Elicio's KRAS vaccine missed its Phase 2 primary endpoint. It's staking a Phase 3 on the subgroup that worked.

What happened

On June 15, 2026, Elicio Therapeutics (Nasdaq: ELTX) disclosed that AMPLIFY-7P, the randomized Phase 2 test of its off-the-shelf mKRAS therapeutic cancer vaccine ELI-002 7P, did not meet its pre-specified primary endpoint of disease-free survival in the intent-to-treat population ^[1]. The randomized Phase 2 enrolled 144 patients across 24 U.S. sites (the NCT05726864 registry lists 158 total across the broader Phase 1/2 program; Prognyx could not reconcile the 144-vs-158 difference from public sources), randomizing ELI-002 7P against observation in patients with resected Stage I–III mKRAS-driven pancreatic ductal adenocarcinoma who had completed surgery and standard locoregional therapy and were radiographically disease-free at entry ^[1][4]. This is a setting with no approved post-locoregional therapy ^[1] — which is exactly why a positive result would have mattered, and why the miss stings.

The company did not release the full-population hazard ratio, confidence interval, or p-value; the public record so far carries only the qualitative "did not meet" ^[1]. Prognyx could not confirm an intent-to-treat DFS hazard ratio from public sources as of this briefing.

What Elicio led with instead was a post-hoc subgroup. In R0 patients — completely resected, lower residual disease, roughly 84% of those enrolled — DFS favored the vaccine with a hazard ratio of 0.65 (p=0.048; n=121), median DFS of 23.8 months versus 12.8 months for observation ^[1]. Absolute recurrence at 18 months was 9.5% lower in the ELI-002 arm ^[1]. Landmark analyses showed a consistent ~14% absolute DFS benefit during active treatment at both 3 and 6 months, with arm separation holding through 9 months ^[1].

Why the subgroup story is fragile

A positive post-hoc subgroup inside a failed ITT analysis is the oldest hopeful story in oncology, and the one regulators and reviewers penalize hardest. A p-value of 0.048 leaves essentially no margin. Elicio's mechanistic argument is that an imbalance in surgical margins, not an inert drug, sank the headline: the ELI-002 arm carried nearly twice the share of R1-resected patients — tumor at or within 1 mm of the margin, an adverse prognostic factor for recurrence — at 19% versus 10% in the observation arm, which the company says weighed against the treatment arm ^[1]. The pre-specified stratification factor, nodal status, was balanced between arms ^[1].

That argument has a biological anchor. mKRAS-specific T-cell responses correlated tightly with improved DFS (HR 0.22, p<0.0001, n=90) ^[1] — evidence the vaccine does what it is designed to do in patients who mount a response. Safety was described as favorable, which the company frames as supporting longer dosing and combinations ^[1]. None of this is nothing. But a margin imbalance large enough to break randomization is also an argument that the Phase 2 was underpowered or unlucky — and the fix Elicio proposes (enrich for R0, add dosing) is precisely the post-hoc-to-prospective leap that an FDA end-of-Phase-2 meeting exists to stress-test.

Who is exposed

ELI-002 is the lead clinical test of the off-the-shelf KRAS-vaccine thesis: one fixed Amphiphile construct — Amph-CpG-7909 adjuvant plus seven lipid-conjugated peptide antigens — aimed at the shared mutations that drive a large fraction of pancreatic and colorectal tumors ^[4]. It built on the two-peptide ELI-002 2P, whose completed 25-patient Phase 1 (NCT04853017) posted safety and biomarker-clearance data ^[5]; the ELI-002 program is catalogued in ChEMBL as CHEMBL4303334 ^[8]. An ITT miss in the cleanest possible setting — adjuvant, minimal residual disease, no competing approved therapy — raises the bar for every shared-antigen KRAS vaccine and personalized-neoantigen program that has pointed to adjuvant PDAC as proof-of-concept ground. The read-through is not fatal — the T-cell correlation is real — but it shifts the burden of proof toward whoever can show a survival effect in a randomized ITT population, not in responders alone.

The approved KRAS franchise is insulated, because it is a different modality. The small-molecule inhibitors sotorasib and adagrasib are approved, with salirasib in Phase 2 ^[9]. Those drugs hit the mutant protein directly and do not depend on a host immune response; AMPLIFY-7P says little about them. If anything, a credible vaccine stumble narrows the near-term competitive set in adjuvant PDAC immunotherapy rather than widening it.

Elicio itself is the most exposed party. Cash and equivalents are guided to fund operations only into Q4 2026 ^[2], and the company is openly "evaluating multiple strategic financing and partnering opportunities" to fund what comes next ^[1]. Endpoints News reported shares fell on the news as the company sought funds to advance to a larger Phase 3 regardless ^[3]. A Phase 3 in PDAC is a multi-year, multi-hundred-patient commitment; the distance between a runway that ends in Q4 2026 and a trial that has not been designed, sized, or registered is the whole ballgame.

What to watch next

• June 15, 2026, 8:30 AM ET — Elicio's conference call ^[1], the first chance to hear the ITT hazard ratio, overall-survival maturity, and Phase 3 contours. Prognyx could not confirm any overall-survival figures from public sources as of this briefing.
• The FDA end-of-Phase-2 outcome. Elicio says it "looks forward to discussing the findings with regulators" ^[1]; whether the agency accepts an R0-enriched registrational Phase 3 built off a post-hoc subgroup is the gating event. The company had guided to an end-of-Phase-2 meeting in H2 2025 and Phase 3 readiness around Q1 2026 ^[6][7] — timelines this readout resets.
• Phase 3 registration. NCT05726864 remains ACTIVE_NOT_RECRUITING with primary completion dated November 2026 and no results posted ^[4]; a new Phase 3 protocol on ClinicalTrials.gov, with sample size and the "additional dosing" schedule, would convert intent into plan.
• A financing or partner. With runway only into Q4 2026 ^[2], terms — or their absence — by late 2026 will decide whether the Phase 3 happens at all.

The now-what

Three options are on the table, and they are not mutually exclusive.

1. Run the R0 Phase 3 alone. Cleanest scientifically if the FDA blesses the subgroup, but the funding math is brutal against a Q4 2026 runway ^[2] — it likely requires a raise that the post-readout share reaction ^[3] just made more dilutive.
2. Partner the asset. A larger oncology developer could fund the Phase 3 in exchange for rights; the favorable safety profile ^[1] and the off-the-shelf, multi-indication framing — including mKRAS-positive lung and other mKRAS cancers ^[2] — are the pitch. The leverage is weak after an ITT miss, but the T-cell biomarker data ^[1] give a believer something to underwrite.
3. Pivot to combinations or a biomarker-selected design. The immune-response correlation ^[1] argues for selecting or boosting responders rather than enrolling all-comers — but that is a longer scientific road, and the cash clock is the binding constraint.

For a competitor running a KRAS-directed or shared-antigen vaccine program, the actionable read is narrow: AMPLIFY-7P shows on-target immunogenicity translating into DFS in responders, and it shows how easily a small randomized adjuvant trial can be knocked off by a margin imbalance. Stratify your own pivotal by resection margin, not just nodal status.

Verdict. Threat level to the off-the-shelf KRAS-vaccine thesis: elevated but not terminal — the immunogenicity signal held even as the ITT endpoint missed. Window to act for Elicio: roughly two quarters, bounded by a Q4 2026 cash runway ^[2] and an unregistered Phase 3 ^[4]. The next data point that matters is not another subgroup; it is whether a regulator or a check-writer agrees with this one.