A peptide vaccine aimed at one of cancer's most common mutations missed the main goal of its mid-stage trial. The way it missed is the part worth reading.

Elicio Therapeutics reported on June 15 that ELI-002 7P, an experimental vaccine that trains the immune system to recognize mutated KRAS, did not beat plain observation on disease-free survival in the Phase 2 AMPLIFY-7P trial ↗. Disease-free survival is just the time a patient lives after surgery without the cancer coming back. Across all 144 patients who had their pancreatic tumors removed and were then randomized across 24 U.S. sites, the vaccine arm and the watch-and-wait arm came out statistically even. By the rule a trial is judged on, that is a miss.

Then the arithmetic of who landed in which group starts to matter.

The two arms were not carrying the same load

Pancreatic surgery comes in two grades. An R0 resection means the surgeon got clean margins, with no cancer cells left at the cut edge. An R1 resection means microscopic disease was left behind, and those patients relapse sooner no matter what comes after. The two arms of AMPLIFY-7P did not start even: 19 percent of the vaccine patients were R1, against 10 percent in the observation arm. Nearly twice the share of the harder-to-treat patients sat in the group that had to prove the drug worked.

Strip those patients out and the picture flips. Among the cleanly resected R0 patients, who made up about 84 percent of the trial (121 people), the vaccine group went a median of 23.8 months before the cancer returned, against 12.8 months for observation. That is a hazard ratio of 0.65 (p=0.048), or roughly a one-third lower risk of recurrence at any given moment, with the median time to relapse nearly doubled. The early checkpoints moved in the same direction: at three months, 90.3 percent of vaccinated R0 patients were still disease-free versus 76.6 percent (p=0.022); at six months, 75.7 versus 61.7 percent (p=0.056, just shy of the usual significance line).

What the shot actually is

ELI-002 7P is not a conventional drug. It is a set of seven short peptides matching the seven most common KRAS mutations, the ones found in about a quarter of all solid tumors, paired with an immune-stimulating CpG adjuvant. Both pieces are chemically tagged so they hitch a ride on albumin, the blood's most abundant carrier protein, which carries them straight to the lymph nodes where T cells are trained. Elicio calls that tagging trick its Amphiphile platform. The aim is to put a faint tumor signal in front of the immune system efficiently, rather than letting the injected material scatter before it reaches the right cells.

The honest reading

None of this undoes the headline. A subgroup that looks good after the overall trial misses is a hypothesis, not a verdict. The R0 p-values sit right at the edge of significance, and the imbalance explanation, while plausible, is a story told after the fact rather than something the trial was built to test. The cautious version is that ELI-002 may work in cleanly resected patients and the trial could not see it through the noise of an uneven randomization. Elicio is betting on that version: its planned Phase 3 would enroll only R0 patients, add more doses beyond the first course, and keep disease-free survival as the primary endpoint, contingent on financing.

KRAS has been one of the hardest targets in oncology to drug, and peptides keep turning up in the attempts, from oral KRAS-binding molecules to designed binders pulled out of large virtual screens, several of which sit in our anticancer coverage ↗. ELI-002 comes at the same mutation from the immune side rather than the chemistry side.

The molecule that missed and the molecule worth running again may be the same molecule. This trial just could not tell them apart.