Discrepancies between preoperative assessment and final pathological criteria in early gastric cancer

Abstract

BACKGROUND

Inappropriate selection of patients with early gastric cancer (EGC) for endoscopic submucosal dissection (ESD) may lead to non-curative resection, necessitating additional gastrectomy. Conversely, inappropriate selection for gastrectomy may result in overtreatment, adversely affecting patients’ quality of life. Few have systematically evaluated the concordance between therapeutic indications under current Japanese guidelines and pathological criteria in EGC. To minimize noncurative resection risks while sparing unnecessary surgery for low-risk patients’, we specifically assess the suitability of Japanese guidelines in non-Japanese populations. This work aims to optimize clinical practice by refining endoscopic treatment criteria for adoption beyond Japan.

AIM

To evaluate EGC clinical decision accuracy by comparing therapeutic indication with postoperative pathological criteria and analyzing factors influencing discrepancies.

METHODS

A retrospective analysis was conducted on 796 EGC cases diagnosed at Peking University Third Hospital between January 2010 and December 2022. Cases were categorized into three groups: Same-estimated (preoperative therapeutic indication with postoperative pathological criteria matched), underestimated (preoperative ESD indication but postoperative surgical criteria), and overestimated (preoperative surgical indication but postoperative ESD criteria). The rate of discrepancy and associated risk factors were assessed.

RESULTS

The accuracy rates of preoperative evaluation for ESD and gastrectomy indications were 73.0% (321/430) and 76.0% (278/366), respectively. The overall discrepancy rate was 25.6% (204/796). Multivariate analysis identified tumor location in the upper-third stomach (odds ratio = 2.158, 95% confidence interval: 1.373-3.390, P = 0.001) was significantly associated with a higher likelihood of being underestimated and undifferentiated histologic type on preoperative biopsy (odds ratio = 2.005, 95% confidence interval: 1.036-3.879, P = 0.039) was more likely to be overestimated. Significant differences were observed in tumor diameter (P < 0.001), depth of infiltration (P < 0.001), ulcerative findings (P < 0.001), and histologic type (P < 0.001) between preoperative and postoperative evaluations.

CONCLUSION

The accuracy of preoperative EGC indications is 74.4%. Upper-third stomach and undifferentiated histology are primary discrepancy predictors. Upper-third tumors are prone to underestimation, while undifferentiated tumors are prone to overestimation.

Key Words: Early gastric cancer; Endoscopic submucosal dissection; Gastrectomy; Therapeutic indication; Pathological criteria

Core Tip: Current Japanese guideline-based therapeutic indications for early gastric cancer show 74.4% concordance with postoperative pathological criteria in this cohort of 796 patiensts (2010-2022). Key discordance risk factors are tumors in the upper third stomach (prone to pathological underestimation) and undifferentiated histology (prone to overestimation). These findings necessitate heightened preoperative vigilance: endoscopic reevaluation with mapping biopsies for proximal lesions to address technical challenges and avoid non-curative and secondary resections.

INTRODUCTION

Endoscopic resection (ER) is a minimally invasive alternative to gastrectomy for mucosal and submucosal neoplastic lesions. For early gastric cancer (EGC) with lymph node metastasis (LNM) risk < 1%, endoscopic submucosal dissection (ESD) offers advantages, such as reduced trauma and preserved gastric function, making it the preferred treatment in select cases[1-3]. ER is a localized treatment limited to lesions with low incidence of LNM. However, preoperative imaging cannot definitively exclude LNM, necessitating the reliance on post-ER pathology for accurate assessment[4]. Discrepancies between preoperative indications and postoperative criteria are common[5,6]. While some EGC patients undergo gastrectomy but are later deemed ineligible based on postoperative findings, improper patient selection for surgery can result in unnecessary interventions and compromised quality of life[7]. Conversely, certain EGC patients initially meeting the criteria for ESD may ultimately require additional surgical management due to unexpected pathological outcomes[8,9].

The current Japanese Gastric Cancer Treatment Guidelines (2021; 6^th edition)[4] (hereinafter, Japanese Guidelines) provide a key framework for managing EGC. However, the evidence grades and recommendation levels for ER - particularly for undifferentiated EGC (UD-EGC) - differ across Japan, China, South Korea, and Western countries due to regional heterogeneity in EGC pathological characteristics. Recently, ER indications have expanded globally. As non-Japanese guidelines lag slightly behind, we aim to validate the Japanese Guidelines’ applicability in non-Japanese populations and discuss their adaptation to promote ER for UD-EGC outside Japan[10]. Growing evidence[11-13] confirms significant interregional variations in EGC pathology and treatment response, underscoring the need for ongoing validation and state-level customization of clinical guidance based on real-world practice.

This study aimed to evaluate the applicability of the Japanese Gastric Cancer Treatment Guidelines 2021 (6^th edition)[4] in 796 ESD and surgical specimens of EGC by assessing the concordance rate between preoperative endoscopic biopsy and final histopathological diagnosis, analyzing factors contributing to diagnostic discrepancies. We evaluated these discrepancies and their contributing factors to optimize clinical decision-making. While previous studies explored associations between tumor size, location, ulceration, and noncurative resection, few systematically evaluated concordance between preoperative indications and postoperative pathological criteria in EGC. Although Jeon et al[8] identified significant endoscopic-pathological discrepancies, their study excluded surgical candidates. To minimize noncurative resection risks while sparing low-risk patients from unnecessary surgery, we identified clinically useful predictors of indication-pathology discordance before treatment initiation. Furthermore, because Song et al[14] validated expanded indications globally via multinational cohorts, we also assessed Japanese guideline suitability in non-Japanese populations. This work sought to optimize clinical practice by refining endoscopic criteria for adoption beyond Japan.

MATERIALS AND METHODS

Data sources and collection

This retrospective study analyzed patients diagnosed with EGC at Peking University Third Hospital from January 2010 to December 2022. The cohort comprised 446 patients treated with ER and 380 patients that were surgically managed. Among 826 initially identified lesions, 28 were excluded: 13 cases of gastric adenocarcinoma of fundic gland type, 2 neuroendocrine carcinomas, and 15 cases with incomplete preoperative or postoperative pathological data. The final analysis included 796 histologically confirmed EGC lesions (Figure 1). Fundic gland type gastric adenocarcinomas and neuroendocrine carcinomas were excluded given their pathological distinction from gastric adenocarcinomas, fundamentally divergent biological behavior, and unique therapeutic principles.

Figure 1 Flowchart of the study. The cohort was comprised 446 patients treated with endoscopic submucosal dissection, and 380 were managed surgically for early gastric cancer (EGC) diagnosis from January 2010 to December 2022. Fundic gland type gastric adenocarcinoma and neuroendocrine carcinomas were excluded given their pathological distinction from gastric adenocarcinomas, fundamentally divergent biological behavior, and unique therapeutic principles. A cohort of 796 EGCs was subsequently analyzed to evaluate EGC clinical decision accuracy by comparing therapeutic indication with postoperative pathological criteria and analyzing factors influencing discrepancies. EGC: Early gastric cancer; ESD: Endoscopic submucosal dissection.

Study design and measurements

This study evaluated 796 EGC cases through preoperative examination and categorized them into two therapeutic indication groups according to the Japanese guidelines[4]: 537 lesions (67.5%) met ESD indications and 259 (32.5%) required surgical intervention. We then evaluated pathological outcomes according to the endoscopic curability (eCura) classification system. Final pathological evaluation of resected specimens further classified cases into ESD criteria (eCura A/B); and surgical criteria (eCura C-2), while expanded criteria cases (eCura C-1) received treatment based on follow-up findings. Based on the concordance between preoperative therapeutic indications and final pathological criteria, cases were stratified into three groups: Same-estimated, underestimated, and overestimated. For same-estimated cases, postoperative management aligned with preoperative assessment, underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively. Comparative analyses examined baseline characteristics, endoscopic features, and pathological outcomes across these groups.

Key definitions

Histological discrepancy: Pathologic types were classified into two types according to the World Health Organization classification of gastric cancer as epithelial tumors and carcinoid tumors[15]. For analytical purposes, tumors were categorized as differentiated (well or moderately differentiated adenocarcinoma), undifferentiated (poorly differentiated adenocarcinoma, signet ring cell carcinoma, or mixed-type carcinoma).

Comorbidity: Presence of ≥ 1 systemic diseases (hypertension, diabetes, hyperlipidemia, cardiovascular disease, cerebrovascular disease, chronic renal failure or cirrhosis).

Tumor characteristics: Macroscopic morphological classification according to the Paris endoscopic classification system[16] categorizing lesions as elevated (types 0-I and 0-IIa), flat (type 0-IIb) or depressed (types 0-IIc and 0-III) and detailed dimensional assessment incorporating tumor size measurements and ulceration status evaluation through a combination of endoscopic ultrasound (EUS) for depth analysis, white light imaging for initial characterization, and magnifying narrow-band imaging (NBI) for precise margin delineation and surface pattern analysis. And comprehensive metastasis evaluation using contrast-enhanced computed tomography (CT) combined with EUS for assessment of nodal involvement and distant metastases. The diagnostic standard for confirming EGC in this study adhered to the Vienna classification[17].

Statistical analysis

All statistical analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, United States) with rigorous methodology implemented as follows: Continuous variables were assessed for normality using Shapiro-Wilk tests. Normally distributed data are presented as mean ± SD; non-normally distributed data are expressed as median (interquartile range). Categorical variables were analyzed using χ² tests, with Fisher’s exact test applied when any expected cell frequency was ≤ 5. Data are presented as frequencies (percentages) for categorical variables. Our analytical approach proceeded in two stages: First, univariate analysis (χ² or Fisher’s exact test) identified potential risk factors using a significance threshold of P < 0.05. Subsequently, all significant variables were entered into multivariate logistic regression models to determine independent predictors. Results are reported as odds ratios (ORs) with 95% confidence intervals (CIs) and corresponding regression coefficients. Throughout all analyses, statistical significance was defined as two-tailed P < 0.05.

RESULTS

Indications and criteria

This retrospective analysis evaluated 796 patients with histologically confirmed EGC. The cohort exhibited a male predominance (70.0%, n = 557) and the age showed right-skewed distribution (Shapiro-Wilk P < 0.01) with a median 64 years (interquartile range 57-71). Treatment modalities were distributed between ER (54.0%, n = 430) and surgical intervention (46.0%, n = 366). Comparative analysis between preoperative endoscopic assessment and postoperative pathological evaluation revealed significant stratification: (1) Concordant group (same-estimated): 74.4% (n = 592); (2) Therapeutic upgrade group (underestimated lesions requiring treatment intensification): 20.4% (n = 162); and (3) Therapeutic de-escalation group (overestimated lesions eligible for less invasive management): 5.3% (n = 42). These findings demonstrate that while three-quarters of cases were accurately staged preoperatively, one-quarter exhibited clinically significant discrepancies that could substantially impact therapeutic decision-making. The substantial proportion of underestimated cases (20.4%) underscores the need for improved preoperative staging accuracy in EGC management.

Concordance analysis between preoperative therapeutic indication and pathological criteria

The cohort comprised 796 cases with the following preoperative treatment allocation: ESD indications (67.5%, 537/796) and surgical indications (32.5%, 259/796) (Tables 1 and 2). In our study of 430 endoscopically treated patients, 130 underwent non-curative resections after ESD, yielding a non-curative resection rate of 30.23%. Following ESD, the additional surgery rate was 6.15% (8/130). Subsequent management included additional ESD in 34 patients and regular follow-up in 88 patients. Among the 85 patients that did not receive any further treatment, reasons included patient preference (n = 38), advanced age (n = 7), comorbidities (n = 5), and unknown factors (n = 35). Notably, among 42 cases meeting expanded ESD criteria (eCura C-1), 81.0% (34/42) achieved curative resection through ESD, while 19.0% (8/42) required surgical conversion. Of these converted cases, 62.5% (5/8) were subsequently deemed pathologically suitable for ESD, suggesting potential overtreatment.

Table 1 Discrepancy between the pre-endoscopic submucosal dissection indications and the post-endoscopic submucosal dissection pathological criteria, n (%).

Variable	Total	Post-ESD criteria
		Absolute criteria	Expand criteria	Surgery criteria
Pre-ESD indication
Absolute indication	407	256 (62.9)	40 (9.9)	109 (26.8)
Surgery indication	25	2 (8.0)	2 (8.0)	21 (84.0)
Total	430	258 (59.7)	44 (9.8)	130 (30.1)

ESD: Endoscopic submucosal dissection.

Table 2 Discrepancy between the pre-surgery indications and the post-surgery pathological criteria, n (%).

Variable	Total	Post-surgery criteria
		ESD criteria	Surgery criteria
Pre-surgery indication
Surgery indication	234	38 (10.40)	196 (53.60)
ESD indication	132	82 (22.40)	50 (3.60)
Total	366	120 (32.80)	246 (67.20)

ESD: Endoscopic submucosal dissection.

Results of univariate analysis for diagnostic discrepancy predictors

Univariate analysis of the three diagnostic discrepancy groups (same-estimated, underestimated, and overestimated) revealed distinct patterns (Tables 3 and 4). Among baseline characteristics, neither age (P = 0.350) nor sex distribution (P = 0.847) showed significant associations. Similarly, lifestyle factors including alcohol consumption (P = 0.255) and established risk factors such as family history of gastric cancer (P = 0.109) and Helicobacter pylori infection status (P = 0.200) were not significantly different between groups. Similarly, tumor morphological features including maximum diameter (P = 0.072), macroscopic appearance (P = 0.698), and ulceration presence (P = 0.415) were not significantly different across groups.

Table 3 Comparison of baseline characteristics according to assessment discrepancy status, n (%).

Variable	Same (n = 592)	Under (n = 162)	Over (n = 42)	χ2 value	P value
Age				2.097	0.350
≤ 60	228 (38.5)	52 (32.7)	14 (33.3)
> 60	364 (61.5)	107 (67.3)	28 (66.7)
Sex				0.332	0.847
Male	411 (69.4)	116 (71.6)	30 (71.4)
Female	181 (30.6)	46 (28.4)	12 (28.6)
History of smoking				6.215	0.044
No	391 (66.0)	90 (55.6)	27 (64.3)
Yes	201 (34.0)	72 (44.4)	15 (35.7)
History of alcohol consumption				2.732	0.255
No	432 (73.0)	110 (67.9)	26 (65.0)
Yes	160 (27.0)	52 (32.1)	14 (35.0)
Family history				4.436	0.109
No	510 (86.1)	148 (91.4)	34 (81.0)
Yes	82 (13.9)	14 (8.6)	8 (19.0)
Helicobacter pylori infection				3.220	0.200
No	421 (71.1)	104 (64.2)	31 (73.8)
Yes	171 (28.9)	58 (35.8)	11 (26.2)

For same-estimated cases, postoperative management aligned with preoperative assessment, underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively. Same: Same-estimated; Under: Under-estimated; Over: Over-estimated.

Table 4 Univariate analysis of factors associated with preoperative-postoperative assessment discrepancies, n (%).

Variable	Same (n = 592)	Under (n = 162)	Over (n = 42)	χ2 value	P value
Pre-endoscopic findings
Location				8.132	< 0.001
Upper	111 (18.8)	65 (38.3)	12 (30.0)
Middle	232 (39.2)	51 (31.5)	14 (32.5)
Lower	249 (42.1)	49 (30.2)	16 (37.5)
Size (cm)				8.590	0.072
≤ 2 cm	401 (67.7)	112 (69.1)	23 (57.5)
> 2 cm and ≤ 3 cm	122 (20.6)	37 (22.8)	9 (22.5)
> 3 cm	69 (11.7)	12 (8.0)	10 (20.0)
Gross type				2.204	0.698
Elevated	282 (47.6)	74 (45.7)	16 (38.1)
Flat	220 (37.2)	65 (40.1)	20 (47.6)
Depressed	90 (15.2)	23 (14.2)	6 (14.3)
Presence of ulcers				1.761	0.415
No	423 (71.5)	117 (72.2)	34 (81.0)
Yes	169 (28.5)	45 (27.8)	8 (19.0)
Pathology				53.839	< 0.001
Differentiated	411 (69.9)	156 (96.3)	24 (57.1)
Undifferentiated	178 (30.1)	6 (3.7)	18 (42.9)

For same-estimated cases, postoperative management aligned with preoperative assessment, underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively. Same: Same-estimated; Under: Under-estimated; Over: Over-estimated.

However, three variables emerged as significantly associated with diagnostic discrepancies: Smoking history (P = 0.044), anatomical tumor location (P = 0.001), and histopathological classification (P < 0.001). These suggest that tumor location, particularly in the upper gastric third, and undifferentiated histopathology represent important determinants of assessment accuracy, while traditional demographic and morphological parameters show limited predictive value. The borderline significance of smoking history (P = 0.044) requires cautious interpretation given multiple comparisons, suggesting a need for validation in larger cohorts.

Multivariate analysis of predictive factors for diagnostic discrepancies

Multivariate logistic regression analysis revealed two independent predictors of discordance between preoperative assessments and final pathology (Tables 5 and 6): Upper gastric third location (OR = 2.158, 95%CI: 1.373-3.390, P = 0.001), with 38.3% of these lesions being significantly underestimated (P < 0.001), and undifferentiated histopathology (OR = 2.005, 95%CI: 1.036-3.879, P = 0.039), where 42.9% of cases demonstrated significant overestimation (P < 0.001). These findings underscore the distinct diagnostic challenges posed by proximal gastric lesions and poorly differentiated histopathology, necessitating improved preoperative staging techniques for upper gastric tumors and optimized biopsy protocols for undifferentiated carcinomas to guide precision treatment planning.

Table 5 Multivariate analysis of factors associated with preoperative-postoperative assessment discrepancies.

Variable		P value	Exp(B)	95% confidence interval
				Lower	Upper
Under-estimated
History of smoking	Yes	0.108	1.357	0.935	1.969
	No	Reference
Location	Upper	< 0.001	2.158	1.373	3.390
	Middle	0.536	1.151	0.738	1.794
	Lower	Reference
Pathology	Undifferentiated	< 0.001	0.104	0.045	0.241
	Differentiated	Reference
Over-estimated
History of smoking	Yes	0.871	0.945	0.483	1.846
	No	Reference
Location	Upper	0.790	1.984	0.888	4.435
	Middle	0.851	0.916	0.436	1.923
	Lower	Reference
Pathology	Undifferentiated	0.416	2.005	1.036	3.879
	Differentiated	Reference

The variable reference category is the same-estimated. For same-estimated cases, postoperative management aligned with preoperative assessment, underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively.

Table 6 Multivariate analysis of factors associated with preoperative-postoperative assessment discrepancies.

Variable		P value	Exp(B)	95% confidence interval
				Lower	Upper
Under-estimated
History of smoking	Yes	0.331	1.436	0.692	2.980
	No	Reference
Location	Upper	0.851	1.087	0.454	2.602
	Middle	0.592	1.257	0.545	2.896
	Lower	Reference
Pathology	Undifferentiated	< 0.001	0.052	0.018	0.146
	Differentiated	Reference

The variable reference category is the over-estimated. Underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively.

Concordance analysis between preoperative biopsy and final surgical pathology

In this comparative analysis of 796 EGC cases, clinically significant discrepancies emerged between preoperative biopsies and final pathology (Table 7). Critical diagnostic variances occurred in tumor diameter measurement (13.9%, 111/796), invasion depth assessment (25.3%, 201/796), and histopathological classification (17.1%, 136/796). Pathology reclassification analysis demonstrated frequent diagnostic upgrades (69.9%, 95/136). Notably, ulceration evaluation errors occurred in 4.9% (39/796) of cases, with biopsy-based overestimations constituting 84.6% (33/39) of cases.

Table 7 Comparison of preoperative biopsy and postoperative pathological findings, n (%).

Postoperative pathology	Pretreatment diagnosis	n = 796	Misdiagnosis rate (%)
Tumor diameter
≤ 3 cm	≤ 3 cm (accurate)	626 (78.6)
	> 3 cm (misdiagnosis)	33 (4.2)	14.0
> 3 cm	≤ 3 cm (misdiagnosis)	78 (9.8)
	> 3 cm (accurate)	59 (7.4)
Depth
Mucosal invasion	Mucosal invasion or submucosal invasion: < 500 μm (accurate)	577 (72.2)
	Submucosal invasion: ≥ 500 μm (misdiagnosis)	37 (4.7)	25.3
Submucosal invasion	Mucosal invasion or submucosal invasion: < 500 μm (misdiagnosis)	164 (20.6)
	Submucosal invasion: ≥ 500 μm (accurate)	20 (2.5)
Ulcerative findings
Absence	Absence (accurate)	217 (27.1)
	Presence (misdiagnosis)	33 (4.1)	4.9
Presence	Absence (misdiagnosis)	6 (0.8)
	Presence (accurate)	542 (68.0)
Main histology
Differentiated type	Differentiated type (accurate)	499 (62.7)	17.1
	Undifferentiated type (misdiagnosis)	41 (5.2)
Undifferentiated type	Differentiated type (misdiagnosis)	95 (11.9)
	Undifferentiated type (accurate)	161 (20.2)

These findings underscore three critical clinical challenges: (1) The 25.3% depth underestimation rate reflects limitations in current staging modalities; (2) The predominant histopathological upgrading (69.9%) suggests inherent sampling limitations in tumor heterogeneity; and (3) The ulceration overestimation bias (84.6%) may prompt unnecessarily aggressive interventions. Collectively, these data highlight the need for enhanced preoperative diagnostic accuracy in EGC management.

Univariate analysis of pathological discrepancies

Further analysis of pathological parameters among the three diagnostic groups (same-estimated, underestimated, and overestimated) revealed significant variations in multiple critical features (Table 8). Lesions in the underestimated group typically exhibited deeper submucosal invasion (51.9% vs 18.1%), while overestimated cases were predominantly characterized by undifferentiated histology (35.7% vs 7.4%) and more frequent ulceration (17.5% vs 3.5%). These findings underscore three critical clinical considerations: Tumor dimensional assessment and ulcer status serve as essential parameters for pretreatment evaluation accuracy, inherent histological heterogeneity represents a major contributor to diagnostic discordance, and existing staging protocols may need optimization, particularly for undifferentiated and deeply invasive subtypes.

Table 8 Univariate analysis of pathological features among three preoperative assessment groups, n (%).

Variable	Same (n = 592)	Under (n = 162)	Over (n = 42)	χ2 value	P value
Diagnosis of tumor diameter				15.067	< 0.001
Accurate diagnosis	526 (88.9)	126 (77.8)	33 (82.5)
Misdiagnosis	66 (11.1)	36 (22.2)	9 (21.4)
Diagnosis of depth				76.933	< 0.001
Accurate diagnosis	485 (81.9)	78 (48.1)	32 (76.2)
Misdiagnosis	107 (18.1)	84 (51.9)	10 (23.8)
Diagnosis of ulcerative findings				16.047	< 0.001
Accurate diagnosis	571 (96.5)	151 (93.2)	35 (82.5)
Misdiagnosis	21 (3.5)	11 (6.8)	7 (17.5)
Diagnosis of main histology
Accurate diagnosis	548 (92.6)	85 (52.5)	27 (64.3)	155.227	< 0.001
Misdiagnosis	44 (7.4)	77 (47.5)	15 (35.7)

Same: Same-estimated; Under: Under-estimated; Over: Over-estimated. For same-estimated cases, postoperative management aligned with preoperative assessment, underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively.

Multivariate analysis of pathological discrepancy predictors

The multivariate logistic regression analysis identified several significant factors associated with diagnostic discrepancies between preoperative assessment and postoperative findings. Compared with the same-estimated, tumor diameter misclassification (OR = 2.555, 95%CI: 1.515-4.309, P < 0.001), depth of infiltration underestimation (OR = 4.641, 95%CI: 3.049-7.064, P < 0.001), ulcerative findings discrepancy (OR = 3.065, 95%CI: 1.346-6.979, P = 0.008) and histological type misdiagnosis (OR = 11.220, 95%CI: 7.030-17.908, P < 0.001) were significantly common in the underestimated group. Therefore, during preoperative evaluation for ESD, tumor diameter, depth of infiltration, ulcerative findings, and main histology type should be carefully assessed (Table 9). The analysis revealed statistically significant differences in misdiagnosis rates between overestimated and accurate cases, particularly for tumor size (OR = 2.339, 95%CI: 1.044-5.242, P = 0.039), ulcerative findings (OR = 6.805, 95%CI: 2.612-17.729, P < 0.001) and main histology type (OR = 7.691, 95%CI: 3.774-15.797, P < 0.001). Consequently, special attention should be paid to the preoperative evaluation of tumor diameter, ulcerative findings, and main histology type when planning gastrectomy procedures (Table 10).

Table 9 Multivariate analysis of pathological discrepancies among three preoperative assessment groups.

Variable		P value	Exp(B)	95% confidence interval
				Lower	Upper
Under-estimated
Diagnosis of tumor diameter	Accurate diagnosis	Reference
	Misdiagnosis	< 0.001	2.555	1.515	4.309
Diagnosis of depth	Accurate diagnosis	Reference
	Misdiagnosis	< 0.001	4.641	3.049	7.064
Diagnosis of ulcerative findings	Accurate diagnosis	Reference
	Misdiagnosis	0.008	3.065	1.346	6.979
Diagnosis of main histology	Accurate diagnosis	Reference
	Misdiagnosis	< 0.001	11.220	7.030	17.908
Over-estimated
Diagnosis of tumor diameter	Accurate diagnosis	Reference
	Misdiagnosis	0.039	2.339	1.044	5.242
Diagnosis of depth	Accurate diagnosis	Reference
	Misdiagnosis	0.372	1.414	0.660	3.030
Diagnosis of ulcerative findings	Accurate diagnosis	Reference
	Misdiagnosis	< 0.001	6.805	2.612	17.729
Diagnosis of main histology	Accurate diagnosis	Reference
	Misdiagnosis	< 0.001	7.691	3.774	15.797

Table 10 Multivariate analysis of pathological discrepancies among three preoperative assessment groups.

Variable		P value	Exp(B)	95% confidence interval
				Lower	Upper
Under-estimated
Diagnosis of tumor diameter	Accurate diagnosis	Reference
	Misdiagnosis	0.838	1.092	0.469	2.544
Diagnosis of depth	Accurate diagnosis	Reference
	Misdiagnosis	0.003	3.281	1.504	7.156
Diagnosis of ulcerative findings	Accurate diagnosis	Reference
	Misdiagnosis	0.135	0.450	0.158	1.282
Diagnosis of main histology	Accurate diagnosis	Reference
	Misdiagnosis	0.307	1.459	0.707	3.010

The variable reference category is the over-estimated. Underestimated and overestimated groups represented cases requiring therapeutic escalation or de-escalation, respectively.

DISCUSSION

In this study, we observed a discrepancy rate of 25.6% between preoperative indications and postoperative pathologic criteria for EGC treatment. Specifically, 27.7% (112/405) of lesions initially deemed suitable for ESD were postoperatively classified as non-curative, necessitating additional gastrectomy. Conversely, 16.2% (42/259) of cases preoperatively assessed as gastrectomy candidates were ultimately found to be free of LNM and eligible for endoscopic indications.

Further analysis of actual treatment modalities revealed critical insights

Among ESD-treated patients, 25 had been preoperatively classified as gastrectomy indications, with 84% (21/25) confirmed as non-curative resections after ESD. Notably, 20 of these cases underwent ESD at the insistence of patients or families, despite gastrectomy with lymph node dissection, likely offering superior oncologic outcomes. This underscores the importance of thorough preoperative counseling regarding the risks of non-curative resection and potential need for additional surgery. However, in 5 high-risk patients (due to comorbidities or advanced age), ESD achieved curative resection in 4 cases, suggesting that carefully selected gastrectomy-indicated lesions may still benefit from ESD in surgically unfit individuals. Among gastrectomy-treated patients, 132 had been preoperatively assessed as ESD candidates, yet 62.1% (82/132) were ultimately found to be free of LNM and eligible for endoscopic indications. Six opted for gastrectomy due to patient preference, highlighting the need for improved clinician-patient communication regarding ESD’s advantages in organ preservation and minimally invasive outcomes. Additionally, technical challenges (11 cases) and initial surgical clinic referrals (13 cases) contributed to gastrectomy selections, emphasizing the importance of multidisciplinary collaboration in optimizing treatment strategies. Consistent with prior studies[18-20], tumors in the upper stomach and undifferentiated histology were independent risk factors for discrepancy.

Previous studies have demonstrated that the upper gastric location is a risk factor for nonradical ESD resection in EGC[21-26]. This is attributed to the anatomical complexity of the cardia and gastric fundus, where both the gastric wall and submucosa are thinner than in other regions, with small lymphocapillaries directly above the muscularis mucosae[27]. Consequently, these areas represent technically challenging sites for endoscopic procedures, leading to higher rates of deep submucosal invasion and positive margins in upper-third tumors. Thus, in our study, tumors located in the upper third were more frequently underestimated due to noncurative resection.

If upper gastric cancer is underestimated, ESD may be incorrectly selected, resulting in non-curative resection, the need for additional surgery, or progression of residual tumors to advanced stages. Besides, the thin wall of the upper gastric segment, adjacent to large blood vessels and esophagus, is easy to perforate or bleed intraoperatively. If the extent of tumor is underestimated before surgery, the difficulty of surgery increases abruptly, the rate of intermediate open abdomen rises, and the risk of anastomotic fistula increases after surgery[27].

Multiple studies have confirmed that UD-EGC is a risk factor for noncurative resection[21-26]. These tumors exhibit greater propensity for deep invasion and extensive intramural spreading. Consequently, the actual tumor size in resected specimens frequently exceeds endoscopic estimates[22,24,28]. Preoperative diagnosis of mixed-type EGC remains challenging due to the absence of standardized criteria[29]. Initial biopsies may misclassify mixed-type EGCs as purely undifferentiated[21]. When biopsy reveals UD-EGC but suspected differentiated components are endoscopically visible, clinicians should strongly consider mixed-type histology. This recognition could prevent unnecessary gastrectomy in eligible cases. Therefore, in our study, undifferentiated tumors were more frequently overestimated. As further supported by Jeon et al[8], biopsy-based undifferentiated pathology is an independent predictor of discordance between endoscopic indications for ESD and pathological criteria in clinical practice. Furthermore, some misdiagnoses may stem from inadequate biopsy targeting, either diagnostically or technically failing to sample from within the cancerous area.

Because of overestimation of tumor pathologic type, total gastrectomy may be chosen instead of distal resection, resulting in significant impairment of quality of life. Overestimation of staging may lead to overuse of adjuvant chemotherapy with increased chemotherapy toxicity and no survival benefit. Undifferentiated carcinoma that is overestimated as a deep infiltration may result in loss of opportunity for ESD, with the patient being forced to undergo a more invasive surgical procedure, repeated tests and over-intervention. Overestimation of lesion extent may lead to multiple biopsies, repeated imaging, and even secondary surgery, resulting in increased diagnostic costs for patients.

We suggest that for upper gastric lesions, mandatory EUS combined with magnifying NBI should be used for in-depth evaluation, and biopsy must follow the multisite sampling scheme of “center-margin-junction”, supplemented with immunohistochemistry to confirm the differentiation status when indolent cell carcinoma is suspected. Advanced imaging techniques enable real-time histological analysis of living tissue during routine endoscopy and can reduce the number of biopsies needed for correct diagnosis, thereby minimizing the risk of sampling error[30]. For controversial cases (e.g., technical difficulties in the upper third of gastric tumors or unclear borders of undifferentiated carcinoma), multidisciplinary collaboration help avoid decision-making bias to ultimately achieve precise individualized treatment.

In this study, the depth of infiltration and pathologic types were the most common reasons for the differences. However, pre-treatment screening assessments had a limited role in predicting tumor infiltration depth. Depth of infiltration is highly correlated with LNM, and routine assessment using EUS did not improve the accuracy of predicting depth of invasion compared to endoscopy[31]. Diagnosis can often only be confirmed by pathological evaluation of the resected specimen. In predicting tumor infiltration depth, the overall accuracy by conventional endoscopy was 73%-82%, with endoscopy accurately detecting mucosal carcinomas and EUS effectively salvaging endoscopically overrated lesions. Previous studies have shown that the combined use of endoscopy and EUS results in a combined accuracy that can be greater than 85%[31,32]. Regarding LNM assessment, preoperative evaluation by contrast-enhanced CT alone identified 76 patients with suspected lymph node involvement in this study. In contrast, EUS combined with contrast-enhanced CT identified 53 patients with suspected involvement. According to postoperative pathology criteria, only 39 cases were ultimately confirmed to have LNM. Thus, the EUS + CT combination achieved 90.45% diagnostic accuracy (vs 73.59% for CT alone). Notably, in this non-blinded clinical series, some enlarged lymph nodes initially suspected of regional metastasis were pathologically confirmed as reactive lymphoid hyperplasia. Previous studies have demonstrated that EUS has superior sensitivity to CT for detecting regional LNM[33]. Therefore, combining EUS with CT reduces regional LNM rates, a direction warranting further exploration. These limitations in preoperative evaluation are the main reason for the discrepancy between pre-EGC indications and postoperative pathologic criteria. If possible, further studies are needed on more accurate methods to predict the presence of LNM.

This study benefits from a large cohort of ESD and gastrectomy-treated EGCs, with comprehensive analysis of indication-pathology discrepancies and associated risk factors - key data for refining therapeutic decision-making. However, this study was a single-center retrospective study, and it is indeed possible that there was selection bias. While we acknowledge that regional variations in medical practice may introduce potential selection bias, our findings remain applicable to EGC patients receiving care in tertiary hospitals within China. Further studies are needed to validate our preliminary findings. Beyond this, limitations include potential selection bias, variability in preoperative assessments (e.g., inconsistent use of NBI/EUS), and inter-endoscopist differences in lesion evaluation.

CONCLUSION

While preoperative endoscopy correctly identified EGC indications in 74.4% of cases, discrepancies were strongly associated with upper-third tumors and undifferentiated histology. To optimize outcomes, we recommend heightened vigilance for upper-third lesions, including preoperative counseling about possible non-curative ESD, and second-line assessment of undifferentiated tumors to avoid unnecessary gastrectomy when ESD may suffice.

ACKNOWLEDGEMENTS

We sincerely thank Yan-Yan Shi for invaluable guidance and critical review of the manuscript. We are grateful to Ying-Tong Chen for guiding us in manuscript writing. Special thanks to Jing-Jin Tan, Ji-Li Wang, Wei-Yang Shu, Xin Huang and my family for their unwavering support.