Endoscopic diagnosis and treatment of early gastric cardia neoplasms: Efficacy and management of non-curative resection

Abstract

BACKGROUND

Gastric cardia cancer is a distinct clinicopathological entity with an increasing global incidence. Because of its location at the esophagogastric junction and complex submucosal architecture, early gastric cardia neoplasms pose significant challenges for accurate endoscopic diagnosis and depth assessment. These difficulties may lead to pathological underestimation and high rates of noncurative resection (NCR) after endoscopic submucosal dissection (ESD). Although additional surgery is generally recommended for NCR, this strategy may lead to overtreatment in selected low-risk patients. Therefore, evaluating the diagnostic performance of endoscopic modalities and exploring individualized management strategies after ESD are important clinically.

AIM

To evaluate ESD efficacy and safety for early gastric cardia neoplasms and explore individualized management for patients undergoing NCR.

METHODS

We retrospectively analyzed 155 patients with 163 early gastric cardia mucosal lesions treated by ESD between March 2014 and June 2025. Diagnostic accuracy of white-light endoscopy, magnifying endoscopy with narrow-band imaging, and endoscopic ultrasonography was evaluated using postoperative pathology as the reference standard. ESD outcomes, management of NCR, and long-term survival were assessed. Logistic regression identified factors associated with NCR, and survival was analyzed using Kaplan-Meier and Cox models.

RESULTS

Most lesions were classified as E = G type (Nishi classification) and located on the posterior wall of the cardia. White-light endoscopy showed a diagnostic accuracy of 70.6% (κ = 0.21), compared with 84.3% (κ = 0.54) for magnifying endoscopy with narrow-band imaging. Endoscopic ultrasonography demonstrated an accuracy of 69.5% (κ = 0.12) for assessing invasion depth. All lesions were removed en bloc; histologically complete resection was achieved in 82.8% of cases, and 73.0% met the e-Cura criteria for curative resection. Complications were rare (bleeding 1.8%). The 5-year overall survival and disease-free survival rates were 97.0% and 94.5%, respectively. Of the 44 patients with NCR (27.0%), 7 underwent additional surgery (with no residual tumor or nodal metastasis found) and 37 opted for surveillance. The 5-year disease-free survival was 85.7% in the surgery group and 93.5% in the surveillance group.

CONCLUSION

ESD is effective for early gastric cardia neoplasms. Individualized surveillance for low-risk NCR patients helps reduce overtreatment and improve clinical prognosis.

Key Words: Cardia; Stomach neoplasms; Endoscopic submucosal dissection; Overtreatment; Endoscopy; Digestive system

Core Tip: Endoscopic submucosal dissection is safe and effective for early gastric cardia cancer. Magnifying endoscopy with narrow-band imaging has superior diagnostic accuracy compared with white-light endoscopy (84.3% vs 70.6%). Despite a 27.0% noncurative resection rate due to the anatomical complexity of the cardia, selected low-risk noncurative resection patients can be safely managed with strict endoscopic surveillance.

INTRODUCTION

Gastric cardia cancer (GCC), located at the esophagogastric junction (EGJ), is a distinct clinicopathological entity. Although the global incidence of gastric cancer (GC) has declined, the incidence of GCC continues to increase[1-3]. The 5-year survival rate for patients with advanced GC remains below 20%, whereas early GC (EGC) achieves a 5-year survival rate exceeding 90% following effective treatment[4]. GCC is typically defined as a tumor whose center lies within 2 cm above or below the EGJ[5]. According to World Health Organization classification, early neoplastic gastric mucosal lesions include low-grade intraepithelial neoplasia (LGIN), high-grade intraepithelial neoplasia, intramucosal carcinoma, and submucosal carcinoma. Most researchers agree that EGC follows Correa’s cascade[6], providing a critical window for endoscopic intervention.

As a transitional zone between squamous and columnar epithelium, the cardia has a complex anatomical structure. Its submucosal architecture and glandular distribution differ from other gastric regions, conferring distinct physiological and biological characteristics to GCC[1,7]. Early neoplastic lesions in the gastric cardia often demonstrate greater heterogeneity in glandular architecture and microvascular patterns compared with those in noncardia gastric regions. They lack the typical and stable microsurface and microvascular features observed in the gastric body and antrum, rendering existing diagnostic criteria less accurate for GCC. Endoscopic submucosal dissection (ESD) is the primary endoscopic treatment for early GCC (EGCC). Current post-ESD therapeutic assessment primarily relies on the Japanese e-Cura system, which results in a high noncurative resection (NCR) rate, and theoretically carries risks of lymph node metastasis and local recurrence. Clinical guidelines recommend additional surgery for patients with NCR; however, surgery entails risks of overtreatment, postoperative complications, mortality, and compromised quality of life[5]. Thus, refining the therapeutic evaluation system and optimizing post-ESD management for EGCC are crucial.

This study systematically analyzed clinical, endoscopic, and pathological characteristics and diagnostic accuracy of early gastric cardia mucosal tumors; evaluated short-term and long-term outcomes of ESD; and explored risk factors and management strategies for NCR, providing evidence for precision diagnosis, therapeutic assessment, and individualized postoperative management.

MATERIALS AND METHODS

Patient selection

Patients with early gastric cardia tumors who underwent ESD from March 2014 to June 2025 were included. Clinical and procedural information was retrieved from the hospital’s HIS system. All patients met ESD indications and had no contraindications due to metastasis. This study was approved by the Ethics Committee (No. 2025-156). Exclusion criteria included: (1) No ESD treatment; (2) Incomplete clinical or follow-up data; and (3) Severe comorbidities (e.g., heart, liver, or renal failure). In addition, patients with a postoperative pathological diagnosis of poorly differentiated adenocarcinoma (including signet-ring cell carcinoma and mucinous adenocarcinoma) were excluded. The rationale for this exclusion was that poorly differentiated GC exhibits highly aggressive biological behavior, with a propensity for early lymphovascular invasion and lymph node metastasis, which differs substantially from the behavior of well-differentiated or moderately differentiated carcinomas. According to the Japanese Gastric Cancer Treatment Guidelines and current consensus, poorly differentiated carcinoma is not considered an absolute or expanded indication for ESD, and surgical resection is typically preferred to ensure curative intent.

ESD procedure

Preoperative evaluation: Lesion characteristics, extent, margins, and invasion depth were assessed by white-light endoscopy (WLE), magnifying endoscopy with narrow-band imaging (ME-NBI), and endoscopic ultrasonography (EUS). Patients fasted for 24 hours and discontinued contraindicated medications. ESD procedure: ESD was performed under anesthesia by experienced endoscopists following standard procedures: (1) Marking; (2) Submucosal injection; (3) Circumferential incision; (4) Submucosal dissection; (5) Wound management; (6) Specimen fixation; and (7) Documentation.

Postoperative management: After the procedure, patients were placed on strict fasting for 24-48 hours and received routine fluid replacement therapy. Prophylactic antibiotics were administered to prevent infection, and gastric mucosal protectants along with acid-suppressive agents were used to promote wound healing. The average postoperative hospital stay was 2-3 days; during which, patients were closely monitored for vital signs and any postoperative complications (such as hematemesis, melena, fever, or chest pain). Most patients gradually resumed a normal diet 24-48 hours after the procedure and were discharged once their clinical condition was stable, and then entered the follow-up protocol.

Pathological evaluation

The location of GCC was classified according to the Nishi system, which divides tumors into five types along the longitudinal axis based on the position of the tumor center and its distance from the EGJ. Anatomical location was also recorded using the Japanese Gastric Cancer Association classification (anterior/posterior wall, lesser/greater curvature). For the purpose of this analysis, the five Nishi types were further consolidated into three categories: (1) E/EG type (tumor center above the EGJ, predominantly on the esophageal side); (2) E = G type (tumor center precisely at the EGJ, involving both the esophagus and stomach); and (3) GE/G type (tumor center below the EGJ, predominantly on the gastric side).

Tumor invasion depth was defined according to the Japanese Gastric Cancer Association classification: (1) M1: Tumor confined to the epithelium; (2) M2: Tumor confined to the lamina propria without involving the muscularis mucosae; (3) M3: Tumor involving the muscularis mucosae; (4) SM1: Tumor invading the superficial submucosa (< 500 μm from the muscularis mucosae); and (5) SM2: Tumor invading the deep submucosa (≥ 500 μm). Due to the limited number of tumors in the lamina propria in this study, M1 and M2 Lesions were grouped together.

Standardized protocol for pathological specimen processing. To minimize the potential impact of curling artifacts and electrocautery-induced thermal damage on the interpretation of NCR, a standardized protocol was implemented for specimen processing. Immediately after ESD, the specimen was flattened and fixed onto a corkboard with stainless steel pins by the endoscopist and pathology technician to restore its in situ morphology as much as possible and reduce curling artifacts caused by submucosal contraction. The horizontal and vertical margins were marked with India ink, and the specimen was serially sectioned at 2-3 mm intervals to ensure proper orientation and integrity. During pathological evaluation, the pathologist paid special attention to identifying areas of thermal injury (e.g., coagulative necrosis or elongated cell morphology at the margins). When a positive margin overlapped with an area of thermal injury, a comprehensive judgment was made based on the morphology of the surrounding tissue, and this was noted in the pathology report. All sections were independently reviewed by two experienced gastrointestinal pathologists, and any disagreements were resolved through discussion to minimize the influence of processing-related artifacts or thermal damage on NCR assessment.

Post-ESD safety and efficacy

Post-ESD complications mainly include bleeding, perforation, and stricture. The criteria for evaluating resection quality were as follows: (1) En bloc resection: Complete resection of the lesion under endoscopy as a single piece; (2) Complete resection: En bloc resection with pathological confirmation of tumor-free horizontal and vertical margins; (3) Curative resection: Defined according to the eCura system in the Japanese Gastric Cancer Treatment Guidelines, with eCuraA and eCuraB corresponding to curative resection. The detailed criteria were as follows: (1) The eCuraA: Differentiated-type carcinoma without ulceration, meeting all of the following: En bloc resection, any tumor size, pT1a, negative horizontal and vertical margins, and no lymphovascular invasion. Differentiated-type carcinoma with ulceration, meeting all of the following: En bloc resection, tumor diameter ≤ 3 cm, pT1a or pT1b, negative horizontal and vertical margins, and no lymphovascular invasion; (2) The eCuraB: Initial ESD or endoscopic mucosal resection confirmed differentiated-type carcinoma meeting eCuraC1 criteria, followed by intramucosal recurrence, and subsequent ESD achieved eCuraA criteria; (3) The eCuraC1: Differentiated-type carcinoma partially meeting eCuraA or eCuraB criteria, but with non-en bloc resection or pathologically positive horizontal or vertical margins; and (4) The eCuraC2: All eCuraC resections other than those classified as eCuraC1.

Follow-up

Patients with curative resection underwent regular follow-up according to the Japanese guidelines. Patients with NCR were managed with additional surgery or intensive surveillance (endoscopy + chest/abdominal computed tomography) based on e-Cura C1/C2 risk stratification.

Statistical analyses

Statistical analyses were performed using SPSS Statistics 26.0 (IBM SPSS Statistics, Armonk, NY, United States) and GraphPad Prism 10 (GraphPad Software Inc., San Diego, CA, United States). Continuous variables are expressed as the mean ± SD or median (interquartile range), and categorical variables were presented as n (%). Comparisons between groups were performed using the χ² test or Fisher’s exact test for categorical variables and the independent samples t-test or Mann-Whitney U test for continuous variables. Univariate analyses were performed to identify factors associated with NCR, and variables with P < 0.10 were subsequently entered into a multivariate logistic regression model to determine independent predictors. Diagnostic performance of different endoscopic modalities was evaluated by calculating sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy, and agreement with pathological findings was assessed using the κ coefficient. Long-term outcomes were analyzed using the Kaplan-Meier method and compared using the log-rank test. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

RESULTS

A total of 155 patients with 163 lesions treated with ESD from March 2014 to June 2025 were included in this study.

Differences in clinical characteristics

Significant differences were observed in baseline clinical characteristics among different lesion groups (Table 1). The LGIN group was younger, had higher body mass index, and showed a higher rate of co-occurring Barrett’s esophagus.

Table 1 Differences in clinical characteristics of cardiac mucosal lesions with different degrees of differentiation, n (%)/median (interquartile range).

Item	Low-grade intraepithelial neoplasia (n = 34)	High-grade intraepithelial neoplasia (n = 40)	Intramucosal carcinoma (n = 74)	Submucosal carcinoma (n = 15)	P value
Sex (%)					0.021
Male	24 (70.6)a	37 (92.5)a	66 (89.2)a	11 (73.3)a
Female	10 (29.4)a	3 (7.5)a	8 (10.8)a	4 (26.7)a
Age (years)	56.2 (44.3-68.1)a	64.5 (55.8-73.2)a	63.5 (55.3-71.7)a	69.2 (59.4-79.0)a	< 0.001
Age groups (years)					0.005
≤ 44	4 (11.8)a	1 (2.5)a	1 (1.4)a	0a
45-64	22 (64.7)a	17 (42.5)a	41 (55.4)a	4 (26.7)a
≥ 65	8 (23.5)a	22 (55.0)a	32 (43.2)a	11 (73.3)a
BMI (kg/m2)	25.4 (17.8-29.7)a	23.0 (17.7-30.1)a	23.1 (16.0-28.7)a	23.8 (14.5-26.9)a	0.010
BMI groups					0.049
< 18.5	2 (5.9)a	2 (5.0)a	5 (6.8)a	1 (6.7)a
18.5-23.9	9 (26.5)a	23 (57.50)a	40 (54.1)a	7 (46.7)a
24-27.9	17 (50.0)a	11 (27.5)a	27 (36.5)a	7 (46.7)a
> 28	6 (17.6)	4 (10.0)	2 (2.7)	0
Comorbidities (%)
Diabetes	0	7 (17.5)	7 (9.6)	1 (6.7)	0.056
Hypertension	9 (26.5)	12 (30.0)	18 (24.3)	4 (26.7)	0.934
Cardiovascular disease	3 (8.8)	7 (17.5)	12 (16.4)	2 (13.3)	0.712
Cirrhosis/hepatitis	0a	0a	8 (10.8)a	0a	0.026
Respiratory disease	1 (2.9)	3 (7.5)	3 (4.1)	1 (6.7)	0.650
Reflux esophagitis	4 (11.8)	6 (15.0)	8 (10.8)	2 (13.3)	0.884
Barrett’s esophagus	5 (14.7)a	1 (2.5)a	0a	0a	0.003
Atrophic gastritis	20 (58.8)a	36 (90.0)a	59 (79.7)a	11 (73.3)a	0.014
Helicobacter pylori infection (%)	5 (14.7)	7 (17.5)	10 (13.5)	1 (6.7)	0.830
Smoking history (%)	7 (20.6)	17 (42.5)	31 (41.9)	8 (53.3)	0.085
Alcohol history (%)	7 (20.6)	9 (22.5)	18 (24.3)	4 (26.7)	0.961
Family history (%)	5 (14.7)	1 (2.5)	6 (8.1)	1 (6.7)	0.269

^aP < 0.05.

BMI: Body mass index.

Differences in pathological characteristics

The pathological types differed significantly depending on the cardiac lesion group (Table 2). Lesion progression significantly correlated with increased lesion length and depth of invasion.

Table 2 Differences in pathological characteristics of cardiac mucosal lesions, n (%)/median (interquartile range).

Item	Low-grade intraepithelial neoplasia (n = 34)	High-grade intraepithelial neoplasia (n = 40)	Intramucosal carcinoma (n = 74)	Submucosal carcinoma (n = 15)	P value
Lesion major diameter (cm)	2.5 (1.4-3.6)a	2.9 (1.8-4.0)a	3.2 (1.8-4.6)a	3.0 (1.7-4.3)a	0.021
Lesion location (%)					0.088
E/EG type	9 (26.5)	6 (15.0)	6 (8.1)	2 (13.3)
E = G type	16 (47.1)	20 (50.0)	31 (41.9)	9 (60.0)
GE/G type	9 (26.5)	14 (35.0)	37 (50.0)	4 (26.7)
Lesion depth (%)					< 0.001
M1, M2	34 (100.0)a	34 (85.0)a	37 (50.0)a	0a
M3	0a	2 (5.0)a	37 (50.0)a	0a
SM1	0a	4 (10.0)a	0a	11 (73.3)a
SM2	0a	0	0	4 (26.7)a
Pathological type (%)					< 0.001
Glandular epithelial	17 (50.0)a	34 (85.0)a	73 (98.6)a	15 (100.0)a
Squamous epithelial	12 (35.3)a	6 (15.0)a	1 (1.4)a	0a
Mixed glandular/squamous	5 (14.7)a	0a	0a	0a

^aP < 0.05.

E/EG type: Tumor center above the esophagogastric junction, predominantly on the esophageal side; E = G type: Tumor center precisely at the esophagogastric junction, involving both the esophagus and stomach; GE/G type: Tumor center below the esophagogastric junction, predominantly on the gastric side; M: Tumor is confined to the mucosa; SM1: Tumor invading the superficial submucosa (< 500 μm from the muscularis mucosae); SM2: Tumor invading the deep submucosa (≥ 500 μm).

Differences in WLE features

The LGIN group was dominated by flat and elevated types, while the other three groups were predominantly flat. Mucosal whitening was common in the LGIN group but significantly reduced or absent in higher-grade lesions (Table 3).

Table 3 Differences in white light endoscopy features, n (%).

Item	Low-grade intraepithelial neoplasia (n = 34)	High-grade intraepithelial neoplasia (n = 40)	Intramucosal carcinoma (n = 74)	Submucosal carcinoma (n = 15)	P value
Lesion location (%)					0.892
Greater curvature	5 (17.2)	3 (8.8)	12 (16.4)	1 (6.7)
Lesser curvature	10 (34.5)	9 (26.5)	19 (26.0)	4 (26.7)
Anterior wall	3 (10.3)	2 (5.9)	7 (9.6)	1 (6.7)
Posterior wall	11 (14.4)	20 (58.8)	35 (47.9)	9 (60.0)
Paris classification (%)					0.001
Elevated	13 (38.2)a	3 (7.5)a	7 (9.5)a	1 (6.7)a
Flat	20 (58.8)a	28 (70.0)a	56 (75.7)a	9 (60.0)a
Depressed	1 (2.9)a	2 (5.0)a	2 (2.7)a	1 (6.7)a
Mixed	0a	7 (17.5)a	9 (12.2)a	4 (26.7)a
Roughness (%)	18 (52.9)	23 (57.5)	49 (66.2)	10 (66.7)	0.536
Redness (%)	20 (58.8)	31 (77.5)	55 (74.3)	7 (46.7)	0.057
Whitening (%)	5 (14.7)	0	1 (1.4)	0	0.005
Ulceration or erosion (%)	6 (17.6)	7 (17.5)	15 (20.3)	4 (26.7)	0.876
White fur/coating (%)	4 (11.8)	4 (10.0)	16 (21.6)	4 (26.7)	0.248
Spontaneous bleeding (%)	0	2 (5.0)	4 (5.4)	0	0.573
Fold abnormality (%)	0	0	0	0
Poor peristalsis (%)	0	0	5 (6.8)	1 (6.7)	0.214

^aP < 0.05.

ME-NBI endoscopic features

For the 12 patients with squamous epithelial lesions, significant differences in IPCL classification distribution were observed among the three groups (Table 4). For the 108 patients with glandular epithelial lesions, microvascular abnormalities and demarcation lines were significantly more common in intramucosal carcinoma and submucosal carcinoma groups compared to the LGIN group (Table 5). Microglandular abnormalities were significantly more common in the high-grade intraepithelial neoplasia and intramucosal carcinoma groups compared to the LGIN group.

Table 4 Magnifying endoscopy with narrow-band imaging features (squamous epithelium), n (%).

Item	Low-grade intraepithelial neoplasia (n = 6)	High-grade intraepithelial neoplasia (n = 5)	Intramucosal carcinoma (n = 1)	P value
Intrapapillary capillary loop types (%)				0.031
A	5 (83.3)a	0a	0a
B1	1 (16.7)a	4 (80.0)a	1 (100.0)a
B2	0a	1 (20.0)a	0a
B3	0	0	0

^aP < 0.05.

Table 5 Magnifying endoscopy with narrow-band imaging features (adenomatous epithelium), n (%).

Classification	Low-grade intraepithelial neoplasia (n = 17)	High-grade intraepithelial neoplasia (n = 28)	Intramucosal carcinoma (n = 53)	Submucosal carcinoma (n = 10)	P value
Irregular microvascular pattern	5 (29.4)a	17 (60.7)a	45 (84.9)a	10 (100.0)a	< 0.001
Irregular microsurface pattern	5 (29.4)a	21 (75.0)a	37 (69.8)a	8 (80.0)a	0.006
Demarcation line	7 (41.2)a	22 (78.6)a	47 (88.7)a	10 (100.0)a	< 0.001

^aP < 0.05.

Diagnostic efficacy of different endoscopies for EGCC

Using postoperative pathological diagnosis as the gold standard, we evaluated the diagnostic efficacy of WLE and ME-NBI for EGCC nature, and EUS for EGCC invasion depth.

WLE: The sensitivity, specificity, and accuracy of histological biopsy under WLE for diagnosing EGCC were 74.5%, 52.4%, and 70.6%, respectively, with a κ coefficient of 0.21 (Table 6).

Table 6 Comparison of white light imaging biopsy and postoperative pathology for early gastric cardia cancer diagnosis.

Preoperative biopsy diagnosis	Postoperative pathology		Total
	Positive (HGIN + carcinoma)	Negative (LGIN)
Positive (HGIN + carcinoma)	73	10	83
Negative (LGIN)	25	11	36
Total	98	21	119

HGIN: High-grade intraepithelial neoplasia; LGIN: Low-grade intraepithelial neoplasia.

ME-NBI (MESDA-G algorithm): Using the Magnifying Endoscopy Simple Diagnostic Algorithm for Early Gastric Cancer (MESDA-G) to evaluate glandular lesions: Sensitivity, specificity, and accuracy of MESDA-G for diagnosing EGCC were 85.4%, 79.0%, and 84.3%, respectively, with a κ coefficient of 0.54 (Table 7).

Table 7 Comparison of Magnifying Endoscopy Simple Diagnostic Algorithm for Early Gastric Cancer and postoperative pathology for early gastric cardia cancer diagnosis.

Magnifying Endoscopy Simple Diagnostic Algorithm for Early Gastric Cancer	Postoperative pathology		Total
	Positive	Negative
Positive	76	4	80
Negative	13	15	28
Total	89	19	108

EUS assessment of invasion depth: Analysis of 95 Lesions assessed by preoperative EUS showed an overall accuracy of approximately 69.5% and a κ coefficient of 0.12, suggesting low consistency between EUS diagnosis and pathological results (Tables 8 and 9).

Table 8 Comparison of endoscopic ultrasonography and pathological staging.

Endoscopic ultrasonography invasion depth	Pathological invasion depth			n
	M (pT1a)	SM1 (pT1b)	SM2 (pT1b)
uT1a	60	68	0	68
uT1b	21	27	2	27
Total	81	95	2	95

M: Tumor is confined to the mucosa; pT1a: Pathological stage T1a, tumor confined to the mucosa (M layer); pT1b: Pathological stage T1b, tumor invading the submucosa (SM layer), further divided into SM1 (invasion < 500 μm from the muscularis mucosae) and SM2 (invasion ≥ 500 μm from the muscularis mucosae); SM1: Tumor invading the superficial submucosa (< 500 μm from the muscularis mucosae); SM2: Tumor invading the deep submucosa (≥ 500 μm); uT1a: Endoscopic ultrasonography stage T1a, tumor confined to the mucosa; uT1b: Endoscopic ultrasonography stage T1b, tumor invading the submucosa.

Table 9 Sensitivity analysis of endoscopic ultrasonography for invasion depth.

Invasion depth	Sensitivity	Specificity	Accuracy	Positive predictive value	Negative predictive value
uT1a	74.1	42.9	69.5	88.2	22.2
uT1b	42.9	74.1	69.5	22.2	88.2

uT1a: Endoscopic ultrasonography stage T1a, tumor confined to the mucosa; uT1b: Endoscopic ultrasonography stage T1b, tumor invading the submucosa.

Therapeutic outcomes and safety of ESD

Complete resection rates and curative resection rates decreased significantly with increasing lesion severity (Table 10).

Table 10 Differences in surgical efficacy and safety, n (%).

Item	Low-grade intraepithelial neoplasia (n = 34)	High-grade intraepithelial neoplasia (n = 40)	Intramucosal carcinoma (n = 74)	Submucosal carcinoma (n = 15)	P value
Absolute indication	32 (94.1)a	34 (85.0)a	65 (87.8)a	0a	< 0.001
En bloc resection	34 (100.0)	40 (100.0)	74 (100.0)	15 (100.0)
Complete resection	33 (97.1)a	36 (90.0)a	57 (77.0)a	9 (60.0)a	0.004
Curative resection	31 (91.2)a	33 (82.5)a	50 (67.6)a	5 (33.3)a	< 0.001
Complications	0	2 (5.0)	0	1 (6.7)	0.073
Intraoperative bleeding	0	0	0	0
Delayed bleeding	0	2 (5.0)	0	1 (6.7)	0.281
Perforation	0	0	0	0

^aP < 0.05.

Factors influencing diagnostic accuracy and resection quality

Patient sex and lesion differentiation degree were independent risk factors for pathological underestimation. Lesion size and mucosal redness were independent risk factors for incomplete resection (Tables 11 and 12).

Table 11 Multivariate analysis of pathological underestimation.

Related factors	Categories	P value	Odds ratio	95%CI
Sex	Male	0.017	1	1.271-32.139
	Female		6.391
Differentiation degree	Low-grade intraepithelial neoplasia	0.045	0.086	0.008-0.952
	Submucosal carcinoma		1

Table 12 Multivariate analysis of factors related to incomplete resection.

Related factors	Categories	P value	Odds ratio	95%CI
Lesion size	-	0.044	0.641	0.416-0.988
Mucosal redness	Yes	0.010	0.140	0.032-0.623
	No		1

Clinical characteristics and management outcomes of patients with NCR

This study included 44 cases of NCR, with an incidence of 27.0%. The classification of NCR causes is shown in Table 13. We performed a cross-study comparative analysis of key management indicators for patients with EGCC NCR from previous studies (Table 14)[8-11].

Table 13 Comparison of causes for noncurative resection.

Item	Low-grade intraepithelial neoplasia (n = 3)	High-grade intraepithelial neoplasia (n = 7)	Intramucosal carcinoma (n = 24)	Submucosal carcinoma (n = 10)	Total (n = 44)
Incomplete resection					25
Horizontal margin	1	3	13	4	21
Vertical margin	0	0	4	0	4
Complete resection with high-risk factors
Ulcer	2	2	6	3	13
SM2 invasion	0	1	1	3	5
Diameter > 3 cm	2	3	6	4	15
Lymphovascular invasion	0	0	1	1	2

A patient could have more than one cause for Noncurative resection. Therefore, the sum of subcategories may exceed the total number of patients in groups. SM2: Tumor invading the deep submucosa (≥ 500 μm).

Table 14 Cross-study comparison of key indicators for noncurative resection patient management.

Key indicators	Noncurative resection incidence (%)	Surgical group (%)	Positive postoperative pathology rate (%)
Our study	27.0	15.9	0
Cao et al[11]	19.5	43.0	7.0
Yoshinaga et al[8]	28.0	28.6	0
Jang et al[9]	34.0	30.0	-
Osumi et al[10]	18.4	62.5	0

DISCUSSION

GCC is a distinct subtype of GC. Due to its anatomical location and structural specificity, the epidemiology, biological behavior, endoscopic features, efficacy, and prognosis of endoscopic resection for GCC differ from those of non-GCC. This study systematically evaluated clinicopathological features, endoscopic diagnostic efficacy, and long-term prognosis of ESD for early cardiac mucosal tumors, with a focus on exploring management strategies for NCR.

In this study, the diagnostic accuracy of MESDA-G reached 84.3%, significantly higher than that of WLE (70.6%), which is consistent with previous reports[12,13]. Several studies have reported variable accuracy (64.8%-92%) for EUS in assessing invasion depth in EGC at noncardia sites[14-16]. The overall accuracy of EUS for determining invasion depth in early cardia lesions was 69.5%, suggesting suboptimal performance. We found that the diagnostic accuracy of both the MESDA-G criteria and EUS was significantly lower for early lesions located in the gastric cardia than for those in noncardia sites. The MESDA-G criteria were primarily established based on observations from the single columnar epithelium of the gastric body and antrum, where microsurface and microvascular patterns are uniform and stable. In contrast, the gastric cardia is located at the squamocolumnar junction of the esophagogastric transition zone, where the glandular architecture exhibits marked intrinsic heterogeneity and lacks a consistent normal structural baseline. In addition, this region is particularly susceptible to chronic inflammation related to gastroesophageal reflux, which may induce mucosal microvascular dilation and structural distortion, making it more difficult to distinguish inflammatory alterations from neoplastic changes, and potentially reducing the diagnostic performance of the MESDA-G criteria[17,18]. The ability of EUS to discriminate between mucosal cancer and superficial submucosal invasion in EGC remains limited. The spatial resolution of EUS may not adequately detect subtle disruption of the muscularis mucosae when tumor invasion is minimal, leading to potential staging inaccuracies. Inflammatory reactions, fibrosis, or ulcerative changes around the lesion may also produce hypoechoic areas that mimic submucosal involvement, resulting in overstaging. Technical factors further contribute to diagnostic difficulty in the cardia. The confined anatomical space hampers stable apposition of the endoscope and ultrasound probe, while respiratory and cardiac motion can compromise the stability of ME-NBI and reduce the clarity of EUS-defined wall layers. In addition, the submucosa in this region contains abundant fibrous and vascular tissue, which may obscure the true invasion boundary when involved by tumor infiltration[19-21]. These factors collectively explain the suboptimal diagnostic performance observed in this region.

ESD technology demonstrates excellent safety and efficacy for EGCC. Previous small studies reported curative resection rates for EGCC ESD fluctuating between 66% and 81%[8-10]. A large study in 2020 showed a curative resection rate of 80.5%[11], consistent with our results. We found that lesion diameter and mucosal redness were independent risk factors affecting complete resection. Mucosal redness often corresponds histologically to abnormal capillary proliferation and dense infiltration. This increases intraoperative bleeding risk, obscures the visual field, and implies more active biological behavior of the lesion. This finding effectively links endoscopic appearance with treatment difficulty via pathophysiology. We suggest that clinical endoscopists should regard mucosal redness as a warning sign requiring high vigilance and finer pre-resection planning. We found that curative resection rate decreased with increasing lesion malignancy, suggesting that strictly following absolute indication criteria in the cardiac region is important for ensuring radical cure. This result aligns well with the Japanese Gastric Cancer Treatment Guidelines regarding risk control for cardiac ESD.

The most significant contribution of this study lies in its exploration of individualized management for patients with NCR. Based on high-risk factors, physical condition, and personal preference, patients with NCR adopted two different follow-up strategies. Surgical group: Seven patients (15.9%) underwent additional surgery. Postoperative pathology showed no residual tumor or lymph node metastasis in any case. Follow-up showed 1 patient died of respiratory failure post-surgery. The disease-free survival rate was 85.7%. Surveillance group: 37 patients (84.1%) chose regular endoscopic monitoring due to advanced age, comorbidities, or refusal of surgery. Follow-up revealed one local recurrence (recurrence rate 2.7%), which was radically cured after repeat endoscopic resection. All patients survived, with a disease-free survival rate of 93.5% (Figure 1). The NCR incidence for EGCC in this study (27.0%) is consistent with the range reported in previous literature (18.4%-34.0%)[8-11], confirming the technical difficulty and high NCR risk associated with cardiac ESD. This may be attributed to the unique anatomical and histological characteristics of the cardia. Specifically, the tight adhesion between mucosa and submucosa, frequent submucosal fibrosis due to chronic reflux, and limited submucosal lifting effect increase the difficulty of precise dissection and margin assessment. The need for retroflexed endoscopic positioning in a narrow working space reduces procedural stability and may contribute to positive margins and underestimation of invasion depth[22,23].

Figure 1 Kaplan-Meier curves show overall survival and disease-free survival during follow-up. A: Overall survival; B: Disease-free survival (DFS); C: DFS curve in the curative resection subgroup; D: DFS curve in the non-curative resection subgroup. NCR: Non-curative resection.

Beyond confirming these technical challenges, this study provides exploratory insights into the management of patients with NCR. However, the number of patients undergoing additional surgery was low, and survival outcomes were heavily influenced by a small number of events. Therefore, a reliable comparison between additional surgery and strict endoscopic surveillance is not feasible. Importantly, most patients in the surveillance group represented a low-risk population, characterized by advanced age, multiple comorbidities, and absence of high-risk pathological features such as deep submucosal invasion or lymphovascular involvement. Accordingly, the favorable outcomes observed in this group are more likely attributable to selection bias rather than a true therapeutic advantage of surveillance. These findings should therefore be interpreted as hypothesis-generating rather than practice-changing.

Based on these findings, the outcomes of the surveillance group, which constituted the vast majority, became key evidence. This group demonstrated 100% OS and only 2.7% recurrence rate, with the recurrent case radically cured by repeat resection. In contrast, although the additional surgery group achieved 0% residue, one postoperative death occurred, resulting in lower OS than in the surveillance group. The underlying reason for this apparent contradiction may lie in patient selection bias. Patients accepting surveillance in this study were mostly low-risk NCR cases (e.g., only positive margins without lymphovascular invasion or deep infiltration) with advanced age and multiple comorbidities. Their inherent risk of tumor progression might have been overestimated by the e-Cura system, while surgical trauma could have triggered deterioration of underlying conditions. These comparative results provide strong real-world data supporting the hypothesis that strict endoscopic surveillance can serve as an alternative strategy for low-risk NCR patients. The thermal coagulation effect during ESD may have cleared minute marginal residues, or cardiac specimen curling during fixation may have caused false-positive pathology. Therefore, for patients with only positive horizontal margins (HM0) but no other high-risk factors (e.g., SM2 invasion or lymphovascular invasion), mandatory additional high-risk cardiac surgery may lead to overtreatment.

This study had several limitations. First, as a single-center retrospective study, the overall sample size – particularly within the NCR subgroup – was small, which may have introduced selection bias. Specifically, only 7 patients underwent additional surgery, and the postoperative death of 1 patient had a disproportionate impact on the survival analysis, limiting the statistical power for reliable comparison with the surveillance group. Second, all endoscopic image interpretations and pathological evaluations were performed by experts from our center. Although this ensured internal consistency, the absence of multicenter blinded assessment may limit the external validity of the reported diagnostic performance. Third, owing to the favorable prognosis in this cohort, with few recurrence or mortality events during follow-up, it was not possible to establish a robust long-term prognostic predictive model. In addition, patients with poorly differentiated adenocarcinoma were excluded from this study. Therefore, the findings regarding the diagnostic performance of the MESDA-G criteria and EUS, the safety and efficacy of ESD, and the exploratory observation that strict endoscopic surveillance may represent a potential management option for selected patients with NCR are primarily applicable to well or moderately differentiated early gastric cardia neoplasms. These conclusions should not be directly extrapolated to poorly differentiated lesions, for which surgical resection remains the recommended treatment according to current guidelines.

Future studies should aim to clarify several unresolved issues. First, detailed pathological analyses of cardiac ESD specimens are needed to better understand potential causes of false-positive assessments within the e-Cura system at this anatomical site; particularly factors affecting margin evaluation such as thermal coagulation artifacts and specimen curling. Second, the development of cardia-specific diagnostic models integrating conventional endoscopic morphology and magnifying endoscopy features may help improve preoperative diagnostic accuracy, potentially with the assistance of artificial intelligence. Finally, multicenter prospective studies with larger sample sizes are required to rigorously compare long-term outcomes between additional surgery and strict endoscopic surveillance in low-risk NCR patients with gastric cardia lesions.

CONCLUSION

In summary, although ESD is technically feasible and effective for early gastric cardia neoplasms, its diagnostic and curative performance is limited by the unique anatomical and histological characteristics of this region. Careful preoperative assessment and appropriate patient selection remain crucial to optimize clinical outcomes.