Evaluating the prognostic significance of lymph node yield in resected pancreatic adenocarcinoma: a population-based study Authors: Chia-Jen Tsai, Pei-Yi Lin, Khaa-Hoo Ong, Chung-Han Ho, Yi-Chen Chen, Chien-Feng Li, Ching-Chieh Yang, Chih-Ying Lu Published: 2026 ABSTRACT IntroductionSurgical resection plays a pivotal role in the management of pancreatic cancer by achieving complete tumor removal and regional lymphadenectomy. Although most clinical guidelines recommend examining at least 15 lymph nodes for accurate staging and better prognostic assessment, the optimal threshold remains uncertain and requires validation across population-based datasets.MethodsUsing the Taiwan Cancer Registry and National Health Insurance Research Database, 885 patients with pancreatic adenocarcinoma who underwent curative resection between 2013 and 2019 were identified. Cox proportional hazards regression models were used to evaluate the association between lymph node yield (LNY ≥ 15 vs. <15) and overall survival (OS), adjusting for potential confounders. Several sensitivity analyses were performed to assess the robustness of the findings, including a subgroup analysis restricted to patients with pT1–3 pancreatic head tumors who underwent pancreatoduodenectomy. Stratified analyses were additionally performed according to pathological T and N classifications and treatment modalities.ResultsThe median LNY was 15, with 474 patients (53.6%) having ≥ 15 nodes retrieved. Patients with adequate LNY were more likely to have tumors located in the pancreatic head, undergo pancreatoduodenectomy, and present with more advanced pT and pN stages as well as a higher number of positive nodes. No significant difference in 1- or 3-year overall survival (OS) was observed between patients with LNY ≥ 15 and those with LNY < 15 in the overall cohort. In the predefined subgroup of patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy, however, retrieval of ≥ 15 lymph nodes was associated with significantly improved 1- and 3-year OS compared with retrieval of < 15 lymph nodes. After adjustment for confounders, LNY ≥ 15 remained associated with improved 1- and 3-year OS in this subgroup, whereas no such association was observed in the overall cohort. Stratified analyses by pathological stage and treatment demonstrated generally consistent findings, with the most pronounced survival benefit of LNY ≥ 15 observed among patients with pT1–2 disease.ConclusionsRetrieval of ≥ 15 lymph nodes were not directly associated with improved survival in the overall cohort but was associated with better survival among patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy. These findings suggest that the prognostic value of the conventional 15-node threshold may vary across clinical subgroups and warrants further validation in larger population-based studies.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12957-026-04475-4. Introduction Pancreatic cancer remains one of the most aggressive and lethal malignancies, with a 5-year overall survival rate of approximately 5% [1]. The poor prognosis is largely attributable to the fact that most patients present with either metastatic or locally advanced disease at diagnosis, rendering them ineligible for curative treatment [2]. Surgical resection remains the cornerstone and only potentially curative treatment option, providing the best opportunity for long-term survival. Among patients who undergo curative-intent resection followed by adjuvant therapy, the reported 5-year survival rate ranges from 15% to 25% [3]. However, only approximately 20% of patients present with resectable disease at diagnosis. Therefore, optimizing surgical quality and perioperative management remains essential to improve oncologic outcomes in patients with resectable pancreatic cancer. Among patients undergoing curative-intent resection, several clinicopathological factors have been consistently associated with recurrence and survival, including postoperative serum CA 19 − 9 levels, tumor size, pathological stage, resection margin status, and lymph node involvement [4]. Adequate lymph node evaluation is an essential component of high-quality pancreatic cancer surgery. The number of lymph nodes (LNs) examined, commonly referred to as lymph node yield (LNY), reflects both the extent of surgical lymphadenectomy and the thoroughness of pathological assessment. As a result, LNY may influence staging accuracy and prognostic stratification. In colorectal and gastric cancers, clinical guidelines recommend the examination of at least 12 and 16 LNs, respectively, to ensure reliable pathological staging [5, 6]. Similarly, several pancreatic cancer guidelines advocate the evaluation of at least 15 LNs following resection. However, whether achieving this threshold translates into improved prognostic assessment or represents merely a marker of surgical and pathological quality remains controversial. Therefore, the optimal LNY threshold and its prognostic significance in resected pancreatic adenocarcinoma warrant further investigation. Although inadequate lymph node retrieval may lead to pathological understaging and less accurate prognostic assessment, the clinical significance of LNY remains controversial. Previous studies have reported conflicting results, with some demonstrating improved outcomes associated with higher LNY, whereas others found no significant survival benefit [7–10]. These inconsistencies may be related to differences in patient characteristics, disease stage, tumor location, and treatment strategies. Therefore, we analyzed data from the Taiwan Cancer Registry (TCR) to evaluate the prognostic significance of LNY in patients undergoing curative-intent resection for pancreatic adenocarcinoma. Using the International Study Group on Pancreatic Surgery (ISGPS)-recommended threshold of 15 examined LNs, we aimed to assess whether this benchmark remains an appropriate indicator of staging adequacy and prognostic stratification in contemporary practice [11]. Materials and methods Study design and participants The clinicopathological data of patients with pancreatic cancer were obtained from the Taiwan Cancer Registry (TCR) and the National Health Insurance Research Database (NHIRD) [12]. Established in 1979, the TCR underwent substantial enhancements following the implementation of the Cancer Control Act in 2003. Since then, it has provided a highly comprehensive and reliable nationwide dataset, recognized for its excellent completeness and data quality, comparable to other leading national cancer registries [13–15]. Patients diagnosed with pancreatic cancer between January 2013 and December 2019 were identified from the Taiwan Cancer Registry using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) site codes C25.0, C25.1, C25.2, C25.3, C25.7, C25.8, and C25.9, and histology/behavior codes 8140/8500. Our study period spans years in which both the AJCC 7th and 8th editions were in use (the AJCC 8th edition was implemented in clinical practice starting in 2018). Because the staging definitions differ between the two editions, direct replacement with a single staging system for all patients could introduce misclassification. Therefore, to ensure consistency across the entire cohort, we used harmonized stage groupings based on shared pathological features (pT1–2 vs. pT3–4 and pN0 vs. pN1-2), allowing unified analysis while preserving comparability across calendar years. Follow-up was calculated from the date of pancreatic cancer diagnosis to December 31, 2021. The dataset included comprehensive clinicopathological variables such as age, sex, tumor location, histologic grade, surgical procedure, clinical/pathological stage, margin status, lymph node count, comorbidities, receipt of chemotherapy or radiotherapy, and personal habits. Comorbidity severity was assessed using the Charlson Comorbidity Index (CCI), as previously described [16, 17]. Thus, a. total of 4,603 patients with pancreatic adenocarcinoma and available treatment information were initially screened. Among these, patients who did not undergo surgical resection were excluded, and 1,025 patients who received surgery were identified. We further excluded patients with incomplete treatment records, unclear pathological classification, or missing information on lymph node retrieval (n = 135), as well as those lacking data on smoking and alcohol consumption (n = 5). Additionally, patients with a prior malignancy or evidence of metastatic disease at diagnosis were excluded. After applying these criteria, a total of 885 patients who underwent curative-intent resection for pancreatic cancer and had available lymph node data were included in the final analysis. As shown in Fig. 1, these patients were subsequently categorized based on the number of LNs examined into two groups: <15 LNs (n = 411) and ≥ 15 LNs (n = 474). Fig. 1Flow chart of study patients Flow chart of study patients Statistical analysis This retrospective, population-based study investigated the impact of LNY on survival among patients with resected pancreatic cancer, comparing groups with ≥ 15 and < 15 retrieved nodes. Overall survival (OS) probabilities were evaluated for patients with a minimum follow-up of 1 year and up to 3 years. Categorical variables were summarized as frequencies with percentages and compared between groups using Pearson’s chi-square test. Diagnosed age and the number of positive lymph node was expressed as the median with interquartile range (IQR) and compared using the Wilcoxon rank-sum test. Cox proportional hazard regression was performed to estimate the relative risk ratio of survival associated with LNY (≥ 15 vs. <15), adjusted for potential confounders [18]. Collinearity diagnostics demonstrated no significant multicollinearity among the covariates included in the multivariable model, with all variance inflation factor (VIF) values below 5 (Supplementary Table 2). For the effect of LNY when adjusting for multiple, potentially more influential variables, including age group, sex, location, grade, pT classification, pN classification, smoking, alcohol, neoadjuvant therapy and adjuvant therapy, margin positive, and CCI group, the model showed good discriminative power with a C-index of 0.70. Furthermore, it supported the stability of the model’s predictive accuracy (ROC of one-year = 0.7336, ROC of 3-year = 0.7264). The survival rate in the two groups over time was performed using the Kaplan–Meier method with a log-rank test to determine the differences. The proportional hazards assumption was evaluated using Schoenfeld residuals, which indicated no evidence of violation (global test p = 0.142). Several sensitivity analyses were performed to evaluate the robustness of the findings. Because LNY may be influenced by tumor characteristics, anatomical location, and surgical procedure, a subgroup analysis was included as a sensitivity test. This analysis restricted patient with pT1–3 tumors located in the pancreatic head who underwent pancreatoduodenectomy, which minimize the residual bias. Additional stratified analyses were performed according to pathological T and N classifications and adjuvant treatment status. All statistical analyses were conducted using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA) and Stata (15, StataCorp LLC, College Station, TX). A two-sided p-value of < 0.05 was considered statistically significant. Results Clinicopathological characteristics Clinicopathological characteristics of patients with resected pancreatic cancer are summarized in Table 1 and Supplement Table 1. Between 2013 and 2019, a total of 885 patients were included, comprising 496 men (56%) and 389 women (44%). The median age at diagnosis was 64 years (Q1–Q3: 57–71), and the median follow-up duration was 1.30 years (Q1–Q3: 0.79–1.96). Tumors were most frequently located in the pancreatic head (77.6%), followed by the tail (8.5%), body (7.3%), and other sites (6.5%). Overall, 104 patients (11.8%) received neoadjuvant therapy, and 571 (64.5%) underwent adjuvant chemotherapy, with or without radiotherapy. The median number of examined LNs was 15 (Q1–Q3: 9–22). Of the entire cohort, 474 patients (53.6%) had ≥ 15 retrieved LNs, whereas 411 (46.4%) had < 15 LNs. Compared with patients who had fewer than 15 LNs examined, those with adequate LNY (≥ 15) were more likely to have tumors located in the pancreatic head, to have undergone pancreatoduodenectomy, and to present with more advanced pathological T/N classifications, and greater numbers of positive LNs. Table 1Characteristics of patients with pancreatic cancer undergoing surgery between groups based on the number of lymph nodes (LNs) removed, n = 885The number of LNs removed, n (%)< 15, n = 411≧ 15, n = 474P-valueAge, years Median (Q1-Q3)65 (58–73)64 (57–71) < 65202 (49.15)249 (52.53)0.315 ≧ 65209 (50.85)225 (47.47)Sex Male246 (59.85)250 (52.74) 0.034 Female165 (40.15)224 (47.26)Location Head284 (69.10)403 (85.02) < 0.001 Body48 (11.68)27 (5.70) Tail52 (12.65)13 (2.74) Others27 (6.57)31 (6.54)Grade Well43 (10.46)33 (6.96)0.060 Moderately293 (71.29)338 (71.31) Poorly67 (16.30)82 (17.30) Not stated8 (1.95)21 (4.43)Operation Pacreatoduodenectomy303 (73.72)411 (86.71) < 0.001 Distal pancreatectomy94 (22.87)35 (7.38) Total pancreatectomy/ Other14 (3.41)28 (5.91)pT classification Early (1–2)119 (28.95)186 (39.24) 0.001 Advanced (3–4)292 (71.05)288 (60.76)pN classification 0162 (39.42)126 (26.58) < 0.001 ≧ 1249 (60.58)348 (73.42)AJCC TNM edition- 7th283 (68.86)240 (50.63) 8th128 (31.14)234 (49.37)Neoadjuvant therapy52 (12.65)52 (10.97)0.439Adjuvant therapy250 (60.83)321 (67.72) 0.033 Margin positive123 (29.93)126 (26.58)0.270Smoking history117 (28.47)127 (26.79)0.578Alcohol history85 (20.68)100 (21.10)0.879CCI score 0117 (28.47)160 (33.76)0.187 1121 (29.44)138 (29.11) ≧ 2173 (42.09)176 (37.13)Positive Lymph node count Median (Q1-Q3)1 (0–2)2 (0–4) < 0.001 Lymph node ratio 0-0.25304 (73.97)419 (88.40) < 0.001 0.26–0.5077 (18.73)49 (10.34) 0.56–0.7526 (6.33)6 (1.27) 0.76-1.004 (0.97)0 (0.00)P-value was derived from Pearson’s chi-square test for categorical variables, and Wilcoxon rank sum test for median of continuous variablesBold values indicate statistical significance (p < 0.05) Table 1Characteristics of patients with pancreatic cancer undergoing surgery between groups based on the number of lymph nodes (LNs) removed, n = 885The number of LNs removed, n (%)< 15, n = 411≧ 15, n = 474P-valueAge, years Median (Q1-Q3)65 (58–73)64 (57–71) < 65202 (49.15)249 (52.53)0.315 ≧ 65209 (50.85)225 (47.47)Sex Male246 (59.85)250 (52.74) 0.034 Female165 (40.15)224 (47.26)Location Head284 (69.10)403 (85.02) < 0.001 Body48 (11.68)27 (5.70) Tail52 (12.65)13 (2.74) Others27 (6.57)31 (6.54)Grade Well43 (10.46)33 (6.96)0.060 Moderately293 (71.29)338 (71.31) Poorly67 (16.30)82 (17.30) Not stated8 (1.95)21 (4.43)Operation Pacreatoduodenectomy303 (73.72)411 (86.71) < 0.001 Distal pancreatectomy94 (22.87)35 (7.38) Total pancreatectomy/ Other14 (3.41)28 (5.91)pT classification Early (1–2)119 (28.95)186 (39.24) 0.001 Advanced (3–4)292 (71.05)288 (60.76)pN classification 0162 (39.42)126 (26.58) < 0.001 ≧ 1249 (60.58)348 (73.42)AJCC TNM edition- 7th283 (68.86)240 (50.63) 8th128 (31.14)234 (49.37)Neoadjuvant therapy52 (12.65)52 (10.97)0.439Adjuvant therapy250 (60.83)321 (67.72) 0.033 Margin positive123 (29.93)126 (26.58)0.270Smoking history117 (28.47)127 (26.79)0.578Alcohol history85 (20.68)100 (21.10)0.879CCI score 0117 (28.47)160 (33.76)0.187 1121 (29.44)138 (29.11) ≧ 2173 (42.09)176 (37.13)Positive Lymph node count Median (Q1-Q3)1 (0–2)2 (0–4) < 0.001 Lymph node ratio 0-0.25304 (73.97)419 (88.40) < 0.001 0.26–0.5077 (18.73)49 (10.34) 0.56–0.7526 (6.33)6 (1.27) 0.76-1.004 (0.97)0 (0.00)P-value was derived from Pearson’s chi-square test for categorical variables, and Wilcoxon rank sum test for median of continuous variablesBold values indicate statistical significance (p < 0.05) Characteristics of patients with pancreatic cancer undergoing surgery between groups based on the number of lymph nodes (LNs) removed, n = 885 P-value was derived from Pearson’s chi-square test for categorical variables, and Wilcoxon rank sum test for median of continuous variables Bold values indicate statistical significance (p < 0.05) Analysis for LNs on survival Kaplan–Meier survival curves were generated to compare the 3-year OS by different LNY ( ≧ 15 and < 15) groups. As shown in Fig. 2, while there was no significant difference in the overall survival (OS), there was a significant distinction observed in subgroup patients identified by pT1-3 classification, particularly those located at the head and undergoing pancreatoduodenectomy. Log-rank tests confirmed that subgroup patients with 15 or more LNs had better survival compared to those with fewer than 15 LNs (p = 0.033). After adjustment for clinicopathological and treatment-related factors (Table 2 and Supplementary Table 3), retrieval of ≥ 15 LNs was not associated with improved 1- or 3-year overall survival in the overall cohort (adjusted hazard ratio [aHR] 0.84, 95% CI 0.65–1.08; and aHR 0.88, 95% CI 0.75–1.03, respectively). However, in the predefined subgroup cohort, retrieval of ≥ 15 LNs remained significantly associated with reduced 1- and 3-year mortality (aHR 0.70, 95% CI 0.52–0.94; and aHR 0.82, 95% CI 0.68–0.99, respectively). Stratified analyses by AJCC pT/pN classification and treatment modality (Fig. 3 and Supplement Table 4, and 5) likewise showed no significant survival advantage associated with retrieval of ≥ 15 LNs at 1 or 3 years, except among patients with pT1–2 disease. Fig. 2Probability of 3-years survival rate between groups based on the number of lymph nodes (LNs) removed in (A) patients with pancreatic adenocarcinoma and (B) the subgroup of patients, classified as pT1-3 classification, located at the head, undergoing pancreatoduodenectomy Probability of 3-years survival rate between groups based on the number of lymph nodes (LNs) removed in (A) patients with pancreatic adenocarcinoma and (B) the subgroup of patients, classified as pT1-3 classification, located at the head, undergoing pancreatoduodenectomy Table 2The risk of mortality within 1, and 3 years between groups based on the number of lymph nodes (LNs) removed in patients with pancreatic adenocarcinoma and the subgroup of patient, classified as pT1-3 classification, located at the head, undergoing pancreatoduodenectomyOverall, n = 885Subgroup, n = 645LNs removed, n (%)Adjusted HRaLNs removed, n (%)Adjusted HRa< 15 (reference)≧ 15(95% C.I.)< 15 (reference)≧ 15(95% C.I.)Time to death, years 1 - year mortality128 (31.14)133 (28.06)0.84 (0.65–1.08)96 (35.82)97 (25.73)0.70 (0.52–0.94) 3 - years mortality314 (76.40)350 (73.84)0.88 (0.75–1.03)209 (77.99)278 (73.74)0.82 (0.68–0.99)a The cox proportional hazard regression was used to estimate the relative risk between the number of LNs removed ≧ 15 and < 15 and adjusted the baseline information of patient including, age group, sex, location, grade, pT classification, pN classification, smoking, alcohol, neoadjuvant therapy and adjuvant therapy, margin status, and CCI group Table 2The risk of mortality within 1, and 3 years between groups based on the number of lymph nodes (LNs) removed in patients with pancreatic adenocarcinoma and the subgroup of patient, classified as pT1-3 classification, located at the head, undergoing pancreatoduodenectomyOverall, n = 885Subgroup, n = 645LNs removed, n (%)Adjusted HRaLNs removed, n (%)Adjusted HRa< 15 (reference)≧ 15(95% C.I.)< 15 (reference)≧ 15(95% C.I.)Time to death, years 1 - year mortality128 (31.14)133 (28.06)0.84 (0.65–1.08)96 (35.82)97 (25.73)0.70 (0.52–0.94) 3 - years mortality314 (76.40)350 (73.84)0.88 (0.75–1.03)209 (77.99)278 (73.74)0.82 (0.68–0.99)a The cox proportional hazard regression was used to estimate the relative risk between the number of LNs removed ≧ 15 and < 15 and adjusted the baseline information of patient including, age group, sex, location, grade, pT classification, pN classification, smoking, alcohol, neoadjuvant therapy and adjuvant therapy, margin status, and CCI group The risk of mortality within 1, and 3 years between groups based on the number of lymph nodes (LNs) removed in patients with pancreatic adenocarcinoma and the subgroup of patient, classified as pT1-3 classification, located at the head, undergoing pancreatoduodenectomy a The cox proportional hazard regression was used to estimate the relative risk between the number of LNs removed ≧ 15 and < 15 and adjusted the baseline information of patient including, age group, sex, location, grade, pT classification, pN classification, smoking, alcohol, neoadjuvant therapy and adjuvant therapy, margin status, and CCI group Fig. 3The risk of mortality within 1, and 3 years between groups based on the number of lymph nodes (LNs) removed in patients with pancreatic adenocarcinoma in different (A) pT and (B) pN classification The risk of mortality within 1, and 3 years between groups based on the number of lymph nodes (LNs) removed in patients with pancreatic adenocarcinoma in different (A) pT and (B) pN classification Discussion Surgical resection remains the cornerstone of curative treatment for pancreatic cancer. Accurate pathological staging is essential for prognostic assessment and postoperative treatment planning [19]. Therefore, adequate LNs evaluation remains an important component of high-quality pancreatic cancer surgery. Insufficient LNs retrieval may result in pathological understaging and reduce the accuracy of nodal classification [20]. In the present nationwide cohort, examination of ≥ 15 LNs was not associated with improved overall survival in the entire study population. However, an exploratory subgroup analysis demonstrated a significant association between higher LN yield and lower mortality among patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy. Taken together, our findings suggest that the conventional 15-node benchmark may have greater clinical relevance in selected surgical populations and support its role as an indicator of staging adequacy and surgical quality rather than as an independent determinant of survival. This study has several strengths. First, by utilizing the Taiwan Cancer Registry, which captures more than 95% of cancer cases nationwide, we were able to analyze a large, population-based cohort with comprehensive follow-up, enhancing the generalizability of our findings. Second, the inclusion of patients diagnosed between 2013 and 2019 reflects contemporary treatment practices. In addition, the database provides detailed clinicopathological information, including surgical procedures, lymph node yield, pathological stage, resection margin status, and perioperative treatments, allowing adjustment for important prognostic factors. Finally, linkage with nationwide healthcare databases enabled more complete assessment of comorbidities, treatment patterns, and survival outcomes. These strengths provide valuable real-world evidence regarding the clinical relevance of lymph node yield in resected pancreatic cancer. Several professional organizations have proposed recommendations for adequate lymph node evaluation in pancreatic cancer. The International Study Group on Pancreatic Surgery (ISGPS) recommends retrieval of at least 15 lymph nodes, while both the College of American Pathologists (CAP) and the 8th edition of the American Joint Committee on Cancer (AJCC) staging system recommend examination of at least 12 nodes [11, 21, 22]. The National Comprehensive Cancer Network (NCCN) also acknowledges that evaluation of approximately 11–17 lymph nodes may optimize staging accuracy [23]. Consistent with these recommendations, previous studies have suggested adequate lymph node yields ranging from 10 to 21 nodes [24–27]. Despite some variation, a threshold of approximately 15 lymph nodes has been widely adopted as a benchmark for adequate nodal assessment and an indicator of surgical and pathological quality. Therefore, we selected 15 examined lymph nodes as the predefined cutoff in the present study [26, 27]. Using this predefined threshold, adequate lymph node evaluation (≥ 15 nodes) was more frequently achieved in patients with pancreatic head tumors, those undergoing pancreatoduodenectomy, and those with more advanced pathological stages. Although a higher lymph node yield was associated with the identification of more positive lymph nodes, no significant survival advantage was observed in the overall cohort. Nevertheless, an exploratory subgroup analysis demonstrated that examination of ≥ 15 lymph nodes was associated with lower 1-year and 3-year mortality among patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy. These findings suggest that the primary value of adequate lymph node retrieval may lie in improving staging accuracy and quality assessment, although a potential prognostic association cannot be excluded in selected patient subgroups. Importantly, the purpose of this study was not to identify a new optimal lymph node yield threshold, but rather to validate the commonly recommended benchmark of examining at least 15 LNs in a nationwide real-world cohort. By applying a predefined threshold derived from existing guidelines and consensus recommendations, our findings can be directly compared with prior studies while avoiding the bias associated with data-driven cut-off selection. Nevertheless, despite multivariable adjustment, differences in baseline characteristics between patients with higher and lower LNY remained. Therefore, residual confounding related to tumor location, surgical procedure, pathological assessment, institutional practice patterns, and treatment selection cannot be completely excluded. Accordingly, the observed findings should be interpreted primarily as supporting the role of LNY as an indicator of staging adequacy and surgical quality, rather than evidence of a direct causal effect on survival. Several studies have reported an association between higher lymph node yield and improved survival following pancreatic cancer resection, particularly among node-negative patients [28–31]. These findings have been attributed primarily to improved staging accuracy rather than a direct therapeutic effect of more extensive lymph node dissection. In contrast, our nationwide analysis did not demonstrate a significant survival advantage associated with examination of ≥ 15 lymph nodes in the overall cohort, although a potential association was observed in a selected subgroup of patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy. The discrepancy between our findings and previous reports may reflect differences in patient populations, tumor characteristics, surgical procedures, pathological assessment, and the increasing use of multimodality treatment. Taken together, our results suggest that lymph node yield may be more appropriately viewed as a marker of staging adequacy and surgical quality rather than an independent determinant of survival. Although no significant association between lymph node yield and overall survival was observed in the overall cohort, an exploratory association was identified in selected subgroups. Similar findings have been reported in previous studies, which suggested that the prognostic relevance of lymph node yield may be more evident in patients with early-stage or node-negative disease [27, 32, 33]. These observations support the possibility that the clinical impact of lymph node retrieval is not uniform across all patients and may depend on tumor characteristics and surgical context. However, given the exploratory nature of these subgroup findings, further validation is warranted before definitive conclusions can be drawn. An additional consideration is the potential influence of temporal changes in staging systems and treatment practices during the study period. In sensitivity analyses stratified by AJCC staging era, LNY ≥ 15 was not significantly associated with either 1-year mortality (adjusted HR 0.96, 95% CI 0.69–1.34) or 3-year mortality (adjusted HR 0.92, 95% CI 0.75–1.13) during the AJCC 7th edition period. In contrast, during the AJCC 8th edition period, LNY ≥ 15 was associated with lower 1-year mortality (adjusted HR 0.63, 95% CI 0.42–0.94) and 3-year mortality (adjusted HR 0.71, 95% CI 0.51–0.93). These findings suggest that the prognostic relevance of the 15-node benchmark may be more apparent in contemporary practice. The observed differences may reflect stage migration, improvements in pathological assessment, enhanced surgical quality, and evolving multimodality treatment strategies. Therefore, the clinical significance of LNY should be interpreted within the context of changing pancreatic cancer management and staging systems. Limitations This study has several limitations. First, although the Taiwan Cancer Registry is a nationwide, population-based database, detailed center-level information regarding surgical techniques, vascular reconstruction, and pathological processing was unavailable. Nevertheless, more than three-quarters of patients were treated at medical centers, suggesting a relatively consistent standard of surgical and pathological care. Second, certain treatment-related variables, including specific chemotherapy regimens, radiotherapy details, and salvage treatments, were not captured in the registry and may have contributed to residual confounding. Third, information regarding preoperative anatomical resectability was unavailable; therefore, pathological stage was used as a surrogate measure of disease extent. In addition, cancer-specific mortality data were not available, precluding cancer-specific survival analyses. Although age, CCI, and other clinical covariates were adjusted for in the multivariable model, residual influence from non-cancer mortality cannot be completely excluded. Finally, additional subgroup analyses according to surgical procedure, treatment modality, or comorbidity burden were not performed. While such analyses may provide further insight, extensive stratification could reduce statistical power and yield unstable estimates. Moreover, because lymph node yield is inherently influenced by surgical technique and pathological assessment, subgroup-specific findings would require cautious interpretation. Despite these limitations, our study provides valuable real-world evidence regarding the clinical significance of lymph node yield in resected pancreatic adenocarcinoma. Further prospective studies are warranted to validate these findings and determine whether the optimal extent of lymph node evaluation varies across different clinical settings. Conclusion In conclusion, examination of ≥ 15 LNs was not associated with improved overall survival in the overall cohort. However, patients with pT1–3 pancreatic head tumors undergoing pancreatoduodenectomy have better survival. These findings support the role of LNY as an indicator of staging adequacy and surgical quality rather than an independent determinant of survival. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.