Pancreatic cancer (PC) remains one of the most lethal malignancies, with a 5-year survival rate of less than 10% despite advances in surgical techniques and systemic therapies (1, 2).

Lymph node status has long been recognized as a critical prognostic factor in pancreatic cancer. The presence of lymph node metastasis is consistently associated with poor survival, and the extent of nodal involvement is incorporated into current staging systems (3-7). Enlarged lymph nodes on preoperative imaging are generally interpreted as evidence of metastatic spread, reinforcing the perception that nodal enlargement predicts adverse outcomes (8).

However, lymph node enlargement does not always reflect metastatic disease. Enlarged nodes can result from reactive hyperplasia or immune cell infiltration in response to tumor antigens. In laryngeal carcinoma, patients with clinically enlarged but histologically reactive lymph nodes showed better survival than those without lymph node enlargement, suggesting that palpable benign nodes may represent host immune activation with favorable prognostic significance (9). Similarly, statistical analyses in other disease contexts have found that patterns of lymph node “reactivity” or stimulation, rather than depletion, predict improved long-term survival outcomes (10). Moreover, systemic anti-tumor immune responses involving regional lymph nodes, beyond the tumor microenvironment, have emerged as important contributors to therapeutic response, implying that nodal enlargement may indeed reflect beneficial immune dynamics (11).

Nevertheless, to date, the prognostic significance of lymph node size itself has not been well studied in PC. In this study, we focused on the No. 8 lymph node (N8LN), which is a regional node around the common hepatic artery, routinely assessed during pancreaticoduodenectomy (PD) (12) that can be usually identified on preoperative imaging studies. We aimed to clarify whether the size of the N8LN is associated with patient prognosis and to explore the underlying immunological basis by quantifying CD8⁺ T-cell infiltration.

Patient selection. Between May 2009 and December 2019, 134 patients who underwent PD for PC at our hospital were screened; 35 were excluded for the reasons shown in Figure 1, and the remaining 99 were enrolled in this study. Specifically, eleven patients with positive metastasis in the N8LN, nine patients with non-invasive intraductal papillary mucinous carcinoma, eight patients in whom the N8LN could not be identified on preoperative multidetector-row computed tomography (MDCT), and seven patients for other reasons. This study was approved by the ethics committee of The Jikei University School of Medicine (approval number 27-177), and the requirement for informed consent was waived as anonymized data were used. The study was conducted in accordance with the Declaration of Helsinki (1964) and its later amendments.

Definition of N8LN size. Based on axial and coronal slices obtained from baseline MDCT performed at the initial presentation, the long and short axes of N8LN were measured on axial sections, and the vertical dimension was assessed on coronal sections, using 1.5‑mm slice images with SYNAPSE software (FUJIFILM Corporation, Tokyo, Japan) and its attached measurement tools (Figure 2). The minimum value among these measurements was defined as the representative N8LN size. Each parameter was measured independently by the first author (T.S.) and two additional evaluators (Y.Ya. and Y.Yh.). The two additional evaluators (surgeons) were blinded to survival outcomes, pathological findings, group allocation, and all other study results. The mean of the three measurements was used for analysis. Patients were then divided into two groups according to the median N8LN size: the small group (< median) and the large group (≥ median).

Preoperative blood tests. Preoperative laboratory data, including serum carbohydrate antigen 19-9 (CA19-9) levels, were collected within 1-3 days before surgery. Patients with CA19-9 levels ≤2 U/ml were classified as “CA19-9 non-secretors” and were excluded from analyses involving CA19-9–based comparisons.

Immunohistochemistry. Formalin-fixed, paraffin-embedded lymph node specimens were sectioned at a thickness of 4 μm and subjected to immunohistochemical staining for CD8. Staining was performed using a rabbit monoclonal anti-CD8 antibody (clone SP239, Roche, Basel, Switzerland) according to the manufacturer’s instructions. The stained sections were scanned at high resolution using the Aperio AT2 system (Leica Biosystems, Wetzlar, Germany) to generate whole-slide images (Figure 3A).

Evaluation of CD8⁺ T-cell density. For each case, the largest cross-section of the lymph node was selected for immunohistochemical evaluation. CD8⁺ T-cell density was quantified using QuPath software (version 0.4.4) (13-16). The number of CD8⁺ cells across the entire lymph node cross-section was divided by the total tissue area, and expressed as cells/mm² (Figure 3B).

Postoperative follow-up. Following radical resection, patients received adjuvant chemotherapy for 6 months with gemcitabine up to 2013 and for 6 months with S-1 after 2014. Patients underwent physical examinations and blood tests, including CA19-9 levels every 1 to 2 months, and MDCT of the chest and abdomen or magnetic resonance imaging (MRI) scan of the abdomen every 3-4 months to check for recurrence. Patients who developed relapse were treated with S-1, gemcitabine, nab-paclitaxel plus gemcitabine, FOLFIRINOX, or best supportive care, depending on their physical condition or medical history. The median follow-up period was 21 months (range=3-162 months).

Statistical analysis. Data were analyzed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) (17). More precisely, EZR is a modified version of R commander (version 2.7-1) designed to add statistical functions frequently used in biostatistics. Continuous data are expressed as medians with interquartile ranges and were compared using the Mann–Whitney U-test. Categorical data were expressed as frequencies with percentages and were compared using Fisher’s exact test. Disease-specific survival (DSS) was calculated from the date of the first treatment (neoadjuvant therapy or surgery) to the date of death, and progression-free survival (PFS) was defined as the time from the date of surgery to the date of MDCT or MRI confirmation of recurrence. For PFS, radiological recurrence and death without documented recurrence were both considered events. Patients alive without recurrence at last follow-up were censored. The use of two different starting points is acknowledged as a limitation. The Kaplan–Meier method was used to estimate cumulative survival probabilities, and the log-rank test was performed to compare survival distributions between groups. Cox proportional-hazards regression analyses were performed to identify factors associated with PFS and DSS. Variables with p<0.10 in univariable analyses and clinically relevant covariates were included in the multivariable models. Reference categories were defined as follows: resectable disease (R), small N8LN size (<6.2 mm), negative lymph node status, well differentiation, absence of biliary drainage, and no PV/SMV resection. Hazard ratios (HRs) and 95% confidence intervals (CIs) were reported for all variables.

Small and large groups. The size of the N8LN ranged from 2.8 to 13.2 mm, with a median of 6.2 mm. Patients were divided into two groups based on N8LN size: the small group (<6.2 mm) and the large group (≥6.2 mm).

Patient characteristics. A total of 99 patients were analyzed (small group, n=49; large group, n=50). Baseline characteristics were generally balanced between groups (Table I). Notably, the large group had a greater number of retrieved lymph nodes (median 26 vs. 21, p=0.016). There was also a trend toward a higher frequency of preoperative biliary drainage in the large group (58% vs. 39%, p=0.071). In addition, use of adjuvant chemotherapy was more common in the large group (88% vs. 69%, p=0.028). Distributions of clinical stage and nodal status were comparable [Union for International Cancer Control (UICC) stage, p=0.783; pathological N category, p=0.762). C-reactive protein (CRP) tended to be higher preoperatively in the large group (0.2 vs. 0.1 mg/dl, p=0.054).

Survival analysis by N8LN size. Using the median minimum diameter of the No. 8 lymph node (6.2 mm) as the cut-point, Kaplan–Meier analysis showed significantly longer PFS and DSS in the large group compared with the small group (median PFS 19.7 vs. 9.4 months, log-rank p=0.008; median DSS 41.7 vs. 17.1 months, p<0.001) (Figure 4). In multivariable Cox models, large N8LN size remained an independent favorable prognostic factor for both PFS [hazard ratio (HR)= 0.41, 95% confidence interval (CI)=0.25-0.67, p<0.001] and DSS (HR= 0.35, 95%CI=0.20-0.60, p<0.001) (Table II).

Other prognostic covariates. Borderline resectable or unresectable locally advanced (BR/URLA) status at initial evaluation was independently associated with shorter PFS (HR= 2.44, 95%CI=1.32-4.49, p=0.004). Pathological lymph node positivity was independently associated with worse PFS and DSS (PFS: HR=2.90, 95% CI=1.63-5.15, p<0.001; DSS: HR=3.03, 95% CI=1.62-5.64, p<0.001). Poor/moderate differentiation (vs. well) was also independently adverse (PFS: HR=1.96, 95% CI=1.12-3.42, p=0.018; DSS: HR=2.19, 95% CI=1.21-3.94, p=0.009). Variables such as portal vein resection and biliary drainage showed significance only on univariable analyses but not after adjustment. Adjuvant chemotherapy use was not independently associated with DSS (HR=1.49, 95% CI=0.81-2.77, p=0.204) (Table II).

CD8⁺ T-cell density in lymph nodes. Whole-slide quantitative immunohistochemistry revealed no significant difference in CD8⁺ T-cell density between groups [large vs. small: median=3,915 cells/mm², interquartile range (IQR)=2,280-5,277 vs. 3,452 cells/mm², IQR=2,096-4,729; p=0.35] (Table III).

Although various prognostic markers for pancreatic cancer have been proposed, none have utilized the size of regional lymph nodes (18, 19). In this study, we demonstrated that enlargement of the N8LN was associated with favorable prognosis in patients undergoing PD for PC, independent of major clinicopathologic factors. Although lymph node enlargement is typically regarded as a radiologic sign of metastasis (8), our findings indicate that non-metastatic nodal enlargement may reflect biological processes distinct from tumor progression.

We initially hypothesized that enlarged N8LNs may indicate a heightened host immune response to tumor-derived antigens. While CD8⁺ T-cell density did not differ significantly between groups, cytotoxic T cells represent only one component of lymph node immunobiology. Lymphoid activation involves a broad range of immune and stromal elements, including dendritic cell maturation, B-cell expansion, germinal center activity, and tertiary lymphoid structures (20-22) that were not evaluated in this study. Thus, although our findings do not provide direct immunologic evidence, they remain compatible with the possibility that N8LN enlargement reflects aspects of immune stimulation not captured by CD8⁺ quantification alone.

Patients in the large N8LN group showed higher rates of preoperative biliary drainage and a trend toward elevated CRP levels, raising the possibility that inflammation contributes to nodal enlargement. Such inflammation may derive from cholangitis or stent-related irritation. However, inflammatory stimuli also activate antigen-presenting pathways and lymphoid hyperplasia, potentially enhancing host immune readiness. The large group also had a higher frequency of adjuvant chemotherapy; although this difference may introduce treatment-related confounding, N8LN size remained independently associated with outcome after adjustment. Established prognostic variables, including nodal metastasis and tumor differentiation, also remained significant, supporting the robustness of our findings (23). Taken together, these observations suggest that inflammation or treatment intensity may influence nodal size, but they do not negate, and may even support, the broader possibility that enlarged nodes reflect an immunologically active host environment.

Importantly, patients with pathologically confirmed metastasis to the N8LN were excluded because the goal of this study was not to establish N8LN size as a practical preoperative prognostic tool. Rather, we aimed to explore whether enlargement of non-metastatic nodes conveys biological information relevant to patient outcomes. Therefore, our findings should not be interpreted as evidence that N8LN size can discriminate metastatic from benign nodes on imaging. Instead, we propose that non-metastatic nodal enlargement may serve as a surrogate marker of host–tumor interactions.

These findings align with emerging evidence that regional lymph nodes function as active immunologic hubs rather than passive filters. For example, a study in stage III colorectal cancer reported that more extensive removal of regional lymph nodes was paradoxically associated with worse outcomes (24), suggesting that loss of immunologically active nodes may impair systemic anti-tumor immunity. Our results may represent a similar phenomenon in pancreatic cancer.

Study limitations. First, it was a retrospective, single-institution study with a relatively small sample size and a long enrollment period during which perioperative treatment strategies changed. The N8LN cutoff was derived from the cohort median and has not been externally validated, and interobserver reproducibility of the imaging measurements was not formally assessed. In addition, patients with pathologically metastatic N8LN were excluded; therefore, the findings apply only to pathologically non-metastatic nodes and should not be interpreted as establishing N8LN size as a preoperative diagnostic or prognostic marker. Differences in biliary drainage, inflammatory status, and adjuvant chemotherapy between groups may also have resulted in residual confounding.

Second, immunophenotyping was limited to CD8-positive T-cell density in a single lymph node section, whereas other immune and stromal components and other regional lymph nodes were not comprehensively evaluated. Imaging and inflammatory variables were incompletely characterized, and some clinical variables had missing data. Consequently, the observed association between larger N8LN size and survival does not establish causality or an immune-mediated mechanism and should be regarded as hypothesis-generating pending independent validation with broader immunologic, radiologic and inflammatory assessment.

Enlargement of the N8LN was independently associated with favorable survival in patients with PC undergoing PD. Although CD8⁺ T-cell density did not differ between groups, N8LN enlargement may still reflect underlying host biological responses, including, but not limited to, immune activation. While our findings do not establish lymph node size as a preoperative prognostic tool, they suggest that non-metastatic nodal enlargement may serve as a surrogate marker of host-tumor interactions. Further studies incorporating broader immunologic and inflammatory assessments are required to clarify the biological basis of this phenomenon and its potential implications for pancreatic cancer management.

The Authors have no related conflicts of interest to declare in relation to this study.

TS conceived the study concept and design, collected clinical data, conducted the immunohistochemistry, QuPath and statistical analyses, and drafted the manuscript. TG, MT, YS, NO, RH, YT, TA, KO and KF provided critical revisions to the study design, contributed to interpretation of oncologic outcomes, and reviewed the manuscript for important intellectual content. YYa and YYh contributed to the measurements of N8LN size. TI supervised the entire project, provided guidance on methodology and statistical interpretation. All Authors read and approved the final manuscript.

This work was supported by JSPS KAKENHI Grant Number JP25K11943.

During the preparation of this manuscript, a large language model (Microsoft Copilot) was used solely for language editing and stylistic improvements in select paragraphs. No sections involving the generation, analysis, or interpretation of research data were produced by generative AI. All scientific content was created and verified by the authors. Furthermore, no figures or visual data were generated or modified using generative AI or machine learning–based image enhancement tools.