Mediastinal lymphadenectomy strategy for Siewert II adenocarcinoma of the esophagogastric junction: a retrospective cohort study

Abstract

Background

Siewert type II adenocarcinoma of the esophagogastric junction (AEG) is characterized by complex, bidirectional lymph node metastasis. This retrospective study evaluates recurrence and mortality risks associated with varying extents of mediastinal lymph node dissection.

Methods

A retrospective analysis was conducted on 221 Siewert type II AEG patients who underwent radical surgery (McKeown or Left Thoracotomy [LT] approach) between June 2016 and October 2023. Patients were grouped according to the extent of lymphadenectomy: complete mediastinal and abdominal lymph node dissection (CMAD) or middle/lower mediastinal and abdominal lymph node dissection (MLMAD). To ensure comparability, patients in the CMAD group were matched 1:1 with those in the MLMAD group using propensity score matching (PSM). Prognostic outcomes were analyzed using Kaplan–Meier curves, and Cox regression analysis was performed to identify risk factors.

Results

CMAD was associated with significantly improved overall survival (OS) (HR = 0.37, 95%CI:0.23–0.60; P < 0.001) and disease-free survival (DFS) (HR = 0.36, 95%CI:0.22–0.58; P < 0.001) compared to MLMAD before PSM; these benefits persisted after PSM (OS HR = 0.44, P = 0.01; DFS HR = 0.44, P = 0.02). Upper mediastinal metastasis was observed in 4.3% of patients. Dissection of upper mediastinal stations, particularly 2R, 4R, and 8U, contributed to the survival benefit. Removal of more than 11 lymph nodes was associated with improved survival. The McKeown approach enabled a more comprehensive upper mediastinal and abdominal lymphadenectomy compared to LT (P < 0.001), without increasing complication rates or hospital stay.

Conclusion

The findings suggest that a surgical strategy involving the comprehensive removal of mediastinal and abdominal lymph nodes, when applied to selected patients in a high-volume center, is associated with superior survival outcomes compared to more limited dissections.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-026-04252-3.

Introduction

The global incidence of adenocarcinoma of the esophagogastric junction (AEG) is increasing significantly, and its anatomical and biological characteristics present challenges for diagnosis and treatment [1–3]. The Siewert classification, which guides surgical management, defines type II AEG as tumors with their center located 1 cm above to 2 cm below the esophagogastric junction (EGJ) [4]. Due to its location, Siewert type II AEG exhibits bidirectional lymphatic drainage to both the abdomen and thorax, resulting in complex lymph node metastasis patterns. This complexity has led to ongoing debate regarding the necessity and extent of mediastinal lymphadenectomy in thoracic surgery [5, 6]. Traditional approaches, based on gastric cancer management, often prioritize abdominal lymph node dissection and consider mediastinal dissection as secondary [7]. However, evidence shows that metastasis to middle, lower, and upper mediastinal lymph nodes in Siewert type II AEG is not uncommon. Limiting dissection to the abdomen may miss micrometastatic disease, potentially compromising oncological outcomes [8, 9]. Some studies support transthoracic esophagogastrectomy, which allows for more extensive lymph node dissection without increasing postoperative complications [10, 11]. Therefore, transthoracic esophagogastrectomy is considered a viable option for most Siewert type II AEG patients, and lymph node metastasis remains an independent risk factor for long-term survival [12]. International guidelines lack consensus on the optimal mediastinal lymph node dissection in Siewert type II AEG. This study retrospectively analyzes the clinicopathological Siewert type II AEG who underwent surgery, aiming to clarify mediastinal lymph node metastasis patterns and optimal dissection strategies from a thoracic surgery perspective. We hypothesized that complete mediastinal lymphadenectomy, including upper mediastinal stations, would result in superior survival outcomes compared to middle/lower mediastinal dissection alone. This is a theoretical basis for mediastinal lymph node dissection strategies in Siewert type II AEG.

Method

Patients

This single-center, retrospective cohort study has been reported in accordance with the Strengthening the Reporting of Cohort, Cross-sectional and Case-Control Studies in Surgery (STROCSS) criteria [13] and is registered with the research registry (https://www.researchregistry.com). The study was conducted at the Department of Thoracic Surgery, West China Hospital, a high-volume tertiary care center specializing in esophageal and gastric surgery. Our institution performs over 500 esophagogastric resections annually, ensuring substantial surgical experience and standardized perioperative management. We retrospectively analyzed Siewert type II AEG patients who underwent radical surgery in our thoracic surgery department from June 2016 to October 2023. Siewert type II AEG was defined based on preoperative computed tomography, endoscopy, intraoperative observation, and postoperative pathological examination. Patients were divided into two groups according to the extent of lymph node dissection. The Complete Mediastinal and Abdominal Dissection (CMAD) group: This encompassed a D2 abdominal lymphadenectomy (including stations 16, 17, 18, 19, and 20) plus a comprehensive mediastinal dissection. The mediastinal dissection included all middle/lower mediastinal stations (7, 8 M, 8 L, 9, 15) and extended cranially to systematically clear the upper mediastinal stations (2R, 2 L, 4R, 4 L, and 8U). The Middle/Lower Mediastinal and Abdominal Dissection (MLMAD) group: This involved the same D2 abdominal lymphadenectomy but, which was limited to the middle and lower mediastinum only (stations 7, 8 M, 8 L, 9, 15), specifically excluding the upper mediastinal stations (2R, 2 L, 4R, 4 L, 8U). To minimize selection bias, patients with clinically suspicious upper mediastinal or cervical lymph nodes on preoperative imaging were excluded from the analysis. Specifically, only patients without radiographic evidence of metastasis in stations 2R, 2 L, 4R, 4 L, or cervical nodes were included in the present study. Consequently, any upper mediastinal lymph node metastases found postoperatively in our cohort were occult. Inclusion criteria included: (1)Patients diagnosed with Siewert type II AEG; (2)The primary surgical approaches were the McKeown esophagectomy and the trans-left thoracic approach; (3)No distant metastasis was detected. Exclusion criteria included: (1)Patients with unresectable cancer involving adjacent structures or distant metastasis; (2)those with concomitant malignancies; (3)Palliative resection. The relevant Ethics Committee provided approval for this study, and the need for patient consent was waived.

Definition of lymph node station

Tumor and lymph node pathological features were defined and classified according to the 8th edition of the TNM staging system [14]. The definition of each lymph node station was numbered according to AJCC (2R= right upper paratracheal lymph nodes; 4R= right lower paratracheal lymph nodes; 2 L= left upper paratracheal lymph nodes; 4 L= left lower paratracheal lymph nodes; and 8U= upper paraesophageal lymph nodes, 7 = subcarinal lymph nodes; 8 M= middle thoracic paraesophageal lymph nodes; 8 L= lower thoracic paraesophageal lymph nodes; 9 = pulmonary ligament lymph nodes; 10 = main bronchus lymph nodes; 15 = supradiaphragmatic lymph nodes; 16 = paracardial lymph nodes; 17 = lymph nodes at the root of the left gastric artery; 18 = lymph nodes along the common hepatic artery; 19 = lymph nodes along the splenic artery; 20 = lymph nodes along the celiac artery) [15]. 2 L was considered equivalent to 106-recL in the Japanese Esophageal Cancer Classification (JCEC) staging system, while station 2R was equivalent to 106-recR. n addition, we conducted subgroup analysis to explore differences in lymph node dissection methods according to different tumor T stages.

Treatment procedures

To evaluate resectability, a standard baseline clinical examination, routine blood analyses, contrast-enhanced computed tomography (CT), Electrocardiogram (ECG), and Cardiopulmonary function testing were performed for all patients. Certain patients underwent external ultrasound of the cervical region and positron emission tomography/computed tomography (PET/CT). Among the 221 enrolled patients, 178 (80.5%) underwent preoperative PET/CT scans, and 205 (92.8%) underwent cervical ultrasound examinations. None of the included patients showed evidence of cervical or upper mediastinal lymph node metastasis on these imaging studies. Patients underwent left thoracotomy(LT) and McKeown surgery with proximal gastrectomy by experienced thoracic surgeons. The choice of surgical approach and the extent of lymphadenectomy (CMAD vs. MLMAD) was primarily based on a combination of tumor characteristics (e.g., tumor location, clinical T/N stage assessed via preoperative imaging), patient factors (e.g., cardiopulmonary fitness, body habitus), and surgeon’s preference and experience. During the study period, there was no strict institutional protocol mandating a particular approach; thus, the decision was individualized after multidisciplinary discussion. The choice of surgical approach was a complex decision made by a multidisciplinary team (MDT) after a comprehensive evaluation of each patient. The selection criteria for the McKeown approach over the LT approach included: (1) Tumor factors: The size of the tumor has been evaluated and can be removed minimally invasively, and/or advanced clinical nodal (cN) stage. (2) Patient factors: Better cardiopulmonary reserve and overall fitness to tolerate a more extensive thoracic procedure. Conversely, the LT approach was preferentially selected for patients with: (1) Tumor factors: The tumor is too large to be removed through minimally invasive surgery. (2) Patient factors: Comorbidities that increased the perioperative risk of a longer, more complex procedure (e.g., poor pulmonary function, advanced age), or a body habitus that favored a unilateral approach. This selection paradigm explains why the LT group in our unmatched cohort presented with a higher burden of disease (larger tumors, more advanced pN stage), as these patients were often deemed less optimal candidates for the maximalist McKeown/CMAD approach. We explicitly acknowledge and have statistically addressed this inherent selection bias through PSM.For patients with no evidence of cervical lymph node metastasis based on preoperative CT and ultrasound, a routine two-field lymph node dissection is performed; otherwise, a three-field lymph node dissection is performed. The anastomosis was located in the neck. In evaluating the surgical quality of esophagectomy procedures. We utilized a video assessment scale that emphasized technical performance and oncological quality, Surgeons provided short video clips to allow evaluation by the assessor (Supplementary video clip), with a specific focus on lymphadenectomy completeness and exposure of anatomic landmarks. To maximize the accuracy of classifying tumors as Siewert type II, we employed a rigorous multi-modal approach: (1)The epicenter of the tumor was determined primarily by esophagogastroduodenoscopy (EGD) and CT. On EGD, the location of the tumor’s center relative to the esophagogastric junction (EGJ) was carefully documented, with the EGJ defined as the level of the proximal margin of the gastric folds. CT scans were used to corroborate the endoscopic findings and assess the bulk of the tumor mass; (2)The endoscopic localization was confirmed by surgical visualization during the procedure; (3)The final classification was based on the pathological examination of the resected specimen, which is considered the gold standard. The pathologist measured the distance from the tumor’s macroscopic center to the EGJ (histologically defined as the squamocolumnar junction). Only tumors with their center confirmed to be within 1 cm above and 2 cm below the EGJ were definitively included as Siewert type II in this study.

Follow-up

Patient demographics, clinical characteristics, pathological information, intraoperative index, postoperative major complications, follow-up data were collected. Data of patients who were alive or lost to follow-up were censored at the date of their last follow-up. Follow-up assessments were conducted every 3 months during the first 2 years, and then every 6 months thereafter. The last follow-up was in September 2025, which served as the cutoff date for the survival analysis.

Assessment of complications

Postoperative complications occurring within 30 days after surgery were systematically recorded. The severity of complications was graded according to the Clavien-Dindo classification [16]. Grade I-II complications were considered minor, and Grade III-V complications were considered major.

Statistical analysis

Continuous data were expressed as interquartile ranges, and categorical data were compared using the χ² test or Fisher’s exact probability test. Cox proportional hazards regression models were used for both univariate and multivariate analyses. Variables with a P-value < 0.10 in the univariate analysis were subsequently included in the multivariate Cox regression model using the Enter method, where all selected variables are entered into the model simultaneously. The results are presented as hazard ratios (HRs) with 95% confidence intervals (CIs). Multicollinearity among the variables included in the multivariate Cox regression model was assessed using variance inflation factors (VIF). All VIF values were below 1.38, indicating that multicollinearity was not a significant concern in our analysis. Adjusted HR was used to represent differences in variables after PSM. To minimize selection bias and the influence of confounding factors, PSM was performed in a 1:1 ratio using the nearest-neighbor matching algorithm with a caliper width of 0.02 times the standard deviation of the logit of the propensity score. The propensity score was estimated using a multivariable logistic regression model, where the dependent variable was the assignment to either the CMAD or MLMAD group. The choice of matching variables was based on clinical relevance and their potential to influence both the selection of surgical approach and the oncological prognosis. We included the following covariates in the propensity score model: sex, age, BMI, pT stage, pN stage, tumor differentiation (grade), and the administration of neoadjuvant therapy. The balance of covariates between groups before and after matching was assessed using standardized mean differences (SMD), with an SMD < 0.10 indicating good balance. Overall Survival (OS) was defined as death from any cause. Given that the primary objective was to assess the oncologic efficacy of the surgical intervention, and that non-cancer-related deaths were relatively infrequent in our cohort (4.1% of all deaths), we utilized the standard Kaplan-Meier method for survival curve estimation and the Cox proportional hazards model for univariable and multivariable analyses. While this approach does not formally account for competing risks, it remains the conventional and widely reported method for comparing surgical cohorts in oncology. Disease-free survival (DFS) analysis, which considers any recurrence or death, inherently incorporates competing risks. All survival analyses were right-censored at the date of last follow-up for patients who were alive or lost to follow-up. X-Tile determines the optimal cut-off values for the number of lymph node stations and the number of removed lymph nodes in the survival stratification by analyzing the minimum P-value obtained from the log-rank test. All statistical analyses were performed using SPSS Statistics (version 24, IBM, Armonk, NY) and the R programming language (version 3.6.3, Vienna, Austria). Statistical significance was defined as a two-sided P-value less than 0.05. Figure 1 provides a flow chart of the enrolled patients and the statistical analysis.

Fig. 1.

The patient inclusion flowchart and statistical analysis schema

Results

Patient characteristics

A total of 221 Siewert type II AEG patients were included in this study. Among them, 82 patients underwent complete mediastinal and abdominal lymph node dissection, and 139 patients underwent middle/lower mediastinal and abdominal lymph node dissection. Table 1 lists the baseline characteristics of the patients. Overall, 4100 lymph nodes were dissected, of which 684 (16.7%) were positive. 11 patients had complete mediastinal and abdominal lymph node metastasis, 27 patients had middle/lower mediastinal and abdominal lymph node metastasis, and 95 patients had only abdominal lymph node metastasis. In order to further reduce the influence of confounding factors on the results, we conducted a PSM analysis, including factors such as pT stage, pN stage, and whether receiving neoadjuvant and adjuvant therapy, and finally analyzed 70 pairs (140 cases) of patients; All key prognostic variables (including pT stage, pN stage, and neoadjuvant therapy) were well-balanced after matching (all SMD < 0.10)(Table 2). Although the SMD for Age (-0.206), BMI (-0.240), and Gender (-0.179) slightly exceeded the ideal threshold, they represented a substantial improvement compared to the pre-match imbalance and were considered acceptable for subsequent analysis given.

Table 1.

Baseline characteristics in total cohort

Variable	Total cohort
	Total (n = 221)	CMAD (n = 82)	MLMAD (n = 139)	Statistic	P	SMD
Age, M (Q₁, Q₃)	65.00 (59.00, 70.00)	63.50 (58.25, 70.00)	66.00 (59.00, 70.00)	Z=-1.377	0.168	0.195
BMI, M (Q₁, Q₃)	22.98 (21.16, 25.26)	23.41 (21.50, 25.72)	22.59 (20.90, 24.73)	Z=-2.172	0.030	-0.341
Differentiation, n (%)				χ²=6.474	0.039
G1	8 (3.62)	4 (4.88)	4 (2.88)			-0.120
G2	105 (47.51)	47 (57.32)	58 (41.73)			-0.316
G3	108 (48.87)	31 (37.80)	77 (55.40)			0.354
Gender, n (%)				χ²=0.751	0.386
Female	30 (13.57)	9 (10.98)	21 (15.11)			0.115
Male	191 (86.43)	73 (89.02)	118 (84.89)			-0.115
Tumor Deposit, n (%)				χ²=0.476	0.490
No	201 (90.95)	76 (92.68)	125 (89.93)			-0.092
Yes	20 (9.05)	6 (8.57)	6 (8.57)			0.092
pT, n (%)				χ²=7.668	0.006
pT1-2	67 (30.32)	34 (41.46)	33 (23.74)			-0.417
pT3-4	154 (69.68)	48 (58.54)	106 (76.26)			0.417
pN, n (%)				χ²=2.335	0.506
pN0	88 (39.82)	37 (45.12)	51 (36.69)			-0.175
pN1	46 (20.81)	17 (20.73)	29 (20.86)			0.003
pN2	51 (23.08)	18 (21.95)	33 (23.74)			0.042
pN3	36 (16.29)	10 (12.20)	26 (18.71)			0.167
Neoadjuvant Therapy, n (%)				χ²=0.495	0.482
No	128 (57.92)	45 (54.88)	83 (59.71)			0.099
Yes	93 (42.08)	37 (45.12)	56 (40.29)			-0.099
Adjuvant, n (%)				χ²=1.884	0.170
No	113 (51.13)	37 (45.12)	76 (54.68)			0.192
Yes	108 (48.87)	45 (54.88)	63 (45.32)			-0.192

BMI body mass index, pT pathological T stage, pN pathological N stage, CMAD complete mediastinum and abdominal lymph node disscton, MLMAD middle and lower mediastinum and abdominal lymph node disscton

Bold values indicate statistical significance at P < 0.05

Table 2.

Baseline characteristics after propensity score-matched

Variable	After PSM
	Total (n = 221)	CMAD (n = 70)	MLMAD (n = 70)	Statistic	P	SMD
Age, M (Q₁, Q₃)	64.00 (58.00, 70.00)	64.00 (60.00, 70.75)	65.00 (57.25, 67.75)	Z=-0.897	0.370	-0.206
BMI, M (Q₁, Q₃)	22.62 (21.06, 24.61)	23.01 (21.34, 24.82)	22.22 (20.76, 24.58)	Z=-1.338	0.181	-0.240
Differentiation, n (%)				-	0.678
G1	8 (5.71)	4 (5.71)	4 (5.71)			0.000
G2	69 (49.29)	37 (52.86)	32 (45.71)			-0.143
G3	63 (45)	29 (41.43)	34 (48.57)			0.143
Gender, n (%)				χ²=1.301	0.254
Female	23 (16.43)	9 (12.86)	14 (20.00)			0.179
Male	117 (83.57)	61 (87.14)	56 (80.00)			-0.179
Tumor Deposit, n (%)				χ²=0.000	1.000
No	128 (91.43)	64 (91.43)	64 (91.43)			0.000
Yes	12 (8.57)	6 (8.57)	6 (8.57)			0.000
pT, n (%)				χ²=0.033	0.856
pT1-2	45 (32.14)	23 (32.86)	22 (31.43)			-0.031
pT3-4	95 (67.86)	47 (67.14)	48 (68.57)			0.031
pN, n (%)				χ²=0.720	0.869
pN0	65 (46.43)	30 (42.86)	35 (50.00)			0.143
pN1	28 (20.00)	15 (21.43)	13 (18.57)			-0.073
pN2	30 (21.43)	16 (22.86)	14(20.00)			-0.071
pN3	17 (12.14)	9 (12.86)	8 (11.34)			-0.045
Neoadjuvant Therapy，n (%)				χ²=0.726	0.394
No	79 (56.43)	42 (60.00)	37 (52.86)			-0.143
Yes	61 (43.57)	28 (40.00)	33 (47.14)			0.143
Adjuvant, n (%)				χ²=0.718	0.397
No	65 (46.43)	30 (42.86)	35 (50.00)			0.143
Yes	75 (53.57)	40(57.14)	35 (50.00)			-0.143

BMI body mass index, pT pathological T stage, pN pathological N stage, CMAD complete mediastinum and abdominal lymph node disscton, MLMAD middle and lower mediastinum and abdominal lymph node disscton

To evaluate potential temporal trends in surgical practice, we compared the proportion of patients undergoing CMAD and the McKeown approach between early (2016–2019) and late (2020–2023) study periodsAnalysis of surgical practice over time revealed a non-significant increase in the use of CMAD from 41.3% (24/58) in the early period to 57.3% (47/82) in the late period (P = 0.065). The utilization of the McKeown approach remained stable (50.0% vs. 52.4%, P = 0.874). In multivariable analysis, the year of surgery was not an independent prognostic factor for OS (P = 0.17).

Lymph node dissection and metastasis patterns

The optimal cutoff value of > 11 lymph nodes for survival stratification was determined using X-Tile software (version 3.6.1, Yale University, USA) (Supplementary Fig. 1). Table 3 lists the frequency of lymph node dissection and metastasis after PSM. It can be seen that among the upper mediastinal lymph nodes, the most frequently dissected and most frequently metastatic lymph nodes were 8U (30%) and 2R (10%), respectively. Among the middle and lower mediastinal lymph nodes, the most frequently dissected and most frequently metastatic lymph node was station 15 (82.9% and 15.5%). Among the abdominal lymph nodes, the most frequently dissected and most frequently metastatic lymph node was station 17 (90.7% and 44.9%). Upper mediastinal lymph node metastasis occurred in 4.3% of the study population.Among the MLMAD group (which did not undergo routine UMLND), 18.6% of patients who experienced recurrence developed isolated or component upper mediastinal nodal recurrence. This figure starkly contrasts with a rate of 7.1% in the CMAD group (P = 0.036). Although systematic dissection of 2R, 2 L, 4R, 4 L, and 8U stations was planned for all patients in the CMAD group, actual dissection rates varied due to factors such as anatomical variation, tumor location, and intraoperative feasibility. In the CMAD group after PSM, the dissection frequency for each upper mediastinal station is presented in Supplementary Table 1.

Table 3.

Frequency of lymph node dissection and metastasis after PSM

Lymph node station	Dissection Frequency(n = 140,%)	Metastatic Frequency (in dissected patients,%)	Metastatic Frequency (in total cohort, n = 140,%)
UMN
2R	40(28.6)	4/40(10)	4/140(2.9)
2 L	36(25.7)	1/36(2.8)	1/140(0.7)
4R	32(22.9)	2/32(6.3)	2/140(1.4)
4 L	2(1.4)	0	0
5	6(4.3)	0	0
8U	42(30.0)	2/42(4.8)	2/140(1.4)
M/LMN
7	105(75.0)	6/105(5.7)	6/140(4.3)
8 M	90(64.3)	7/90(7.8)	7/140(5.0)
9	74(52.9)	6/74(8.1)	6/140(4.3)
10	84(60.0)	1/84(1.2)	1/140(0.7)
8 L	103(73.6)	8/103(7.8)	8/140(5.7)
15	116(82.9)	18/116(15.5)	18/140(12.9)
Abdominal
16	120(85.7)	41/120(34.2)	41/140(2.9)
17	127(90.7)	57/127(44.9)	57/140(40.7)
18	89(63.6)	17/89(19.1)	17/140(12.1)
19	64(45.7)	18/64(28.1)	18/140(12.9)
20	53(37.9)	15/53(28.3)	15/140(10.7)

Abbreviation: UMN upper mediastinal lymph node, M/LMN middle/lower mediastinal lymph node

2R: Right upper paratracheal lymph nodes

2L: Left upper paratracheal lymph nodes

4R: Right lower paratracheal lymph nodes

4L: Left lower paratracheal lymph nodes

5: Lymph nodes below the aortic arch

8U: Upper thoracic paraesophageal lymph nodes

8 M: Middle thoracic paraesophageal lymph nodes

8 L: Lower thoracic paraesophageal lymph nodes

7: Subcarinal lymph nodes

9: Pulmonary ligament lymph nodes

15: Supradiaphragmatic lymph nodes

16: Paracardial lymph nodes

17: Lymph nodes at the root of the left gastric artery

18: Lymph nodes along the common hepatic artery

19: lymph nodes along the splenic artery

20: lymph nodes along the celiac artery

Survival comparison

The median follow-up time in this study was 61 months (range, 13–111), and the median overall survival for the entire cohort was 51 months. Survival analysis suggested that the CMAD group achieved better OS (HR = 0.37, 95%CI:0.23–0.60; P < 0.001) (Fig. 2a). The median OS for the CMAD group was not reached, while the median OS for the MLMAD group was 37 months. The Kaplan-Meier estimated 5-year overall survival rate was dramatically higher in the CMAD group (71% vs. 34.6%). In the short term, the CMAD group also showed a higher 3-year OS rate (78.9% vs. 51.5%). After PSM, similarly, the results showed that the overall survival of patients who underwent complete mediastinal lymph node dissection (HR = 0.44, 95%CI:0.26–0.74; P = 0.01) (Fig. 2b) was better, with the estimated 5-year OS rate remaining higher at 60.0% compared to 34.4% in the MLMAD group. The 3-year OS rate was also higher (71.1% vs. 52.7%).

Fig. 2.

Difference in survival between the CMAD group and the MLMAD group before and after PSM(A) OS differences of patient receiving CMAD and MLMAD before PSM; (B) OS differences of patient receiving CMAD and MLMAD after PSM; (C) DFS differences of patient receiving CMAD and MLMAD before PSM; (D) DFS differences of patient receiving CMAD and MLMAD after PSM

Survival analysis suggested that the CMAD group achieved longer DFS (HR = 0.36, 95%CI:0.22–0.58; P < 0.001) (Fig. 2c).The median DFS for the CMAD group in the entire cohort was not reached, while the median DFS for the MLMAD group was 35 months. In addition, The estimated 5-year DFS rate was 65.2% for the CMAD group and 34.7% for the MLMAD group. Additionally, the 3-year DFS rate was higher (77.6% vs. 49.7%). After PSM, the results showed that the DFS of patients who underwent complete mediastinal lymph node dissection (HR = 0.44, 95%CI:0.26–0.75; P = 0.02) (Fig. 2d) was better, with the estimated 5-year OS rate remaining higher at 62.2% compared to 32.5% in the MLMAD group. The 3-year OS rate was also higher (73.6% vs. 51.6%).

We further explored the importance of upper mediastinal lymph node stations and found that after PSM, complete mediastinal lymph node dissection including 8U, 2R, 2 L, and 4R groups showed significantly better overall survival (8U: HR = 0.45, 95%CI:0.23–0.89; P = 0.02; 2R: HR = 0.42, 95%CI:0.21–0.84; P = 0.01; 2 L: HR = 0.47, 95%CI:0.23–0.97; P = 0.04; 4R group: HR = 0.44, 95%CI:0.21–0.94; P = 0.03) compared to middle and lower mediastinal lymph node dissection (Fig. 3), while the survival difference among the five upper mediastinal lymph node groups was not significant (P = 0.75).

Fig. 3.

The overall survival difference between CMAD group including 8U or 2R, 2 L, 4R and MLMAD group after PSM.(A) Station 8U (Upper paraesophageal); (B) Station 4R (Right lower paratracheal); (C) Station 2R (Right upper paratracheal); (D) Station 2 L (Left upper paratracheal)

Subgroup analysis

To investigate the risk factors for Siewert type II AEG patients, Cox regression analysis results suggested that tumor T-stage and CMAD were independent risk factors for patients (Table 4). To further verify the influence of tumor infiltration depth on upper mediastinal lymph node metastasis and lymph node dissection methods, we calculated the upper mediastinal lymph node metastasis rate at different pT stages and compared the differences in survival prognosis between CMAD and MLMAD at these stages.

Table 4.

Cox regression analysis of patient with SiewertII AEG after PSM

	Univariate Analysis		Multivariate Analysis
Variables	P	95%CI	P	95%CI
LND
MLMAD		1(Reference)		1(Reference)
CMAD	0.002	0.44(0.26-0.74)	0.016	0.48(0.27-0.87)
Differentiation
G1
G2	0.723	0.77(0.18-3.31)	0.399	0.52(0.12-2.36)
G3	0.078	3.61(0.87-15.02)	0.471	1.73(0.39-7.62)
Age
≤60		1(Reference)
>60	0.641	1.13(0.67-1.91)
pN
pN0		1(Reference)
pN1	0.123	1.86(0.85-4.08)	0.826	1.11(0.45-2.72)
pN2	<0.001	3.59(1.81-7.11)	0.323	1.53(0.66-3.54)
pN3	<0.001	5.50(2.63-11.50)	0.051	2.36(1.00-5.59)
pT
pT1-2		1(Reference)		1(Reference)
pT3-4	<0.001	3.60(1.83-7.09)	0.039	2.37(1.04-5.37)
Approaches
Mckeown		1(Reference)		1(Reference)
LT	0.045	1.69(1.01-2.81)	0.990	0.85(0.55-1.80)
Gender
Female		1(Reference)
Male	0.488	0.78(0.38-1.58)

LND lymph node dissection, CMAD complete mediastinum and abdominal lymph node disscton, MLMAD middle and lower mediastinum and abdominal lymph node disscton, pT pathological T stage, pN pathological N stage, LT left thoracic

The results showed that in patients with pT1-2 stage disease, the upper mediastinal lymph node metastasis rate was 2.2%. The OS (P = 0.004) and DFS (P = 0.006) in the CMAD group were better than those in the MLMAD group. In patients with pT3-4 stage disease, the upper mediastinal metastasis rate was 9.6%. The OS (P = 0.04) and DFS (P = 0.05) in the CMAD group were better than those in the MLMAD group (Fig. 4).

Fig. 4.

Differences of CMAD and MLMAD in survival outcomes in different pT stages: (A) Differences in OS between the CMAD and MLMAD groups in pT1-2 staging; (B) Differences in DFS between the CMAD and MLMAD groups in pT1-2 staging; (C) Differences in OS between the CMAD and MLMAD groups in pT3-4 staging; (D) Differences in DFS between the CMAD and MLMAD groups in pT3-4 staging

Surgical approaches comparison

Table 5 lists the lymph node dissection status for each station before and after PSM for the two surgical approaches. For upper mediastinal lymph nodes, the McKeown approach had a higher dissection frequency for stations 2R, 2 L, 4R, and 8U than the LT approach (P < 0.001). In 4 L and 5 lymph node dissections, the LT approach had a higher and statistically significant dissection frequency (P < 0.001). For middle and lower mediastinal lymph node dissection, the McKeown approach had a higher dissection frequency for subcarinal lymph nodes than the LT approach (before PSM: P = 0.018; after PSM: P = 0.014), while the LT approach had a higher and statistically significant dissection frequency for station 9 lymph nodes (P < 0.001). Overall, in abdominal lymph node dissection, the McKeown approach showed a higher dissection frequency for stations 16 and 18 compared to the LT approach.

Table 5.

Comparison of lymph node dissection frequency between the two surgical methods before and after PSM

Lymph node station	Before PSM			After PSM
	LT(n = 122)	Mckeown(n = 99)	P	LT(n = 74)	Mckeown(n = 66)	P
UMN
2R	1	47	<0.001	0	41	<0.001
2 L	2	40	<0.001	2	35	<0.001
4R	0	37	<0.001	0	31	<0.001
4 L	2	0	<0.001	2	0	<0.001
5	7	0	0.018	5	0	<0.001
8U	13	38	<0.001	10	32	<0.001
M/LMN
7	78	78	0.018	55	60	0.014
8 M	81	61	0.483	53	39	0.154
9	80	37	<0.001	50	24	<0.001
10	67	52	0.787	44	41	0.863
8 L	91	70	0.545	57	48	>0.999
15	96	80	0.739	61	55	>0.999
Abdominal
16	93	90	0.004	60	62	0.025
17	100	94	0.003	62	62	0.068
18	63	77	<0.001	38	47	0.002
19	59	45	0.686	34	26	0.495
20	49	34	0.404	33	20	0.116

Abbreviation: UMN upper mediastinal lymph node, M/LMN middle/lower mediastinal lymph node

2R: Right upper paratracheal lymph nodes

2 L: Left upper paratracheal lymph nodes

4R: Right lower paratracheal lymph nodes

4 L: Left lower paratracheal lymph nodes

5: Lymph nodes below the aortic arch

8U: Upper thoracic paraesophageal lymph nodes

8 M: Middle thoracic paraesophageal lymph nodes

8 L: Lower thoracic paraesophageal lymph nodes

7: Subcarinal lymph nodes

9: Pulmonary ligament lymph nodes

15: Supradiaphragmatic lymph nodes

16: Paracardial lymph nodes

17: Lymph nodes at the root of the left gastric artery

18: Lymph nodes along the common hepatic artery

19: lymph nodes along the splenic artery

20: lymph nodes along the celiac artery

Bold values indicate statistical significance at P < 0.05

The detailed profile of postoperative complications, classified by type and Clavien-Dindo grade, is presented in Supplementary Table 2. After PSM, the overall complication rate was comparable between the CMAD and MLMAD groups (80.0% vs. 77.1%, P = 0.837). The rates of major complications (Clavien-Dindo grade III-V) were similarly not significantly different (12.9% vs. 15.7%, P = 0.810). Specifically, the incidence of common and serious complications such as anastomotic leak (5.7% vs. 2.9%, P = 0.681), pneumonia (18.6% vs. 18.6%, P > 0.999), and recurrent laryngeal nerve palsy (10.0% vs. 7.1%, P = 0.764) was all statistically comparable between the two groups. The 90-day postoperative mortality was identical in both groups (4.3%).

Recognizing the significant selection bias between the two surgical approaches, we directly compared survival outcomes. In the overall unmatched cohort, the McKeown approach was not associated with superior OS (HR = 1.45, 95%CI:0.98–2.14; P = 0.06) or DFS (HR = 1.45, 95%CI:0.99–2.14; P = 0.06). After PSM was applied to balance key prognostic variables (including tumor size, clinical nodal stage, and comorbidity indices), the survival analysis revealed a different picture. The McKeown approach was then associated with significantly improved OS (HR = 0.59, 95%CI:0.36–0.99; P = 0.04). A trend towards improved DFS was also observed, though it did not reach statistical significance (HR = 0.61, 95%CI:0.37–1.02; P = 0.06).

Discussion

This retrospective study demonstrates that complete mediastinal and abdominal lymphadenectomy (CMAD) is associated with superior survival in patients with Siewert type II AEG, with the dissection of upper mediastinal stations contributing to this benefit. The McKeown approach proved to be an effective technical strategy for achieving this comprehensive lymph node dissection.

In our cohort of clinically node-negative patients, the incidence of occult upper mediastinal lymph node metastasis was 4.3%. While this rate is lower than the 10.5% reported in a large mapping study that included all Siewert types [17], the difference likely reflects the lower propensity for cephalad spread in Siewert type II tumors compared to type I. Importantly, the presence of these occult metastases carries clear therapeutic implications, as failing to address them could adversely affect outcomes for these patients. This is particularly relevant for stations such as 2R and 4R, where metastasis has been associated with a poorer prognosis [18]. Our finding that CMAD, which includes clearance of the upper mediastinum, was associated with a significant survival benefit over MLMAD provides strong support for considering an extended dissection field in the surgical management of Siewert type II AEG.

The choice of surgical approach is pivotal for achieving the oncologic goal of complete lymphadenectomy. Our comparative analysis suggests that the McKeown approach is a more reliable technical strategy for performing CMAD compared to LT. This is primarily because the three-field exposure of the McKeown procedure provides superior and consistent access to the critical upper mediastinal nodal basins (stations 2R, 2 L, 4R, and 8U), which were associated with survival benefit in our study, as well as to key abdominal stations. While the LT approach offers distinct advantages for dissecting stations around the aortic arch and in the left inferior pulmonary ligament (e.g., stations 4 L and 9), these areas are of lesser prognostic significance for occult metastasis in Siewert type II tumors. Importantly, after rigorously controlling for confounding via propensity score matching, the McKeown approach was associated with a significant overall survival advantage, likely mediated by this more complete lymphadenectomy. Therefore, for surgeons aiming to implement a CMAD strategy, our data support the McKeown approach as the technically superior option for optimizing nodal clearance.

Our findings, which demonstrate a survival benefit for transthoracic CMAD, contribute to the ongoing discussion regarding the optimal surgical approach for Siewert type II AEG [11]. They suggest that for a select subgroup of patients, the oncologic benefit of addressing the upper mediastinum via an approach like McKeown may be justified, highlighting the need for individualized decision-making based on tumor characteristics and patient risk.

Our data shifts the clinical question to dissect the upper mediastinum in which patients this dissection is most warranted. Refining preoperative selection for CMAD remains a challenge. Until more reliable predictive tools are developed, the decision should be individualized, involving multidisciplinary discussion that weighs potential survival benefits against patient-specific operative risks.

The observed survival benefit of CMAD in patients with pT1-2 tumors, despite a low rate of upper mediastinal metastasis (2.2%), is an intriguing finding that merits cautious interpretation. This may be related to more accurate staging or the therapeutic effect of extended lymph node dissection reported in related articles [19, 20]. Conversely, the relatively attenuated benefit in pT3-4 disease may reflect the increasing influence of systemic disease burden in advanced stages. These observations are exploratory and require validation in future studies.

Although we used PSM to balance known prognostic factors, including the use of neoadjuvant and adjuvant therapy, unmeasured variations in specific treatment regimens, cycles, or patient compliance could persist as residual confounders. Nevertheless, the balanced distribution of therapy between groups strengthens the argument that the observed survival benefit is primarily attributable to the extent of lymphadenectomy.

This study has important limitations inherent to its retrospective, single-center design. First and foremost, despite the use of propensity score matching, selection bias and unmeasured confounding cannot be fully excluded. Factors such as surgeon preference, subtle patient characteristics, and evolving clinical practices over the 7-year study period may have influenced both treatment allocation and outcomes. Second, the execution of lymphadenectomy, even within the CMAD group, was subject to practical variability due to anatomical constraints and intraoperative judgment, reflecting the real-world challenge of standardizing a maximalist surgical strategy. Third, our findings require careful consideration of external validity. All procedures were performed by experienced surgeons at a high-volume referral center, and outcomes may differ in other settings with varying surgical philosophies (e.g., centers favoring a transhiatal approach).

Consequently, our results demonstrate a strong association but cannot establish causality. They should be interpreted as hypothesis-generating, providing the rationale for future prospective, multi-center studies to definitively establish the optimal extent of lymphadenectomy and surgical approach for Siewert type II AEG.

Conclusion

The results of this study demonstrate that in patients without clinically apparent upper mediastinal lymph node involvement, CMAD, which includes the clearance of upper mediastinal stations, is associated with significantly improved survival compared to a more limited dissection. This benefit is likely driven by the removal of occult micrometastatic disease and more accurate staging. The McKeown surgery is superior to left thoracotomy for dissecting critical upper mediastinal and abdominal lymph nodes without compromising perioperative safety.

Abbreviations

AEG

Adenocarcinoma of the Esophagogastric Junction

CMAD

Complete Mediastinal and Abdominal Dissection

MLMAD

Middle/Lower Mediastinal and Abdominal Dissection

PSM

Propensity Score Matching

OS

Overall Survival

DFS

Disease-Free Survival

HR

Hazard Ratio

CI

Confidence Interval

LT

Left Thoracotomy

UMLND

Upper Mediastinal Lymph Node Dissection

CT

Computed Tomography

PET/CT

Positron Emission Tomography/Computed Tomography

EGD

Esophagogastroduodenoscopy

ECG

Electrocardiogram

MDT

Multidisciplinary Team

STROCSS

Strengthening the Reporting of Cohort, Cross-sectional and Case-Control Studies in Surgery

AJCC

American Joint Committee on Cancer

JCEC

Japanese Classification of Esophageal Cancer

TNM

Tumor, Node, Metastasis

VIF

Variance Inflation Factor

SMD

Standardized Mean Difference

ASA

American Society of Anesthesiologists

Authors’ contributions

Bo-Wei Liu: Conceptualization, Formal analysis, Methodology, Software, Validation, Visualization, Writing – original draft; Qi-Xin Shang: Formal analysis, Methodology, Software, Validation, Visualization, Writing – original draft; Wei-Peng Hu: Methodology, Supervision, Validation, Writing – review & editing; Long-Qi Chen: Conceptualization, Supervision, Resources, Writing – review & editing; Yu-Shang Yang: Methodology, Yan-Ru Feng: Developing the Conceptualization and Analysis Methods, Supervision; Bo-Liao: Developing the Conceptualization and Analysis Methods, Supervision.

Funding

The authors have no Fundings to declare.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Patient consent was waived due to retrospective study. This study is a retrospective clinical analysis, which does not involve scientific research on animals, humans, human derived samples, and human embryonic stem cells. Ethical proof is not required.

Consent for publication

Not applicaple.

Competing interests

The authors declare no competing interests.