Lymphatic metastasis in pancreatic cancer: from bedside to bench and back

Abstract

Pancreatic cancer is an extremely aggressive and lethal malignant cancer; almost half of the patients have distant metastasis when first diagnosed, and many challenges arise during treatment. Lymph node metastasis is a vital indicator of disease progression in pancreatic cancer and directly influences patient prognosis and survival rate. The cancer cells invade the lymphatics, promote the growth of lymphatics, propagate to lymph nodes, and finally settle down in lymph nodes. Aiming at this process, treatment modalities involving surgery, targeted therapy, and nanodelivery of chemotherapy are tested regarding their efficacy in both local tumor management and lymph node metastasis. As lymph node metastasis presents a poor prognosis and higher recurrence rate, precise assessment of lymph node metastasis is imperative for clinical management in pancreatic ductal adenocarcinoma. Precise evaluation of lymph node metastasis will not only help in formulating personalized treatment but also provide more accurate prognostic information to the patients. This review aims at synthesizing current knowledge of the molecular mechanisms, treatment options, clinical implications, and evaluation techniques regarding lymph node metastasis in pancreatic cancer, hence allowing a focused perspective for clinicians and researchers in striving for more effective management strategies against lymph node metastasis of pancreatic cancer.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is considered as one of the most difficult cancers worldwide due to early lymphatic metastasis, high recurrence rate, and poor prognosis. It is estimated that there will be about 66,440 new cases of PDAC and approximately 51,750 deaths due to PDAC in 2024 [1]. It has one of the poorest 5-year survival rates among different cancers. The 5-year relative survival of PDAC patients in the USA, according to the Surveillance, Epidemiology and End Result Program, was 12.8% from 2014 to 2020. Such poor prognosis is due to most patients already being at late stage when firstly diagnosed. It was reported that more than 65% of the patients had lymph node metastasis (LNM) at the time of diagnosis, and LNM has been confirmed to be one of the major predictive factors for PDAC [2–5]. Until 2010, according to the American Joint Committee on Cancer (AJCC) 7^th edition, the N stage of PDAC is divided based on whether lymph nodes have metastasized or not, where N0 means negative metastasis and N1 indicates positive metastasis. In recent years, however, many large-sample clinical studies have demonstrated that the count of metastatic lymph nodes is an important factor affecting prognosis. In consideration of these research outcomes, according to the number of metastatic regional lymph nodes, AJCC 8^th edition divided the original N1 into N1 and N2 in 2016, with N1 corresponding to 1–3 nodes and N2 corresponding to 4 or more nodes.

Since the patients without LNM (N0) have significantly better disease-free survival (DFS) and overall survival (OS) compared to the ones with LNM (N1/N2) [6], positive lymph nodes have been found to potentially serve as a reservoir for cancer cells to seed into distant organs according to some studies [7–9]. Therefore, it is critical for us to understand the lymphatic metastatic mechanisms of PDAC and find strategies to increase the survival rate of the patients.

As noted above, LNM is an important step for tumors to affect the prognosis of the patient. Generally, cancer cells attracted by the gradient of chemokines or interstitial fluid pressure invade the lymphatic vessels first. Then, with the flow of lymphatic fluids, they reach the tumor-draining lymph node (TDLN) and create an immunosuppressive niche in TDLN [10–12]. In this review, we present a summary of the mechanisms of LNM, the clinical significance of LNM for patients, therapeutic strategies targeting LNM, and newly developed approaches for the detection and prediction of LNM in PDAC.

Mechanisms of lymph node metastasis in PDAC

As LNM is very common for PDAC, understanding how the tumor cells migrate from the primary site to the TDLNs and how they have escaped the immunosurveillance originally presenting in lymphatic ducts and lymph nodes is very important. This section will discuss this complicated process with several recent research progresses.

Entry into the lymphatic vessels

Function of interstitial fluid pressure gradient

LNM of PDAC is strongly associated with the unique role of interstitial fluid pressure (IFP) gradient [10] (Fig. 1). It has been proven that the elevated IFP is a characteristic feature in the central region of PDAC [13]. There is much evidence that high IFP is related to LNM in PDAC models [14, 15]. By using a computer-based pancreatic model, researchers found small cancer‐cell clusters, which continuously detached from the primary tumor under high IFP, creating a natural basis for aggressive invasion into the lymph nodes and the venous system [14]. The high permeability of abnormal blood vessels within the tumor leads to the continuous accumulation of interstitial fluid, significantly increasing the interstitial fluid pressure in the tumor center [16]. However, the key factor driving metastasis is not the uniformly high pressure in the core, but the steep IFP gradient formed at the tumor margin [17]: the high pressure in the core causes the lymphatic vessels to collapse, losing their transport function, while the interstitial fluid in the margin flows along the hyaluronic acid (HA) gel network towards the peripheral low-pressure areas, creating a directional driving force that directly propels tumor cells into the lymphatic vessels [18].

Fig. 1.

Mechanism of interstitial fluid pressure gradient promoting lymph node metastasis in pancreatic cancer

This process is accompanied by multiple synergistic mechanisms—interstitial flow carries chemokines such as CCL21, forming a chemical gradient that guides tumor cells towards lymphatic vessels [19]; under the mechanical stress of the IFP gradient, lymphatic endothelial gaps widen, providing a channel for cell invasion, and the low-flow environment (< 1 µm/s) within the lymphatic vessels significantly reduces shear stress on tumor cells, enhancing their survival; high IFP also activates the PI3K-AKT pathway to induce macrophage polarization towards the M2 phenotype, which secretes factors like CCL2 and IL-10 to suppress T cell function, while the hypoxic microenvironment cooperates to upregulate PD-L1, forming an immune escape “umbrella” and promoting tumor cell invasion and metastasis [20]; furthermore, the viscoelasticity of HA gel activates the YAP/TAZ signaling axis in tumor cells through integrin receptors, driving epithelial-mesenchymal transition (EMT) and an invasive phenotype [21]. Targeting the multidimensional properties of IFP, many therapeutic strategies are developed (Fig. 1). First, excess stromal HA is a major driver of high IFP in PDAC, collapsing vessels and impeding drug delivery. In a preclinical study, Provenzano et al. showed that enzymatic HA depletion by using PEGPH20 (a HA-degrading enzyme) normalized IFP, reopened tumor vessels, and dramatically improved chemotherapy efficacy and survival in PDAC models [22]. However, the HALO-301 phase III trial found no improvement in overall survival by adding PEGPH20 to gemcitabine/nab-paclitaxel (median ~ 11.2 vs. 11.5 months, HR = 1.0) [23]; possible reasons could be the positive effects of tumor stroma to restrict the growth and progression of tumor despite removing stroma that could promote the delivery of drugs. Second, anti-angiogenic therapy was hypothesized to normalize vessels and reduce IFP. Classic experiments by Jain et al. showed that a single dose of a VEGFR-2 antibody (DC101) in tumor-bearing mice could reduce the interstitial fluid pressure within days, temporarily restoring perfusion and improving response to radiation [24]. Despite this rationale, clinical trials adding bevacizumab (anti-VEGF) to chemotherapy in PDAC failed to prolong survival—a meta-analysis found no OS benefit (HR ~ 1.01) from anti-VEGF across studies [25]. This could be the excessive vessel pruning aggravating hypoxia in tumors, leading to more tumor aggressiveness [26]. Finally, high IFP and matrix stiffness activate mechanosignaling in both cancer cells and stromal cells, such as YAP/TAZ signaling. Verteporfin, a YAP-TEAD complex inhibitor, which in PDAC cell lines (e.g., PANC-1, SW1990) induces G1-phase arrest, downregulates Cyclin D1/E1, upregulates Bax, cleaves PARP, and induces apoptosis, and in xenograft models reduces tumor growth [27]. These interventions aim to disrupt the physical-biochemical network promoting metastasis by reducing interstitial viscoelasticity, repairing vascular leakage, destroying chemotaxis guidance, or reversing immune suppression. Future research should further explore the contribution of HA and collagen to the heterogeneity of IFP, the dynamic changes of the gradient, and the interaction mechanisms with lymphangiogenesis to optimize the spatiotemporal precision of targeted therapies.

Chemokines

Many chemokines are implicated in the LN metastasis of PDAC cells that are originally used by leukocytes homing in the local microenvironment [10] (Fig. 2). CCR7 was significantly upregulated in CD133+ pancreatic cancer stem-like cells, promoting its interaction with CCL21 produced by lymphatic endothelial cells (LEC), thereby increasing the LNM [19]. The CXCR4/CXCL12 (on PDAC cells/produced by LECs, respectively) axis is also observed in PDAC to increase the invasiveness to LNs [28, 29]. LECs also secrete various other chemokine ligands like CXCL10, CXCL1, and CCL5, which bind to CXCR3, CXCR2, and CCR5, respectively, on tumor cells, thus influencing tumor cell migration through lymphatic vessels to LNs in many other cancer types [30]. Additionally, the entry of tumor cells into LNs is controlled by the LEC located at the lymph node subcapsular sinus (SCS). The CCL1 protein, which is specifically expressed in the SCS but not peripheral lymphatics, interacts with CCR8 strongly expressed on melanoma cells; inhibiting CCR8 causes retention of cancer cells in the collecting vessels before the SCS area [31].

Fig. 2.

Lymphangiogenic factors and how cancer cells damage LECs. Cancer cells invade the adjacent lymphatic through the interstitial fluid flow and express ALOX15/COX-2, which catalyze arachidonic acid into 12S-HETE/PGE-2, to destroy the integrity of LECs. Once cancer cells got into lymphatics, the upregulation of CCR7/CXCR4 enabled them binding to the corresponding ligands, CCL21/CXCL12, respectively, which aids cancer cells migrate through lymphatic lumen. Many lymphangiogenic factors can be released by cancer cells, such as VEGFC, PDGF, and FGF, resulting in LEC proliferation and migration

Lymphatic permeability

The loss of lymphatic permeability also plays a contributory role in the metastasis of PDAC (Fig. 2). Through distinct mechanisms, lymphatic vessel dilation disrupts LEC integrity, which helps tumor cells penetrate nearby lymphatic vessels [32]. It has been found that lymphatic vessels draining the tumors have an enlarged and dilated size [33]. This is partly due to increased lymph flow resulting from high IFP, but COX-2 also plays a role. Overexpression of COX-2 in PDAC cells leads to high levels of prostanoids within the tumor micro-environment; among prostanoids, prostaglandin E2 (PGE-2) is known to elicit tumorigenic effects related to EC migration [34]. PGE-2 can also be secreted from LECs themselves following VEGF-D stimulation, causing peritumoral lymphatic capillary dilation [35]. Tumor cells can directly damage LEC integrity via the tumor-derived arachidonic acid metabolite 12[S]-hydroxy-eicosatetraenoic acid (12S-HETE), synthesized by 15-lipoxygenase-1 (ALOX15) expressed in human mammary carcinoma cells. The knockdown or drug suppression of ALOX15 can restore LEC integration [36]. We suppose that pancreatic cancer cells may also possess a similar ability to directly damage LEC.

Lymphangiogenesis

Lymphangiogenesis is an indispensable part of the LNM process in PDAC. It has been widely reported that lymphatic vessel density may be considered a useful predictor for poor prognosis and the occurrence of lymph node metastasis [37, 38]. Many lymphangiogenic factors are produced by pre-metastatic tumors to induce the growth of lymphatic vessels.

The vascular endothelial growth factor (VEGF) family consists of secreted glycoproteins that are essential to the proliferation and migration of endothelial cells (Fig. 2). Among them, VEGF-A has been shown to be mainly associated with angiogenesis [39]. VEGF-C and its receptor, VEGFR3, expressed on LECs, form an important axis in promoting lymphangiogenesis in PDAC; suppression of this VEGF-C/VEGFR3 axis has inhibited the lymphangiogenesis and early lymphatic dissemination in PDAC [40–42]. Beyond that, VEGF-C released by tumor cells can cause structural changes in lymphatic vessels, such as LEC hypertrophy and increased vessel lumen diameter in peripheral lymphatics and sentinel lymph nodes (SLN) [43]. Lymphangiogenesis can also be induced by platelet-derived growth factor-BB, secreted by tumor cells, through its binding to PDGFR, which activates LECs via MAPK pathways, leading to elevated levels of phosphorylated Src, Erk, and Akt in LECs [44]. Fibroblast growth factor (FGF) can upregulate intracellular HK2 expression via the c-MYC-dependent pathway, while VEGF-C does so through the PI3K/Akt/mTOR/HIF-α axis, resulting in lymphangiogenesis. Double inhibition of FGFR and VEGFR has been shown to suppress progression and metastasis in intrahepatic cholangiocarcinoma [45]. The similar function of PDGF/PDGFR and FGF/FGFR in PDAC needs more further researches.

Survive in sentinel lymph node

Formation of premetastatic niche

The sentinel lymph node (SLN) is the first lymph node to which cancer cells are most likely to spread from a primary tumor. Before the tumor successfully infiltrates the SLN, it secretes numerous soluble tumor-derived factors, including proteins and extracellular vesicles. These factors alter the SLN microenvironment, creating a supportive niche that facilitates the proliferation and subsequent dissemination of tumor cells [10] (Fig. 3).

Fig. 3.

Cancer cells survive in the SLN. Cancer cells release many factors, such as extracellular vesicles and VEGFC, to construct pre-metastatic niche before arriving in LNs. When cancer cells entered LN, descending expression of MHC-I and elevating of PD-L1 could facilitate them evade immune clearance caused by CD8+ T cell. M2 Macrophage can enhance the progression of cancer cells by IL-1β self-amplifying loop. Cancer cells can also impair FRC directly, leading to the apoptosis of T cell due to decreasing survival signal released by FRC. Some resident DC in LN could participate in formation of immunosuppressive niche by training T cells to Tregs

Many VEGFs produced by primary tumor cells play a role in establishing the premetastatic niche. As discussed, VEGFs contribute to lymphangiogenesis and similarly influence the SLN by expanding its lymphatic networks. Cancer cell-derived extracellular vesicles, including exosomes, also participate in premetastatic niche formation. For instance, exosomes secreted by pancreatic cancer cells downregulate the expression of the intracellular long non-coding RNA (lncRNA) ABHD11-AS1 in lymphatic endothelial cells (LECs), promoting tubular formation and lymphangiogenesis [46]. In melanoma, cancer-derived extracellular vesicles (EVs) spread through the lymphatic system and selectively bind to CD169(+) macrophages in the subcapsular sinus (SCS) of TDLNs in both mice and humans, which initially restricts EV dissemination. However, as the tumor progresses, this macrophage barrier is impaired, allowing EVs to propagate into the TDLN cortex, thereby contributing to the formation of a premetastatic niche [47].

High endothelial venules (HEV) are specialized blood vessels found within lymph nodes, playing a key role in the immune response by guiding naïve lymphocytes from the bloodstream into the lymph nodes [48]. In SLNs, HEVs have been shown to be crucial in establishing a premetastatic niche [49]. The HEV endothelial cells are typically thick and tall [50]. However, under tumorous conditions—even before cancer cells arrive—these endothelial cells thin out, enlarging the HEV lumen. This morphological change has been shown to increase blood flow, preparing for tumor cell arrival, thus shifting HEV function from an immune mediator to a facilitator of tumor metastasis [51, 52]. In PDAC, MECA79-coated Taxol nanoparticles (MECA79-Taxol-NPs), which target the peripheral node addressin (PNAd) in HEVs, significantly reduced tumor size in primary sites when administered intravenously and could also have similar effects on lymph node sites but need further exploration [52, 53]. Other studies also underscore HEVs’ importance in facilitating the arrival of cancer cells in SLNs [54–56].

Tumor cells interact with tumor microenvironment to escape immunosurveillance

PDAC is notorious for its robust immune evasion mechanisms within both the tumor microenvironment and TDLNs, which leads to limited use of immune checkpoint inhibitors (ICI) in the clinic. Here, we will detail how PDAC cells shape an immunosuppressive microenvironment to survive in TDLNs.

LNs contain numerous immune and stromal cells that provide strong immune pressure to prevent tumor metastasis. However, these cells can be reprogrammed into an immune-suppressive environment that permits tumor growth and survival (Fig. 3). Pancreatic cancer cells, aiming to circumvent immune surveillance, have developed mechanisms to evade elimination by lymphocytes. One such strategy involves the loss of major histocompatibility complex (MHC) expression. It is common to find downregulation of MHC-I expression in pancreatic cancer cells [57, 58], partly because MHC-I molecules are directed toward lysosomal degradation via an autophagy-dependent pathway [59]. Research has shown that glucocorticoid receptor (GR) signaling upregulates PD-L1 expression and downregulates MHC-I expression in PDAC through transcriptional regulation. Inhibiting GR reduces PD-L1 levels and increases MHC-I expression, subsequently enhancing cytotoxic T cell infiltration and activity. This strengthens anti-tumor immunity and helps overcome resistance to immune checkpoint blockade (ICB) therapy [60]. Interestingly, MHC-II expression on PDAC cell surfaces has been found to increase CD4+ and CD8+ T cell cytotoxicity against these cells, highlighting a potential target for neo-antigen-based immunotherapy [61]. PDAC cells also exhibit high PD-L1 expression, which leads to T cell exhaustion and regulatory T cell development [62]. PDAC patients with high PD-L1 expression tend to have shorter DFS and OS, and increased PD-L1+ T cell infiltration in the tumor microenvironment is associated with advanced N and TNM stages [63, 64].

In SLN, macrophages can be polarized to an M2-like, anti-inflammatory phenotype, known as tumor-associated macrophages (TAMs), which enhance cancer cell viability. TGF-β-induced protein (TGFBI), identified through single-cell RNA sequencing as a factor released by TAMs, is strongly associated with PDAC growth. Depleting TGFBI via siRNA in vitro or using a Cre-Lox strategy in vivo has been shown to inhibit macrophage polarization to the M2 phenotype, thus promoting an anti-tumor immune response against PDAC [65]. Exosomes derived from PDAC cells containing FGD5-AS1 can also induce M2 macrophage generation through the STAT3/NF-κB pathway, contributing to PDAC malignancy. FGD5-AS1 interacts with p300, leading to STAT3 acetylation, which enhances STAT3/NF-κB nuclear localization and transcriptional activity [66]. By using single-cell and spatial genomics, researchers identified a subset of TAMs expressing interleukin-1β (IL-1β), induced by local prostaglandin E2 (PGE2) and tumor necrosis factor (TNF). These TAMs, located near PDAC cells, produce IL-1β, which targets IL-1R on the surface of nearby PDAC cells, promoting a pro-inflammatory response that releases TNF, PGE2, and other factors, reinforcing IL-1β+ TAM formation. This self-amplifying loop contributes to pancreatic tumorigenesis and progression [67].

Dendritic cell (DC), as professional antigen-presenting cells (APCs), plays a crucial role in the tumor microenvironment (TME). In PDAC, DC deficiency is associated with impaired immune surveillance, while restoring DC antigen-presenting function in advanced PDAC has been shown to revive tumor-restraining immunity [68]. In studies with caspase-recruitment domain-containing protein 9 (CARD9) knockout mice transplanted with pancreatic cancer cells, a decrease in DC levels was observed due to the suppression of creatine transporter SLC6A8 transcription. Reduced creatine transport into DCs resulted in DC immaturity and weakened anti-tumor immunity [69]. In SLN of breast cancer patients, an immunosuppressive environment is linked to a suppressed state of LN-resident DCs. This suppressed state promotes immune tolerance by recruiting myeloid-derived suppressor cells (MDSCs) and inducing T cell anergy, leading to an increase in regulatory T cells (Tregs) and exhausted T cells [70].

Other immune cells within the tumor microenvironment (TME) have also been shown to play significant roles in PDAC. A subset of B lymphocytes residing in lymph nodes, known as regulatory B cells (Bregs), has been identified for its immunosuppressive function. These Bregs release anti-inflammatory cytokines such as IL-10, IL-35, and TGF-β, which impair the anti-tumor immune response and promote tumor growth and metastasis in PDAC [71]. Additionally, natural killer (NK) cells infiltrating PDAC exhibit signs of dysfunction. This is marked by reduced cytotoxic activity, downregulation of surface markers CD16 and CD57, and decreased expression of activating proteins DNAM-1 and NKP30, which further diminish their ability to combat the tumor [72].

Beyond that, in SLN stroma, especially fibroblastic reticular cell (FRC), fibroblastic reticular cells (FRCs) play a critical role in supporting tumor survival. FRCs are specialized fibroblasts within lymph nodes (LNs) that produce extracellular matrix (ECM) proteins, forming a fibrillary mesh known as the conduit network system. FRCs are categorized based on their location: T cell zone FRCs (TRCs), B cell follicle and germinal center FRCs (also called follicular DCs or fDCs), medullary FRCs (MedRCs), and perivascular FRCs (PRCs). In healthy conditions, FRCs create a reticular network that maintains LN structure, facilitates antigen transport through the conduit system to DCs and lymphocytes, and exports antibodies and other molecules from the lymphoid compartment [73]. In TDLN affected by melanoma, chronic inflammation disrupts the FRC network, hindering native T cell and migratory DC entry, which allows the tumor to evade immune destruction [74]. In TDLNs from colon cancer, tumor cells can directly damage FRCs, reducing their secretion of IL-7. This decline in IL-7 weakens survival signals for T cells, resulting in a decreased T cell population and compromised immune surveillance in TDLNs [75]. However, research related to the role of FRC in PDAC is still limited, and further investigation is required to confirm if FRCs in PDAC function similarly to those in melanoma and colon cancer.

Therapeutic strategies for lymph node metastasis in PDAC

PDAC remains one of the deadliest cancers with persistently low long-term survival rates. However, recent innovations have led to better outcomes and changes in treatment practices over the past decade [76]. LNM is a defining feature of local tumor invasion, acting as an intermediate step in systemic dissemination. Thus, effectively targeting and treating LNM is essential to preventing distant metastasis in pancreatic cancer. The complex anatomy around pancreatic tumors poses challenges in surgically removing all potentially metastatic lymph nodes. Additionally, conventional imaging techniques, such as CT and MRI, often fail to detect metastatic lymph nodes in PDAC. In this context, postoperative lymph node–targeted therapies may help eradicate residual lesions and thereby improve survival in patients with PDAC. Accordingly, this section discusses current and emerging strategies for managing LNM in PDAC.

Lymphadenectomy

Currently, surgical resection remains the most effective treatment for PDAC. With advancements in recent years, the 5-year survival rate has risen to approximately 30% for patients who undergo tumor resection and complete adjuvant chemotherapy successfully [76, 77]. However, a critical question persists: to what extent should surgeons remove lymph nodes during surgery? This section will outline the consensus on lymphadenectomy and discuss the ongoing debate surrounding extended lymphadenectomy (EL).

In 1973, Fortner JG introduced the concept of regional pancreatectomy to improve the long-term outcomes of the traditional Whipple procedure. This approach involves complete tumor removal along with the base of the transverse mesocolon and an extensive margin of soft tissue around the pancreas, including regional lymph nodes [78]. Around the same time, Japanese surgeons developed a similar pancreatectomy procedure that incorporated extended lymphatic tissue clearance. Japanese researchers at the time believed that this radical pancreatectomy with thorough lymph node (LN) clearance was essential for achieving better 5-year survival rates [79, 80]. However, in 1998, a multicenter, prospective, randomized study found that EL did not significantly improve overall survival (OS) compared to standard lymphadenectomy (SL). EL slightly increased the operation time, but not to a statistically significant degree (397 ± 50 min vs. 372 ± 50 min, p > 0.05) [81]. Subsequent clinical reports have supported these findings [82]. More recently, two randomized controlled trials (RCTs) also concluded that EL does not provide survival benefits over SL. In 2021, an RCT reported a 2-year OS rate of 39.5% for EL vs. 25.3% for SL (p = 0.034) [83]. Another RCT in 2022 showed a 3-year survival rate of 27.16% for EL vs. 24.72% for SL (p = 0.717), further suggesting that EL may not provide additional OS benefits compared with SL [84]. Notably, the 2021 RCT reported a higher 2-year OS rate for SL compared to EL. Despite this, EL tends to yield a higher number of positive lymph nodes than SL (2.34 ± 3.46 vs. 1.41 ± 2.12, p = 0.035), which may aid in more accurate lymph node staging for tumors [1, 84]. The TRIANGLE trial (Germany, 270 patients) is now testing whether a radical dissection—“Triangle” procedure, which includes level 3 dissection along the superior mesenteric (SMA) and celiac artery (CA) plus all soft and lymphatic tissue in the triangular region between CA, SMA, and the mesenterico-portal axis—can improve disease-free survival compared with conventional SL under current guidelines [85]. The detailed information about the clinical trials related to comparison SL with EL are listed in Table 1.

Table 1.

Studies compared SL to EL

Study	Status	Type	Extent of lymphadenectomy		Examined lymph nodes		Patients		Postoperative mortality/morbidity		Survival statistics
			SL	EL	SL	EL	SL	EL	SL	EL	SL	EL
Pedrazolli et al. [81]	Completed	Multicenter RCT	12a,b 13a,b 14a,b 17a,b	SL + 8a,p 9 12c,p,h 14a,b,c,d,v 16a2,b1	13.3 ± 8.3	19.8 ± 15.1	40	41	2/18	2/14 (ns)	Mean survival: 552 days (CI: 364~741)	589 days (CI: 453~724) (ns)
Farnell et al. [86]	Completed	Single-center RCT	12b,c 13a,b 14a,b 17a,b 8a	SL + 8p 9 12a,p 14c,d,v 16	15 (3~31)	36 (6~74)	40	39	0/27	1/43 (ns)	Median survival: 26 months	Median survival: 19 months (ns)
Nimura et al. [87]	Completed	Multicenter RCT	13a,b 17a,b	SL + 8a,p 9 14p,d 16a,b 12a,b,p	13.3 (4~30)	40.1 (15~81)	51	50	0%/19.6%	2.0%/22.0%	5-year OS: 15.7% 5-year DFS: 11.8%	5-year OS: 6.0% (ns) 5-year DFS: 6.1% (ns)
Jang et al. [88]	Completed	Single-center RCT	5,6 8a/p 12a 13,17	SL + 8a/p 9,14 12p/b 16a1,a2,b1	13.6 ± 7.9	25.3 ± 11.8	34	32	1/10	2/20 (ns)	Median RFS: 18.7 months Median survival: 22.1 months	Median RFS: 11.3 months (ns) Median survival: 15.5 months (ns)
Wang et al. [84]	Completed	Single-center RCT	5,6 8a 12b,c 13,17 14a,b	SL + 9 12p 14c,d 16a2,b1	18 (16~19)	24 (22~26)	79	74	3/31	2/33 (ns)	2-year OS: 39.5% 2-year DFS: 28.25%	2-year OS: 25.3% (p = 0.034) 2-year DFS: 19.32% (p = 0.046)
Wang et al. [89]	Completed	Multicenter RCT	5,6 8a 12b,c 13a,b 14a,b 17a,b	SL + 8p 12a,p 14c,d 16	16.41 ± 7.53	26.39 ± 9.77	81	89	0/29	2/40 (ns)	3-year survival rate: 27.16% Median survival time: 18 months	3-year survival rate: 24.72% (ns) Median survival time: 15 months (ns)
TRIANGLE Trial [85]; Heger et al. (2023)	Ongoing	Multicenter RCT	SL as guidelines	Triangle region*	-	-	135	135	-	-	Primary endpoints: DFS Secondary endpoints: margin status, local recurrence, perioperative morbidity	-

*Triangle region: extended dissection (Inoue level 3) along the SMA and CA, removing all lymphatic and soft tissue in the triangular space between the CA, SMA, and the mesenterico-portal axis

In 2013, the International Study Group on Pancreatic Surgery (ISGPS) reached a consensus on the region for standard lymph node clearance. During this conference, the nomenclature for lymph node stations, established by the Japan Pancreas Society (JPS), and the definition of SL were widely accepted. It was agreed that EL is not recommended for routine pancreaticoduodenectomy (PD) or distal pancreatectomy (DP) procedures [90].

For pancreatic head carcinoma, the SL during PD includes specific LN regions: peripyloric (LN5,6), anterior of the common hepatic artery (LN8a), along the bile duct and cystic duct (LN12b1, 12b2, 12c), peri-head of the pancreas (LN13a, 13b, 17a, 17b), and the right lateral side of the superior mesenteric artery (SMA) (LN14a, 14b). There is ongoing debate on whether to clear the posterior of the common hepatic artery (LN8p), the entire SMA region (LN14a-d), and the para-aortic region (LN16). Some medical centers opt to resect these lymph nodes in select, fit patients if positive nodes are identified during surgery. To reduce local recurrence, improve R0 resection rates, and ensure better clearance of retropancreatic lymph nodes, the total mesopancreas excision (TMPE) technique was introduced for PD in 2007 [91]. However, evidence supporting TMPE’s impact on long-term survival remains controversial [92, 93], and some studies even question the anatomical existence of the mesopancreas [94].

For pancreatic body or tail carcinoma, the SL during DP includes lymph nodes at the splenic hilum (LN10), along the splenic artery (LN11), and along the inferior margin of the pancreas (LN18). For tumors confined to the body of the pancreas, peri-celiac artery lymph nodes (LN9) are also recommended for removal. DP can be performed as either a standard distal pancreatectomy with splenectomy (SDP) or a more extensive radical antegrade modular pancreatosplenectomy (RAMPS) [95]. Similar to TMPE in PD, the RAMPS procedure was proposed for DP to achieve better margins and lymph node clearance, especially along the retroperitoneum, and involves a more comprehensive lymphadenectomy. This procedure often includes the removal of additional tissues, such as parts of the adrenal gland, Gerota’s fascia, and a broader lymph node dissection, to achieve cleaner margins [95, 96]. The impact of RAMPS on clinical outcomes, however, remains uncertain. Jun et al. reported that RAMPS showed minimal improvement in overall survival (OS) and recurrence-free survival (RFS) compared to SDP [97]. Similarly, Feng et al. found no significant differences in OS, disease-free survival (DFS), or recurrence rates between the two approaches [98]. However, both studies support that RAMPS has advantages in terms of lymph node harvest and R0 resection rates compared to SDP.

Conclusively, the evidence we have gathered suggests that introducing EL in PDAC surgery is not advisable at this time, given the lack of survival benefit and the increased surgical duration. The reasons why EL does not provide significant survival advantages over SL remain unclear, but several studies may offer some insights. First, TDLN that are not invaded by the tumor can adapt to an anti-tumor immune function after primary tumor resection, meaning that EL may unintentionally compromise this immune response in lymph nodes. In head and neck squamous cell carcinomas, progenitor exhausted CD8+ T cells (Tpex) were found to accumulate in uninvolved LNs. Upon anti-PD-L1 immunotherapy, these Tpex cells activate, becoming intermediate-exhausted CD8+ T cells (Tex-int), then circulate in the bloodstream and eventually convert to terminal Tex cells within the tumor, where they exert anti-tumor immunity [99]. In mouse models of colon cancer, TDLNs were involved in immune activation after PD-1/PD-L1 checkpoint therapy, and removing all LNs in surgery could impair this therapy-induced immune response [82]. Second, EL impacts postoperative recovery and fitness of PDAC patients and affects the feasibility and outcomes of adjuvant therapy. The RCT conducted by Wang et al. showed that postoperative recovery and adjuvant therapy suffered in the EL group. All patients received adjuvant chemotherapy, but significantly fewer EL patients could complete all planned cycles (failure to complete in 31.5% EL vs. 17.3% SL, p = 0.032) [89]. The increased surgical burden associated with EL resulted in slower postoperative recovery and a postponed start of chemotherapy. On average, the EL group waited 2 days longer than the SL group (8.79 ± 8.03 vs. 6.77 ± 3.85, p = 0.037) [89]. Another RCT demonstrated significantly better survival (OS and DFS) with SL at 2 years (mentioned above) [84]. Among patients who received adjuvant chemotherapy, survival outcomes diverged even more: the 2-year OS was 60.7% with SL vs. 37.1% with EL (p = 0.021). Notably, the overall incidence of postoperative complications did not differ significantly between EL and SL (p = 0.502), suggesting that, in experienced hands, EL did not greatly increase typical postoperative morbidity. However, the investigators emphasized that EL’s lack of benefit (and trend toward worse survival) might be due to immune suppression—a slower recovery of blood lymphocyte counts after EL. These evidences suggested that trauma of EL may impair recovery/immune function, then influencing the effect of adjuvant chemotherapy, potentially offsetting the oncologic gain. A single-center retrospective study showed the EL group experienced higher operative burden (longer surgery time and more blood loss) and a significantly higher incidence of postoperative diarrhea (attributable to more extensive nerve plexus removal)—18.3% of EL patients had chronic diarrhea vs. 5.0% in the standard group (22 vs. 10 cases, p = 0.001). Notably, this study did observe that in the subset of patients with borderline-resectable tumors, EL was associated with improved survival compared to standard surgery (suggesting a possible benefit in patients with more locally advanced disease). However, for the typical resectable PDAC patient, no overall survival advantage was seen with EL, but there was added risk of GI dysfunction (e.g., diarrhea) that could impair postoperative quality of life and recovery [100]. Overall, the recent evidence indicates EL may in fact affect postoperative recovery and subsequent treatment. Future advancements in preoperative LNM prediction may allow for selective EL in pancreatic cancer patients at high risk for LNM, helping to clarify EL’s impact on patient outcomes. We look forward to continued research in this area to improve treatment strategies for PDAC.

Impact of neoadjuvant therapy on LNM

In recent years, numerous studies have evaluated whether neoadjuvant therapy can downstage nodal disease or reduce the incidence of LNM. The phase III PREOPANC trial compared neoadjuvant gemcitabine-based chemoradiation followed by surgery vs. immediate surgery for resectable/borderline PDAC. While the primary overall survival difference was not significant, patients receiving preoperative chemoradiotherapy had a lower rate of LNM, fewer cases of perineural and venous invasion, and a higher R0 resection rate compared to the immediate surgery group [101]. Several other trials and analyses support the trend of fewer positive lymph nodes and higher rates of R0 resection after neoadjuvant treatment in resectable PDAC patients [102–104]. Besides, neoadjuvant therapy can downstage nodal disease in initially node-positive PDAC. A focused NCDB study by Portuondo et al. examined patients who were clinically node-positive (cN+) before treatment. They reported that neoadjuvant therapy achieved pathologic nodal downstaging (to node-negative) in up to ~ 38% of such patients. Importantly, patients who converted to node-negative status had significantly improved survival compared to those who remained node-positive (hazard ratio ~ 0.61 for death) [105]. On the immunologic front, neoadjuvant therapy appears to remodel the tumor and nodal microenvironments in ways that could impede metastasis. Carmen Mota Reyes et al. analyzed resected PDAC specimens with and without neoadjuvant treatment and found that neoadjuvant therapy reverses immunosuppression by depleting pro-tumor immune cells. Treated tumors had a selective loss of immunosuppressive cell populations—notably regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs) [106]. Studies in PDAC mouse models further illustrate the impact of neoadjuvant approaches on nodal metastases. In an immunocompetent PDAC mouse model, Piper et al. combined radiotherapy with an immunotherapeutic IL-2/PD-1 pathway agonist as a neoadjuvant-like treatment. The treated mice had significantly fewer Tregs in TDLN, along with lower FoxP3 expression and IL-10 levels, indicating a less immunosuppressive nodal environment [107]. Although this specific strategy (radio-immunotherapy) is experimental, it reinforces the concept that altering the preoperative therapy can modulate the lymph node microenvironment to be less permissive for metastasis. While neoadjuvant therapy may provide multiple benefits and destroy the LNM of PDAC, it still faces many clinical challenges. First, a significant challenge is identifying which patients can safely tolerate neoadjuvant therapy. Many PDAC patients are older and have comorbidities; frailty or poor baseline performance status can make intensive chemotherapy or chemoradiotherapy risky. Studies in elderly PDAC patients show therapy dropout rates as high as 21–44% due to toxicity or disease progression, meaning the need for stringent selection based on fitness (ECOG performance status) and comorbidities rather than age alone [108]. Second, determining tumor resectability is critical yet often uncertain. PDAC resectability is defined by thin-slice CT imaging and other modalities (MRI, EUS), but subtle vascular invasion or occult metastases may be missed. Even with modern criteria (e.g., NCCN definitions of resectable vs. borderline), imaging alone cannot always predict which tumors are truly removable with negative margins. For instance, microscopic perineural spread along the superior mesenteric artery may escape detection before surgery and result in R1 resections even when imaging suggests resectability. Moreover, some patients radiologically staged as resectable are later found to have metastatic disease or unresectable extent at exploration, whereas others classified as borderline resectable may in fact benefit from upfront surgery [109]. Third, neoadjuvant therapy carries substantial toxicity, which can limit dosing and compromise treatment delivery. The multi-agent therapies associated with improved efficacy (e.g., FOLFIRINOX or gemcitabine/nab-paclitaxel) also cause high rates of grade ≥ 3 adverse events. For example, modified FOLFIRINOX, while effective, is known for significant hematologic toxicity (neutropenia, febrile neutropenia) and gastrointestinal side effects [110]. Fourth, by introducing several months of preoperative therapy, neoadjuvant therapy carries the risk that some patients will never make it to surgery. Aggressive tumors may progress despite therapy, and vulnerable patients may clinically deteriorate. For instance, in a multicenter study of borderline resectable cases, about one-quarter of patients developed disease progression during neoadjuvant and were never eligible for resection [110]. Finally, accurate assessment of lymph node status following neoadjuvant therapy is difficult. Residual nodal metastases are common even after robust neoadjuvant therapy. In one large study of 546 neoadjuvantly treated PDAC patients, 42.5% had pathologically positive nodes, and the presence of any nodal metastasis was associated with significantly shorter RFS [111]. Conventional imaging has limited sensitivity for nodal response. Treated lymph nodes often shrink to normal size on CT/MRI, yet may still harbor microscopic tumor deposits that imaging cannot detect. For pancreatic tumors, radiologic complete responses are rare and often false—one series found 92% of cases with no visible tumor on post-neoadjuvant CT still had residual cancer on histology [112]. In conclusion, while neoadjuvant strategies hold promises in reducing LNM, their optimal integration into clinical practice requires careful balancing of benefits against limitations.

Targeted therapy

Targeted therapy of LNM is very significant in PDAC. First, it can prevent cancer cells from spreading systemically through lymphatics. Second, it can further eliminate residual lesions in the lymph nodes after surgery in pancreatic cancer patients. Given the multiple mechanisms involved in LNM, as previously discussed, numerous potential therapeutic targets are under investigation for clinical application.

Targeting at epithelial-mesenchymal transition of PDAC cells

An increasing amount of evidence suggests that epithelial-mesenchymal transition (EMT) enhanced the lymph node invaded and metastatic capabilities of cancer cells [113–118]. The activation of the ERK-related pathway is a common feature during EMT [113]. Neuroglobin (NGB) slows PDAC progression by directly binding to GNAI1 and EGFR, suppressing their expression and negatively affecting the GNAI1/EGFR/AKT/ERK axis [119]. Pancreatic stellate cells (PSCs) release IL-6, which promotes pancreatic cancer progression by increasing ERK, NF-κB, and EMT-related factors in PDAC cells. Curcumin (CUR) counters this by suppressing IL-6 secretion in PSCs, thereby reducing the expression of E-cadherin, vimentin, and MMP-9 in cancer cells, and mitigating the effects of ERK and NF-κB [120]. TGF-β-induced EMT is also commonly observed in many cancers. Metformin disrupts TGF-β-induced activation of the Smad2/3 and Akt/mTOR pathways, inhibiting EMT in PDAC cell lines, although the exact mechanisms remain unclear [121]. Riboflavin, shown to directly bind to TGF-β receptor 1 (TβR1), blocks TGF-β signaling and suppresses EMT in PDAC both in vitro and in vivo [122]. Indomethacin also inhibits TGF-β-induced EMT by increasing E-cadherin expression while reducing N-cadherin and Snail levels, as confirmed by Raman spectroscopy [123].

Targeting at lymphangiogenesis

Lymphangiogenesis and the interaction between cancer cells and lymphatic endothelial cells (LECs), as previously discussed, are critical steps in lymph node metastasis (LNM). Several molecular pathways offer potential targets for drug intervention, including VEGF-C/D-VEGFR3, CCR7-CCL21, and CXCR4-CXCL12.

Currently, few drugs specifically target the VEGF-C-VEGFR3 axis, with many still undergoing clinical trials. Numerous agents, however, focus on the VEGF-A-VEGFR1/2 pathway, such as bevacizumab, ranibizumab, and ramucirumab, or act as multi-kinase inhibitors like sunitinib, sorafenib, cediranib, and vandetanib. These drugs primarily aim to inhibit angiogenesis but have also shown potential in targeting lymphangiogenesis. VEGFR2 and VEGFR3 kinase inhibitors, for instance, can suppress both angiogenesis and lymphangiogenesis [124]. In preclinical studies, sorafenib combined with docetaxel demonstrated significant anti-tumor activity in mouse models and PDAC cell lines [125], while cediranib reduced PDAC cell proliferation and migration [126]. However, clinical trials have generally failed to show improvements in overall survival (OS) when these kinase inhibitors are combined with standard chemotherapy in PDAC patients [127, 128]. The limited success of these inhibitors may stem from their low specificity for VEGFR3, which dilutes their effect on lymphangiogenesis [129]. Moreover, other lymphangiogenic factors, such as PDGF-BB and EGF, can drive lymphangiogenesis independently of VEGFR3, further complicating treatment outcomes [130].

CCR7 plays a pivotal role in facilitating cancer cell migration through lymphatic vessels, effectively hijacking the natural lymphatic trafficking mechanisms. In PDAC, a lack of CCR7 expression significantly reduces the rate of LNM [131]. In melanoma, the LNM of CCR7-positive B16 melanoma cells was blocked using neutralizing anti-CCL21 antibodies [132]. In melanoma, the LNM of CCR7-positive B16 melanoma cells was blocked using neutralizing anti-CCL21 antibodies [133, 134]. CXCR4 is another key target in LNM. The CXCR4 antagonist EPI-X4 effectively inhibits the CXCR4-CXCL12 axis, suppressing tumor growth and even shrinking tumors in pancreatic cancer models in vivo [135]. Additionally, the CXCR4 inhibitor AMD3100, when administered to target cancer-associated fibroblasts (CAFs), prompts a rapid accumulation of T cells within PDAC tumors. Furthermore, combining AMD3100 with anti-PD-L1 agents significantly reduces tumor burden in mouse models [136]. In a phase IIa clinical trial, the CXCR4 antagonist BL-8040 combined with the anti-PD-1 inhibitor pembrolizumab increased CD8+ effector T cell infiltration and reduced the presence of myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs) in tumors. This combination enhanced the efficacy of chemotherapy in PDAC patients [137]. These findings highlight the therapeutic potential of targeting CCR7 and CXCR4 pathways in managing LNM and improving immunotherapy outcomes in PDAC.

Drugs by nanodelivery

The unique anatomical and hydrodynamic properties of the lymphatic system make it difficult to achieve high drug concentrations in lymph nodes (LNs) through conventional delivery methods. To overcome this, researchers have developed nanocarrier-modified agents to increase molecular size, preventing their absorption by blood vessels (as smaller drugs are more easily absorbed) while facilitating lymphatic entry [138]. Nanodelivery systems can be classified into two categories based on their modification techniques: passive and active [139].

Passive delivery techniques rely on modifying nanomaterials’ size, shape, surface charge, and chemical composition to enhance lymphatic targeting. Nanoparticles (NPs) larger than 100 nm face challenges diffusing through the interstitium and are quickly cleared by phagocytes, while those smaller than 5–10 nm tend to enter the bloodstream. Therefore, the optimal size for lymphatic uptake is between 10 and 100 nm [140, 141]. Interstitial glycosaminoglycans, which are negatively charged, repel molecules with neutral or negative charges, pushing them toward lymphatic vessels. In contrast, positively charged particles tend to be trapped in the interstitium [142]. As for shape, the ideal particle shape for lymphatic uptake remains under investigation [143]. P@-Gem-HSA-NP is a compound consisting of pyropheophorbide-a (P@), gemcitabine (Gem), and human serum albumin (HSA). This nanoparticle can trace lymphatic vessels due to its optimal size, near-infrared imaging capability, and visible staining, offering a promising approach for diagnosing and treating LNM in PDAC [144]. PEG-Based Micelles (DACHPt/m) are developed for melanoma; these micelles encapsulate platinum-based anti-cancer agents and accumulate in LNs to prevent tumor LNM [145].

Active delivery techniques utilize specific ligands or targeting moieties to bind lymphatic receptors, effectively directing drugs to LNs. LyP-1 is a 9-amino acid cyclic peptide that specifically binds to P32, a protein overexpressed on the surface of tumor-associated lymphatic endothelial cells (LECs) or tumor cells [146]. Research by Luo, Guopei et al. demonstrated that LyP-1-conjugated nanoparticles (LyP-1-NPs) are taken up by metastatic lymph nodes at a rate approximately eight times higher than normal NPs [147]. LyP-1-functionalized multi-walled carbon nanotubes (fMWNTs) deliver MBD1 siRNA directly to tumors, significantly reducing PDAC cell proliferation both in vitro and in vivo [148]. In melanoma models, LyP-1-conjugated liposomes with doxorubicin (DOX) effectively suppress LNM by inhibiting both tumor metastasis and tumor-associated lymphangiogenesis [149]. The RGD motif, consisting of arginine (R), glycine (G), and aspartic acid (D), is commonly found in proteins that interact with αvβ3 integrin, which is overexpressed in many tumor cells. RGD-HAS-GEM/CUR-NPs loaded with gemcitabine (GEM) and curcumin (CUR) show a higher concentration in tumor sites than in peripheral organs, enhancing therapeutic efficacy [150]. RGD-modified nanoparticles with doxorubicin demonstrated the ability to inhibit hilar LNM and reduce tumor growth more effectively than non-RGD drugs [151]. These advanced delivery systems hold significant potential for enhancing drug concentration in LNs and improving the diagnosis and treatment of LNM in PDAC.

Immunotherapy in PDAC

As mentioned above, PDAC tumors commonly downregulate MHC class I on malignant cells and upregulate checkpoint ligands like PD-L1, blunting cytotoxic T lymphocyte recognition. The TME is heavily infiltrated by immunosuppressive cells—including FoxP3⁺ Tregs, regulatory B cells, M2-polarized TAMs, MDSCs, and others—which secrete tolerogenic factors (e.g., IL-10, TGF-β) that inhibit DC priming and directly suppress effector T cells. Concurrently, conventional DCs in PDAC are often functionally impaired or scarce, which impairs anti-tumor T cell activation. These multiple barriers help explain why immune checkpoint inhibitor (ICI) monotherapy has shown minimal efficacy in PDAC. Recent studies are therefore exploring combination strategies to overcome PDAC’s immune defenses.

First, one approach is the TME modulation. Research found that double antagonism of CCR2/CCR5 (CCR2/5i) can reduce the abundance of M2-TAM, MDSC, and Tregs infiltrating in PDAC models, enhance the function of CD8+ T cells, and sensitize the tumor to anti-PD-1 therapy and radiation therapy [152]. DC targeted therapy can enhance antigen presentation and T cell priming to promote anti-tumor immunity. For example, administration of Flt3 ligand to expand DC progenitors with agonistic CD40 antibodies to “license” DCs has led to increased intratumoral CD8+ T cell activation and prolonged survival synergized with radiation in PDAC models [68]. Reprogramming TAMs is likewise a promising strategy. CD40 agonists can convert macrophages to a tumoricidal phenotype [153], and blocking the macrophage phagocytosis checkpoint CD47 (the “don’t eat me” signal) is being tested to enhance cancer cell clearance (phase 1/2 clinical trial NCT05482893) [154]. Notably, one study demonstrated that antagonizing cIAP1/2 (cellular inhibitors of apoptosis proteins) exerted macrophage phagocytic activity against MHC-I^low PDAC cells, effectively eliminating these invisible tumor cells in a T cell–dependent manner [155].

Second, targeting immunosuppressive cytokine pathways is another approach. IL-6 produced by pancreatic stellate cells and TAMs in PDAC stroma drives STAT3-mediated upregulation of PD-L1 in immune cells and impedes immunity, so combining IL-6 blockade with PD-1/PD-L1 inhibitors has shown synergistic anti-tumor effects (dual IL-6/PD-L1 antibody therapy boosted CD8⁺ T cell but not CD4+ T cell infiltration and reduced the α-SMA⁺ fibroblasts to limit the tumor regression in PDAC models) [156].

Third, at the tumor-cell level, recent work identified the glucocorticoid receptor (GR) as a transcriptional hub of immune evasion in PDAC: GR activation in cancer cells upregulates PD-L1 while repressing MHC-I, thereby paralyzing T cells, whereas genetic or pharmacologic GR inhibition restores antigen presentation, improves CD8⁺ T cell infiltration, and sensitizes PDAC to ICIs in preclinical PDAC models [157].

Finally, emerging lymph node–focused interventions aim to remodel the immunosuppressive TDLN stroma. A strong example is the lymph-node‐targeted vaccine ELI-002 2P, which uses amphiphile-modified mutant KRAS (G12D/R) peptides plus a TLR9 agonist to accumulate in draining lymph nodes via albumin hitchhiking. In the AMPLIFY-201 phase 1 trial, this vaccine achieved mKRAS-specific CD4+ and CD8+ T cell responses in ~ 84% of patients with minimal residual PDAC disease, with biomarker reductions and prolonged relapse-free intervals correlating with the magnitude of T cell induction [158]. Another approach used intratumoral delivery of microspheres encapsulating immunotherapy agents in preclinical PDAC models; these microspheres not only modulate the local tumor microenvironment but also drain to TDLNs, promoting lymph node antigen presentation and enhancing systemic anti-tumor immunity. These examples underscore that directing antigen/adjuvant to TDLNs—either via vaccine design or by delivery systems that traffic to lymph nodes—can help overcome poor antigen presentation and T cell priming deficiencies, which are key obstacles in PDAC immunotherapy [159].

Prognosis of lymph node metastasis in PDAC

According to SEER statistics, 29% of PDAC patients are initially diagnosed at the regional stage, where the cancer has spread to regional lymph nodes, with a 5-year relative survival rate of 26.2%. This contrasts with a survival rate of 44.0% for those diagnosed at the localized stage (tumor confined to the primary site) and just 3.1% for patients with distant metastases (tumor spread to other organs). The prognosis for PDAC patients with lymph node metastasis (LNM) depends on several factors, including the pattern of LNM (specific sites of nodal involvement), methods for calculating the number of positive lymph nodes, and the way of LNM (whether through direct invasion or migration via lymphatic pathways).

Pattern of LNM

The frequency of pancreatic cancer metastasis to different lymph nodes varies, influencing prognosis based on the specific lymph node site involved [160]. Several studies show that PALN involvement is associated with a poorer prognosis. For instance, Schwarz, L. et al. reported a median overall survival (OS) of 9.7 months for PALN-positive (PALN+) patients versus 28.5 months for PALN-negative (PALN−) patients (p = 0.012). Median disease-free survival (DFS) was similarly shorter in PALN+ patients at 8.2 months compared to 12.9 months (p = 0.041) [161]. These findings emphasize the need for PALN sampling during surgery to better assess prognosis. More studies showed the strong relation between PALN+ and poor prognosis [162–170]. However, whether to proceed with radical surgery for PALN+ patients remains debated. Earlier in 2007, Japanese studies reported that PALN+ patients had a median survival < 1 year and therefore suggested that surgery may not be beneficial in this subgroup [162]. Similarly, Paiella et al. suggested that PALN involvement may reflect an early systemic disease rather than a purely localized stage [163]. Marchese U et al. showed that terminating pancreatectomy upon detecting PALN+ during surgery resulted in better 3-year OS rates compared to completing the surgery (58.3% vs. 25%) [164]. Additionally, some studies indicate that PALN dissection during pancreaticoduodenectomy (PD) does not improve outcomes and increases complications, such as infections [165]. While most earlier studies emphasized poor prognosis, more recent evidence has challenged this notion. Korean researchers in 2020 reported that PALN+ patients who received postoperative chemotherapy had survival rates comparable to patients with other LN metastases, suggesting that PALN+ should not be an automatic contraindication for surgery if adjuvant chemotherapy is planned [166]. Rupaly et al. supported this, finding no significant differences in OS and DFS between PALN+ patients treated with surgery and chemotherapy and PALN− patients (OS: 23.4 vs. 20.6 months, p = 0.192; DFS: 23.9 vs. 20.5 months, p = 0.718) [167]. Moreover, a 2023 study revealed that radical resection improved OS compared to surgical exploration alone (EXP) in PALN+ patients [168]. Conclusively, the decision to prohibit surgery in all PALN+ patients may be overly conservative. Certain patients could benefit from surgery combined with adjuvant therapy. Further research is essential to identify which PALN+ patients might experience improved outcomes with surgical intervention.

Beyond PALN involvement, LNM in other regions also significantly influences the prognosis of PDAC. Seung Ki et al. classified LN stations into three groups: G1 (no LN metastasis), G2 (peripancreatic/PP), and G3 (pericholedochal/PC, periduodenal/PD, hepatogastric/HG, and superior mesenteric artery/SMA). Their study found that G2 and G3 were identified as independent risk factors for overall survival (OS), with G3 having a worse prognosis than G2. Specifically, G3-SMA and G3-PC showed poorer OS compared to G2-PP, while G3-HG did not significantly impact OS in TNM N1 status [169]. Similar findings have been observed in other studies. Hepatoduodenal LNs (corresponding to G3-PC) and SMA LNs (G3-SMA) are associated with shorter disease-free survival (DFS) (p = 0.001 and p = 0.017, respectively). Hepatoduodenal LNs have been reported as independent prognostic factors for mortality, reducing OS in PDAC patients [170]. Although several studies have suggested that G2 and G3 involvement predicts worse OS, findings are not entirely consistent, particularly regarding hepatic artery lymph nodes (HALNs, corresponding to G3-HG). A study in 2013 found that HALN+ patients had significantly lower OS and DFS (Kaplan-Meier analysis, p = 0.017 and p = 0.013, respectively, using log-rank test) [171]. A meta-analysis in 2016 showed that HALN+ patients had a survival of 15 ± 3 months compared to 19 ± 3 months in HALN− patients after undergoing pancreaticoduodenectomy (PD) (p = 0.02) [172]. However, other studies contradict these findings. Two separate investigations concluded that among PDAC patients with LN metastasis, survival after PD was similar regardless of HALN status. For instance, one study showed no significant difference in OS between HALN+ and HALN− patients (18.4 months vs. 19.7 months, p = 0.659) [173, 174]. Therefore, the impact of LNM patterns on PDAC prognosis, particularly concerning HALNs, remains inconsistent. Further large-scale, multicenter studies are necessary to clarify the prognostic significance of different LN involvement patterns in PDAC.

Different lymph node staging systems

There are several systems for calculating positive LNs, each influencing the prognosis of PDAC. Examined lymph node (ELN) count is one such system, with the recommended minimum number of ELNs during surgery set at 15 by the ISGPS [90] and 11–17 by the NCCN [175]. However, these benchmarks require further refinement. Studies have shown that a higher ELN count improves nodal staging accuracy, which is linked to better outcomes [176]. Various ELN cutoffs have been proposed to stratify prognosis, showing associations with improved recurrence-free survival (RFS) and overall survival (OS) across different studies [176–180]. The number of positive lymph node (NPLN), lymph node ratio (LNR, defined as NPLN/ELN), and log odds of positive lymph nodes (LODDS, defined as log[(NPLN + 0.5)/(ELN-NPLN + 0.5)]) are different LN classification system impacts on the prognosis. They have been proposed, with many studies suggesting superior prognostic value compared to traditional N staging, although findings are not entirely consistent. LNR has been reported as an independent prognostic factor, complementing traditional TNM staging [179, 181–184]. Slidell MB et al. found LNR to be among the strongest survival predictors for N1 patients, with OS varying based on LNR values: 15 months (LNR 0–0.2), 12 months (LNR 0.2–0.4), and 10 months (LNR > 0.4) (p < 0.001). For N0 patients, ELN stratified OS into 16 months (ELN 1–11) versus 23 months (ELN > 12) (p < 0.001) [181]. La Torre M et al. reported that LNR > 0.2 was strongly associated with survival (p = 0.001), whereas N staging lost prognostic significance when the mean ELN was 19.8 (3–26) [182]. He C et al. also suggested that LNR remained an independent prognostic factor for OS, even after accounting for patient age and gender (p = 0.013). However, they found that LNM staging did not correlate with prognosis, with the mean number of ELN being 8 (range: 1 to 32) [183]. Similarly, Mirkin KA et al. concluded that a higher LNR was still a strong independent predictor of mortality, even when the ELN count was suboptimal (ELN ≤ 6) [179]. Joliat GR et al. highlighted the prognostic significance of LNR in N2 patients, reporting an OS of 19 ± 2 months for those with LNR ≥ 0.225, compared to 27 ± 8 months for LNR < 0.225 (p = 0.001). They advocated for its use alongside TNM staging [184]. Several studies indicated that LNR and LODDS may provide superior prognostic discrimination compared to traditional N staging in predicting prognosis and stratifying PDAC patients [185, 186]. Prassas D et al. evaluated the discriminative abilities of LNR and LODDS using Cox regression. They found that both LNR and LODDS classifications demonstrated statistically significant HR that increased across subcategories, outperforming N staging [185]. Similarly, Riediger H et al. suggested that LNR and LODDS provided superior survival discrimination, with relative risks ranging from 1.78 to 2.16, compared to 1.43 to 1.51 for N staging [186]. However, whether LODDS offers advantages over LNR remains a subject of debate. Regarding the number of positive lymph nodes (NPLN), it has frequently been shown to perform similarly to LNR in predicting survival [187, 188]. Elshaer M et al. in a meta-analysis of 19 studies involving 4883 patients concluded that NPLN was significantly associated with OS, with 11 out of 12 studies supporting this finding [187]. Liu ZQ et al. also demonstrated that NPLN was an independent predictor of survival in N1 patients, with HRs increasing from 1 for NPLN 1–2 to 2.17 for NPLN ≥ 3 (p = 0.002) [188]. In conclusion, an adequate number of ELNs is crucial for accurate staging in PDAC. Both NPLN and LNR are valuable tools for prognosis, while the role of LODDS requires further investigation.

Should direct invasion be classified as LNM?

It is important to distinguish between lymphatic metastasis and direct tumor invasion of lymph nodes. This distinction differentiates classical LNM from tumor extension into peritumoral lymph nodes (PTLNs). A 2011 study highlighted this difference, showing no significant difference in overall survival (OS) between patients with direct LN invasion and those without LNM (21 vs. 30 months, p = 0.609) [189]. However, patients without LNM (LNM−) had significantly better OS than those with LNM (LNM+) (30 vs. 15 months, p < 0.001). Additionally, patients with a combination of direct LN invasion and 1–2 true LNMs had worse OS than those with only direct LN invasion, though the difference was not statistically significant. Subsequent studies have supported these findings [190, 191]. A group demonstrated that peritumoral lymph node direct invasion (PTLNI) was associated with significantly better disease-free survival (DFS) compared to regional lymph node metastasis (RLNM) or a combination of PTLNI and RLNM (CLNM) (21 vs. 11 vs. 12 months, p = 0.003). Importantly, PTLNI showed no significant difference in DFS compared to N0 patients (21 vs. 23 months, p = 0.999) [191]. Taken together, current evidence suggests that direct LN invasion may resemble N0 status rather than true nodal metastasis. This highlights the importance of caution to avoid overstaging N status, and several studies have proposed that direct invasion might not be equivalent to LN+ in staging.

New detection methods for lymph node metastasis in PDAC

In the management of pancreatic cancer, identifying LNM plays a pivotal role in determining the most appropriate treatment plan and predicting patient outcomes. LNM serves as a key marker of disease progression, directly affecting the feasibility of surgical resection and influencing long-term survival rates. Accurate detection and evaluation of LNM are crucial for optimizing surgical strategies and selecting the most effective non-surgical treatments, ultimately enhancing overall outcomes for patients. Advances in technology have led to the development of several new methods for more precise LNM assessment, both preoperatively and intraoperatively. In the following sections, we will explore some of the latest techniques being applied in the evaluation of LNM in PDAC.

Radiomics analysis helps staging before surgery

Radiomics, a specialized field within medical imaging, involves extracting a wide range of quantitative features from medical images using advanced data-characterization algorithms. These features capture information on shape, texture, intensity, and spatial relationships within the images, revealing patterns that are often imperceptible to the human eye. Today, radiomics combined with deep learning algorithms presents significant potential for enhancing the diagnosis, prognosis, and treatment strategies for lymph node metastasis (LNM) in pancreatic ductal adenocarcinoma (PDAC). A 2024 meta-analysis highlighted the efficacy of CT-based radiomics and deep learning models in preoperative LNM assessment for PDAC patients. The results demonstrated promising performance, with a sensitivity of 0.83, specificity of 0.76, and an area under the curve (AUC) of 0.85. The study also reported an average Radiomics Quality Score (RQS) of 12.00 ± 3.89 and an average Methodological Radiomics Score (METRICS) of 63.30 ± 10.88, finding a strong correlation between RQS and METRICS (r = 0.810, p = 0.016) [192]. In 2023, dual-transformation-guided contrastive learning for prediction of LNM showed better sensitivity (SE), specificity (SP), and AUC than the other three models included (SE: 0.740, SP: 0.750, AUC: 0.791) [193]. At the same year, using a modified Multiview-guided two-stream convolution network (MTCN) radiomics model accompanied with clinical factors was discovered to be superior to MTCN without clinical factors or radiologist judgement (Train cohort AUC 0.823 vs. 0.793 vs. 0.592, Train cohort accuracy 0.763 vs. 0.744 vs. 0.567) [48]. When comparing artificial intelligence (AI) models to traditional radiomics and clinical prediction models, Bian et al. found that AI models offered the best performance for detecting LNM in PDAC, with an AUC of 0.91 compared to 0.71 for radiomics and 0.76 for clinical models [194]. Furthermore, Gu et al. noted that different body compositions could affect radiomics’ predictive capabilities. For localized PDAC patients, female adipose tissue signatures demonstrated the highest predictive accuracy (AUC: 0.895), outperforming male adipose (AUC: 0.735) and male muscle tissues (AUC: 0.773). This predictive capability improved further when muscle radiomics features were included, raising the AUC to 0.924 [195]. In 2022, a study revealed that deep learning radiomics (DLR) applied to dual-energy CT (DECT) provided the best performance compared to conventional CT methods. Adding clinical factors further enhanced the model’s predictive accuracy. Specifically, test cohort AUCs were as follows: 0.92 for clinical + 100 + 150 kV DLR, 0.87 for only 100 + 150 kV DLR, 0.84/0.80 for 100/150 kV DLR, and 0.76 for virtual monoenergetic image 40 kV DLR (p < 0.05) [196]. Also in 2022, Bian et al. developed a radiomics nomogram for predicting LNM, which outperformed CT-reported LN status, with an AUC of 0.81 compared to 0.63 (p = 0.02) [197]. Radiomics and deep learning models, particularly when combined with clinical features, provide highly efficient and accurate non-invasive methods for preoperatively predicting LNM in PDAC. Relevant details of the studies mentioned are summarized in Table 2.

Table 2.

Basic characteristics of the mentioned research

Authors and year of publication	Design methods	Sample size	Imaging equipment	Phase for analysis	Segmentation features for radiomics studies	Research type
Chen et al. [193] 2023	Retrospective Single center	300 patients with PDAC from China as training datasets 79 with HNSCC datasets from Netherlands as external validation	CT Technical parameters unknown	Arterial and venous	\	Deep learning
Fu et al. [48] 2023	Retrospective Single center	Total 363 PC patients 7/10 as training cohort and 3/10 as test cohort. Another 28 PC patients as external validation	CT 120 kV; thickness: 1 mm	Arterial and venous	\	Deep learning
Bian et al. [194] 2023	Retrospective Single center	545 patients as training cohorts 189 patients as test cohorts	CT 120 kV; thickness: 1 mm	Arterial and venous	\	Deep learning
Gu et al. [195] 2023	Retrospective Single center	196 patients	CT 120 kV; thickness: 2 mm	Arterial and venous	3D ROI with 3D slicer software 1688 features	Radiomics score
An et al. [196] 2022	Retrospective Single center	113 patients as training cohorts 35 patients as test cohorts	Dual-energy CT (DECT) 100/150 kV; thickness: 1 mm	Venous	\	Deep learning
Bian et al. [197] 2022	Retrospective Single center	118 patients as training cohorts 45 patients as validation cohorts	CT 120 kV; thickness: 1 mm	Arterial	3D ROI with 3D slicer software 1029 features	Radiomics nomogram

Lymph node mapping during surgery

Lymph node (LN) mapping during surgery is critical for detecting tumor micro-metastases, which significantly impact the prognosis and treatment of pancreatic cancer. While LN mapping has been widely adopted in breast cancer, gastric cancer, and melanoma [198], its application in pancreatic cancer remains limited. Conventional fluorescent probes, such as indocyanine green (ICG), are commonly used clinically but suffer from drawbacks, including contamination of the surgical field and short retention times in lymph nodes [199]. In pancreatic cancer, studies have highlighted the limitations of classical fluorescent probes. In 2007, using methylene blue for sentinel lymph node (SLN) mapping found it inadequate, as only 4 of 14 patients with LNM showed blue-stained nodes, rendering the method impractical for pancreatic cancer [200]. In 2011, Hutteman M et al. employed near-infrared (NIR) imaging with ICG during pancreaticoduodenectomy (PD). However, they noted that ICG failed to provide a clear boundary between the tumor and normal pancreatic tissue, except around the common bile duct [201]. Compared to traditional fluorescent probes, advances in nanoprobes and organic probes offer promising alternatives. In 2012, a study using Alexa488-antiCEA conjugates (CEA targeted fluorescent agents) on BxPC-3 cells in a mouse model of pancreatic cancer showed significantly higher sensitivity for detecting metastatic lesions compared to traditional bright light surgery (sensitivity: 96% vs. 40%, p < 0.001) [202]. In 2014, graphene oxide with iron oxide nanoparticles (GO-IONP) was tested as a nanotheranostic agent for LNM in pancreatic cancer. MRI imaging demonstrated superior visualization of the regional lymphatic system and better differentiation of lymph nodes in blood-stained surgical fields compared to carbon nanoparticles [39]. Deuteporfin, a second-generation photosensitizer, was found in 2016 to accumulate more in metastatic lymph nodes than in normal pancreas tissue (p < 0.05), confirming its potential for targeted LN mapping via NIR fluorescent imaging [203]. Different from the fluorescent probes mentioned above, the use of radiotracer in lymphoscintigraphy remains challenging and still needs further exploration. In 2016, researchers used Tc(99m)-labelled nanocolloid radiotracer for preoperative injections and detected LNs intraoperatively using a gamma probe. However, the technique was limited, as only two patients completed intraoperative detection, and SLN consistency with histopathological analysis was confirmed in just one patient [21]. A 2020 study employed 99mTc-sodium phytate with a gamma probe for SLN detection. Although histopathological assessments confirmed SLNs postoperatively, the detection rate was only 64%, with a false negative rate as high as 60%, highlighting its inaccuracy for LN mapping [204]. While conventional fluorescent probes have limited utility in pancreatic cancer, emerging nanoprobes and organic probes show promise for more effective lymph node mapping. However, radiotracer-based methods remain in their infancy and require further refinement to improve accuracy and reliability in clinical practice.

Conclusions and prospectives

By integrating current research findings, our review hopes to provide comprehensive knowledge for clinicians and researchers, including lymph node metastasis in pancreatic cancer, the role of LNM in prognosis, new strategies for the detection and prediction of LNM, and multiple treatment approaches to improve survival rates and quality of life for patients with pancreatic cancer. This also implies that the detection of LNM, prognosis evaluation based on LNM, and targeting of specific molecular pathways to eliminate LNM in pancreatic cancer are also some of the areas that future research must focus on. The new therapeutic agents which may inhibit these pathways or reverse the changes associated with metastasis could result in clinical benefits. In this respect, combination therapy-evaluating modulation with agents added to existing treatment modalities may enhance their effectiveness and delay progression.

The other critical direction of future work will be towards personalized medicine. That is, judging different LNM patterns in differing patients precisely and developing specific treatment strategies matched to these differing LNM states. Innovation in diagnostic technologies, such as liquid biopsy, and further development of imaging techniques can allow real-time, minimally invasive assessment of the spread of metastasis. Integration of multi-omics data and artificial intelligence might further provide fine-tuning in the detection and staging of LNM. Treatment plans, as for LNM, are designed in relation to the peculiar characteristics of each patient. Personalized treatment strategies need to consider not just the extent and location of lymph node involvement but also the molecular profile of the metastatic cells. By classifying the patients into groups based on their LNM profile, clinicians are able to propose individualized treatment schemes to maximize therapeutic benefit and minimize unnecessary treatment-related side effects.

We summarized the progress of studies on LNM in pancreatic cancer and highlighted future perspectives. It will focus on an accurate LNM prediction and assessment of prognosis, judgment of individual patients’ health conditions according to their LNM status, and establishment of a personalized treatment strategy. The knowledge would then be brought back into clinical application, where the collaboration between researchers, clinicians, and healthcare professionals would increase overall survival rates and improve quality of life for patients with PDAC.

Abbreviations

PDAC

Pancreatic ductal adenocarcinoma

SEER

Surveillance, Epidemiology and End Result Program

LNM

Lymph node metastasis

AJCC

American Joint Commission on Cancer

DFS

Disease-free survival

OS

Overall survival

EMT

Epithelial-mesenchymal transition

TDLN

Tumor-draining lymph node

GC

Gastric cancer

TGF-β

Transforming growth factor-β

OSCC

Oral squamous cell carcinoma

DC

Dendritic cells

LEC

Lymphatic endothelial cells

IFP

Interstitial fluid pressure

CAF

Cancer-associated fibroblast

RFS

Recurrence-free survival

PGE-2

Prostaglandin E2

LVD

Lymphatic vessel density

VEGF

Vascular endothelial growth factor

SLN

Sentinel lymph nodes

FGF

Fibroblast growth factor

CRC

Colorectal cancer

PDX

Patient-derived xenograft

HIF

Hypoxia-inducible factors

HRE

Hypoxia response element

PI3K

Phosphatidylinositol-3 kinase

TSS

Translation start site

ET-1

Endothelin-1

AP-1

Activator protein-1

RIPK1

Receptor-interacting protein kinase 1

CCA

Cholangiocarcinoma

lncRNA

Non-coding RNA

EV

Extracellular vesicles

SCS

Subcapsular sinus

HEV

High endothelial venules

MHC

Histocompatibility complex

GR

Glucocorticoid receptor

TAM

Tumor-associated macrophage

TGFBI

TGF-β-induced proteins

IL-1β

Interleukin-1β

APC

Antigen-presenting cells

TME

Tumor microenvironment

Breg

Regulatory B cell

FRC

Fibroblastic reticular cell

WES

Whole exome sequencing

EL

Extended lymphadenectomy

SL

Standard lymphadenectomy

ISGPS

International Study Group on Pancreatic Surgery

JPS

Japan Pancreas Society

PD

Pancreaticduodenectomy

DP

Distal pancreatectomy

SMA

Superior mesenteric artery

TMPE

Total mesopancreas excision

SDP

Standard pancreatosplenectomy

RAMPS

Radical antegrade modular pancreatosplenectomy

CUR

Curcumin

PEG

Polyethylene glycol

NP

Nanoparticle

PALN

Para-aortic LN

PP

Peripancreatic

PC

Pericholedochal

HG

Hepatogastric

HALN

Hepatic artery lymph node

ELN

Examined lymph node

NPLN

Number of positive lymph node

LNR

Lymph node ratio

LODDS

Log odds of positive lymph nodes

PTLNI

Peritumoral lymph nodes direct invasion

RLNM

Regional LNM

SE

Sensitivity

SP

Specificity

AUC

Area under curve

MTCN

Multiview-guided two-stream convolution network

DLR

Deep learning radiomics

DECT

Dual-energy CT

ICG

Indocyanine green

SLN

Stain lymph node

NIR

Near-infrared

GO-IONP

Graphene oxides modified with iron oxide nanoparticles

Author contributions

WW conceived, designed the review and gave constructive guidance. SHY, CQ and YZ collected related literature and finished the manuscript. BZ, ZL, TL, LH participated in the design of this review. All authors read and approved the final version of the manuscript.

Funding

Weibin Wang received support from the National Natural Science Foundation of China (No. 82173074, No. 82573412), Beijing Natural Science Foundation (No. 7232127), Capital’s Funds for Health Improvement and Research (No. 2024-2-4017), the National High Level Hospital Clinical Research Funding (No. 2022-PUMCH-D-001, No. 2022-PUMCH-B-004), and the CAMS Innovation Fund for Medical Sciences (CIFMS) (No. 2021-I2M-1-002, 2023-I2M-2-002). Cheng Qin is supported by the Postdoctoral Fellowship Program of CPSF under Grant Number GZC20240146.

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.