This Pancreatic Cancer Vaccine Did Everything but Work. Then the Explanation Changed

A peptide cancer vaccine missed the endpoint it was built to hit, and the company has now explained that miss two different ways, four months apart.

Elicio Therapeutics presented full results from its randomized Phase 2 AMPLIFY-7P trial at the ESMO Congress in Madrid on October 23, 2026.¹ The trial tested camupepimut, also called ELI-002 7P, against observation alone in people whose pancreatic cancer carries a mutated KRAS gene. Disease-free survival in the intent-to-treat population came out at a hazard ratio of 0.85, and the pre-specified primary endpoint was not met.¹ Figures on this page were checked on October 23, 2026.

That part is not new. The company announced the same failure on June 15, 2026, and its shares fell 47% the following day.² What changed in October is the explanation.

What was actually tested?

A peptide vaccine. That is why the trial sits on a peptide site at all.

Camupepimut is built from short mutant KRAS peptides joined to a lipid tail. The tail binds albumin in the body, and albumin drains to the lymph nodes, so the construct delivers its peptides to the place where T cells are primed instead of letting them disperse from the injection site. That lymph-node targeting is the whole design idea, and the compound's name in the trial, amphiphile vaccine, describes the two-ended molecule doing it.¹

This is the second time the compound has been written up in the same journal. A single-arm Phase 1, AMPLIFY-201, treated 25 patients with pancreatic or colorectal cancer and published final results in Nature Medicine in 2025 under Wainberg and colleagues.³ It reported T-cell responses in 84% and tumor biomarker responses in 84%, with complete ctDNA clearance in 6 of 25.³ Those numbers are why anyone funded a randomized trial.

The randomized trial enrolled 144 patients across 24 US sites after standard local treatment, randomizing 96 to the vaccine and 48 to observation.¹ Pancreatic ductal adenocarcinoma recurs in most patients after surgery, and nothing in the adjuvant setting reliably prevents that. Unlike most of what this site covers, this compound went through a randomized trial at all. That gap is the subject of our guide on why most peptides have no human evidence. An immune approach aimed at a mutation present in the great majority of these tumors is a reasonable thing to test.

What did the trial find?

Disease-free survival came out at 0.85 on the hazard ratio, in favor of the vaccine but not by enough.¹

Elicio's release gives that figure without a confidence interval and without a p-value.¹ A figure of 0.85 with no interval around it tells you the point estimate leaned the right way and nothing about how firmly. For a trial of this size, the interval is the part that would tell you whether 0.85 is a near miss or noise, and it is not in the release.

Six-month disease-free survival, written DFS in most trial reporting, was 75.7% with the vaccine against 61.7% with observation.¹ Trade coverage of the June announcement reported that same comparison with a p-value of 0.056.⁴

The whole trial fits in one table. Everything below the first row is a subgroup, and the p-values on those rows are nominal.

Analysis	Vaccine vs observation	Median DFS	HR	p
All randomized patients, the primary endpoint	96 vs 48	Not reported	0.85	Not reported
ctDNA-negative at baseline	62 vs 31	Not reached vs 21.2 months	0.51	0.0280
ctDNA-positive at baseline	20 vs 6	4.3 vs 1.9 months	0.57	0.0489
No prior radiation	73 vs 33	23.8 vs 10.4 months	0.59	0.0221
Adjusted for baseline factors	Not a randomized comparison	Not applicable	0.47	0.0086
Strong vs weak T-cell responders	Not a randomized comparison	Not reached vs 6.08 months	0.22	<0.0001

The top row is the trial.¹ Every row under it was chosen once the result was known, and the bottom two are not comparisons between randomized groups at all.

Why does the company point at ctDNA?

Because the two arms were not balanced on it. Circulating tumor DNA is among the strongest predictors of recurrence there is.

Baseline ctDNA positivity ran at 20.8% in the vaccine arm against 12.5% in the observation arm, so more of the patients most likely to recur were assigned to the vaccine.¹ The size of that effect inside this trial is not in doubt. Pooling both arms, median disease-free survival was 4.2 months for ctDNA-positive patients against 24.0 months for ctDNA-negative patients. The ratio between them is 6.56.¹ An imbalance in a variable that powerful can move a result on its own.

Adjusting for it in a multivariable model put the vaccine's hazard ratio at 0.47.¹ That is a reasonable analysis to run and a reasonable thing to report. It is also not the trial's result. A randomized comparison answers the question it was randomized to answer, and an adjusted model answers a different question using assumptions the randomization was supposed to make unnecessary.

What changed between June and October?

The company changed which imbalance it blames and which subgroup it leads with, on the same trial.

	June 15, 2026 announcement	October 23, 2026 presentation
Primary endpoint	Not met24	Not met, HR 0.851
Imbalance blamed	R1 resection margins, 19% of the vaccine arm against 10% of observation24	Baseline ctDNA positivity, 20.8% against 12.5%1
Lead subgroup	R0-resected patients, about 84% of those enrolled, HR 0.6524	ctDNA-negative HR 0.51, ctDNA-positive HR 0.57, no prior radiation HR 0.591
Phase 3 plan	Focused on R0-resected patients2	Stratified on ctDNA and other prognostic factors1

The October release does not contain the letters R0 or R1 anywhere, nor the figure 0.65, nor any reference to the June announcement.¹ The subgroup that the company described in June as the basis of its Phase 3 design has been replaced by a different one.

Look at how thin one of the new subgroups is. The ctDNA-positive comparison reaches a p-value of 0.0489 on 20 vaccinated patients against 6 observed ones.¹ Six people is a number where one patient moving changes the answer. A p-value sitting that close to 0.05 in a group that size is not a finding. It is an arithmetic coincidence waiting to be checked.

There is a defensible reading of this. ESMO and a journal both demand a fuller analysis than a topline release, ctDNA data takes time to mature, and a company that finds a better explanation four months on is entitled to present it. Margin status and ctDNA positivity also measure overlapping things, since both are proxies for how much disease was left behind.

The difficulty is what the switch does to the statistics. Each explanation was chosen after the result was known. A subgroup picked once the data are in is a hypothesis, and a second subgroup picked after the first one is a second hypothesis, and the nominal p-values attached to them do not account for either choice. Elicio describes the October subgroups as pre-specified.¹ Trade coverage of the June announcement described the R0 analyses as post-hoc.⁴ Both framings cannot be doing the same work, and this page could not resolve which analyses sat in the statistical analysis plan, because that plan is not public.

What held up regardless of the endpoint?

Two things held up: the immune response and the safety profile. Neither was ever the thing in question.

Mutant KRAS-specific T-cell responses appeared in 99% of evaluable treated patients, with a median 44.3-fold rise from baseline, and 85% of those assessed produced both CD4 and CD8 responses.¹ Responses persisted at least 46 weeks in 95.5% of evaluable patients.¹ A vaccine that reliably generates the T cells it was designed to generate has cleared a real bar, and plenty of cancer vaccines never do.

Grade 3 or higher adverse events ran at 10.4% on the vaccine against 14.6% on observation, with serious adverse events at 6.3% against 8.3%, and no adverse event caused withdrawal of study medication.¹ Both rates came in lower in the treated arm than in the untreated one. That sounds stranger than it is, since the comparator here was no treatment at all, not another drug.

The exploratory analysis that gets quoted most is the one tying response size to outcome: patients above a T-cell response threshold did not reach median disease-free survival, against 6.08 months below it, giving 0.22.¹ That is a striking number and it is also the weakest kind of evidence on the page, because patients who mount strong immune responses differ from patients who do not in ways that have nothing to do with the vaccine.

What is not confirmed?

Three things. The first is the one most likely to get repeated as fact.

Elicio states that the results published simultaneously in Nature Medicine under the title "Lymph node-targeted, mKRAS-specific amphiphile vaccine in pancreatic cancer: the randomized phase 2 AMPLIFY-7P trial."¹ This page could not locate that article. No DOI, author list or publication record was found, and none is constructed here. The claim may well be correct and simply not indexed yet.

No overall survival data appears anywhere in the release.¹ In resected pancreatic cancer that absence matters more than it would elsewhere, because disease-free survival is a stand-in for the thing patients care about and this disease has a long record of treatments that delayed recurrence without extending life. Follow-up may simply be too short, and the company does not say.

The release gives no ClinicalTrials.gov registration number, and no NCT identifier is asserted on this page.¹

No independent journalism on the October 23 presentation was found. What circulated instead was wire syndication of the release itself and automated stock-news summaries, two of which read it in opposite directions on the same day: one headlined that the trial missed its primary endpoint, another headlined the results as positive.⁵ Both were describing one document.

Is this one trial or a pattern?

A pattern. The second half of it landed eight weeks ago.

BioNTech and Genentech terminated their Phase 2 of autogene cevumeran on August 28, 2026, in patients with ctDNA-positive resected colorectal cancer.⁶ An independent data monitoring board found a numerical imbalance in overall survival between the arms, judged that continuing was unlikely to change the efficacy outcome, and recommended stopping; the futility boundary had been crossed the previous October.⁶ That program is the most watched individualized cancer vaccine in development, and its colorectal trial is the one that reported an overall survival imbalance in the wrong direction.

Both programs followed the same route. A small single-arm study showed immune responses in most patients and a striking association between response size and outcome, and the randomized test that followed did not reproduce the benefit. In the Phase 1, patients above the T-cell response threshold came out at 0.12 against those below it.³ In the Phase 2, the same comparison gave 0.22.¹ Neither is a randomized comparison. Patients who mount strong immune responses are different people from patients who do not, in ways that predict survival on their own.

BioNTech's pancreatic trial, IMcode003, is unaffected by the colorectal termination and continues.⁶ That study pairs the vaccine with chemotherapy and a checkpoint inhibitor instead of testing it alone. Elicio named the same direction when it listed combinations among its Phase 3 considerations.¹

What happens next?

Elicio says it will take the full findings to the FDA to inform a possible Phase 3 adjuvant trial, with ctDNA and other prognostic factors used for stratification, along with longer treatment and possible combinations.¹

Whether that trial runs is partly a financing question. At the June announcement the company said its cash was expected to last into the fourth quarter of 2026, and it raised $15 million in a registered direct offering that closed on July 6.²⁷

A Phase 3 stratified on ctDNA would be the right experiment. It would take the explanation Elicio has offered for this trial's failure and put it where an explanation belongs, in the design of the next study instead of the analysis of the last one.

Common questions

Did the AMPLIFY-7P trial succeed?

What is camupepimut?

Does a hazard ratio of 0.85 mean the vaccine did nothing?

Why do the subgroup hazard ratios look so much better?

What is ctDNA and why did it matter here?

Was the imbalance between the arms the company's fault?

Was the vaccine safe?

Have other cancer vaccines done better?

Is this peptide available anywhere?

Was it published in Nature Medicine?

Does this tell us anything about peptides generally?

Sources

1. Elicio Therapeutics. "Elicio Therapeutics Presents Late-Breaking Phase 2 AMPLIFY-7P Results at ESMO Congress 2026 and Announces Simultaneous Publication in Nature Medicine." Press release, October 23, 2026, 09:00 ET, Boston. globenewswire.com. Tier: read at source, and it is a company document rather than independent reporting. 144 patients at 24 US sites, 96 to camupepimut and 48 to observation, in mKRAS-driven pancreatic ductal adenocarcinoma after standard locoregional therapy. ITT disease-free survival HR 0.85, reported with no confidence interval and no p-value; the release does not use the phrase "primary endpoint," saying instead that the study did not meet its pre-specified primary disease-free survival endpoint in the ITT population. Six-month DFS 75.7% against 61.7%. Pre-specified subgroups with nominal p-values and no confidence intervals: ctDNA-negative at baseline, 62 against 31, median DFS not reached against 21.2 months, HR 0.51, p = 0.0280; ctDNA-positive, 20 against 6, 4.3 against 1.9 months, HR 0.57, p = 0.0489; no prior radiation, 73 against 33, 23.8 against 10.4 months, HR 0.59, p = 0.0221. Recurrence or death in ctDNA-negative patients 27.4% against 48.4%. Pooled across arms, median DFS 4.2 months ctDNA-positive against 24.0 months ctDNA-negative, HR 6.56, p < 0.000001. Baseline ctDNA positivity 20.8% (20/96) against 12.5% (6/48), which the release calls "a baseline imbalance in ctDNA positivity, a strong predictor of recurrence." Multivariable adjusted HR for camupepimut 0.47, p = 0.0086; elevated CA19-9 3.16, p < 0.0001; positive baseline ctDNA 5.99, p < 0.0001. T-cell responses in 99% of evaluable treated patients, median 44.3-fold increase, both CD4 and CD8 in 85% of those assessed, persisting at least 46 weeks in 95.5%. Exploratory threshold analysis, 74 of 90 above threshold, median DFS not reached against 6.08 months, HR 0.22, nominal p < 0.0001. Grade 3 or higher adverse events 10.4% against 14.6%, serious adverse events 6.3% against 8.3%, no withdrawals of study medication. No overall survival data. No ClinicalTrials.gov number, no confidence intervals anywhere, and no reference to the June 2026 announcement. Checked against the same release for the strings R0, R1, resection margin and margin status, and for the hazard ratio 0.65; none appears. Nature Medicine title given as "Lymph node-targeted, mKRAS-specific amphiphile vaccine in pancreatic cancer: the randomized phase 2 AMPLIFY-7P trial," with no authors listed. ESMO described as a late-breaking Proffered Paper oral session with no abstract number, session name or presenter given. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17 ↩18 ↩19 ↩20 ↩21 ↩22 ↩23 ↩24 ↩25 ↩26 ↩27 ↩28
2. Jacobus N. "Elicio Therapeutics' Shares Drop Following Announcement of Phase II Trial Results." Pharmaceutical Executive. pharmexec.com. Tier: read at source. Trade coverage of a company release dated June 15, 2026; the underlying June release was not read at source for this page. Reports the ITT primary endpoint as missed, R0-resected subgroup HR 0.65 with no p-value given, R1 resection in 19% of the vaccine arm against 10% of observation and described as a key confound, landmark DFS 90.3% against 76.6% at three months (p = 0.022) and 75.7% against 61.7% at six months (p = 0.056), T-cell fold-change subgroup HR 0.22 (p < 0.0001, n = 90), shares down 47%, cash expected to last into Q4 2026, and a Phase 3 plan focused on R0-resected patients with extended treatment duration. The page header reads June 16, 2026 while the page metadata gives a published time of October 24, 2026; this page could not resolve that conflict and treats the June date as correct because the article is written around a June 15 release and because a second outlet carries the same June dating. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6
3. Elicio Therapeutics. "Elicio Therapeutics Announces Publication of ELI-002 Updated AMPLIFY-201 Phase 1 Follow-up Data in Nature Medicine for Minimal Residual Disease ("MRD") Positive, Adjuvant-Stage Patients." Press release, August 12, 2025, 08:00 ET. elicio.com. Tier: read at source, and it is a company document. Gives the publication as "Lymph-node-targeted, mKRAS-specific amphiphile vaccine in pancreatic and colorectal cancer: Phase 1 AMPLIFY-201 trial final results," cited by the company as Wainberg et al., Nature Medicine, 2025; no DOI is given in the release and none is constructed here, and the article itself was not opened for this page. Data through September 24, 2024. 25 adjuvant-stage patients, all MRD-positive for mKRAS after locoregional treatment, 20 pancreatic and 5 colorectal. Direct ex vivo mKRAS-specific T-cell responses in 21 of 25 (84%); both CD4 and CD8 responses in 71%, and in 6 of 6 at the recommended Phase 2 dose. Tumor biomarker responses in 21 of 25 (84%), with complete ctDNA clearance in 6 of 25 (24%), 3 pancreatic and 3 colorectal. At 19.7 months median follow-up, median relapse-free survival 16.33 months across the cohort; above the T-cell efficacy threshold median RFS not reached against 3.02 months below it, HR 0.12. Median overall survival 28.94 months. This page uses the Phase 1 only for lineage and for the fact that the same responder-versus-non-responder comparison was run in both studies. ↩1 ↩2 ↩3
4. Cancer Network Staff. "ELI-002 7P Misses Primary End Point in Phase 2 AMPLIFY-7P PDAC Study." CancerNetwork. cancernetwork.com. Tier: read at source. Trade coverage of the same June 15, 2026 release, fact-checked per the byline by Tim Cortese and Roman Fabbricatore. Reports the ITT primary endpoint as missed with no ITT hazard ratio given, describes the R0 analyses as post-hoc, R0 subgroup HR 0.65 (P = .048) in about 84% of patients (n = 121), R1 as a recurrence factor HR 1.56 (P = .181), T-cell fold change above 9.17 times baseline HR 0.22 (P < .0001, n = 90), landmark DFS 90.3% against 76.6% at three months (P = .022) and 75.7% against 61.7% at six months (P = .056), median DFS in R0 patients 23.8 against 12.8 months, and 144 patients at 24 US sites. Does not mention ctDNA. Same date conflict as note 2: header June 16, 2026, metadata October 23, 2026. The two outlets agree on every figure they both report, which is why the June framing is treated as reliable here despite neither being the primary document. ↩1 ↩2 ↩3 ↩4 ↩5
5. Automated stock-news summaries of the October 23, 2026 release, consulted to establish whether independent reporting existed. Tier: headlines read in search results, articles not opened, and nothing from them is used as a factual source on this page. StockTitan headlined the release "Elicio Phase 2 trial misses primary endpoint"; BiopharmaWatch headlined it "Positive AMPLIFY-7P Results for Elicio." Both describe the same company release on the same day. Searches on October 23, 2026 returned no coverage of the presentation by an independent newsroom; what else appeared was wire syndication of the release itself. ↩1
6. BioNTech SE. "BioNTech Provides Update on Phase 2 Clinical Trial of Autogene Cevumeran in Resected Colorectal Cancer." Statement, August 28, 2026, Mainz, Germany. biontech.com. Tier: read at source, and it is a company document. Concerns trial BNT122-01, NCT04486378, testing autogene cevumeran (BNT122/RO7198457) as adjuvant monotherapy in ctDNA-positive resected Stage II high-risk or Stage III colorectal cancer, co-developed with Genentech. BioNTech terminated the trial in consultation with Genentech, stating that "The decision follows a new recommendation from the independent Data Safety Monitoring Board" that treatment be discontinued. The board found a numerical imbalance in overall survival between arms and judged that continuing was unlikely to change the efficacy outcome; the futility boundary had been crossed in October 2025, when the board considered the data too immature for reliable conclusions and the trial continued per protocol. No new safety signals. The statement gives no survival figures, no size of the imbalance and no enrollment number, so none is quoted here. It states that the adjuvant pancreatic trial IMcode003, NCT05968326, combining the vaccine with checkpoint inhibition and chemotherapy, "is not affected and continues as planned." ↩1 ↩2 ↩3
7. Elicio Therapeutics. "Elicio Therapeutics Announces Closing of $15 Million Registered Direct Offering." Press release, July 6, 2026. elicio.com. Tier: title and date read at source from the company's press release index; the release body was not opened, and only the fact, amount and closing date are used here. ↩1

Citing this page. Peptide Decoding. This Pancreatic Cancer Vaccine Did Everything but Work. Then the Explanation Changed Trial figures from Elicio Therapeutics' October 23, 2026 press release; June 2026 framing from Pharmaceutical Executive and CancerNetwork coverage of the company's June 15, 2026 release; Phase 1 lineage from Elicio's August 12, 2025 release; comparator program from BioNTech's August 28, 2026 statement. Reviewed October 23, 2026. https://peptidedecoding.com/news/amplify-7p-kras-vaccine