Chen T, Feng AQ, Zhang WF, Dang NY, Ye F, Zhang GX. Narrow-band imaging-modified Kyoto score predicts pathological upgrading after endoscopic submucosal dissection in gastric intraepithelial neoplasia. World J Gastroenterol 2026; 32(45): 123704 [DOI: 10.3748/wjg.123704]
Corresponding Author of This Article
Guo-Xin Zhang, PhD, Department of Gastroenterology, The First Affiliated Hospital of Nanjing Medical University, No. 300 Guangzhou Road, Nanjing 210029, Jiangsu Province, China. nanjingmedical@163.com
Research Domain of This Article
Gastroenterology & Hepatology
Article-Type of This Article
research-article
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Tao Chen, An-Qiang Feng, Department of Gastroenterology, Xuzhou Central Hospital Affiliated to Southeast University, Xuzhou 221009, Jiangsu Province, China
Wei-Feng Zhang, Ni-Yi Dang, Feng Ye, Guo-Xin Zhang, Department of Gastroenterology, The First Affiliated Hospital of Nanjing Medical University, Nanjing 210029, Jiangsu Province, China
Author contributions: Chen T and Zhang GX designed the research study; Chen T, Feng AQ, Zhang WF, Dang NY, and Ye F performed the research and collected the data; Chen T and Feng AQ analyzed the data and they contributed equally to this manuscript as co-first authors; Chen T drafted the manuscript; Zhang GX supervised the study. All authors have read and approved the manuscript.
AI contribution statement: ChatGPT was used during manuscript revision for language refinement, formatting assistance, structural organization, and preparation of responses to reviewers. It was not used for study design, data generation, or statistical analysis. All AI-assisted content was reviewed and verified by the authors, who take full responsibility for the accuracy, originality, and integrity of the manuscript.
Supported by the National Natural Science Foundation of China, No. 82470593; and General Project of the Development Fund of the Affiliated Hospital of Xuzhou Medical University, No. XYFM202334.
Institutional review board statement: Ethical approval for the study protocol was granted by the Institutional Review Board of Xuzhou Central Hospital Affiliated to Southeast University (approval No. XZXY-LK-20191223-0156).
Clinical trial registration statement: This prospective observational cohort study did not involve assignment of participants to an intervention and was not registered as an interventional clinical trial.
Informed consent statement: All study participants, or their legal guardian, provided informed written consent prior to study enrollment.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
CONSORT 2010 statement: The authors have read the CONSORT 2010 Statement, and the manuscript was prepared and revised according to the CONSORT 2010 Statement.
Data sharing statement: The data supporting the findings of this study are available from the corresponding author upon reasonable request.
Corresponding author: Guo-Xin Zhang, PhD, Department of Gastroenterology, The First Affiliated Hospital of Nanjing Medical University, No. 300 Guangzhou Road, Nanjing 210029, Jiangsu Province, China. nanjingmedical@163.com
Received: May 29, 2026 Revised: July 10, 2026 Accepted: July 28, 2026 Published online: December 7, 2026 Processing time: 142 Days and 2.3 Hours
Abstract
BACKGROUND
Pathological upgrading after endoscopic submucosal dissection (ESD) remains clinically challenging in patients with gastric intraepithelial neoplasia (GIN). Existing prediction approaches mainly focus on lesion-specific features, whereas background gastric mucosal status, particularly intestinal metaplasia, is less systematically assessed. Conventional white-light endoscopy has limited accuracy and interobserver agreement in evaluating background gastric mucosal status, particularly intestinal metaplasia.
AIM
To evaluate the predictive performance of a narrow-band imaging (NBI)-modified Kyoto Gastritis Score for pathological upgrading in GIN.
METHODS
This prospective observational cohort study consecutively enrolled 225 patients with GIN undergoing ESD from 2020 to 2024. All patients underwent white-light and NBI endoscopic examinations. Conventional and NBI-modified Kyoto Gastritis Scores were assessed, with post-resection pathology as the reference standard. Receiver operating characteristic curve analysis was performed to evaluate and compare the predictive performance of the two scores. The area under the curve, sensitivity, specificity, and negative predictive value were calculated.
RESULTS
Compared with the conventional white-light Kyoto Gastritis Score, the NBI-modified score achieved a significantly higher area under the curve of 0.862 vs 0.769 (P = 0.002). The modified system improved overall diagnostic accuracy from 74.67% to 80.44%, with sensitivity increasing from 62.0% to 76.1%. At the cutoff score of ≥ 4, the positive and negative predictive values were 76.47% and 83.22%, respectively. Multivariate analysis incorporating Helicobacter pylori (H. pylori) infection status and other factors revealed that the NBI-modified Kyoto Gastritis Score was an independent risk factor for pathological progression after ESD. The predicted performance of the NBI-modified Kyoto Gastritis Score did not differ significantly between patients with current H. pylori infection and those with past/never infection.
CONCLUSION
In this single-center cohort, the NBI-modified Kyoto Gastritis Score improved prediction of pathological upgrading after endoscopic submucosal dissection and may support preoperative risk stratification, pending external validation.
Core Tip: This prospective study evaluated a narrow-band imaging-magnifying endoscopy-modified Kyoto Gastritis Score for predicting pathological upgrading after endoscopic submucosal dissection in gastric intraepithelial neoplasia. The modified score replaces white-light assessment of intestinal metaplasia with narrow-band imaging-magnifying endoscopy features, improving discrimination over the conventional score (area under the curve: 0.862 vs 0.769) and remaining independently associated with upgrading. A cutoff of ≥ 4 provided a negative predictive value of 83.22%, supporting low-risk identification and surveillance planning across Helicobacter pylori infection subgroups, although multicenter validation remains necessary.
Citation: Chen T, Feng AQ, Zhang WF, Dang NY, Ye F, Zhang GX. Narrow-band imaging-modified Kyoto score predicts pathological upgrading after endoscopic submucosal dissection in gastric intraepithelial neoplasia. World J Gastroenterol 2026; 32(45): 123704
GLOBOCAN 2022 estimated 968350 new cases of gastric cancer and 659853 related deaths worldwide, placing the disease fifth for both cancer incidence and cancer mortality[1]. Early diagnosis with timely treatment is crucial for improving patient survival[2]. For early gastric cancer (EGC) and gastric precancerous lesions, endoscopic submucosal dissection (ESD) is now a standard therapeutic approach that enables en bloc removal while preserving gastric integrity[3]. With the expanding use of ESD, discordance between preoperative biopsy findings and the definitive histopathology of resected specimens has become an increasingly relevant clinical concern. Li et al[4] observed pathological upgrading in up to 40.9% of patients with gastric intraepithelial neoplasia (GIN) undergoing ESD. In another recent retrospective study, Ding et al[5] reported an overall upgrading rate of 59.4%, with rates of 51.2% among patients with low-grade intraepithelial neoplasia (LGIN) and 80.2% among those with high-grade intraepithelial neoplasia (HGIN). Such pathological upgrading may lead to inadequate treatment, additional surgical intervention, and inaccurate prognostic assessment. Accordingly, stratifying the likelihood of pathological upgrading before ESD is clinically important.
Current research on the prediction of pathological progression after ESD primarily focuses on local lesion characteristics. Multiple studies have confirmed that lesion size > 2 cm is the strongest and most consistent independent risk factor[4-10]. Local morphological features such as surface erythema[4,9-12], surface nodules[4,6,7], surface ulcers/erosions[9,12], and Paris classification[5] have also been widely reported. Yang et al[13] further showed that a positive narrow-band imaging magnifying endoscopy (NBI-ME) assessment independently predicted histological upgrading. Xiao et al[14] showed that features including demarcation line, microvascular pattern irregularity, and white opaque substance (WOS) under NBI-ME were significant risk factors. Using magnifying endoscopy, Lyu et al[15] identified irregular or absent microvascular and microsurface patterns, light blue crest (LBC), intestinal metaplasia (IM), and atrophy as major predictors of pathological progression. These studies suggest that while local lesion characteristics are critical, the status of the background gastric mucosa, particularly atrophy and IM, serves as the precancerous basis for gastric cancer. However, its potential as a predictive marker for pathological upgrading after ESD has rarely been systematically explored. The Kyoto Gastritis Score is an internationally validated endoscopic scoring system based on five endoscopic findings: Atrophy, IM, enlarged folds, nodularity, and diffuse redness, with a total score ranging from 0 to 8. Originally, developed to evaluate Helicobacter pylori (H. pylori) infection status, the Kyoto score has subsequently been validated for predicting gastric cancer risk[16]. Xu et al[17] recently examined the white-light Kyoto score in the context of post-ESD pathological upgrading and reported an area under the curve (AUC) of 0.667. Although this work established the Kyoto score as a quantitative risk stratification tool for pathological upgrading after ESD, the assessment was based on conventional white-light endoscopy which has well-recognized limitations in evaluating IM[17,18]. IM represents a late, largely irreversible stage in the Correa cascade of gastric carcinogenesis, yet its detection under white-light endoscopy suffers from limited accuracy and suboptimal interobserver agreement. The Endoscopic Grading of Gastric IM (EGGIM) system provides a validated endoscopic framework for grading the extent of IM, and its extent-based approach offers an important methodological reference for endoscopic IM assessment. However, EGGIM focuses specifically on IM, whereas the Kyoto classification captures a broader background mucosal phenotype by incorporating atrophy, IM, diffuse redness, enlarged folds, and nodularity[19]. Kim[20] argued that this grading system cannot accurately reflect the gastric cancer risk in patients with previous H. pylori infection because H. pylori eradication is followed by disappearance of enlarged gastric folds, nodularity, and diffuse redness. Dohi et al[21] reviewed that magnifying endoscopy with NBI-ME significantly improves the diagnostic accuracy and reproducibility of H. pylori-related precancerous conditions, including IM.
Based on the above background, we hypothesized that incorporating precise IM assessment via NBI-ME would overcome the limitations of white-light endoscopy while preserving the multidimensional structure of the Kyoto classification. Therefore, this prospective cohort study was designed to compare the predictive performance of the narrow-band imaging (NBI)-modified Kyoto score with that of the conventional white-light score, assess the score’s performance across patients with different H. pylori infection statuses, and evaluate its predictive value for pathological upgrading after ESD. This study aimed to refine the Kyoto score for preoperative risk stratification of post-ESD pathological upgrading using a practical NBI-based IM assessment.
MATERIALS AND METHODS
Study design and patients
This prospective observational cohort study was conducted at Xuzhou Central Hospital Affiliated to Southeast University from January 2020 to December 2024. The study ultimately included 225 patients. Inclusion criteria were as follows: (1) Age ≥ 18 years; (2) Have low- or high-grade GIN confirmed by preoperative endoscopic forceps biopsy; and (3) Underwent ESD at our institution. Exclusion criteria were as follows: (1) More than 3 months had elapsed between endoscopic biopsy and ESD; (2) History of gastric surgery; (3) Advanced gastric cancer; (4) Concurrent malignancies; and (5) Inadequate endoscopic image quality for evaluation. Baseline clinical and endoscopic characteristics, including sex, age, body mass index, H. pylori infection status, lesion size, lesion location, morphology, gastric mucosal nodules status, surface redness, and NBI observations, were collected for all enrolled patients. The sample size was determined a priori to compare the AUC between the NBI-modified and conventional Kyoto scores. Assuming a conservative pathological upgrading rate of 50%, an expected AUC of 0.65 for the conventional score (based on recent literature), and an anticipated AUC of 0.85 for the NBI-modified score, we required approximately 200 patients to detect a difference of 0.20 with an α of 0.05 and 80% power. Considering a 10% potential dropout rate, the target enrollment was set at 220 patients. Ultimately, 225 patients were enrolled and included in the final analysis. The study protocol received approval from the Institutional Review Board of Xuzhou Central Hospital Affiliated to Southeast University (approval No. XZXY-LK-20191223-0156). All participants provided written informed consent before enrollment, and the study was conducted in accordance with the Declaration of Helsinki.
Endoscopic procedures and image evaluation
All endoscopic examinations were performed using white-light followed by magnifying endoscopy with NBI-ME (GIF-H290Z, Olympus, Tokyo, Japan). Prior to the procedure, patients received standard preparation including 8 hours of fasting and oral administration of a defoaming agent (simethicone, 30 mL) and mucolytic agent (streptoproteinase granules, 20000 U) 30 minutes before gastroscopy, ensuring clear visualization of the gastric mucosa. Propofol (2 mg/kg) and fentanyl (2 μg/kg) were administered 5 minutes prior to the examination to effectively suppress visceral traction response and body movement, ensuring optimal endoscopic inspection by the physician. All procedures were recorded and stored in the institutional picture archiving and communication system. Two experienced endoscopists with over ten years of experience and more than 5000 endoscopic procedures, blinded to clinical and pathological information, independently reviewed all endoscopic images. In cases of disagreement, a third senior endoscopist was consulted to reach a consensus. Interobserver agreement for NBI feature assessment between the two endoscopists was evaluated using weighted kappa statistics. Kappa coefficients were categorized as poor (< 0.40), moderate (0.40-0.60), good (0.61-0.80), or excellent (0.81-1.00).
Conventional and NBI-modified Kyoto classification score
The conventional Kyoto classification score was assessed under white-light endoscopy according to the original criteria. The assessment comprised five components: Atrophy, IM, diffuse redness, enlarged folds, and nodularity. After white-light examination, NBI-ME was performed to evaluate IM defined by the identification of LBC, WOS, and tubulovillous or villous microsurface patterns according to the Sakaki classification. The modified Kyoto score used LBC, WOS, and microsurface patterns collectively to identify IM and graded IM solely according to extent (focal ≤ 30% vs extensive > 30%), with the original Kyoto classification framework (0-2). This 30% threshold for defining focal vs extensive IM follows the convention established by the EGGIM system[19]. The other four components (atrophy, diffuse redness, enlarged folds and nodularity) were scored according to the original criteria. Diffuse redness was categorized according to the visibility of the regular arrangement of collecting venules (RAC): A score of 0 indicated no diffuse redness with clearly visible RAC; 1 indicated mild redness with partially visible or blurred RAC; and 2 indicated severe redness with absent or invisible RAC. Thus, the total score of the NBI-modified Kyoto score ranged from 0 point to 8 points, identical to the original scoring system, ensuring direct comparability and clinical familiarity. The two scoring criteria are summarized in Table 1.
Table 1 Components of both the conventional and narrow-band imaging-modified Kyoto classification scores.
Component
Conventional score
NBI-modified score
Scoring range
Atrophy
0 = none or closed-type (C1); 1 = closed-type (C2 or C3); 2 = open-type (O1-O3)
Same as the conventional Kyoto score
0-2
Intestinal metaplasia
0 = none; 1 = confined to the antrum; 2 = extending to the corpus
IM identified by LBC, WOS, and/or tubulovillous or villous microsurface patterns under NBI-ME: 0 = none; 1 = focal involvement (≤ 30%); 2 = extensive involvement (> 30%)
0-2
Diffuse redness
0 = absent diffuse redness with clearly visible RAC; 1 = mild diffuse redness with partially visible or blurred RAC; 2 = severe diffuse redness with absent or invisible RAC
H. pylori status was classified using three diagnostic sources: (1) A 13C-urea breath test with a cutoff of 2.5‰; (2) Rapid urease test on biopsy specimens; and (3) Histopathological examination of gastric biopsies with hematoxylin and eosin or Giemsa staining. Patients were classified as having current infection when at least one test was positive and there was no previous history of eradication therapy. Past infection was defined as documented history of successful eradication therapy with confirmed negative post-eradication tests. Never infected was defined as negative results on all tests with no history of eradication. Because the never-infected subgroup was small, patients with past and never infection were combined as the past/never infection group for analysis.
Histopathological evaluation
Endoscopic biopsy samples and ESD resection specimens were placed in 10% formalin, cut at 2-mm intervals, and processed for paraffin embedding. Serial sections were prepared and stained with hematoxylin and eosin for histological assessment. Two gastrointestinal pathologists, each with more than 10 years of experience, independently examined the specimens without knowledge of the endoscopic findings or clinical information. Diagnoses were assigned according to the World Health Organization classification of gastric tumors, and discrepant interpretations were resolved through joint review and consensus. Pathological upgrading was defined as a higher histological grade in the ESD specimen than in the corresponding preoperative biopsy. This included progression from LGIN to HGIN or EGC, as well as progression from HGIN to EGC.
Statistical analysis
Statistical analyses were conducted with SPSS version 27.0 (IBM Corp., Armonk, NY, United States) and MedCalc version 20.0 (MedCalc Software, Ostend, Belgium). Distributional normality of continuous variables was examined using the Shapiro-Wilk test, after which continuous data were summarized as mean ± SD. Categorical data were reported as n (%). Between-group comparisons were performed using the independent-samples t test for continuous variables and the χ2 test or Fisher exact test, as appropriate, for categorical variables.
Univariable and multivariable logistic regression analyses were performed to identify factors associated with pathological upgrading. Variables with clinical relevance and/or statistical significance in univariable analysis were entered into the multivariable model using the Enter method. The conventional Kyoto score was not included in the same multivariable model as the NBI-modified Kyoto score because the two scores share several components and would introduce conceptual redundancy.
Discrimination of the conventional and NBI-modified Kyoto scores was examined by receiver operating characteristic curve analysis and summarized using the AUC with 95% confidence intervals (CIs). The cutoff maximizing the Youden index was selected as the optimal threshold. Sensitivity, specificity, positive predictive value, and negative predictive value (NPV) were then calculated at this threshold. The AUCs of the two scoring systems were compared using the DeLong test.
A hierarchical sensitivity analysis was performed to assess the incremental predictive value of the NBI-modified Kyoto score beyond conventional clinical and lesion-specific factors. Variables other than the Kyoto scores that were significantly associated with pathological upgrading in univariable analysis were entered into the base model. The NBI-modified Kyoto score was then added to form the extended model. Multicollinearity was assessed using tolerance and variance inflation factor, with tolerance > 0.20 and variance inflation factor < 5 considered acceptable. A two-sided P < 0.05 was considered statistically significant.
RESULTS
Baseline characteristics of patients stratified by pathological upgrading status
The final study cohort comprised 225 patients (Figure 1), of whom 162 had LGIN and 63 had HGIN on preoperative endoscopic biopsy. The pathological results of endoscopic biopsy showed LGIN in 162 cases and HGIN in 63 cases. Post-ESD pathological upgrading occurred in 92 patients (40.88%), including 43 cases (19.11%) of LGIN to HGIN, 18 cases (8%) of LGIN to EGC, and 31 cases (13.78%) of HGIN to EGC. Table 2 presents the baseline characteristics according to pathological upgrading status. Compared with patients without upgrading, those with upgrading more frequently had lesions ≥ 20 mm, Paris type IIc morphology, surface redness, and current H. pylori infection (all P < 0.05). Both the conventional Kyoto score and the NBI-modified Kyoto score were significantly higher in the upgrading group compared to the non-upgrading group (P < 0.001). The IM subscore showed a marked difference between groups (P < 0.001), suggesting a difference in IM extent between patients with and without pathological upgrading.
Univariable and multivariable logistic regression analysis for pathological upgrading
In univariable analysis, lesion size ≥ 20 mm, Paris type IIc morphology, H. pylori status, lesion surface redness, and the conventional and NBI-modified Kyoto score were significantly associated with upgrading (P < 0.05), consistent with the baseline comparison in Table 2. However, in multivariable analysis, lesion size ≥ 20 mm, Paris type IIc morphology, H. pylori status, surface redness, were no longer significant, while the NBI-modified score remained an independent predictor [odds ratio (OR) = 14.781, 95%CI: 6.621-32.997, P < 0.001]. This suggests that the effect of other factors on pathological upgrading is largely captured by the NBI-modified score (Table 3). Collinearity diagnostics showed that all variables included in the final multivariable model had tolerance values > 0.20 and variance inflation factor values < 5, indicating no substantial multicollinearity (Supplementary Table 1).
Table 3 Univariable and multivariable logistic regression analysis for pathological upgrading.
In the hierarchical sensitivity analysis, model 1 included lesion size ≥ 20 mm, Paris type IIc morphology, lesion surface redness, and current H. pylori infection, with an AUC of 0.669 (95%CI: 0.597-0.714). After the NBI-modified Kyoto score ≥ 4 was added in model 2, the AUC increased to 0.870 (95%CI: 0.821-0.920), and the NBI-modified Kyoto score ≥ 4 remained associated with pathological upgrading (OR = 9.612, 95%CI: 5.108-18.084, P < 0.001) (Supplementary Table 2).
Predictive performance of the conventional and NBI-modified Kyoto score
The predictive performance of the conventional and NBI-modified Kyoto scores is compared in Table 4 and Figure 2. The NBI-modified score showed greater discrimination than the conventional score, with AUCs of 0.862 and 0.769, respectively (P = 0.002 by Delong test). Using ≥ 4 as the optimal threshold, the modified score had a sensitivity of 76.1%, specificity of 83.5%, positive predictive value of 76.47%, and NPV of 83.22%.
Figure 2 Receiver operating characteristic curves of the two scoring systems for predicting pathological upgrading.
A: Receiver operating characteristic curve of the narrow-band imaging-modified Kyoto score; B: Receiver operating characteristic curve of the conventional Kyoto score.
Table 4 Predictive performance of the conventional and narrow-band imaging-modified Kyoto scores, % (95% confidence interval).
Efficiency of NBI-modified score in different H. pylori infection status
The performance of the NBI-modified score was further examined separately in patients with current H. pylori infection and those classified as having past/never infection (Table 5 and Figure 3). The AUC was 0.837 in patients with current infection and 0.878 in those with past/never infection. AUCs were compared between subgroups using the Delong test. The NBI-modified score demonstrated no significant difference in predictive performance across H. pylori status subgroups (P = 0.213).
In this prospective study, we evaluated the predictive performance of a novel NBI-modified Kyoto Gastritis Score for pathological upgrading after ESD in patients with GIN. Our primary findings demonstrated that the NBI-modified score significantly outperformed the conventional white-light Kyoto score in predicting pathological progression (AUC: 0.862 vs 0.769, P = 0.002). Furthermore, multivariable analysis confirmed that the NBI-modified score was independently associated with pathological upgrading. Notably, applying an optimal cutoff score of ≥ 4 yielded a high NPV of 83.22%, suggesting potential value for identifying patients at relatively lower risk of pathological upgrading in this cohort. Importantly, this predictive efficacy showed no significant difference across H. pylori infection status subgroups.
Pathological upgrading after endoscopic resection represents a significant clinical challenge that impacts treatment outcomes and prognosis. While previous studies have utilized the conventional Kyoto classification for long-term gastric cancer risk stratification[22,23], and Xu et al[17] first explored the relationship between Kyoto score and pathological upgrading after ESD. However, their model relied on white-light assessment of IM, a component with recognized diagnostic limitations. The EGGIM system is a well-validated endoscopic tool for assessing the extent of IM, and its extent-based approach provides an important methodological reference for endoscopic IM grading. However, EGGIM focuses specifically on IM, whereas the Kyoto classification offers a broader assessment of the gastric mucosal background by incorporating atrophy, IM, diffuse redness, enlarged folds, and nodularity. Therefore, we used the EGGIM-derived extent threshold to refine the IM component while preserving the multidimensional Kyoto framework. In this context, the present study evaluated an NBI-modified Kyoto classification score using a simplified NBI-based IM assessment protocol. Compared with the white-light Kyoto score model reported by Xu et al[17], the NBI-modified score showed numerically higher predictive performance, with a higher AUC, sensitivity, and specificity. In the present cohort, the NBI-modified score also outperformed the conventional white-light Kyoto score (AUC: 0.862 vs 0.769, P = 0.002). Additionally, the modified score achieved a high NPV of 83.22%, which may help identify patients with a lower probability of pathological upgrading. Moreover, the predictive performance of the NBI-modified score was not significantly different across subgroups with different H. pylori infection statuses (P = 0.213), suggesting similar predictive performance across H. pylori status subgroups in this cohort. While H. pylori infection was associated with upgrading in the univariable analysis (P < 0.05), this association was attenuated and became non-significant after the inclusion of the NBI-modified score in the multivariable model. This finding suggests that the effect of H. pylori-related mucosal injury may be partly reflected by background mucosal changes, particularly IM, captured by the NBI-modified score. Once IM is established, it serves as a durable marker of gastric cancer risk regardless of current H. pylori status, consistent with previous reports[24,25]. These findings support the rationale for integrating NBI-based IM assessment into the Kyoto classification.
Consistent with existing literature, our univariable analysis identified lesion size ≥ 20 mm, Paris type IIc morphology, and surface redness as significant predictors of pathological upgrading. However, in the multivariable analysis, these classical lesion-specific features lost their statistical significance, whereas the NBI-modified score emerged as the sole independent predictor (OR = 14.781, 95%CI: 6.621-32.997, P < 0.001). To address the possibility of overlap between the score and conventional lesion-specific predictors, we performed a hierarchical sensitivity analysis. After lesion size, Paris type IIc morphology, surface redness, and H. pylori status were included in the base model, addition of the NBI-modified Kyoto score increased the AUC from 0.669 to 0.870, and the NBI-modified score remained associated with pathological upgrading. Collinearity diagnostics also showed no substantial multicollinearity among variables in the final model. These results support the robustness of the association between the NBI-modified score and pathological upgrading. Nevertheless, because the score incorporates background mucosal features that may be biologically related to some lesion-specific findings, the attenuation of other variables should be interpreted cautiously rather than as evidence that local lesion features are unimportant. These two categories of risk factors are not mutually exclusive but rather reflect distinct aspects of gastric carcinogenesis. In this ESD-selected cohort, the NBI-modified background mucosal score appeared to provide additional discriminatory information beyond conventional lesion-specific factors.
Our study advances this knowledge in several important ways. First, we achieved an improvement in predictive performance of the Kyoto Gastritis Score. This enhancement can be attributed to the superior diagnostic accuracy of NBI-ME for IM which was the Kyoto Gastritis Score component with the lowest interobserver agreement under white-light endoscopy[26-29]. The LBC, WOS, and microsurface patterns used in our protocol have demonstrated high sensitivity for IM detection, enabling more precise identification and extent grading of IM than white-light endoscopy alone. Moreover, we prioritized clinical practicality. Our scoring protocol used multiple NBI features collectively to identify IM and graded it solely by extent, thereby preserving the original Kyoto score’s 0-8 framework, leveraging NBI technological advantages. Endoscopists need only switch to NBI mode during routine preoperative endoscopy examination, assess the presence and extent of IM based on well-established signs, and apply the same 0-2 scale familiar from the conventional Kyoto classification. No additional equipment or complex calculations are needed. This simple scoring approach may facilitate clinical translation after further validation. Second, the consistent predictive performance of the NBI-modified score across H. pylori status subgroups has important clinical implications. The NBI features including LBC, WOS, and tubulovillous/villous patterns incorporated in our score are all direct correlates of histologically confirmed IM, a late and largely irreversible alteration in the Correa cascade of gastric cancer[30-33]. Once established, IM does not regress after H. pylori eradication, unlike reversible inflammatory features such as diffuse redness, enlarged folds, or nodularity[34,35]. Toyoshima et al[36] prospectively demonstrated that the total Kyoto scores significantly decreased after successful eradication (from 3.90 to 2.78, P < 0.001), and this reduction was driven entirely by improvements in reversible components, while atrophy and IM scores remained unchanged. Takayama et al[37] further reported that patients with map-like redness or absence of RAC, both markers of irreversible mucosal damage, remained at significantly increased risk for metachronous gastric cancer even after eradication (hazard ratio: 2.561 and 3.131, respectively). Our score therefore may capture cumulative mucosal damage related to past or present H. pylori infection rather than transient inflammatory activity alone. This may explain its similar performance across H. pylori status subgroups.
From a clinical perspective, the NBI-modified Kyoto score may provide supportive information for preoperative risk stratification. At the cutoff of ≥ 4 points, the high NPV observed in this cohort suggests that patients with lower scores may have a relatively lower probability of pathological upgrading. However, treatment selection, ESD planning, and postoperative surveillance should not be based on this score alone until further validation is available. For patients with higher scores, the score may remind clinicians to maintain heightened awareness of possible pathological upgrading, even when the initial biopsy indicates low-grade dysplasia.
This study has several limitations. First, the single-center design and moderate sample size (n = 225), while adequate for the primary analysis, limit generalizability, and multicenter external validation is warranted. Second, although a hierarchical sensitivity analysis was added, this study did not include internal or external validation of the prediction model. Therefore, the clinical use of the score should be considered exploratory at this stage. Third, the limited number of events within each upgrading subtype precluded subtype-specific receiver operating characteristic analyses. Larger studies are needed to determine whether the score performs consistently across different upgrading patterns. Fourth, subjectivity in NBI feature assessment cannot be completely eliminated. Although we employed independent evaluations by two endoscopists with good interobserver agreement (weighted kappa 0.76), future studies may explore artificial intelligence-assisted image analysis to enhance objectivity and reproducibility[38].
CONCLUSION
This study demonstrates that an NBI-modified Kyoto score incorporating precise IM assessment showed improved predictive performance for pathological upgrading after ESD for GIN compared with the conventional Kyoto score. The simplified scoring protocol retains diagnostic accuracy while maintaining clinical practicality. The score’s similar predictive performance across different H. pylori status subgroups suggests that it may be applicable across patients with different infection backgrounds. However, given the single-center design and lack of model validation, the NBI-modified Kyoto score should be considered a supportive tool for preoperative risk stratification rather than a definitive basis for clinical decision-making. Further studies, including internal and external validation in multicenter cohorts, are warranted.
Aumpan N, Vilaichone RK, Nunanan P, Chonprasertsuk S, Siramolpiwat S, Bhanthumkomol P, Pornthisarn B, Uchida T, Vilaichone V, Wongcha-Um A, Yamaoka Y, Mahachai V. Predictors for development of complete and incomplete intestinal metaplasia (IM) associated with H. pylori infection: A large-scale study from low prevalence area of gastric cancer (IM-HP trial).PLoS One. 2020;15:e0239434.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 3][Cited by in RCA: 16][Article Influence: 2.7][Reference Citation Analysis (5)]
Kawamura M, Uedo N, Koike T, Kanesaka T, Hatta W, Ogata Y, Oikawa T, Iwai W, Yokosawa S, Honda J, Asonuma S, Okata H, Ohyauchi M, Ito H, Abe Y, Ara N, Kayaba S, Shinkai H, Shimokawa T. Kyoto classification risk scoring system and endoscopic grading of gastric intestinal metaplasia for gastric cancer: Multicenter observation study in Japan.Dig Endosc. 2022;34:508-516.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 58][Cited by in RCA: 60][Article Influence: 15.0][Reference Citation Analysis (26)]
Kawamura M, Uedo N, Yao K, Koike T, Kanesaka T, Hatta W, Ogata Y, Iwai W, Yokosawa S, Honda J, Asonuma S, Okata H, Ohyauchi M, Ito H, Abe Y, Ara N, Kayaba S, Shinkai H, Kanemitsu T. Endoscopic and histological risk stratification for gastric cancer using gastric intestinal metaplasia.J Gastroenterol Hepatol. 2024;39:1910-1916.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 11][Reference Citation Analysis (0)]
Otani K, Watanabe T, Kosaka S, Matsumoto Y, Nakata A, Nadatani Y, Fukunaga S, Hosomi S, Tanaka F, Kamata N, Taira K, Nagami Y, Tanigawa T, Kimura T, Fukumoto S, Kawada N, Fujiwara Y. Utility of Kyoto Classification of Gastritis in subjects with a high-negative titer of anti-Helicobacter pylori antibody during a medical check-up.J Clin Biochem Nutr. 2020;67:317-322.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 3][Cited by in RCA: 6][Article Influence: 1.0][Reference Citation Analysis (0)]
Takayama S, Dohi O, Horie R, Yasuda T, Ochiai T, Iwai N, Imamoto E, Takagi T, Handa O, Konishi H, Ando T, Naito Y, Takemura T, Itoh Y. Kyoto Classification-Based Predictive Factors Associated with the Development of Gastric Cancer After Helicobacter pylori Eradication: A Prospective Multicenter Observational Study.Diagnostics (Basel). 2025;15:2376.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in RCA: 1][Reference Citation Analysis (14)]
