The relationship between examined lymph nodes and oncological outcome of proximal gastrectomy for Siewert II/III adenocarcinoma of the esophagogastric junction Authors: Haikuo Wang, Xinxin Shao, Yiming Lu, Haitao Hu, Yujuan Jiang, Wangyao Li, Quan Xu, Yantao Tian Published: 2025 ABSTRACT ObjectiveThe objective of this study was to summarize the oncological outcomes and long-term prognosis of patients with Siewert II/III adenocarcinoma of the esophagogastric junction (AEG) who underwent proximal gastrectomy (PG) at the China National Cancer Center (CNCC), with particular focus on the relationship between the number of examined lymph nodes (ELNs) and prognosis.MethodsWe retrospectively analyzed 242 patients with Siewert II/III AEG who underwent PG between 2017 and 2021 at the CNCC. Lymphadenectomy followed D1 + principles for early-stage and D2 for advanced-stage disease. Recurrence-free survival (RFS) and overall survival (OS) were assessed using Kaplan–Meier curves, and prognostic factors were identified through Cox regression. Restricted cubic spline (RCS) modeling was applied to explore the dose–response relationship between ELNs and OS.ResultsOf the 242 patients, 163 (67.4%) did not achieve textbook outcome with ≥ 30 ELNs (TO-30LN) and had significantly worse survival (OS: HR = 1.85, 95% CI: 1.08–3.18; RFS: HR = 1.88, 95% CI: 1.06–3.34; both P < .05). Multivariable analysis identified pN + status and Clavien–Dindo grade ≥ II complications as independent adverse prognostic factors. RCS analysis revealed a nonlinear association between ELN count and OS in cT2–3 disease, with an optimal range of 28–53 nodes, whereas no survival benefit was observed in cT4 disease.ConclusionFor Siewert II/III AEG following PG, achieving TO-30LN improved survival. RCS analysis revealed an optimal ELN threshold in cT2–3 patients, but no benefit in cT4, indicating limited suitability of PG for this subgroup.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12957-025-04062-z. Introduction Although the incidence of gastric cancer (GC) has gradually declined over the past two decades, its global burden remains considerable. In 2022, an estimated 968,784 new cases and 660,175 deaths were reported worldwide, ranking GC fifth in incidence and mortality, and third in cancer-related mortality in China [1]. A 2018 analysis of global cancer registry data for gastric and esophageal cancers indicated that East Asia accounted for 67.1% of all new cardia cancer cases worldwide [2]. Furthermore, a recent study from China reports that gastroesophageal junction cancers now constitute approximately 38.2% of all gastric cancer cases (2017–2023), a marked increase compared to 22.3% in 1988–1992 and 35.7% in 2008–2012, indicating a rising trend in AEG incidence in recent years [3]. Total gastrectomy (TG) has long been the standard surgical approach for Siewert type II/III AEG, but the low early diagnosis rate in China and the inevitable nutritional complications after TG have led to an increasing number of advanced cases undergoing PG in clinical practice [4, 5]. The central controversy surrounding PG is whether it can achieve oncological radicality, particularly with respect to adequate lymphadenectomy. The number of examined lymph nodes (ELNs) is a cornerstone for both accurate pathological staging and reliable prognostic assessment However, the actual yield of ELNs is subject to multiple influencing factors, including the surgeon’s technique and extent of lymphadenectomy, the pathologist’s diligence in specimen retrieval, and patient-related characteristics such as obesity or prior abdominal surgery [6]. According to the Japanese Gastric Cancer Association (JGCA) and the TNM system, a minimum of 16 ELNs is required for accurate staging [7], whereas the Chinese Society of Clinical Oncology (CSCO) recommends examining preferably 30 nodes [8]. A large study integrating SEER data with a Chinese cohort demonstrated that higher ELN counts were associated with improved staging accuracy and survival, and proposed 18 as an optimal cutoff for evaluating dissection quality and prognostic value [9]. However, the extent of lymph node dissection differs between total gastrectomy and gastric-preserving procedures, leading to variations in the ideal ELN threshold. Moreover, textbook outcome (TO) has emerged as an important composite indicator of surgical quality [10], in which an adequate number of examined lymph nodes (ELNs) is a key determinant. Most previous studies have adopted 15–16 examined lymph nodes (ELNs) as the ideal threshold, yet their prognostic value remains uncertain [11, 12]; moreover, this cutoff may not fully capture the adequacy of lymphadenectomy. To further explore the association between ELN count and OS, we retrospectively analyzed patients with Siewert II/III AEG from the CNCC database (2017–2021), with particular focus on evaluating TO definitions based on different ELN thresholds and determining the optimal cutoff for PG, as well as identifying stage-specific ELN ranges associated with the greatest survival benefit. Methods Inclusion and exclusion criteria Figure 1 presents the inclusion and exclusion criteria and the analysis process of this study. From our institutional case database, 299 patients who underwent PG at the CNCC from January 2017 to December 2021 were identified. Among them, 21 patients with proximal GC not classified as Siewert type II/III, 8 with gastrointestinal stromal tumors (GIST), 8 with non-adenocarcinoma histology, 5 with distant metastasis, and 15 who were lost to follow-up or had incomplete data were excluded. Ultimately, 242 patients with Siewert II/III AEG were included in the final analysis. Fig. 1Flowchart of patient selection, exclusion criteria and analytical methods Flowchart of patient selection, exclusion criteria and analytical methods The study protocol was approved by the Ethics Committee of the Cancer Hospital, Chinese Academy of Medical Sciences (25/116–5062), and conducted in accordance with the STROCSS 2021 guidelines [13]. Surgical procedure In our department, only AEG tumors located within 2 cm distal to the esophagogastric junction (EGJ) were treated, and all resections were performed via a transabdominal approach. For patients with advanced disease (cT3–4 and/or N+), treatment strategies included either neoadjuvant therapy (NAT) followed by surgery or upfront surgery with adjuvant chemotherapy, depending on clinical assessment. The indications for PG were: (i) preservation of more than half of the stomach remnant was anticipated; (ii) preoperative radiological evaluation confirmed the absence of lymph node metastasis at stations No. 4 d, 5, 6, and 12a, and PG was considered adequate for achieving oncological lymphadenectomy; and (iii) patients expressed a strong preference for gastric function preservation. All postoperative specimens were reviewed, and lymph nodes classified by two dedicated surgeons. The extent of lymphadenectomy was tailored according to the Japanese Gastric Cancer Association (JGCA) guidelines [4, 14, 15], and the patient’s clinical status, as follows: for cT1N0, D1 + dissection was performed (including stations No. 1, 2, 3, 4sa, 4sb, 7, 8a, 9, and 11p); for cT1N + and cT2 + N0/+, D2 dissection was performed (D1 + plus No. 11 d); and when the tumor invaded the esophagus, No. 19 and 20 were also dissected. Definitions and research variables Siewert type II/III classification was determined by integrating findings from preoperative endoscopy and postoperative specimen evaluation. Clinical tumor staging was assessed preoperatively using endoscopic ultrasound (EUS) and contrast-enhanced computed tomography (CT). The definition of textbook outcome (TO) was adopted from the Dutch Upper Gastrointestinal Cancer Audit (DUCA, 2017) [16], which requires fulfillment of the following ten criteria: curative-intent tumor resection, R0 resection, retrieval of at least 16 lymph nodes, absence of intraoperative adverse events, no severe postoperative complications, no reintervention, no unplanned ICU admission, postoperative hospital stay shorter than 14 days, no readmission due to complications within 30 days, and no mortality within 30 days postoperatively. According to the 2021 CSCO guidelines, at least 30 lymph nodes should ideally be examined to ensure optimal N-staging accuracy [8]. Therefore, in this study, the lymph node requirement in the TO definition was modified from 16 nodes (TO-16LN) to 30 nodes, and the revised criteria were designated as TO-30LN. Severe postoperative complications were defined as Clavien–Dindo grade II or higher [17]. Reintervention was defined as any additional surgery, endoscopic or radiologic intervention, or invasive procedure under local anesthesia (abdominal or thoracic puncture). Follow‑up In this study, three follow-up staff conducted postoperative follow-ups with patients by reviewing outpatient records and through telephone follow-ups to assess patient prognosis. The last follow-up was conducted on October 1, 2024. Our institution’s follow-up protocol recommends that patients return for a check-up every 3 months at a local hospital or our institution during the first two years after surgery, and every 6 months within five years. Follow-up evaluations include chest-abdomen-pelvis CT scans, endoscopy, and tumor markers. RFS was defined as the time from initial treatment to tumor recurrence, metastasis, loss to follow-up, or death. OS was defined as the time from initial treatment to death or loss to follow-up. Restricted cubic spline analysis To investigate the potential nonlinear relationship between ELNs and survival outcomes, we performed restricted cubic spline (RCS) analysis within a Cox proportional hazards regression framework. Four knots were placed at prespecified percentiles (5th, 35th, 65th, and 95th) of the ELNs distribution, as commonly recommended in methodological literature [18]. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated across the range of ELNs, and the nonlinear association was formally tested using the Wald χ² test for the spline terms. RCS curves were plotted to visualize the adjusted relationship between ELNs and OS, both in the overall cohort and in the subgroup of patients who underwent upfront surgery without NAT. Statistical analysis The primary endpoint was OS, with secondary endpoints including RFS and the optimal range of ELN associated with OS. Between-group variables were compared using the t-test and chi-square test. Survival curves were constructed using KM analysis. Cox regression analysis was employed to assess risk factors for survival and recurrence, and variables with a p-value < 0.05 in univariate analysis were entered into the multivariate model. A two-sided p-value < 0.05 was considered statistically significant. All data analyses were conducted using R software version 4.4.0 (http://www.r-project.org/). Results Baseline characteristics Table 1 presents the baseline characteristics of the study cohort. A total of 242 patients were included in this study, of whom 194 were male, accounting for 80.2%, and 48 were female, accounting for 19.8%. There were 143 patients with Siewert II, accounting for 59.1%, and 99 patients with Siewert III, accounting for 40.9%. There were 48 patients with a clinical T1 stage, constituting only 19.8% of the total. Laparoscopic surgery was performed in 158 patients, accounting for 65.3%. The primary esophagogastrostomy technique was end-to-side esophagogastrostomy in 184 patients (76%), followed by tubular gastroesophagostomy in 36 patients (14.9%). Table 1Baseline characteristics of the study cohortVariableOverall (N = 242)Gender F48 (19.8%) M194 (80.2%)Age (y.o.)63.0 [21.0, 84.0]BMI (kg/m2)24.1 [17.0, 35.5]ASA 1 + 2217 (89.7%) 325 (10.3%)NAT No219 (90.5%) Yes23 (9.5%)R0 No4 (1.7%) Yes238 (98.3%)Tumor Size (cm) <4138 (58.0%) ≥ 4100 (42.0%)Esophageal invasion No140 (57.9%) Yes102 (42.1%)Siewert II143 (59.1%) III99 (40.9%)Silver ring cell carcinoma No218 (90.1%) Yes24 (9.9%)Lauren Diffuse48 (19.8%) Intestinal119 (49.2%) Mixed75 (31.0%)LVI No121 (50.0%) Yes121 (50.0%)PNI No111 (45.9%) Yes131 (54.1%)Previous abdominal surgery No203 (83.9%) Yes39 (16.1%)Histopathological type Medium137 (56.6%) Poor97 (40.1%) Well8 (3.3%)cT stage T148 (19.8%) T2-3121 (50.0%) T473 (30.2%)cN stage N0125 (51.7%) N+117 (48.3%)(y)pT stage T0-173 (30.2%) T2-3102 (42.1%) T467 (27.7%)(y)pN stage N0130 (53.7%) N+112 (46.3%)Type Laparoscopy158 (65.3%) Open84 (34.7%)Abbreviations: F Female. M Male. y.o. Years Old, BMI Body Mass Index, ASA American Society of Anesthesiologists, CCI Charlson Comorbidity Index, LVI Lymphovascular Invasion, PNI Perineural Invasion, (y)pT stage Pathological T Stage (after Neoadjuvant Therapy), (y)pN stage Pathological N Stage (after Neoadjuvant Therapy), NAT Neoadjuvant Therapy Table 1Baseline characteristics of the study cohortVariableOverall (N = 242)Gender F48 (19.8%) M194 (80.2%)Age (y.o.)63.0 [21.0, 84.0]BMI (kg/m2)24.1 [17.0, 35.5]ASA 1 + 2217 (89.7%) 325 (10.3%)NAT No219 (90.5%) Yes23 (9.5%)R0 No4 (1.7%) Yes238 (98.3%)Tumor Size (cm) <4138 (58.0%) ≥ 4100 (42.0%)Esophageal invasion No140 (57.9%) Yes102 (42.1%)Siewert II143 (59.1%) III99 (40.9%)Silver ring cell carcinoma No218 (90.1%) Yes24 (9.9%)Lauren Diffuse48 (19.8%) Intestinal119 (49.2%) Mixed75 (31.0%)LVI No121 (50.0%) Yes121 (50.0%)PNI No111 (45.9%) Yes131 (54.1%)Previous abdominal surgery No203 (83.9%) Yes39 (16.1%)Histopathological type Medium137 (56.6%) Poor97 (40.1%) Well8 (3.3%)cT stage T148 (19.8%) T2-3121 (50.0%) T473 (30.2%)cN stage N0125 (51.7%) N+117 (48.3%)(y)pT stage T0-173 (30.2%) T2-3102 (42.1%) T467 (27.7%)(y)pN stage N0130 (53.7%) N+112 (46.3%)Type Laparoscopy158 (65.3%) Open84 (34.7%)Abbreviations: F Female. M Male. y.o. Years Old, BMI Body Mass Index, ASA American Society of Anesthesiologists, CCI Charlson Comorbidity Index, LVI Lymphovascular Invasion, PNI Perineural Invasion, (y)pT stage Pathological T Stage (after Neoadjuvant Therapy), (y)pN stage Pathological N Stage (after Neoadjuvant Therapy), NAT Neoadjuvant Therapy Baseline characteristics of the study cohort Abbreviations: F Female. M Male. y.o. Years Old, BMI Body Mass Index, ASA American Society of Anesthesiologists, CCI Charlson Comorbidity Index, LVI Lymphovascular Invasion, PNI Perineural Invasion, (y)pT stage Pathological T Stage (after Neoadjuvant Therapy), (y)pN stage Pathological N Stage (after Neoadjuvant Therapy), NAT Neoadjuvant Therapy Follow-up Figure 2 illustrates the survival curves for the overall cohort and each staging group. The median follow-up time for the entire cohort was 67.1 months (95% CI: 63.7–71). The 5-year OS rate for the entire cohort was 69.5%, and the 3-year OS rate was 83.5%. Survival in cT4 patients was significantly worse than in cT1 patients [cT4 vs. cT1 OS: HR = 2.26 (1.11–4.63), p =.03] (Fig. 2A). The 5-year RFS rate was 72.3%, and the 3-year RFS rate was 78.7%. The risk of recurrence in the entire cohort was significantly associated with increasing cT stage, with the recurrence risk for advanced AEG being 4.55-fold that of the cT1 stage [HR = 4.55 (1.66–12.52)], the recurrence risk for cT2-3 being 3.74-fold that of cT1 [HR = 3.74 (1.33–10.57)], and the recurrence risk for cT4 being 5.99-fold that of cT1 [HR = 5.99 (2.11–17.02)] (Fig. 2B). Fig. 2Survival outcomes by clinical T stage in patients with Siewert type II/III AEG. A Kaplan–Meier overall survival (OS) curves for different cT stages; B Kaplan–Meier recurrence-free survival (RFS) curves for different cT stages Survival outcomes by clinical T stage in patients with Siewert type II/III AEG. A Kaplan–Meier overall survival (OS) curves for different cT stages; B Kaplan–Meier recurrence-free survival (RFS) curves for different cT stages Prognostic value of textbook outcome Figure 3 illustrates the achievement of TO within the study cohort and its association with patient prognosis. Using the TO concept defined by the DUCA, 144 out of 242 patients (59.5%) achieved TO-LN16. The two primary reasons for not achieving TO-LN16 were a length of hospital stay exceeding 14 days (57 patients, 23.6%) and the occurrence of postoperative complications classified as Clavien–Dindo grade 2 or higher (50 patients, 20.7%) (Fig. 3A). However, there was no significant association between TO and RFS or OS, with RFS [HR = 1.12 (0.69–1.82), p =.64] (Fig. 3B) and OS [HR = 1.23 (0.78–1.93), p =.38] (Fig. 3C). Fig. 3Comparison of textbook outcomes (TO) and their association with OS and RFS in Siewert type II/III AEG patients. A Percentage of patients achieving each parameter of TO based on 16 examined lymph nodes (TO-16LN). B Kaplan–Meier OS curves for TO-16LN versus non-TO-16LN. C Kaplan–Meier RFS curves for TO-16LN versus non-TO-16LN. D Percentage of patients achieving each parameter of TO based on 30 examined lymph nodes (TO-30LN). E Kaplan–Meier OS curves for TO-30LN versus non-TO-30LN. F Kaplan–Meier RFS curves for TO-30LN versus non-TO-30LN Comparison of textbook outcomes (TO) and their association with OS and RFS in Siewert type II/III AEG patients. A Percentage of patients achieving each parameter of TO based on 16 examined lymph nodes (TO-16LN). B Kaplan–Meier OS curves for TO-16LN versus non-TO-16LN. C Kaplan–Meier RFS curves for TO-16LN versus non-TO-16LN. D Percentage of patients achieving each parameter of TO based on 30 examined lymph nodes (TO-30LN). E Kaplan–Meier OS curves for TO-30LN versus non-TO-30LN. F Kaplan–Meier RFS curves for TO-30LN versus non-TO-30LN To explore an optimal lymphadenectomy extent, we redefined the TO concept to require the examination of more than 30 ELN. In this cohort, 120 out of 242 patients (49.6%) had more than 30 ELN (Fig. 3D). Under this revised definition, only 79 patients (32.6%) achieved TO-30LN. Grouping based on the new TO definition (TO-LN30) revealed a significant prognostic impact. Compared to patients who achieved TO-30LN, those who did not had an 85% increased risk of all-cause mortality [HR = 1.85 (1.08–3.18), p =.03] (Fig. 3E) and an 88% increased risk of recurrence [HR = 1.88 (1.06–3.34), p =.03] (Fig. 3F). Summary of Cox regression analysis outcomes for OS and RFS in AEG patients Table 2 and Stable1 present the results of the risk factor analyses for OS and RFS, respectively. In univariate analysis for OS, advanced age (> 65 years), tumor diameter greater than 4 cm, positive lymphovascular invasion (LVI), positive perineural invasion (PNI), clinical T4 stage, clinical N + stage, pathological T4 stage, pathological N + stage, and the occurrence of Clavien–Dindo grade ≥ 2 complications (CD ≥ 2) were associated with poorer long-term survival. However, in multivariate analysis for OS, only pathological N + stage and CD ≥ 2 remained as independent risk factors threatening long-term survival. Patients with pathological N + stage had a 2.27-fold higher risk of all-cause mortality compared to N0 patients (95% CI: 2.08–5.56, p =.016), and those who experienced CD ≥ 2 had a 2.36-fold higher risk of all-cause mortality than others (95% CI: 1.45–3.83, p =.001) (Table 2). Table 2Summary of Cox regression analysis outcomes for OSVariableUN-HR (95%CI, p)MU-HR (95%CI, p)GenderM191 (80.3%)F47 (19.7%)0.58 (0.30–1.10, p =.094)Age (y.o.)64-139 (58.4%)65+99 (41.6%)1.61 (1.02–2.53, p =.040)1.33 (0.83–2.13, p =.235)SiewertII141 (59.2%)III97 (40.8%)1.09 (0.69–1.72, p =.711)LaurenIntestinal117 (49.2%)Diffuse46 (19.3%)0.92 (0.49–1.73, p =.800)Mixed75 (31.5%)1.12 (0.67–1.86, p =.670)SigNo215 (90.3%)Yes23 (9.7%)0.82 (0.36–1.89, p =.645)Size (cm)<4138 (58.0%)≥ 4100 (42.0%)2.09 (1.32–3.30, p =.002)1.49 (0.88–2.53, p =.138)NATYes22 (9.2%)No216 (90.8%)0.59 (0.30–1.15, p =.121)Histopathological typeWell/Medium142 (59.7%)Poor96 (40.3%)1.11 (0.70–1.75, p =.665)LVINo117 (49.2%)Yes121 (50.8%)1.94 (1.22–3.10, p =.005)1.37 (0.81–2.32, p =.234)PNINo108 (45.4%)Yes130 (54.6%)1.98 (1.23–3.19, p =.005)1.05 (0.55–2.00.55.00, p =.883)cT stageT147 (19.7%)T2-3119 (50.0%)1.58 (0.78–3.19, p =.203)0.84 (0.34–2.07, p =.705)T472 (30.3%)2.26 (1.11–4.63, p =.025)0.79 (0.29–2.12, p =.634)cN stageN0123 (51.7%)N+115 (48.3%)2.08 (1.30–3.31, p =.002)1.71 (0.99–2.96, p =.056)(y)pT stageT0-170 (29.4%)T2-3102 (42.9%)1.73 (0.92–3.26, p =.092)0.97 (0.37–2.50, p =.941)T466 (27.7%)3.10 (1.65–5.84, p <.001)1.29 (0.44–3.76, p =.636)(y)pN stageN0127 (53.4%)N+111 (46.6%)3.40 (2.08–5.56, p <.001)2.27 (1.17–4.41, p =.016)CD ≥ 2No165 (69.3%)Yes73 (30.7%)2.01 (1.27–3.18, p =.003)2.36 1.45–3.83, p =.001)Abbreviations: HR Hazard Ratio, CI Confidence Interval, MU Multivariate, UN Univariate, OS Overall Survival, y.o. Years Old, Sig Signet Ring Cell Component, NAT Neoadjuvant Therapy, LVI Lymphovascular Invasion, PNI Perineural Invasion, cT stage Clinical T Stage, cN stage Clinical N Stage, (y)pT stage Pathological T Stage (Post-Neoadjuvant Therapy), (y)pN stage Pathological N Stage (Post-Neoadjuvant Therapy), CD ≥ 2 Complications of Clavien–Dindo Grade 2 or Higher Table 2Summary of Cox regression analysis outcomes for OSVariableUN-HR (95%CI, p)MU-HR (95%CI, p)GenderM191 (80.3%)F47 (19.7%)0.58 (0.30–1.10, p =.094)Age (y.o.)64-139 (58.4%)65+99 (41.6%)1.61 (1.02–2.53, p =.040)1.33 (0.83–2.13, p =.235)SiewertII141 (59.2%)III97 (40.8%)1.09 (0.69–1.72, p =.711)LaurenIntestinal117 (49.2%)Diffuse46 (19.3%)0.92 (0.49–1.73, p =.800)Mixed75 (31.5%)1.12 (0.67–1.86, p =.670)SigNo215 (90.3%)Yes23 (9.7%)0.82 (0.36–1.89, p =.645)Size (cm)<4138 (58.0%)≥ 4100 (42.0%)2.09 (1.32–3.30, p =.002)1.49 (0.88–2.53, p =.138)NATYes22 (9.2%)No216 (90.8%)0.59 (0.30–1.15, p =.121)Histopathological typeWell/Medium142 (59.7%)Poor96 (40.3%)1.11 (0.70–1.75, p =.665)LVINo117 (49.2%)Yes121 (50.8%)1.94 (1.22–3.10, p =.005)1.37 (0.81–2.32, p =.234)PNINo108 (45.4%)Yes130 (54.6%)1.98 (1.23–3.19, p =.005)1.05 (0.55–2.00.55.00, p =.883)cT stageT147 (19.7%)T2-3119 (50.0%)1.58 (0.78–3.19, p =.203)0.84 (0.34–2.07, p =.705)T472 (30.3%)2.26 (1.11–4.63, p =.025)0.79 (0.29–2.12, p =.634)cN stageN0123 (51.7%)N+115 (48.3%)2.08 (1.30–3.31, p =.002)1.71 (0.99–2.96, p =.056)(y)pT stageT0-170 (29.4%)T2-3102 (42.9%)1.73 (0.92–3.26, p =.092)0.97 (0.37–2.50, p =.941)T466 (27.7%)3.10 (1.65–5.84, p <.001)1.29 (0.44–3.76, p =.636)(y)pN stageN0127 (53.4%)N+111 (46.6%)3.40 (2.08–5.56, p <.001)2.27 (1.17–4.41, p =.016)CD ≥ 2No165 (69.3%)Yes73 (30.7%)2.01 (1.27–3.18, p =.003)2.36 1.45–3.83, p =.001)Abbreviations: HR Hazard Ratio, CI Confidence Interval, MU Multivariate, UN Univariate, OS Overall Survival, y.o. Years Old, Sig Signet Ring Cell Component, NAT Neoadjuvant Therapy, LVI Lymphovascular Invasion, PNI Perineural Invasion, cT stage Clinical T Stage, cN stage Clinical N Stage, (y)pT stage Pathological T Stage (Post-Neoadjuvant Therapy), (y)pN stage Pathological N Stage (Post-Neoadjuvant Therapy), CD ≥ 2 Complications of Clavien–Dindo Grade 2 or Higher Summary of Cox regression analysis outcomes for OS Abbreviations: HR Hazard Ratio, CI Confidence Interval, MU Multivariate, UN Univariate, OS Overall Survival, y.o. Years Old, Sig Signet Ring Cell Component, NAT Neoadjuvant Therapy, LVI Lymphovascular Invasion, PNI Perineural Invasion, cT stage Clinical T Stage, cN stage Clinical N Stage, (y)pT stage Pathological T Stage (Post-Neoadjuvant Therapy), (y)pN stage Pathological N Stage (Post-Neoadjuvant Therapy), CD ≥ 2 Complications of Clavien–Dindo Grade 2 or Higher In the univariate analysis for RFS, female gender and the absence of NAT were associated with a lower risk of recurrence, while tumor diameter greater than 4 cm, positive perineural invasion (PNI), clinical T + stage, clinical N + stage, pathological T + stage, pathological N + stage, and CD ≥ 2 were associated with a higher risk of recurrence. In the multivariate analysis for RFS, female gender emerged as an independent protective factor against recurrence [HR = 0.42 (95% CI: 0.20–0.89, p =.023)]. Clinical N + stage was associated with a 2.10-fold higher risk of recurrence compared to cN0 (95% CI: 1.15–3.85, p =.016), pathological N + stage with a 1.98-fold higher risk compared to pN0 (95% CI: 1.08–3.64, p =.028), and CD ≥ 2 with a 2.55-fold higher risk compared to others (95% CI: 1.55–4.20, p <.001) (Stable1). In summary, key prognostic factors affecting survival and recurrence include both clinically and pathologically confirmed lymph node metastasis, as well as the occurrence of severe postoperative complications. The relationship between the number of ELNs and prognosis in advanced AEG We used RCS analysis to clarify the relationship between the number of ELN and overall OS. Figure 4A illustrates the distribution of ELNs across the entire cohort, showing a normal distribution pattern, with approximately 85% of AEG surgeries yielding more than 20 ELN. Figure 4B presents the dose-response relationship between ELNs and OS for the full cohort, revealing that the association between the number of ELNs and prognosis in AEG is not purely linear but rather follows an S-shaped curve. In the analysis of AEG across all stages, OS decreases sharply with an increase in ELNs within the range of 21–42, reaching an equilibrium point at 29 ELNs, suggesting that this number may represent the optimal lymph node yield for balanced prognosis. Fig. 4Distribution and prognostic impact of examined lymph nodes (ELNss) on survival. A Histogram showing the distribution of ELNss in the study cohort, with a majority of patients having more than 20 ELNss. B Restricted cubic spline analysis illustrating the relationship between ELNs count and OS across all stages, with key inflection points at ELNs = 21, 29, and 42, indicating potential optimal thresholds for prognostic benefit Distribution and prognostic impact of examined lymph nodes (ELNss) on survival. A Histogram showing the distribution of ELNss in the study cohort, with a majority of patients having more than 20 ELNss. B Restricted cubic spline analysis illustrating the relationship between ELNs count and OS across all stages, with key inflection points at ELNs = 21, 29, and 42, indicating potential optimal thresholds for prognostic benefit However, under the surgical protocols and clinical conditions of our institution, a higher cT stage often corresponds to a higher count of retrieved ELN. Therefore, we conducted an additional RCS analysis specifically for advanced AEG. The results indicated that 30 ELNs serve as a prognostic threshold; within the range of 18–41 ELNs, prognosis improves with an increasing number of ELNs, and there is a significant enhancement in survival between 30 and 41 ELNs (Fig. 5A). Fig. 5Restricted cubic spline analysis of ELNs and OS stratified by cT stage in Siewert type II/III AEG. A Relationship between ELNs count and OS for all advanced stages (cT2-4). B Relationship between ELNs count and OS for cT2-3, showing an optimal range of 28–53 ELNss for survival benefit. C Relationship between ELNs count and OS for cT4, indicating no significant survival benefit with increased ELNss even at the optimal point of ELNs = 39. D Relationship between ELNs count and OS for all advanced stages (cT2-4) in Siewert type II/III AEG without NAT. E Relationship between ELNs count and OS for cT2-3, showing an optimal range of 30–48 ELNs for survival benefit in Siewert type II/III AEG without NAT. F Relationship between ELNs count and OS for cT4, indicating no significant survival benefit with increased ELNs in Siewert type II/III AEG without NAT Restricted cubic spline analysis of ELNs and OS stratified by cT stage in Siewert type II/III AEG. A Relationship between ELNs count and OS for all advanced stages (cT2-4). B Relationship between ELNs count and OS for cT2-3, showing an optimal range of 28–53 ELNss for survival benefit. C Relationship between ELNs count and OS for cT4, indicating no significant survival benefit with increased ELNss even at the optimal point of ELNs = 39. D Relationship between ELNs count and OS for all advanced stages (cT2-4) in Siewert type II/III AEG without NAT. E Relationship between ELNs count and OS for cT2-3, showing an optimal range of 30–48 ELNs for survival benefit in Siewert type II/III AEG without NAT. F Relationship between ELNs count and OS for cT4, indicating no significant survival benefit with increased ELNs in Siewert type II/III AEG without NAT However, when analyzing cT2-3 and cT4 individually, we found no significant survival benefit associated with higher ELNs counts for cT4 AEG (Fig. 5C). Conversely, in cT2-3 AEG, an ELNs range of 28–53 was associated with notable survival improvement (Fig. 5B). These findings suggest that extending lymphadenectomy does not improve prognosis for cT4 AEG, while for advanced AEG that is not cT4, achieving an ELNs count of approximately 30 or more may provide substantial clinical benefits. The relationship between the number of ELNs and prognosis in advanced AEG without NAT Current guidelines recommend NAT for advanced AEG. However, in this study cohort, only 9.2% (23/242) of patients received NAT. Since NAT may influence the ELNs, we conducted an analysis of the relationship between ELNs and prognosis in non-NAT patients. Figure 5D-F presents the relationship between ELNs count and OS in non-NAT AEG patients, which was consistent with the findings in the overall population (Fig. 5A-C). In AEG patients with cT2-4 stage, an ELNs range of 30–48 was consistently associated with significant clinical benefits, with the highest survival benefit observed at ELNs = 42 (Fig. 5D). Similarly, in AEG patients with cT2-3 stage, an ELN range of 31–45 was associated with significant clinical benefits, peaking at ELNs = 42 (Fig. 5E). In AEG patients with cT4 stage, an increasing trend of survival benefit was observed with a higher ELN count, although the association did not reach statistical significance (Fig. 5F). Discussion This study evaluated the prognostic impact of ELNs count and TO in patients with Siewert II/III AEG undergoing PG. Our key finding was that achieving TO-30LN was associated with significant survival benefit. RCS analysis further demonstrated that cT2–3 patients derived prognostic advantage from an ELN range of 28–53, whereas no survival benefit was observed in cT4 disease, suggesting that PG may not be oncologically appropriate for this subgroup. Standard and thorough lymphadenectomy is essential for curative resection in gastric cancer. The central controversy regarding PG versus TG lies in whether PG can achieve adequate nodal clearance, particularly for groups 4 d, 5, 6, and 12a.Yamashita et al. reported in a nationwide retrospective study that for AEG ≤ 4 cm (with the tumor center within 2 cm of the EGJ), distal epigastric lymph node metastasis was exceedingly rare, thereby supporting the safety of transabdominal PG [19]. Similarly, Kurokawa et al. demonstrated in a nationwide prospective multicenter study of 363 patients with cT2–4 AEG that metastasis rates in nodal groups 4 d, 5, and 6 were each below 5% [20]. Lee et al. conducted a propensity score–matched study and found comparable OS (p =.676) and 10-year RFS (p =.634) between PG and TG for proximal GC and Siewert III AEG, with no recurrence observed in groups 4 d, 5, 6, or 12a after PG [21]. Similarly, in China, Peng et al. showed that among patients with pT2–3 gastric cancer ≤ 4 cm and Borrmann I–II type, no metastases were found in nodal groups 5 and 6, and the 5-year survival rate was comparable between PG and TG (p =.275). However, for pT4 or Borrmann III tumors, the rates of metastasis in groups 5 and 6 were significantly higher (9.9% and 10.6%, respectively) [22]. At our center, individualized lymphadenectomy was tailored according to guidelines, and the cohort achieved favorable outcomes, with a 5-year OS of 69.5% overall, 77.6% in early-stage, and 67.3% in advanced-stage patients. Although our findings support the oncological feasibility of PG in carefully selected cT2–3 AEG, TG remains the more radical and widely accepted standard, and PG should be reserved for patients in whom both oncological safety and functional preservation can be reasonably ensured. Comprehensive perioperative indicators such as TO have gained increasing attention. Busweiler et al. first introduced TO in 2017 and reported that only 32.1% of gastric cancer patients achieved TO, with inadequate lymphadenectomy as the main limiting factor [16]. Subsequent studies confirmed that achieving TO improves prognosis in gastric cancer patients [23, 24]. However, the prognostic implications remain debated. In our study, 59.5% achieved TO-16LN, but this was not associated with improved prognosis, echoing the inconsistent findings of earlier research. With advances such as indocyanine green staining, more extensive lymphadenectomy has become feasible. Indeed, at our center, >85% of patients had more than 20 nodes retrieved [25]. Under the revised definition of TO-30LN, patients who failed to achieve it had an 85% higher risk of all-cause mortality (HR = 1.85, p =.03) and an 88% higher risk of recurrence (HR = 1.88, p =.03). This underscores the necessity of thorough lymphadenectomy as a cornerstone of curative treatment. ELN yield may also serve as a surrogate marker for surgical radicality or pathological diligence and thus should be interpreted with caution when considered an independent prognostic factor. Restricted cubic spline analysis clarified the dose-response association between ELNs and survival. Our results showed that 28–30 ELNs represented a pivotal threshold across stages, consistent with the CSCO guideline recommendation of retrieving at least 30 nodes [8]. Importantly, the survival benefit of ELNs was stage dependent. In cT2–3 AEG, retrieving 28–53 nodes were associated with improved survival, whereas in cT4 AEG, even beyond the apparent inflection point of 39 nodes, no significant benefit was observed. These findings suggest that for cT4 disease, more extensive lymphadenectomy does not translate into oncological advantage, reinforcing the limited suitability of PG for this subgroup. It should also be recognized that the apparent survival advantage of higher ELNs counts may in part reflect stage migration, rather than a direct therapeutic benefit of extended lymphadenectomy. Multivariate analysis further demonstrated that pN + disease and Clavien–Dindo grade ≥ 2 complications were independent predictors of recurrence and survival. Patients with pN + status had a 2.40-fold higher recurrence risk, while those with severe complications had a 2.59-fold increased mortality risk. These results align with the findings of Beecher et al., who reviewed the literature and concluded that postoperative complications contribute to worse long-term outcomes [26]. Mechanistically, Coffey et al. suggested that postoperative immunosuppression, which peaks around day 3 and persists for up to two weeks, may create an environment favorable for micro metastatic tumor growth [27]. This study has several limitations. First, its retrospective single-center design may have introduced selection bias and limits the generalizability of our findings. Second, the proportion of patients who received NAT was relatively low, which constrains the ability to assess the prognostic value of PG in the context of modern multimodality treatment and may not fully reflect contemporary clinical practice. Third, variations in reconstruction methods between 2017 and 2021 could have acted as confounding factors, particularly regarding functional outcomes such as reflux, although the primary focus of this study was on oncological results. Finally, while higher ELNs counts were associated with better survival, these findings should be interpreted cautiously, as they may partly reflect stage migration or surgical/pathological diligence rather than a direct therapeutic effect of extended lymphadenectomy. Conclusion In advanced AEG with PG, achieving TO-30LN was associated with improved long-term survival. Restricted cubic spline analysis further suggested an optimal examined lymph node threshold for overall survival in cT2–3 patients, whereas no clear survival benefit was observed in cT4 patients. These findings indicate that proximal gastrectomy may be oncologically appropriate for carefully selected cT2–3 AEG, while its role in cT4 disease remains limited. Supplementary Information Supplementary Material 1. Supplementary Material 1.