survival of 4.4 years versus 3.5 years in a comparison group, but that comparison group was a propensity-matched set of past patients from the same institution, not a randomized control arm, and the difference carried a log-rank p-value of 0.23. In plain terms, roughly a one-in-four chance the gap is noise. The trials were designed to answer whether you can provoke the immune response, and they did. They were not designed to prove people live longer, and they do not.
That gap is the whole story. The field has spent a decade proving that neoantigen vaccines can make T cells. The harder, slower question is whether those T cells change the course of the disease, and only a randomized trial with survival as the prespecified endpoint can answer it.
The same tension ran through an off-the-shelf peptide vaccine that stabilized one patient's Ewing sarcoma ↗ reported last month. That vaccine targeted a mutation shared across many patients; these target mutations unique to one person. Opposite ends of the personalization spectrum, same unfinished sentence: the immune system noticed, and whether noticing is enough remains the next trial's job.