Immuno-Oncology in Pancreatic Ductal Adenocarcinoma: Biological Rationale, Clinical Progress, and Future Directions

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal solid malignancies, characterised by rising incidence, poor long-term survival, and profound resistance to systemic therapy. Its poor prognosis reflects late diagnosis, early metastatic spread, and a uniquely desmoplastic, immunosuppressive tumour microenvironment that limits both drug delivery and effective antitumour immunity. Although cytotoxic chemotherapy remains the therapeutic backbone, its clinical benefit is modest. In parallel, molecular profiling has revealed a limited but expanding set of actionable vulnerabilities; however, immuno-oncology has not yet replicated in PDAC the transformative impact observed in other tumour types. This review examines the evolving immuno-oncology landscape in PDAC, focusing on four principal domains: immune checkpoint blockade in MSI-H/dMMR disease, tumour microenvironment-directed therapies, adoptive cellular therapies, and therapeutic cancer vaccines. Taken together, available evidence suggests that progress in PDAC immuno-oncology will depend on biomarker-guided patient selection, earlier disease settings, minimal residual disease-directed approaches, and rational combinations that address immune exclusion, stromal and myeloid barriers, impaired T-cell priming, and adaptive resistance. While durable clinical benefit remains limited and most strategies remain in early-phase evaluation, emerging data from molecularly selected cohorts, adoptive cellular therapies, and personalised or shared-antigen vaccines provide a compelling rationale for continued refinement of immune-based approaches in PDAC.

1. Introduction

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal solid malignancies worldwide. Globally, PDAC accounts for approximately 2.6% of all cancer diagnoses but a disproportionate share of all cancer-related mortality (4.8% of cancer-related deaths), with over 510,000 new cases and 467,000 deaths reported annually [1]. Incidence rates continue to rise and long-term survival remains poor: the 5-year overall survival (OS) rate is approximately 13%, varying substantially by stage from 44% for localised disease to 17% for regional disease and 3% for distant disease [2,3]. The majority of patients present with advanced or metastatic disease, and recurrence rates remain high even after curative intent surgery, reflecting persistent challenges in early detection and effective treatment [4].

Chemotherapy remains the backbone of PDAC treatment, with regimens such as FOLFIRINOX (5-fluorouracil, leucovorin, irinotecan and oxaliplatin) or gemcitabine-based combinations providing modest survival benefits at the cost of significant toxicity and inevitable therapeutic resistance [5,6]. In metastatic PDAC trial populations, outcomes remain poor with a median OS of 11.1 months and progression-free survival (PFS) of 6.4 months in the first-line setting (PRODIGE 4/ACCORD 11, NCT00112658), and a median OS of 6.1 months at best in the second-line setting (NAPOLI-1, NCT01494506) [7].

Although immuno-oncology in PDAC has been reviewed extensively, the field has changed materially in recent years. Emerging data from personalised mRNA neoantigen vaccines, minimal residual disease (MRD)-directed KRAS vaccines, KRAS-specific TCR-engineered T cells and CLDN18.2-directed cellular therapies, together with increasingly biomarker-integrated trial designs, have shifted the central question from whether PDAC can respond to immunotherapy to which biological subsets and disease settings are most permissive. Accordingly, this review does not simply catalogue immune-based modalities. It uses both positive and negative clinical studies to identify recurrent mechanisms of resistance, distinguish pharmacodynamic or immunological activity from evidence of clinical efficacy, and consider where emerging strategies may fit across the PDAC disease continuum from perioperative and MRD-positive disease to treatment-refractory metastatic cancer. The principal biological barriers and therapeutic targets within the PDAC tumour microenvironment are illustrated in Figure 1, while the major therapeutic strategies and their current evidence base are summarised in Table 1.

Figure 1. Immuno-oncology targets within the PDAC tumour microenvironment. Schematic overview of key cellular and molecular interactions contributing to immune exclusion and immune suppression in PDAC, including tolerogenic dendritic cells, regulatory T cells, suppressive myeloid populations, tumour-associated macrophages, cancer-associated fibroblasts, endothelial cells, NK cells and CD8⁺ T cells. Annotated therapeutic strategies correspond to approaches discussed in this review. Therapeutic vaccines aim to enhance antigen presentation and T-cell priming; CD40, STING and TLR agonists target antigen-presenting cells and innate immune activation; immune checkpoint inhibitors target inhibitory PD-1/PD-L1 and CTLA-4 signalling; TIL/TCR-T-cell therapies and CAR-T/CAR-NK therapies supply or redirect tumour-reactive effector cells; PEGPH20 and certepetide/CEND-1 target stromal barriers and tumour penetration; CCR2, CCR2/CCR5 and CXCR4 blockade target suppressive myeloid recruitment and chemokine-mediated immune exclusion; and FAK and TGF-β pathway inhibition aim to remodel stromal and immune-suppressive signalling. Abbreviations: Arg-1: arginase-1; CAF: cancer-associated fibroblast; CAR: chimeric antigen receptor; CEA: carcinoembryonic antigen; CSF2: colony-stimulating factor 2; CSF3: colony-stimulating factor 3; CTLA-4: cytotoxic T lymphocyte-associated protein-4; CXCL12: C-X-C motif chemokine 12; DC: dendritic cell; FAK: focal adhesion kinase; FGF2: fibroblast growth factor; IL-1β: interleukin-1β; IL-4: interleukin-4; IL-6: interleukin-6; IL-8: interleukin-8; IL-10: interleukin-10; IL-27: interleukin-27; IL-35: interleukin-35; IDO: indoleamine 2,3-dioxygenase; iNOS: inducible nitric oxide synthase; LAG-3: lymphocyte-activation gene 3; MDSC: myeloid-derived suppressor cell; MMP: matrix metalloprotease; NK: natural killer; NO: nitric oxide; PD-1: programmed cell death protein-1; ROS: reactive oxygen species; TAA: tumour-associated antigen; TAM: tumour-associated macrophage; TCR-T: T-cell receptor-engineered T cell; TIL: tumour-infiltrating lymphocyte; TGFβ: transforming growth factor-β; TLR: Toll-like receptor; TNFα: tumour necrosis factor-α; TRAIL: TNF-related apoptosis-inducing ligand; Treg: regulatory T cell; VEGF: vascular endothelial growth factor. The graphic was created with BioRender.com.

Table 1. Clinical development and evidence level of immuno-oncology strategies in PDAC.

Strategy	Representative Agents/ Platforms	Highest Level of Evidence	Key Clinical Message	Main Limitations
Checkpoint blockade in molecularly selected disease	Pembrolizumab; anti-PD-1 therapy in MSI-H/dMMR disease [8,9,10,11,12]	Phase II PDAC subgroup; retrospective real-world cohorts	MSI-H/dMMR PDAC represents the most clinically actionable immunotherapy subset, with durable responses in some patients, although outcomes vary across cohorts	MSI-H/dMMR are rare; MSI/MMR status alone does not guarantee response; unselected PD-1/PD-L1 or CTLA-4 blockade has limited activity
Stromal depletion and vascular/ penetration enhancement	PEGPH20; CEND-1/certepetide [13,14,15,16,17,18,19]	PEGPH20: phase III negative; CEND-1: early randomised phase II data	PEGPH20 showed early promise in hyaluronan-high disease but failed to improve survival in phase III; CEND-1/certepetide has shown preliminary clinical activity with gemcitabine/nab-paclitaxel	Stromal biology is context-dependent; stromal depletion may not reliably improve immune permissiveness; phase III validation is lacking for newer approaches
Innate immune activation	STING agonists; TLR9 agonists; TLR7/8 agonists; radiotherapy combinations [20,21,22]	Phase I/II	Innate immune agonists can induce pharmacodynamic immune activation and occasional responses, but clinical activity in PDAC remains limited	Delivery challenges, suppressive TME redundancy, modest efficacy, and uncertain optimal combinations with checkpoint blockade or radiotherapy
FAK inhibition and immune-stromal remodelling	Defactinib with pembrolizumab and gemcitabine [23,24,25]	Phase I; ongoing/randomised phase II studies with translational endpoints	FAK inhibition may reduce stromal fibrosis, CAF-associated signalling, suppressive myeloid recruitment and Tregs, with early evidence of increased CD8+ T-cell infiltration	Evidence remains early-phase; mature randomised efficacy data are not yet available
CD40 agonism, myeloid reprogramming and chemokine-axis blockade	CP-870,893; sotigalimab/APX005M; PF-04136309; BMS-813160; BL-8040/motixafortide [26,27,28,29,30,31,32,33]	Randomised phase II/phase I-II depending on platform	CD40 agonists, CCR2/CCR5 blockade and CXCR4 inhibition have shown immune modulation and selected early clinical signals, but efficacy has been inconsistent	No standard role established; optimal combinations and biomarkers remain unresolved; toxicity concerns include pulmonary toxicity with PF-04136309 plus gemcitabine/nab-paclitaxel
TGF-β pathway inhibition	Galunisertib/LY2157299 with gemcitabine or durvalumab [34,35,36]	Randomised phase II	Galunisertib plus gemcitabine produced a modest OS improvement, and exploratory analyses suggest possible biomarker enrichment using baseline CCL3	Benefit is modest; galunisertib plus durvalumab showed limited activity; biomarker selection remains investigational
Microbiome-directed strategies	Gut and intratumoural microbiome modulation [37,38,39]	Preclinical and translational human studies	Intratumoural microbial composition has been linked to immune regulation, survival and gemcitabine resistance, providing a rationale for microbiome-directed strategies	Causal clinical evidence is lacking; substantial methodological challenges include contamination, antibiotic exposure, diet and inter-patient variability
Oncolytic virotherapy	LOAd703 [40]	Early phase I/II clinical development	Oncolytic virotherapy can combine tumour-cell lysis with local immune activation and has demonstrated feasibility and preliminary activity in PDAC	Evidence remains non-randomised and early phase; contribution to combination efficacy and optimal delivery remain uncertain
Bispecific T-cell engagers	IBI389 (CLDN18.2/CD3); HPN536 (mesothelin-directed T-cell engager) [41,42]	CLDN18.2: preliminary phase I PDAC data; mesothelin approaches predominantly preclinical/early clinical development	T-cell engagers provide a mechanistically distinct means of redirecting endogenous T cells against tumour-associated antigens, with preliminary PDAC-specific activity reported for CLDN18.2-directed therapy	Antigen heterogeneity, impaired T-cell trafficking/function, cytokine-mediated toxicity and on-target/off-tumour effects; mature PDAC efficacy data are lacking
Adoptive cellular therapies	Neoantigen-selected TILs; KRAS-directed TCR-engineered T cells; TP53-directed TCR approaches [43,44,45,46,47,48,49]	Early-phase clinical studies and case-level proof-of-concept	TIL and TCR-engineered T-cell approaches provide proof-of-concept that tumour-reactive cellular therapy can produce responses in selected PDAC patients	HLA restriction, antigen heterogeneity, manufacturing complexity, lymphodepletion toxicity and limited PDAC-specific efficacy data
CAR-T and CAR-NK therapies	CLDN18.2 CAR-T; mesothelin CAR-T; HER2 CAR-T; CD133 CAR-T; MUC1 CAR-NK [50,51,52,53,54,55,56]	Early-phase clinical studies; CAR-NK predominantly preclinical/early clinical development	CLDN18.2-directed CAR-T therapy has shown the strongest early clinical signal in refractory metastatic PDAC; other CAR-based approaches remain less mature	Poor trafficking and persistence, antigen heterogeneity, immunosuppressive TME, CRS/haematological toxicity and on-target/off-tumour risk
Therapeutic cancer vaccines	Personalised mRNA neoantigen vaccines; KRAS vaccines including ELI-002 and TG01; GVAX; CRS-207; MUC1, WT1, mesothelin and hTERT vaccines [57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92]	Contemporary neoantigen/KRAS approaches: phase I/II; historical platforms include negative randomised phase III	Vaccine strategies appear most promising in adjuvant, perioperative or MRD settings, particularly with personalised neoantigen or KRAS-directed approaches	Many data remain single-arm or early-phase; metastatic efficacy has been limited; challenges include antigen selection, immune escape, manufacturing timelines and optimal patient selection

The relevant literature was identified through PubMed/MEDLINE and ClinicalTrials.gov searches, supplemented by reference-list review and targeted searches for major guidelines, conference proceedings and recent studies. Searches included the literature published up to March 2026 and combined terms relating to pancreatic ductal adenocarcinoma with immune checkpoint blockade, MSI/dMMR, tumour microenvironment, stromal and myeloid targeting, CD40, STING/TLR, CXCR4, adoptive cellular therapies, cancer vaccines, KRAS, microbiome, bispecific T-cell engagers, oncolytic viruses, radiotherapy, ctDNA and spatial immune profiling. Priority was given to peer-reviewed PDAC-specific clinical and translational studies, with mixed-tumour studies, conference abstracts and trial-registry data included where relevant evidence was limited or rapidly evolving. This was a narrative rather than systematic review; no formal protocol, quality assessment or meta-analysis was undertaken.

2. MSI-H/dMMR PDAC

Microsatellite instability-high (MSI-H) and mismatch repair-deficiency (dMMR) are rare in PDAC, generally reported in approximately 1–3% of cases, with prevalence varying by cohort, assay, and histological composition [93]. Immune checkpoint inhibitors have demonstrated tumour-agnostic activity across MSI-H/dMMR malignancies, leading to tumour-agnostic FDA approval of the anti-PD-1 antibody pembrolizumab; however, PDAC-specific outcomes appear more heterogeneous.

Within the phase II KEYNOTE-158 trial, the PDAC cohort of 22 previously treated patients who received pembrolizumab achieved an overall response rate (ORR) of 18.2%, with a median PFS of 2.1 months and a median OS of 3.7 months [8]. Pembrolizumab therefore represents an established tumour-agnostic treatment option for patients with MSI-H/dMMR disease, although the limited PDAC cohort in KEYNOTE-158 underscores that MSI-H/dMMR status does not confer uniform sensitivity in this tumour type [9].

By contrast, in the real-world AGEO European cohort of advanced dMMR/MSI-H PDAC, immune checkpoint blockade demonstrated substantial clinical activity with a favourable safety profile, largely in the second-line or later-line setting [10]. Among 31 evaluable patients treated predominantly with single-agent anti-PD-1 therapy, the objective response rate was 48.4%. After a median follow-up of 18 months, median PFS was 26.7 months and median OS was not reached, indicating durable benefit in a subset of patients. Disease control was particularly high in less heavily pre-treated patients, with a disease control rate (DCR) of 76.5% among those who had received only one prior line of therapy. Notably, no grade 3–4 treatment-related adverse events were reported, underscoring the tolerability of immune checkpoint inhibitors in this molecularly selected PDAC population.

The variability between KEYNOTE-158 and real-world/retrospective cohorts likely reflects both biological and clinical heterogeneity within MSI-H/dMMR PDAC, including differences in tumour mutational burden, antigen presentation, TME context, line of therapy, performance status and prior treatment exposure. Although MSI-H/dMMR remains the most clinically actionable immunotherapy biomarker in PDAC, its low prevalence and variable PDAC-specific response to checkpoint blockade should be recognised. MSI/MMR assessment may be incorporated within broader molecular profiling, with positive findings potentially informing treatment selection and prompting consideration of germline evaluation for Lynch syndrome where appropriate. Testing may be performed using mismatch repair protein immunohistochemistry, PCR-based MSI testing or validated next-generation sequencing assays. Importantly, MSI-H/dMMR status alone does not guarantee response to PD-1 blockade, highlighting the need for complementary biomarkers of sensitivity and resistance.

Outside molecularly selected MSI-H/dMMR disease, immune checkpoint blockade has shown limited clinical activity in PDAC. Single-agent CTLA-4 blockade with Ipilimumab produced no RECIST objective responses in a phase II trial of locally advanced or metastatic pancreatic adenocarcinoma, while durvalumab monotherapy or durvalumab plus tremelimumab produced ORRs of 0% and 3.1%, respectively, in previously treated metastatic PDAC [11,12]. This limited efficacy likely reflects the low neoantigen burden of most microsatellite-stable PDACs, poor effector T-cell infiltration, myeloid-dominant immunosuppression, and stromal immune exclusion.

For patients who have progressed on immune checkpoint inhibitors, independent CRISPR–Cas9 loss-of-function screens have identified WRN (Werner syndrome RecQ helicase) as an essential gene in MSI-H/dMMR cancer cells [94,95]. Early-phase clinical trials are underway in MSI-H/dMMR advanced solid tumours, including gastrointestinal cancers, using first-in-class WRN inhibitors HRO761 (NCT05838768) and RO7589831/VVD-133214 (study code BP44474; NCT06004245) as monotherapy or in combination regimens. For HRO761, eligible patients must have progressed after or be intolerant to prior standard therapy; and Arms A and C specifically require progression after a prior line of immune checkpoint inhibitor therapy. For RO7589831, eligibility is generally second-line or later, although the study also includes a first-line metastatic colorectal pembrolizumab combination cohort.

3. Tumour Microenvironment (TME) Modifiers

3.1. Targeting Stromal Depletion: PEGPH20

Targeting the dense hyaluronan-rich stroma in PDAC has been pursued as a strategy to reduce interstitial pressure, decompress tumour vasculature, and enhance chemotherapy delivery.

In the phase II HALO-202 trial, conducted in previously untreated unselected metastatic PDAC, PEGPH20 (pegylated recombinant human hyaluronidase PH20) plus gemcitabine/nab-paclitaxel provided only a modest PFS benefit in the overall, unselected population compared with chemotherapy alone (6.0 vs. 5.3 months) [13]. In the prespecified hyaluronan-high subgroup, median PFS was 9.2 versus 5.2 months, ORR was approximately 45%, and median OS was 11.5 versus 8.5 months. However, this signal did not translate into clinical benefit in the subsequent phase III biomarker-enriched HALO-301 trial, in which median OS was 11.2 versus 11.5 months [14].

3.2. Vascular and Penetration Enhancement: CEND-1

CEND-1 (iRGD/LSTA1, now also referred to as certepetide) is a tumour-penetrating peptide designed to enhance intratumoural delivery of co-administered therapies through sequential binding to αV integrins and neuropilin-1 [15]. In a first-in-human phase I trial (NCT03517176), CEND-1 was combined with gemcitabine/nab-paclitaxel as first-line treatment for metastatic PDAC [16]. Among 29 evaluable patients the combination achieved an ORR of 59% (including one complete response) and a median OS of 13.2 months. Given the small, single-arm design and selected trial population, these outcomes cannot establish comparative efficacy relative to historical gemcitabine/nab-paclitaxel cohorts and should be regarded as hypothesis-generating pending randomised validation.

Certepetide is being evaluated in randomised studies in combination with first-line gemcitabine plus nab-paclitaxel. The ASCEND trial (NCT05042128) is a randomised, double-blind, placebo-controlled phase II/IIb study evaluating certepetide versus placebo added to gemcitabine plus nab-paclitaxel in patients with previously untreated metastatic PDAC [17]. Preliminary results from cohort B showed a median PFS of 7.5 versus 4.7 months (HR 0.60; 95% CI 0.34–1.07); however, this comparison did not reach statistical significance (p = 0.08) [18]. The ORR (45.2% vs. 19.0%) and DCR (85.7% vs. 61.9%) were higher with certepetide, but these preliminary findings remain insufficient to establish clinical benefit and OS data have not yet been reported. A separate randomised phase II trial (NCT06261359) is also underway evaluating CEND-1 in combination with first-line gemcitabine plus nab-paclitaxel in patients with locally advanced unresectable or metastatic PDAC. Efficacy results from this study are awaited.

In the phase I/IIa CENDIFOX trial (NCT05121038), neoadjuvant certepetide combined with modified FOLFIRINOX was feasible and well tolerated in patients with resectable or borderline resectable PDAC [19]. In preliminary data from the PDAC cohort (n = 35), most grade ≥ 3 adverse events were consistent with expected mFOLFIRINOX-related toxicities. Ten patients (29%) proceeded to resection of the pancreatic primary, and among these resected/evaluable cases, the reported R0 resection rate was only 50% and the pathologic partial response rate (defined as Tumour Regression Grade 2) was 70%. These findings should be interpreted with caution given the small number of resected patients and the absence of mature survival outcomes.

These data raise the possibility that certepetide may be better suited to selected chemotherapy backbones, although current mechanistic evidence supports enhanced intratumoural delivery of co-administered therapies broadly rather than preferential synergy with FOLFIRINOX or gemcitabine/nab-paclitaxel specifically. Clinically, the gemcitabine/nab-paclitaxel backbone has the stronger evidence base to date, while FOLFIRINOX data remain immature.

3.3. Innate Immune Activation: STING and Toll-like Receptor (TLR) Agonists

Activation of the Stimulator of Interferon Genes (STING) pathway has been explored as a means of converting the immunologically “cold” PDAC TME into an inflamed, T-cell-permissive state. In a first-in-human phase I trial (NCT02675439) of intratumoural MIW815/ADU-S100, a first-generation STING agonist, in patients with pre-treated advanced solid tumours or lymphomas, treatment was generally tolerable and induced pharmacodynamic evidence of immune activation, but antitumour activity was limited, with only one confirmed partial response reported [20].

When MIW815 was evaluated in combination with the anti-PD-1 antibody Spartalizumab/PDR001 in a phase Ib trial (NCT03172936), clinical activity was again modest with an ORR of 10.4% among evaluable solid-tumour patients and no clear evidence of meaningful activity in PDAC [21].

Toll-like receptor (TLR) agonists represent another approach to innate immune priming in PDAC, aiming to enhance antigen presentation and remodel suppressive myeloid populations. The TLR9 agonist SD-101 has been evaluated in a phase I trial (NCT04050085) in combination with the anti-PD-1 antibody Nivolumab and radiotherapy in patients with chemotherapy-refractory metastatic PDAC. However, efficacy outcomes have not yet been reported.

The dual TLR7/8 agonist BDB001 has also been investigated in the phase II AGADIR basket trial (NCT03915678) in combination with the PD-L1 inhibitor atezolizumab and immunogenic SBRT in heavily pre-treated patients lacking suitable standard-of-care therapeutic options. Immunogenic SBRT was given as an immunomodulatory strategy intended to enhance systemic antitumour immunity—including potential abscopal effects. In the PDAC cohort SBRT was directed to at least one metastatic lesion rather than specifically to the pancreatic primary. Among 32 enrolled patients in the PDAC cohort, 21 had undergone central review at the time of reporting; the DCR was 38%, with two confirmed partial responses (9.5%), six cases of stable disease (28.5%), and 13 cases of progressive disease (62.0%) [22].

3.4. Focal Adhesion Kinase (FAK) Inhibition and Immune-Stromal Remodelling

Focal adhesion kinase (FAK) inhibition represents another strategy to remodel the immunosuppressive TME in PDAC. Preclinical studies suggest that FAK inhibition can reduce stromal fibrosis and suppressive immune-cell recruitment, including cancer-associated fibroblast (CAF)-associated signalling, myeloid-derived suppressor cells, tumour-associated macrophages, and regulatory T cells (Tregs), thereby increasing CD8⁺ T-cell infiltration and sensitising tumours to immune checkpoint blockade [23].

In a phase I dose-escalation and expansion study of defactinib, an oral small-molecule inhibitor of focal adhesion kinase (FAK/PTK2) and proline-rich tyrosine kinase 2 (PYK2), plus pembrolizumab and gemcitabine in treatment-refractory advanced/metastatic PDAC, the combination was tolerable and showed preliminary activity [24]. Among 20 treated patients, the DCR was 80%, with one partial response and 15 patients achieving stable disease; median PFS and OS were 3.6 and 7.8 months, respectively. In the maintenance cohort, among 10 evaluable patients, the DCR was 70% with one partial response and six patients achieving stable disease; median PFS and OS on study treatment were 5.0 and 8.3 months, respectively. Paired biopsy analyses demonstrated increased infiltrating CD8⁺ T cells and decreased Tregs after treatment, supporting pharmacodynamic immune modulation.

Defactinib is also being evaluated in a phase II randomised trial (NCT03727880), which investigates pembrolizumab with or without defactinib, following chemotherapy as a neoadjuvant and adjuvant treatment for patients with resectable PDAC. Preliminary translational reporting suggests reduced fibroblast infiltration and increased antitumour immune activation with defactinib, including higher pro-inflammatory (M1) macrophage and CD8⁺ T-cell infiltration [25]. However, mature peer-reviewed clinical efficacy outcomes have not yet been reported.

3.5. CD40 Agonism, Myeloid Reprogramming and Chemokine-Axis Blockade

Beyond stromal depletion, vascular penetration enhancement, innate immune agonism and FAK inhibition, additional TME-directed strategies have focused on modulating the myeloid compartment and chemokine networks that contribute to immune exclusion in PDAC. These approaches include CD40 agonism to activate antigen-presenting cells and remodel macrophage function, CCR2/CCR5-directed strategies to limit suppressive monocyte and macrophage recruitment, and CXCL12-CXCR4 blockade to improve T-cell trafficking into tumour tissue.

CD40 agonists activate antigen-presenting cells, particularly dendritic cells and macrophages, thereby enhancing T-cell priming and promoting myeloid remodelling. Early translational studies showed that CD40 activation could remodel the PDAC stroma and induce macrophage-mediated antitumour activity in preclinical models and patients with PDAC [26]. In a phase I trial (NCT00711191) of 22 chemotherapy-naïve patients with advanced/metastatic PDAC, CP-870,893 (an agonist CD40 monoclonal antibody) combined with gemcitabine was tolerable and produced partial responses in 4 of 21 evaluable patients, with an ORR of 19%, stable disease in 11 patients, median PFS of 5.6 months and median OS of 7.4 months [27]. In the phase Ib PRINCE trial (NCT03214250), sotigalimab/APX005M plus gemcitabine/nab-paclitaxel, with or without Nivolumab, was tolerable and produced responses in 14 of 24 evaluable patients [28]. However, in the randomised phase II PRINCE trial, only Nivolumab plus chemotherapy met the primary 1-year OS endpoint (57.7%; median OS 16.7 months), whereas sotigalimab plus chemotherapy (48.1%; median OS 11.4 months) and the triplet regimen (41.3%; median OS 10.1 months) did not [29]. These data suggest context-dependent activity for CD40 agonism, while highlighting unresolved questions around biomarker selection and optimal combination strategies.

Myeloid-cell targeting represents another rational strategy, given the abundance of tumour-associated macrophages, inflammatory monocytes, neutrophils and myeloid-derived suppressor cells within PDAC. CCR2 inhibition has been explored as a means of limiting recruitment of suppressive inflammatory monocytes. PF-04136309 is a small-molecule CCR2 antagonist that inhibits CCL2–CCR2-mediated recruitment of inflammatory monocytes into the PDAC TME. In a phase Ib trial (NCT01413022) of patients with borderline resectable or locally advanced PDAC, 16 of 33 radiographically evaluable patients achieved an objective response with PF-04136309 plus FOLFIRINOX, corresponding to an ORR of 49%; local tumour control was observed in 32 of 33 patients (97%) [30]. However, in a separate phase Ib trial (NCT02732938) of 21 patients with previously untreated metastatic PDAC, PF-04136309 combined with gemcitabine plus nab-paclitaxel produced an ORR of 23.8%, but was associated with pulmonary toxicity in 24% of patients, including grade 4 hypoxia. Although CD14⁺CCR2⁺ inflammatory monocytes decreased in peripheral blood, the regimen did not show an efficacy signal above gemcitabine/nab-paclitaxel alone and raised concern for synergistic pulmonary toxicity [31]. Dual CCR2/CCR5 blockade has also been investigated using BMS-813160. In a phase Ib/II trial (NCT03184870) of first-line metastatic PDAC, BMS-813160 combined with gemcitabine plus nab-paclitaxel produced an ORR of 26% compared with 28% for gemcitabine plus nab-paclitaxel alone, while the addition of Nivolumab to BMS-813160 plus gemcitabine/nab-paclitaxel produced an ORR of 37% [32]. Median duration of response was longer with BMS-813160-containing regimens, but the ORR was not clearly improved over chemotherapy alone, and the regimen remains investigational.

Chemokine-axis blockade has also been investigated as a method to reverse immune exclusion. CXCL12–CXCR4 signalling has been implicated in restricting T-cell infiltration into PDAC, providing a rationale for CXCR4 inhibition in combination with immune checkpoint blockade. In the phase IIa COMBAT/KEYNOTE-202 trial (NCT02826486), the CXCR4 antagonist BL-8040/motixafortide was evaluated with pembrolizumab, with or without chemotherapy, in metastatic PDAC. In the chemotherapy-resistant cohort treated with BL-8040 plus pembrolizumab, the DCR was 34.5%, including one partial response and nine cases of stable disease, with a median OS of 3.3 months. In the chemotherapy-combination cohort, BL-8040 plus pembrolizumab and chemotherapy produced an ORR of 32%, DCR of 77% and median duration of response of 7.8 months. Treatment was associated with increased CD8⁺ effector T-cell tumour infiltration and reductions in myeloid-derived suppressor cells and circulating regulatory T cells, supporting pharmacodynamic immune modulation [33].

TGF-β pathway inhibition has also been explored as a strategy to reduce stromal and immune-mediated suppression in PDAC. Galunisertib/LY2157299, an oral TGF-β receptor I/ALK5 kinase inhibitor, was evaluated with gemcitabine in the phase Ib/randomised phase II H9H-MC-JBAJ study (NCT01373164) in patients with unresectable locally advanced or metastatic pancreatic cancer. In the randomised phase II portion, galunisertib plus gemcitabine modestly improved median OS compared with gemcitabine alone (8.9 vs. 7.1 months; HR 0.79), with minimal added toxicity [34]. A subsequent exploratory biomarker analysis of the same trial suggested that high baseline plasma CCL3 may identify a subgroup more likely to benefit from galunisertib plus gemcitabine: high CCL3 was associated with a poor outcome in the placebo/gemcitabine arm, but with an improved outcome in the galunisertib/gemcitabine arm [35]. However, subsequent evaluation of galunisertib with durvalumab in recurrent/refractory metastatic PDAC (NCT02734160) demonstrated limited activity, with one partial response, seven cases of stable disease, a DCR of 25.0%, median PFS of 1.87 months and median OS of 5.72 months among 32 patients treated at the recommended phase II dose [36]. These findings suggest that TGF-β inhibition may require biomarker selection and rational combination strategies, rather than broad application in unselected PDAC populations.

3.6. Sensitising Microsatellite-Stable PDAC to Immunotherapy: Integrated and Emerging Strategies

For the vast majority of PDACs that are microsatellite stable and mismatch repair-proficient, resistance to checkpoint blockade is not attributable to a single lesion but to failure across several stages of the cancer–immunity cycle. Low neoantigen burden and inefficient antigen presentation limit productive T-cell priming; dense and heterogeneous stroma and chemokine signalling restrict trafficking; suppressive macrophages, monocytes, myeloid-derived suppressor cells and regulatory T cells inhibit effector function; and metabolic stress and persistent inhibitory signalling promote intratumoural T-cell dysfunction [24]. Sensitisation of MSS PDAC should therefore be viewed as a coordinated strategy rather than a single therapeutic class. Radiotherapy, oncolytic viruses and vaccines may increase antigen release and immune priming; CD40, STING and TLR agonists can activate antigen-presenting cells; stromal-, FAK-, myeloid- and chemokine-directed therapies may improve immune-cell trafficking and reduce suppression; and adoptive cellular therapies or T-cell-engaging antibodies can partially bypass inadequate endogenous priming. To date, however, no such sensitisation strategy has established a routine clinical role in unselected MSS PDAC.

3.6.1. Microbiome Modulation

The gut and intratumoural microbiomes may represent an additional layer of immune and treatment resistance in PDAC. Preclinical work has shown that the pancreatic tumour microbiome can promote innate and adaptive immune suppression and that microbial depletion can remodel the immune microenvironment and reduce tumour progression [37]. In human resected PDAC, greater intratumoural microbial diversity and distinct microbial signatures have been associated with long-term survival and enhanced antitumour immune infiltration [38]. Intratumoural bacteria may also contribute directly to treatment resistance: Gammaproteobacteria expressing bacterial cytidine deaminase can metabolise gemcitabine into an inactive metabolite [39]. These observations provide a rationale for microbiome-informed or microbiome-modulating therapeutic approaches, but causal relationships in patients remain incompletely defined and microbiome-directed treatment is not currently established in PDAC. Prospective studies will require rigorous control for sampling contamination, antibiotic exposure, diet and other host factors and, critically, demonstration that microbiome modulation alters clinically relevant immune or therapeutic outcomes.

3.6.2. Radiotherapy and Oncolytic Virotherapy

Radiotherapy may contribute to immune sensitisation through tumour-cell death, antigen release and inflammatory signalling, providing a rationale for combination with checkpoint blockade or immune-priming strategies. As discussed above, early PDAC studies combining SBRT with innate immune agonists and checkpoint blockade have demonstrated feasibility and preliminary clinical activity, but evidence of systemic clinical efficacy remains limited [22]. More recently, a single-arm phase II study evaluated GVAX, pembrolizumab and SBRT after induction chemotherapy in locally advanced PDAC. Among 54 evaluable patients, median distant metastasis-free survival was 9.8 months and median OS from initiation of immunotherapy was 21.8 months [96]. However, the multimodal, non-randomised design precludes attribution of benefit to any individual component, and the optimal radiotherapy dose, fractionation, sequencing and combination strategy remain unresolved.

Oncolytic viruses offer a complementary strategy by coupling tumour-cell lysis with local immune activation. In the non-randomised phase I/II LOKON001 study, intratumoural LOAd703, an oncolytic adenovirus encoding TMZ-CD40L and 4-1BBL, was combined with gemcitabine/nab-paclitaxel in advanced PDAC and demonstrated feasibility, antitumour responses and treatment-emergent immune activation [40]. These findings provide proof-of-concept for local virotherapy as an immune-priming strategy, but the early-phase design does not establish additive clinical efficacy and subsequent combination strategies remain investigational.

3.6.3. Bispecific T-Cell Engagers

Bispecific T-cell-engaging antibodies represent another approach to bypass inadequate endogenous T-cell priming by simultaneously binding a tumour-associated surface antigen and CD3 on T cells. Targets under investigation in PDAC include CLDN18.2 and mesothelin. Mesothelin-directed T-cell-engaging constructs such as HPN536 have demonstrated T-cell activation and antitumour activity in preclinical models, including pancreatic cancer models, and have progressed to early-phase clinical evaluation in mesothelin-expressing solid tumours [41]. Preliminary PDAC-specific clinical activity has also been reported with the CLDN18.2/CD3 bispecific antibody IBI389 in a phase I study of patients with advanced PDAC [42]. However, these data remain early, and mature evidence of clinical efficacy in PDAC is limited. Potential barriers include heterogeneous target expression, impaired T-cell trafficking and function within the PDAC microenvironment, cytokine-mediated toxicity and on-target/off-tumour effects. Bispecific approaches therefore remain investigational but provide a mechanistically distinct strategy for redirecting endogenous T cells against PDAC.

4. Adoptive Cell Therapies

4.1. Tumour-Infiltrating Lymphocyte (TIL) Therapy

Adoptive Tumour-Infiltrating Lymphocyte (TIL) therapy involves ex vivo expansion of tumour-derived lymphocytes, followed by reinfusion after non-myeloablative lymphodepleting chemotherapy, typically with high-dose interleukin-2 support [43].

In the phase II NCT01174121 trial, neoantigen-selected TIL therapy was evaluated in patients with metastatic, treatment-refractory, mismatch repair-proficient gastrointestinal cancers, including PDAC. In one cohort, 39 patients received neoantigen-selected TIL therapy alone, with an ORR of 7.7% (3/39). In a subsequent cohort, 34 patients received neoantigen-selected TIL therapy plus pembrolizumab, with an ORR of 23.5% (8/34). PDAC comprised a small subset of the treated population (n = 6); all six patients received selected neoantigen-specific TIL therapy, and three also received pembrolizumab. One patient with metastatic PDAC receiving the combination achieved a confirmed RECIST partial response, while other PDAC cases showed early tumour regression that did not translate into confirmed RECIST responses [44]. A PDAC-specific ORR was not reported.

Separate trials, NCT03935893 and NCT05098197, are evaluating autologous TIL therapy with non-myeloablative lymphodepletion and high-dose aldesleukin in locally advanced or metastatic solid tumours, including PDAC; however, no published efficacy outcomes are currently available.

4.2. T-Cell Receptor (TCR)-Engineered T-Cell Therapy

TCR-engineered T cells are generated by genetically modifying autologous T cells ex vivo to express tumour-specific TCRs capable of recognising tumour-derived peptide antigens presented by specific HLA molecules. This approach is therefore both antigen- and HLA-restricted [45,46].

At the molecular level, PDAC is defined predominantly by KRAS-driven oncogenesis, present in approximately 90% of cases [47]. A smaller subset of tumours harbour distinct molecular vulnerabilities, including rare oncogenic fusions and other targetable alterations such as BRCA1/2 mutations, BRAF mutations, HER2 amplification, CLDN18.2 expression and MTAP deletion.

In a proof-of-concept case report, a patient with metastatic PDAC treated with TCR-engineered T cells targeting KRAS G12D presented by HLA-C*08:02 achieved objective regression of metastatic disease, with a 72% partial response by RECIST 1.1 that remained ongoing at 6 months post-treatment [45].

Additional KRAS-directed TCR-engineered T-cell therapy approaches are now in early clinical development. In a first-in-human phase I study, NW-301V, an autologous TCR-engineered T-cell therapy targeting KRAS G12V in HLA-A*11:01-positive tumours demonstrated an ORR of 42.9% across 14 heavily pre-treated patients with unresectable or advanced PDAC or colorectal cancer [48]. Among 10 patients treated at the target dose levels 2–3, the partial response rate was 50%, stable disease was observed in 50%, and median PFS was 5.8 months [49]. No tumour-specific response breakdown was reported. The most common Grade ≥ 3 adverse events were decreased lymphocytes (78.6%), decreased white blood cells (50%) and decreased neutrophils (50%), all of which were related to lymphodepletion. Grade 1–2 cytokine release syndrome (CRS) was reported in six (42.9%) patients, and there were no occurrences of immune effector cell-associated neurotoxicity syndrome (ICANS).

Separately, NCT04146298 is evaluating autologous T cells engineered with a KRAS G12V-specific, HLA-A*11:01-restricted TCR in patients with advanced PDAC. In this single-arm phase I/II study of five patients, one patient with liver metastases achieved a complete response lasting 5.5 months, while two patients experienced short-term stable disease. TCR-T cells were detectable in peripheral blood in four patients at 1 month after the first infusion. These findings provide preliminary evidence of clinical activity but require validation in larger cohorts [46].

Other early-phase TCR-engineered T-cell studies include NCT03745326, a phase I/II study of peripheral blood lymphocytes transduced with an anti-KRAS G12D murine TCR in HLA-A*11:01-positive patients with advanced solid tumours expressing KRAS G12D, and NCT05877599, a phase I study of NT-175, an autologous TCR-engineered T-cell product targeting TP53 R175H in HLA-A*02:01-positive advanced solid tumours, including PDAC. A phase I study of mesothelin-specific TCR-engineered T cells in metastatic PDAC (NCT04809766) was terminated because of the end of funding, and no efficacy outcomes have been publicly reported.

4.3. Chimeric Antigen Receptor (CAR)-T- and NK-Cell Therapies

Chimeric antigen receptor (CAR)-T cells and CAR natural killer (NK) cells are genetically engineered immune effector cells designed to recognise tumour-associated surface antigens in an MHC-independent manner, enabling targeted antitumour cytotoxicity through adaptive (T-cell) or innate (NK-cell) effector mechanisms [50].

In a phase I trial of CD133-directed CAR-T cells in 23 patients with advanced or metastatic CD133-positive malignancies, including PDAC and other gastrointestinal malignancies, three patients achieved partial remission and 14 achieved stable disease, corresponding to a 3-month DCR of 65.2%; median PFS was 5 months [51]. However, these data were derived from a heterogeneous, non-randomised cohort, and PDAC-specific efficacy was not reported.

Satricabtagene autoleucel (CT041) is a CLDN18.2-targeted chimeric antigen receptor T-cell (CAR-T) therapy. In a pooled analysis of two trials (NCT03874897, NCT04581473) the efficacy of CT041 was evaluated in patients (n = 24) with refractory metastatic PDAC [52]. Five patients (20.8%) had previously received one line of therapy, whereas 19 (79.2%) received ≥2 lines of therapy. The ORR and DCR were 16.7% and 70.8%, respectively. The median PFS was 3.3 months and the median OS was 10.0 months. CRS occurred in the majority of patients but was limited to grade 1–2 events. The most common grade ≥ 3 treatment-related adverse events (TRAEs) were haematological toxicities.

Additional CAR-T targets with reported clinical experience in pancreatic cancer include mesothelin and HER2. In a phase I study (NCT01897415) of RNA-engineered mesothelin-specific CAR-T cells (CARTmeso) in six patients with chemotherapy-refractory metastatic PDAC, treatment was feasible and generally tolerable, but antitumour activity was limited; two patients achieved stable disease, with a PFS of 3.8 and 5.4 months, respectively [53]. In another phase I study (NCT01935843) of HER2-directed CAR-T cells in HER2-positive advanced biliary tract and pancreatic cancers, 11 patients were treated; one patient achieved a partial response lasting 4.5 months and five achieved stable disease, with a median PFS of 4.8 months [54]. However, this was a mixed pancreatic/biliary cohort, and PDAC-specific efficacy was not reported.

Additional investigational CAR-T targets in PDAC or PDAC-inclusive solid-tumour studies include CEA (NCT02349724), EpCAM (NCT03013712), EGFR/EGFRvIII (NCT03267173), and MUC1 (NCT02587689, NCT03267173); however, published efficacy data remain unavailable. Overall, outside CLDN18.2-directed Satricabtagene autoleucel, CAR-T clinical activity in PDAC remains modest and based on small early-phase studies.

CAR-NK approaches remain less clinically mature in PDAC. In preclinical studies, allogeneic NK cells have been engineered to express apoptosis-inducing ligands and redirected toward folate receptor α (FRα), thereby initiating death receptor 4/5 (DR4/5)-mediated apoptosis in FRα/DR4/5-positive PDAC models [55]. Mesothelin-directed CAR-NK strategies have also shown preclinical activity in PDAC models [56]. Clinically, CAR-NK development in PDAC remains early. NCT02839954 will evaluate anti-MUC1 CAR-NK cells in MUC1-positive relapsed or refractory solid tumours, including PDAC, but published efficacy data remain unavailable.

5. Vaccines

Therapeutic cancer vaccines in PDAC employ diverse technological platforms, including peptide, dendritic-cell, whole-cell, viral-vector, DNA, and mRNA-based approaches [97]. Available clinical data suggest that antigen selection, immune contexture, tumour burden, and disease setting are likely to be at least as important as the vaccine platform itself in determining biological and clinical activity.

Clinical experience in PDAC suggests that vaccine-based strategies may be more promising in the neoadjuvant, adjuvant or MRD setting, where tumour burden is lower and immune competence is relatively preserved [98]. In the perioperative setting, vaccination strategies enable tumour-specific T cells to be primed in the presence of the TME, facilitating clonal expansion and intratumoural infiltration prior to resection [99]. Conversely, metastatic PDAC is characterised by dense desmoplastic stroma, immune exclusion, suppressive myeloid populations, and T-cell dysfunction, all of which may limit the efficacy of vaccine-based approaches as monotherapy [100].

An important limitation when interpreting early vaccine studies is that associations between treatment-induced immune responses and subsequent survival are correlative rather than necessarily causal. Analyses comparing immune responders with non-responders are susceptible to time-dependent selection or guarantee-time effects, because patients must remain clinically well enough and progression-free for sufficient time to receive vaccination and develop a measurable immune response. They may also be confounded by tumour burden, underlying disease biology, treatment exposure and other prognostic factors. Such analyses provide important proof-of-mechanism but cannot establish that vaccine-induced immunity caused the observed survival difference. Randomised comparisons and appropriately designed landmark or time-dependent analyses are required to demonstrate clinical benefit.

5.1. Personalised Neoantigen Vaccines

Personalised neoantigen vaccines target tumour-specific somatic mutations unique to an individual patient’s cancer, with the aim of eliciting highly specific T-cell responses against tumour-derived neoepitopes. They may be particularly suited to minimal residual disease (MRD) settings, including the adjuvant or early post-resection period, where tumour burden is low and immune competence is relatively preserved. Clinical implementation remains constrained by manufacturing complexity, tumour sequencing requirements, and delays inherent to bespoke vaccine production.

A phase I trial (NCT04161755) evaluated the personalised mRNA neoantigen vaccine autogene cevumeran (RO7198457/BNT122), which encodes up to 20 patient-specific neoantigens in patients with surgically resected PDAC. The treatment sequence consisted of adjuvant atezolizumab followed by autogene cevumeran and modified FOLFIRINOX [57]. The safety profile was favourable, with TRAEs predominantly low-grade. Among patients treated with atezolizumab (n = 19), no grade ≥ 3 adverse events were observed. For autogene cevumeran (n = 16), all patients experienced grade 1–2 adverse events, and one patient (6%) developed grade 3 fever and hypertension; no higher-grade toxicities were reported. Vaccine-induced neoantigen-specific T-cell responses were detected in 8 of 16 vaccinated patients. At a median follow-up of 18 months, median recurrence-free survival (RFS) was not reached in patients with vaccine-induced T-cell responses, whereas patients without such responses had a median RFS of 13.4 months. With extended follow-up at a median of 3.2 years, responders continued to show prolonged RFS, with median RFS not reached compared with 13.4 months in non-responders. Six of eight responders remained disease-free, whereas seven of eight non-responders experienced recurrence [58]. These findings demonstrate durable vaccine-induced immunogenicity and support further evaluation in resected PDAC, but the apparent RFS separation remains hypothesis-generating given the small, single-arm design and post-treatment classification according to immune response.

5.2. Off-the-Shelf Antigen-Specific Vaccines

Off-the-shelf vaccines target shared tumour-associated antigens (TAAs) or recurrent oncogenic driver mutations and offer advantages in scalability and rapid deployment compared with personalised strategies. However, their efficacy may be limited by central tolerance, antigen heterogeneity, HLA restriction, and immune escape, often necessitating combination approaches or careful patient selection.

5.2.1. KRAS-Targeted Vaccines

A notable recent advance in KRAS-targeted vaccination is the phase I AMPLIFY-201 trial (NCT04853017), which evaluated ELI-002 2P, an off-the-shelf lymph-node-targeting amphiphile vaccine incorporating mutant KRAS G12D and G12R peptides with an amphiphile-modified CpG-7909 adjuvant, in 25 patients (20 PDAC and 5 colorectal) with MRD following definitive locoregional therapy [59]. AMPLIFY-201 incorporated an MRD-enriched population, defined by detectable ctDNA and/or persistently elevated serum tumour antigens after locoregional therapy. At a median follow-up of 19.7 months, patients whose mutant KRAS-specific T-cell responses exceeded a predefined 9.17-fold threshold experienced prolonged radiographic relapse-free survival (median not reached vs. 3.02 months; HR 0.12; p = 0.0002), as well as prolonged OS (median not reached vs. 15.98 months; HR 0.23; p = 0.0099). At extended follow-up, 21/25 patients generated mutant KRAS-specific T-cell responses, and patients with responses above the 9.17-fold threshold universally achieved tumour biomarker reductions, including complete ctDNA clearance in 6/6 patients. Both CD4⁺ and CD8⁺ mutant KRAS-specific T-cell responses were detected in 71% of patients, supporting the biological rationale for coordinated helper and cytotoxic T-cell immunity, and antigen spreading to additional tumour antigens was observed in approximately 67%. TRAEs occurred in 48% of patients and were exclusively grade 1–2 in severity, with no grade ≥ 3 vaccine-related toxicities, CRS, or dose-limiting toxicities observed. The most common TRAEs were fatigue (24%), injection-site reactions (16%) and myalgia (12%). Although the relationship between the magnitude of KRAS-specific immunity, biomarker reduction and the clinical outcome is biologically compelling, these responder-defined analyses remain exploratory and cannot establish that the induced immune response caused the observed survival difference.

A randomised phase II trial, AMPLIFY-7P (NCT05726864), is evaluating the seven-peptide formulation ELI-002 7P as adjuvant monotherapy in patients with KRAS/NRAS-mutant PDAC following surgical resection and chemotherapy, with or without radiation. The study has completed enrolment, and preliminary company-reported analyses describe robust mutant KRAS-specific T-cell responses; however, mature peer-reviewed disease-free survival and OS outcomes have not yet been reported.

TG01/GM-CSF is a KRAS-targeted synthetic long peptide vaccine directed against common RAS mutations, including G12C, G12D, G12R, G12V, and G13D, administered with GM-CSF. In a phase I/II study of 32 patients with resected RAS-mutant PDAC, adjuvant TG01/GM-CSF combined with gemcitabine was associated with a median OS of 33.3 months and a 2-year OS rate of 71.9% [60]. In the absence of a randomised comparator, these outcomes support feasibility and further evaluation but cannot establish comparative efficacy to standard adjuvant treatment. TRAEs were predominantly grade 1–2 injection-site and systemic symptoms, including erythema, pruritus, oedema, injection-site pain, fatigue, and flu-like symptoms. Rare severe hypersensitivity reactions occurred, with two grade 4 anaphylactic events and one hypersensitivity reaction attributed to TG01 which resolved within 1–2 h with antihistamines and corticosteroids.

Notably, earlier RAS-targeted vaccine efforts were less successful. GI-4000, a series of heat-killed recombinant Saccharomyces cerevisiae vaccines expressing mutated RAS proteins, was evaluated with adjuvant gemcitabine in a randomised phase II trial (NCT00300950) of patients with resected RAS-mutant PDAC [61]. RFS was not improved compared with placebo (354 vs. 357 days; HR 1.01; 95% CI 0.73–1.41; p = 0.936), although a higher frequency of reduced regulatory T-cell populations was observed among R0/R1 patients treated with GI-4000. Subsequent GI-4000 studies in pancreatic cancer, including combinations with adoptive transfer or other immunotherapeutic approaches, were withdrawn or terminated because of feasibility and accrual limitations.

A separate phase I trial (NCT04117087) evaluated a pooled mutant KRAS synthetic long peptide vaccine targeting G12V, G12A, G12R, G12C, G12D, and G13D, administered with poly-ICLC, Nivolumab, and Ipilimumab in patients with resected PDAC or microsatellite-stable colorectal cancer. In the reported cohort of 12 patients, vaccine-related adverse events were limited to grade 1–2. Significant increases in average mutant KRAS-specific T-cell responses across the six vaccine antigens were observed in 11/12 patients (91.7%), while 10/12 patients (83.3%) developed significant T-cell responses against their tumour-matched KRAS mutation. At a median follow-up of 35.8 months, 4/12 patients (33%) remained disease-free and 8/12 (66.7%) had developed recurrence; median DFS and OS from first vaccination were 6.35 months and 29.59 months, respectively. In exploratory analyses, stronger mutant KRAS-specific T-cell responses were associated with longer DFS (18.8 vs. 2.76 months; HR 0.19; 95% CI 0.04–0.95; p = 0.024) [62].

In a phase I trial (NCT03948763) mRNA5671/V941, a lipid nanoparticle-formulated mRNA vaccine encoding four common KRAS neoantigens (G12C, G12D, G12V, G13D), administered either as monotherapy or in combination with pembrolizumab, was evaluated in patients with KRAS-mutant advanced or metastatic non-small cell lung cancer, colorectal cancer and PDAC; no efficacy results have been reported.

A phase I/II trial (NCT03953235) evaluated a heterologous prime/boost shared neoantigen vaccine regimen, GRT-C903 followed by GRT-R904, using chimpanzee adenoviral and self-amplifying mRNA vectors in combination with Ipilimumab and Nivolumab in patients with advanced or metastatic solid tumours, including PDAC [63]. The vaccine initially encoded up to 20 shared oncogenic neoantigens, including recurrent KRAS mutations such as G12D, G12R, G12V, G13D, and Q61H. In the interim phase I analysis, 19 patients were treated, of whom 18 had KRAS-mutant tumours. The regimen was generally tolerable, with most TRAEs being grade 1–2; however, two patients experienced grade 3/4 serious TRAEs that were also dose-limiting toxicities. No objective responses were reported; stable disease was observed in 8/19 patients (42%), with target-lesion reductions in three patients. Median PFS and OS were 1.9 months and 7.9 months, respectively. Immunologically, robust KRAS G12C-specific CD8⁺ T-cell responses were observed in approximately one-third of treated patients, supporting proof-of-mechanism, although clinical activity was limited and PDAC-specific efficacy was not established.

5.2.2. Other Shared Tumour-Associated Antigens (Mesothelin, MUC1, WT1)

Several vaccines in PDAC target overexpressed shared TAAs such as mesothelin, MUC1, WT1, and CEA. Although these antigens are not tumour-specific, their broad expression has enabled off-the-shelf vaccine development, particularly when combined with immune adjuvants, vector-based platforms, or dendritic-cell priming strategies.

Mesothelin is overexpressed in approximately 70–80% of PDAC cases and has been explored as a shared vaccine target [64]. In a phase I trial (NCT00585845) that included patients with metastatic PDAC, the live-attenuated Listeria monocytogenes vaccines ANZ-100 and mesothelin-expressing CRS-207 induced immune activation. Thirty-seven percent of subjects lived ≥15 months, although this was not specific to PDAC and primarily supported platform feasibility and immune activation [65].

A subsequent randomised phase II trial (NCT01417000) evaluated CRS-207 in combination with GVAX (further detail in Section 5.3.1) and low-dose cyclophosphamide in patients with pre-treated metastatic PDAC, showing improved median OS compared with GVAX and cyclophosphamide alone (6.1 vs. 3.9 months). In a prespecified per-protocol analysis of patients receiving at least three treatment doses, median OS was 9.7 versus 4.6 months, respectively [66]. However, this early signal was not confirmed in the later randomised multicentre ECLIPSE phase IIb trial (NCT02004262), in which GVAX given with cyclophosphamide plus CRS-207 or CRS-207 alone did not improve OS compared with a physician’s choice of chemotherapy in previously treated metastatic PDAC [67].

Mucin 1 (MUC1) is a TAA widely expressed in >80% of PDAC cases and has been explored as a shared vaccine target using dendritic cell-based delivery platforms [68]. In a phase I trial, 42 patients with unresectable or metastatic PDAC who had received at least one prior line of treatment received MUC1-mRNA-transfected dendritic cells (MUC1-DCs) in combination with gemcitabine and adoptively transferred MUC1-specific cytotoxic T lymphocytes (MUC1-CTLs) [69]. One patient achieved a complete response (2.4%), three had partial responses (7.1%), and 22 experienced stable disease (52.4%). The median OS was 13.9 months, increasing to 16.1 months among patients that received more than one vaccination. However, the non-randomised, multimodal design precludes attribution of these outcomes specifically to vaccination.

Wilms tumour 1 (WT1) is a shared TAA overexpressed in 75% of PDAC cases [70]. In a phase I trial, eight patients with resected PDAC received WT1-pulsed DCs with S-1-based chemotherapy, with or without gemcitabine, in the adjuvant setting. WT1-specific cytotoxic T lymphocytes were detected in seven of eight vaccinated patients [71].

In another phase I pilot study, 10 HLA-A*2402-positive patients with chemotherapy-naïve advanced/metastatic pancreatic cancer received WT-1 peptide-pulsed dendritic-cell vaccination combined with gemcitabine as first-line therapy [72]. WT1-specific immune responses were observed in a subset of patients, including delayed-type hypersensitivity (DTH) positivity in three patients. All three DTH-positive patients achieved stable disease, although the small study was not designed to establish clinical efficacy.

In a separate phase I study, 11 HLA-positive patients with advanced PDAC, including 10 with metastatic disease, received dendritic cells pulsed with WT1 peptides in combination with gemcitabine [73]. WT1-specific DTH positivity was observed in four of seven evaluable PDAC patients. DTH positivity was associated with longer OS and PFS, with all three patients showing strong DTH responses achieving a median OS of 717 days; however, this post-treatment responder analysis was exploratory and susceptible to time-dependent selection and prognostic confounding.

PANVAC-VF is a poxvirus-based prime-boost vaccine targeting MUC1 and CEA with costimulatory signalling. In a phase I trial, 10 patients with advanced PDAC were treated with PANVAC-VF. Antigen-specific T-cell responses were detected in five of eight evaluable patients (62.5%), with longer observed OS among patients who developed CEA- and/or MUC1-specific immune responses than among non-responders (15.1 vs. 3.9 months) [74]. Given the very small cohort and responder-defined analysis, this finding should be interpreted as immunological proof-of-concept rather than evidence of clinical efficacy. A subsequent phase I trial (NCT00669734) evaluated EUS-guided intratumoural PANVAC-F with subcutaneous PANVAC-V/F and GM-CSF in patients with unresectable locally advanced or metastatic pancreatic cancer. Intratumoural delivery was feasible, although mature peer-reviewed efficacy outcomes have not been published [75].

Another multi-antigen dendritic-cell approach is MesoPher, an autologous monocyte-derived dendritic-cell vaccine loaded with allogeneic tumour lysate containing antigens shared with PDAC, including mesothelin, WT1, MUC1, and survivin. In a phase I study, 10 patients with resected PDAC and no radiological evidence of recurrence after standard-of-care treatment received MesoPher [76]. Vaccine-induced T-cell activation was observed in all evaluable patients and treatment-induced T-cell reactivity against autologous tumour was detected. At a median follow-up of 25 months, 7 of 10 patients had not experienced disease recurrence or progression.

5.2.3. Human Telomerase Reverse Transcriptase (hTERT)-Targeted Vaccines

Telomerase activity is increased in 85–90% of PDAC cases, enabling sustained tumour proliferation and making human telomerase reverse transcriptase (hTERT) an attractive shared tumour-associated antigen [77]. GV1001 is a 16-amino-acid hTERT-derived peptide vaccine designed to induce hTERT-specific T-cell responses.

In a dose-escalating phase I/II trial of 48 patients with unresectable PDAC, GV1001 combined with GM-CSF was administered as upfront vaccination, with palliative gemcitabine permitted from week 6 in patients without early progression. Vaccine-specific immune responses were detected in a substantial proportion of evaluable patients and were associated with longer survival, with immune responders showing median OS of approximately 7–9 months compared with 3–4 months in non-responders [78]. This responder-defined survival comparison was exploratory and cannot establish a causal relationship between vaccine-induced immunity and the clinical outcome, particularly given the potential for time-dependent selection and prognostic confounding.

A subsequent phase I study combining GV1001, GM-CSF, and gemcitabine in 28 patients with advanced PDAC detected vaccine-specific immune responses in four of six evaluable patients, although these responses were generally weak and transient [79]. Clinical efficacy was not established.

However, in the later randomised phase III TeloVac (NCT00425360) trial, 1062 treatment-naïve patients with locally advanced or metastatic PDAC were assigned to gemcitabine/capecitabine alone, sequential chemoimmunotherapy, or concurrent chemoimmunotherapy with GV1001. The addition of GV1001 did not improve OS compared with chemotherapy alone. These findings indicate that telomerase vaccination with GV1001 is immunogenic, but has not demonstrated clinical benefit in unselected advanced/metastatic PDAC [80].

5.3. Whole Tumour Cell-Based Vaccines

Whole tumour cell-based vaccines deliver a broad repertoire of tumour antigens, aiming to promote epitope spreading, enhance antigen presentation, and modulate the TME.

5.3.1. GVAX

GVAX is composed of two genetically engineered, irradiated allogeneic pancreatic tumour cell lines that secrete granulocyte–macrophage colony-stimulating factor (GM-CSF) which promotes dendritic cell recruitment and antigen presentation.

In a phase I trial involving 14 patients with resected stage II/III PDAC, escalating doses of GVAX were administered in the adjuvant setting [81]. A subset of patients (n = 3) developed DTH responses and remained disease-free for 25–30 months after diagnosis, suggesting an association between vaccine-induced immune activation and prolonged DFS. However, this small responder-defined analysis cannot establish a causal relationship between vaccination and the outcome.

In a perioperative platform trial (NCT02451982) in patients with resectable PDAC, GVAX with low-dose cyclophosphamide was evaluated alone or in combination with Nivolumab, with or without Urelumab, before and after surgical resection [82]. Patients received GVAX/cyclophosphamide alone (Arm A; n = 16), GVAX/cyclophosphamide plus Nivolumab (Arm B; n = 14), or GVAX/cyclophosphamide plus Nivolumab and Urelumab (Arm C; n = 10). The triplet regimen met its primary immunologic endpoint by significantly increasing intratumoral CD8⁺ CD137⁺ T cells compared with GVAX/Nivolumab. Median DFS and OS were 13.90 and 23.59 months in Arm A, 14.98 and 27.01 months in Arm B, and 33.51 and 35.55 months in Arm C, respectively. However, survival differences were not statistically significant because of the small sample size.

In the metastatic setting, a phase I trial (NCT00836407) in previously treated PDAC evaluated Ipilimumab with or without GVAX [83]. Among 30 patients, median OS was numerically longer with GVAX plus Ipilimumab than with Ipilimumab alone (5.7 vs. 3.6 months). However, this was a small early-phase study, and the survival difference was not statistically significant.

In contrast, a randomised phase II trial (NCT01896869) evaluated GVAX in combination with Ipilimumab as maintenance therapy vs. continuation of FOLFIRINOX in patients with metastatic PDAC who achieved stable disease or response after first-line FOLFIRINOX [84]. The study was stopped early for futility after interim analysis; median OS was 9.38 months with GVAX plus Ipilimumab versus 14.7 months with continued FOLFIRINOX, indicating no benefit from this maintenance immunotherapy approach.

5.3.2. Algenpantucel-L

Algenpantucel-L is composed of two irradiated allogeneic human PDAC cell lines (HAPa-1 and HAPa-2) genetically modified by retroviral transduction to express murine α(1,3)-galactosyltransferase (αGT), resulting in surface αGal epitopes and intended induction of hyperacute rejection-type immune responses. Despite its immunological rationale, Algenpantucel-L has not demonstrated a survival benefit in randomised phase III trials in PDAC.

In a phase II trial (NCT00569387) of 70 patients with resected PDAC, Algenpantucel-L added to standard adjuvant therapy yielded a 12-month DFS rate of 62% and a 12-month OS of 86% after a median follow-up of 21 months [85]. However, in the subsequent adjuvant phase III IMPRESS trial (NCT01072981), addition of Algenpantucel-L to standard therapy did not improve OS; median OS was 27.3 months with Algenpantucel-L plus standard therapy versus 30.4 months with standard therapy alone [86].

Similarly, in the phase III HyperAcute-Pancreas trial (NCT01836432) in patients with borderline resectable or locally advanced unresectable PDAC, the addition of Algenpantucel-L to standard neoadjuvant chemotherapy and chemoradiation did not improve outcomes. Median OS was 14.9 months with standard therapy versus 14.3 months with Algenpantucel-L, and median PFS was 13.4 months versus 12.4 months, respectively [87].

5.4. Microenvironment and Stroma-Targeted Vaccine Strategies

Beyond tumour cell-intrinsic antigens, some vaccine strategies in PDAC target components of the TME, including tumour vasculature and stromal support cells with the aim of disrupting tumour-supportive vascular and stromal processes.

DNA-based vaccine platforms have been used to explore this strategy. Preclinical studies in PDAC have investigated targets including fibroblast activation protein-α (FAPα) expressed by CAFs, as well as tumour-associated antigens such as survivin and α-enolase/ENO1 [88,89,90].

Clinical translation of this concept has been most advanced with VXM01, an oral bacterial DNA vaccine consisting of live attenuated Salmonella typhi Ty21a carrying a plasmid encoding vascular endothelial growth factor receptor 2 (VEGFR2). In a randomised, double-blind, placebo-controlled phase I dose-escalation trial (NCT01486329) in patients with locally advanced inoperable or metastatic PDAC, VXM01 was administered with gemcitabine-based standard therapy. VXM01 induced VEGFR2-specific T-cell responses and was associated with a significant reduction in tumour perfusion at day 38, assessed by dynamic contrast-enhanced magnetic resonance imaging [91]. A subsequent prime-boost extension study evaluated repeated VXM01 vaccination [92]. Among 16 patients who received at least one boost vaccination, eight developed pronounced VEGFR2-specific T-cell responses, defined as at least a three-fold increase over baseline, further supporting the immunogenicity of VXM01. However, these studies primarily provide proof-of-mechanism for vascular-targeted vaccination in PDAC, and clinical efficacy has not been established.

6. Next-Generation Biomarkers for PDAC Immuno-Oncology

The development of predictive biomarkers for PDAC immunotherapy will likely require integration of dynamic and spatial measures with conventional tumour genomics. Circulating tumour DNA (ctDNA) is particularly relevant in resected disease, where postoperative detection can identify molecular residual disease and is associated with a high risk of recurrence [101]. Tumour-informed assays may therefore provide a means of enriching adjuvant or MRD-directed immunotherapy studies for patients with residual molecular disease and of measuring molecular response through ctDNA clearance or persistence [102]. Importantly, however, postoperative ctDNA is currently better established as a prognostic than a predictive biomarker in PDAC, and prospective evidence is required before ctDNA status can be used to determine whether a patient should receive a specific immune therapy [101,102].

Spatial profiling may address a complementary limitation of bulk genomic and transcriptomic analyses by preserving the relationships between malignant, stromal and immune compartments that are central to PDAC immune resistance. Single-cell and spatial transcriptomic studies have identified heterogeneous malignant, fibroblast and immune states and treatment-associated changes within the PDAC microenvironment [103]. Multiplex immunofluorescence and spatial proteomic analyses have similarly demonstrated marked regional heterogeneity in lymphoid and myeloid immune populations within PDAC, with spatial immune patterns associated with the clinical outcome [104]. More recent large-scale spatial profiling has identified recurring antigen-presenting-cell, helper-T-cell and cytotoxic-T-cell neighbourhoods associated with T-cell expansion and the clinical outcome [105]. Together, these technologies provide a framework for biomarkers based not only on the presence of immune cells but on their functional state, proximity and organisation within the tumour.

The principal challenge is to translate these high-dimensional measurements into reproducible biomarkers suitable for prospective trials. Future studies should therefore integrate tissue genomics and antigen-presentation status with spatial immune and stromal features and circulating measures such as ctDNA, while prospectively defining assay methodology, thresholds and clinical endpoints. Such integrated biomarkers may ultimately distinguish patients with an immunologically permissive tumour from those requiring additional priming, trafficking or microenvironmental remodelling and may be particularly valuable for mechanism-based early-phase trials.

The potential positioning of immune-based strategies across different stages of the PDAC disease continuum, together with relevant biomarker approaches, is summarised in Figure 2.

Figure 2. Potential positioning of immuno-oncology strategies across the PDAC disease continuum. Immune-based strategies may have different biological rationales according to disease setting and tumour burden. Perioperative and minimal residual disease (MRD) settings may be particularly suitable for vaccine and immune-priming approaches, whereas established metastatic disease may require strategies that overcome stromal exclusion, myeloid suppression, T-cell dysfunction and antigen heterogeneity or directly supply or redirect tumour-reactive effector cells. MSI-H/dMMR disease represents the principal currently actionable biomarker-defined setting for immune checkpoint blockade. Cross-cutting biomarker approaches include tumour genomics and HLA/antigen-presentation status, target-antigen expression, circulating tumour DNA and spatial immune/stromal profiling. Placement within the figure reflects biological rationale and current clinical development and should not be interpreted as a treatment recommendation; except for biomarker-selected checkpoint blockade, the approaches shown remain investigational in PDAC.

7. Lessons from Failed and Emerging Immunotherapy Strategies in PDAC

Across PDAC immuno-oncology studies, repeated therapeutic failures can be mapped to five recurrent and interacting biological barriers: (1) limited antigenicity and low neoantigen burden in most MSS disease; (2) inadequate antigen presentation and productive T-cell priming; (3) physical and chemokine-mediated T-cell exclusion produced by heterogeneous stroma; (4) myeloid-dominant immunosuppression; and (5) intratumoural T-cell dysfunction driven by persistent inhibitory and metabolic signals. These barriers are partially redundant and operate at different stages of the cancer–immunity cycle. Consequently, correcting one component can produce measurable pharmacodynamic activity without restoring effective antitumour immunity or generating durable clinical benefit. This recurring pattern explains why checkpoint blockade, stromal depletion, myeloid targeting, innate immune stimulation and vaccination can alter their intended biological target yet fail clinically. The central translational challenge is therefore not simply to identify additional immune agents, but to determine which biological barrier is dominant in an individual tumour and design biomarker-enriched, mechanism-led combinations accordingly. These lessons and their implications for future PDAC immunotherapy trial design are summarised in Table 2.

Table 2. Conceptual lessons for future immuno-oncology trial design in PDAC.

Conceptual Lesson	Rationale from Prior Studies	Trial-Design Implication
Biomarker selection is necessary but not sufficient	MSI-H/dMMR PDAC demonstrates that molecular selection can identify responders, but responses remain variable and MSI/MMR status alone does not capture antigen presentation, tumour mutational burden, immune contexture or primary resistance mechanisms	Use integrated biomarkers combining genomic, transcriptomic, spatial immune and circulating markers rather than single-analyte selection alone
Disease setting may be as important as therapeutic platform	Vaccines and immune-priming strategies may require time for T-cell expansion and may be disadvantaged in bulky metastatic disease with entrenched immune suppression	Prioritise perioperative, adjuvant and MRD-directed studies, especially where ctDNA or tumour-marker positivity identifies high recurrence risk
Immune resistance in PDAC is redundant and layered	PDAC immune failure reflects overlapping stromal, myeloid, chemokine, metabolic and T-cell-intrinsic barriers, so blockade of a single pathway often produces pharmacodynamic activity without durable clinical benefit	Design mechanism-based combinations that coordinate antigen release, antigen presentation, T-cell priming, trafficking, effector function and resistance prevention
Stromal targeting should remodel rather than ablate	Indiscriminate stromal depletion has not reliably improved survival and may disrupt tumour-restraining stromal functions	Develop stromal-normalising approaches with embedded pharmacodynamic, vascular, spatial and immune endpoints to confirm productive remodelling
Pharmacodynamic activity should not be mistaken for efficacy	Several agents have shown immune activation, altered myeloid populations or increased T-cell infiltration without clear survival benefit	Require early paired-tissue, blood and imaging biomarkers, but advance regimens only when immune modulation aligns with objective clinical endpoints
Cellular therapies need PDAC-specific engineering solutions	TIL, TCR-T and CAR-based approaches can provide proof-of-concept activity, but remain limited by antigen heterogeneity, HLA restriction, trafficking, persistence, manufacturing feasibility and toxicity	Incorporate antigen/HLA selection, trafficking-enhancement strategies, resistance to suppressive TME signals and scalable manufacturing into early trial design
Endpoints should match the biological mechanism	ORR and PFS may underestimate benefit for some immune approaches, while immune-response endpoints alone may overstate clinical relevance	Pair conventional endpoints with mechanism-specific measures such as ctDNA clearance, relapse-free survival, immune-response durability, spatial T-cell infiltration and antigen-spreading analyses
Combination timing and sequencing require prospective testing	Chemotherapy, radiotherapy, stromal remodelling, checkpoint blockade, vaccines and cellular therapies may interact differently depending on sequence, tumour burden and immune state	Build trials around explicit sequencing hypotheses rather than adding immunotherapy empirically to chemotherapy backbones
Negative trials should inform biology, not just efficacy	Failed or modest studies in PDAC often reveal barriers such as compensatory myeloid recruitment, inadequate trafficking, antigen loss or absence of productive T-cell priming	Embed translational endpoints that explain resistance and identify which patients, disease settings or combinations should be pursued next

First, biomarker selection is essential but incomplete. MSI-H/dMMR PDAC illustrates that molecularly selected patients can respond to checkpoint blockade, yet responses remain variable even within this enriched subgroup. MSI/MMR status alone does not capture tumour mutational burden, antigen-presentation competence, immune contexture, tumour burden, prior treatment exposure or mechanisms of primary resistance. Future trials should therefore integrate genomic, transcriptomic, spatial immune and circulating biomarkers, including tumour-informed ctDNA where appropriate.

Second, the disease setting may be as important as the therapeutic platform. Vaccines, immune-priming strategies and cellular therapies may be disadvantaged in bulky metastatic PDAC, where immune suppression is already entrenched. Perioperative, adjuvant and minimal residual disease-directed settings may offer a more favourable therapeutic window, with lower tumour burden, greater immune competence and more time for durable immune surveillance to develop.

Third, single-pathway immune modulation is unlikely to overcome the layered resistance of PDAC. Stromal targeting, innate immune activation, FAK inhibition, CD40 agonism, myeloid-cell targeting, chemokine-axis blockade and TGF-β inhibition have each shown mechanistic rationale and, in some cases, pharmacodynamic activity. However, inconsistent efficacy suggests that targeting one suppressive axis is often insufficient. Future combinations should be mechanism-led, coordinating antigen release, antigen presentation, T-cell priming, trafficking, effector function and resistance prevention.

Fourth, pharmacodynamic activity should not be equated with clinical efficacy. Increased T-cell infiltration, reduced suppressive myeloid populations, stromal remodelling or vaccine-induced T-cell responses are important proof-of-mechanism signals, but they do not guarantee durable tumour control. Translational endpoints should therefore be paired with clinically meaningful outcomes such as ctDNA clearance, recurrence-free survival, response durability, PFS and OS.

Finally, cellular therapies demonstrate both the promise and constraints of directly supplying tumour-reactive immune effectors. TIL, TCR-engineered T-cell, CAR-T and CAR-NK approaches can bypass inadequate endogenous priming and have shown proof-of-concept activity in selected patients. However, broader application in PDAC remains limited by HLA restriction, antigen heterogeneity, trafficking failure, limited persistence, suppressive TME signals, toxicity and manufacturing complexity. Progress will require PDAC-specific engineering and patient-selection strategies rather than simple extrapolation from more immune-responsive malignancies.

Together, these lessons support a shift from empiric immunotherapy combinations toward mechanism-led trial design. Future PDAC immunotherapy studies should prioritise biomarker-enriched populations, earlier disease settings, rational sequencing, pharmacodynamic confirmation of target engagement and endpoints matched to the biological mechanism of the intervention.

8. Conclusions and Future Perspectives

PDAC remains one of the most treatment-refractory solid malignancies, with poor outcomes driven by late presentation, early metastatic dissemination, therapeutic resistance and a profoundly immunosuppressive TME. Although immuno-oncology has transformed the management of several solid tumours, its impact in PDAC has been limited. The clearest exception is the rare MSI-H/dMMR subgroup, in which immune checkpoint blockade can produce clinically meaningful and sometimes durable responses. For the majority of patients with microsatellite-stable disease, however, immune-based therapies remain investigational and have not yet established a routine therapeutic role.

Despite this limited clinical impact, the field has advanced substantially. Trials of checkpoint inhibitors, stromal and vascular modifiers, innate immune agonists, myeloid- and chemokine-directed therapies, adoptive cellular therapies and cancer vaccines have clarified why PDAC is resistant to conventional immunotherapy. Negative studies have been particularly informative, demonstrating that single-mechanism approaches, including checkpoint blockade, stromal depletion or vaccination alone, are unlikely to overcome the layered and redundant immune-resistance networks that characterise PDAC.

Emerging signals nevertheless support cautious optimism. MRD-directed KRAS and personalised neoantigen vaccines, KRAS-directed TCR-engineered T cells, CLDN18.2-directed CAR-T cells, tumour-penetrating agents, FAK-mediated stromal remodelling and selected myeloid or chemokine-axis approaches suggest that immune-based strategies may be most effective when applied in biologically selected patients, lower-burden disease settings and rational combinations. These data argue against abandoning PDAC immunotherapy but strongly support a shift away from empiric combinations toward mechanism-led development.

Future trials should prioritise integrated biomarker strategies that combine tumour genomics, antigen-presentation status, immune contexture, stromal and myeloid features, and circulating markers such as tumour-informed ctDNA. Earlier-stage, perioperative, adjuvant and MRD-positive settings may offer a more favourable therapeutic window than bulky metastatic disease, where antigen heterogeneity, T-cell dysfunction, stromal exclusion and suppressive myeloid populations are more entrenched. Trial endpoints should also be aligned with biological mechanism, incorporating ctDNA clearance, recurrence-free survival, immune-response durability, spatial immune remodelling and conventional efficacy outcomes.

Durable benefit in PDAC is unlikely to be achieved through single-agent immunotherapy alone. Successful strategies will probably need to coordinate antigen release and presentation, tumour-specific T-cell generation or expansion, stromal and myeloid remodelling, immune-cell trafficking, effector function and prevention of adaptive immune escape. Early-phase signals should be interpreted cautiously, particularly when derived from small, single-arm or mixed-tumour cohorts, and require validation in adequately powered studies with embedded translational endpoints.

Ultimately, the successful integration of immuno-oncology into PDAC treatment will depend on matching the right immune strategy to the right biological context, disease stage and biomarker-defined patient population. While major challenges remain, accumulating evidence suggests that immunotherapy may yet assume a meaningful role in PDAC—not as a broadly applicable single-agent approach, but as a selective, biomarker-guided and combination-based component of future treatment strategies.