Why This Study Matters
Surgery is the only curative option in pancreatic ductal adenocarcinoma (PDAC), yet prognosis remains poor, with a 5-year survival of only 15–17.4% as cited by the authors. Mutations in KRAS occur in approximately 90–95% of PDAC, making them an attractive target for off-the-shelf immunotherapy. Camupepimut (ELI-002 7P) combines seven amphiphile-modified mutant KRAS peptide antigens (G12D, G12R, G12V, G12A, G12C, G12S and G13D) with an amphiphile-modified CpG oligonucleotide adjuvant, designed to deliver the vaccine components to lymph nodes.
The earlier phase 1 trial of the two-peptide predecessor (ELI-002 2P) linked larger T cell responses to better outcomes. AMPLIFY-7P is the first randomized test of camupepimut, comparing it with standard-of-care observation.
Study Design
AMPLIFY-7P was a multicenter, open-label, randomized phase 2 trial conducted at 24 centers in the USA. Patients were randomized 2:1 to camupepimut or observation, with randomization stratified by nodal status (node negative vs node positive). The recommended phase 2 dose was Amph-Peptides 7P 4.9 mg with 10.0 mg Amph-CpG-7909 adjuvant, given as six prime and four boost administrations. The primary endpoint was investigator-assessed disease-free survival (DFS) by iRECIST criteria. Secondary endpoints included overall survival (OS), safety and tumor biomarker dynamics; OS data are not yet mature. The trial prespecified a success threshold of HR <0.70 with P < ~0.10 under a two-sided alpha of 0.1.
144 patients randomized (96 camupepimut vs 48 observation) at 24 US centers; 2:1 randomization; median follow-up 19.4 months; six prime plus four boost doses; data cutoff April 20, 2026.
Patient Population
Eligible adults had resected, KRAS-mutated stage I–III PDAC, expressed at least one of the seven targeted mutations, and had no radiographic evidence of disease within 6 months of finishing locoregional therapy. Patients with or without minimal residual disease (MRD) were eligible; prior immune checkpoint inhibitor or cancer vaccine treatment was an exclusion. Enrollment ran from January 8 to December 20, 2024. Among camupepimut-treated patients, median age was 65 years (range 33–86), 54.2% were female, and 31.3%, 42.7% and 26.0% had stage I, II and III disease. All patients had prior chemotherapy and pancreatic resection.
Several baseline factors were numerically imbalanced: ctDNA positivity (20.8% camupepimut vs 12.5% observation), R1 resection (26.0% vs 18.8%) and prior radiation (24.0% vs 31.3%). Nodal status, the stratification factor, was balanced (N1/N2 59.4% vs 60.4%).
Primary Endpoint Results
In the intention-to-treat population, DFS events occurred in 42/96 (43.8%) patients on camupepimut vs 26/48 (54.2%) on observation. Median DFS was 23.8 vs 21.2 months (HR 0.85, 95% CI 0.52–1.41; P = 0.5329), so the prespecified success criterion was not met. A prespecified analysis of the dosing period showed a 6-month DFS of 75.7% vs 61.7% (a 14.0% difference); exploratory DFS analysis from randomization to 6 months gave HR 0.55 (P = 0.0557).
A post hoc multivariate Cox model gave an adjusted HR of 0.47 for camupepimut (P = 0.0086). Adverse prognostic factors were stage III disease (adjusted HR 2.29, P = 0.0127), elevated CA19-9 (adjusted HR 3.16, P < 0.0001) and baseline ctDNA positivity (adjusted HR 5.99, P < 0.0001). Across both arms, median DFS was 4.2 months in ctDNA-positive vs 24.0 months in ctDNA-negative patients (HR 6.56, 95% CI 3.63–11.83). The authors describe the multivariate model as post hoc, exploratory and hypothesis-generating, to be confirmed prospectively.
Subgroup Analyses
Prespecified subgroups examined ctDNA status and prior radiation. Authors note that these analyses were not statistically adjusted for multiplicity and provide supportive insights requiring prospective validation.
| Subgroup | n | Median DFS (camupepimut vs observation) | HR (95% CI) | P | Events |
|---|---|---|---|---|---|
| ITT population | 144 | 23.8 vs 21.2 months | 0.85 (0.52–1.41) | 0.5329 | 42/96 vs 26/48 |
| ctDNA negative at baseline | 93 (64.6% of ITT) | Not reached vs 21.2 months | 0.51 (0.25–1.02) | 0.0280 | 17/62 (27.4%) vs 15/31 (48.4%) |
| ctDNA positive at baseline | 26 | 4.3 vs 1.9 months | 0.57 (0.22–1.48) | 0.0489 | 16/20 (80%) vs 6/6 (100%) |
| No prior radiation | 106 (73.6% of ITT) | 23.8 vs 10.4 months | 0.59 (0.33–1.03) | 0.0221 | 31/73 (42.5%) vs 20/33 (60.6%) |
In the ctDNA-negative subgroup, 70% vs 56% were free from relapse or death at 18 months; in the no prior radiation subgroup, the corresponding figures were 55% vs 44%. Forest plots across baseline characteristics showed generally consistent HRs, including for each commonly mutated KRAS allele.
Prior radiation was associated with lower baseline absolute lymphocyte levels (1.2 vs 1.7 × 10⁹ cells per liter, P < 0.0001) and weaker vaccine-induced T cell responses (15-fold vs 58-fold, P = 0.0011). Fewer patients with prior radiation exceeded the 9.17-fold response threshold (61.9% vs 88.4%; odds ratio 0.21, P = 0.005).
Immune Response
Among 90 evaluable camupepimut-treated patients, 89 (98.9%) generated mKRAS-specific T cell responses, with a median 44.3-fold increase over baseline. A response above the previously defined 9.17-fold threshold occurred in 82.2% (74/90) and correlated with DFS in this exploratory analysis (HR 0.22, 95% CI 0.08–0.63; P < 0.0001), with median DFS not reached vs 6.08 months and events in 26/74 (35.1%) vs 13/16 (81.3%). The authors caution that immunogenicity data come only from the camupepimut arm. Antigen spreading to neoantigens beyond KRAS was seen in 13/15 (86.7%) evaluated patients.
Safety Profile
Camupepimut was generally well tolerated. Treatment-emergent adverse events (TEAEs) were similar between arms, no TEAEs led to withdrawal of study medication, and none led to death. Events related to administration were typical vaccine reactogenicity, such as dizziness and influenza-like illness, of low incidence and limited to grade 1 in severity.
| Adverse event summary | Camupepimut (n = 96) | Observation (n = 48) |
|---|---|---|
| Any TEAE | 70 (72.9%) | 37 (77.1%) |
| Treatment-related TEAE | 23 (24.0%) | – |
| Grade ≥3 TEAE | 10 (10.4%) | 7 (14.6%) |
| Serious TEAE | 6 (6.3%) | 4 (8.3%) |
| TEAE leading to dose modification | 5 (5.2%) | 0 |
| TEAE leading to withdrawal / death | 0 / 0 | 0 / 0 |
| Diarrhea | 12 (12.5%) | 7 (14.6%) |
| Abdominal pain | 10 (10.4%) | 8 (16.7%) |
| Fatigue | 10 (10.4%) | 6 (12.5%) |
| Dizziness | 6 (6.3%) | 0 |
Interpretation and Broader Context
The investigators conclude that AMPLIFY-7P provides randomized proof-of-concept data and a basis for a confirmatory phase 3 trial. The authors propose that a future phase 3 trial exclude patients with prior radiation, stratify by ctDNA status, and test additional maintenance dosing and combination with immune checkpoint blockade (a neoadjuvant combination study, NCT07671339, is cited). Because the overall result was negative and the subgroup findings are hypothesis-generating, these signals should not be read as evidence of clinical benefit until confirmed.
Collectively, results from AMPLIFY-7P randomized phase 2 provide important lessons for future phase 3 trial design in adjuvant PDAC.
— George et al., Nature Medicine 2026, Discussion
Limitations
The authors acknowledge important limitations: a modestly sized, open-label design with two-sided alpha of 0.1; no multiplicity adjustment for subgroup analyses; a high proportion of white patients treated at a limited number of US centers; investigator-assessed DFS by iRECIST; relatively short follow-up; and OS that is not yet mature, with crossover from observation to active treatment permitted by design. The multivariate analysis was post hoc, baseline factors were numerically imbalanced between arms, and immunogenicity data come only from the camupepimut arm.