Abstract
Background
Early detection of gastric cancer (GC) is crucial for improving prognosis. This study aimed to address the paucity of longitudinal evidence by validating and comparing the long-term predictive efficacy of the serology-based Li’s scoring system and the endoscopic Kyoto classification for GC risk stratification in a Chinese population.
Methods
This retrospective, single-center cohort study included 5801 patients who underwent esophagogastroduodenoscopy between April 2017 and December 2023. We conducted a cross-sectional analysis to evaluate the diagnostic accuracy of both systems using receiver operating characteristic (ROC) curves. A longitudinal analysis of 2836 patients with a median follow-up of 4.05 years was performed using Cox proportional hazards models and Kaplan-Meier curves to assess long-term predictive value for GC development.
Results
Cross-sectional evaluation revealed that the Kyoto classification outperformed Li’s scoring system in gastric cancer discrimination, yielding a significantly higher AUC (0.898, 95% CI: 0.859–0.938) than Li’s scoring system (0.760, 95% CI: 0.692–0.827), P = 0.011. These findings suggest that endoscopic-based assessment provides more precise initial risk stratification than serological scoring alone. The primary contribution of this study is the longitudinal analysis, which demonstrated that Li’s scoring system more effectively identified individuals at high long-term risk; the high-risk group had a significantly higher hazard ratio (HR) for developing GC compared to the low-risk group (HR = 4.27, 95% CI: 2.13–8.54, P < 0.001). The annual GC incidence was markedly higher in the high-risk group defined by Li’s scoring system (2.29%) compared to that defined by the Kyoto classification (1.23%).
Conclusion
Both Li’s scoring system and the Kyoto classification are effective for gastric cancer risk assessment, yet they serve potentially complementary clinical roles. The Kyoto classification demonstrates significantly superior immediate discriminative power, while the serology-based Li’s scoring system shows promising prognostic value for longitudinal risk stratification, suggesting its potential utility as a preliminary non-invasive triage tool. However, further prospective multicenter validation studies and formal cost-effectiveness evaluations are needed before an integrated screening strategy can be recommended for large-scale population implementation.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12876-026-04967-3.
Keywords: Gastric cancer, Kyoto classification, Li’s scoring system
Background
Gastric cancer (GC) is a significant global health challenge. According to an estimate by the International Agency for Research on Cancer in its Global Cancer Observatory: Cancer Today, gastric cancer was the fifth most common cancer worldwide in 2022, following lung, breast, colorectal, and prostate cancers [1]. In China, it remains one of the most prevalent tumors. The prognosis of GC is highly dependent on its stage at diagnosis; early-stage gastric cancer has a cure rate exceeding 90% [2]. Consequently, accurate identification of high-risk individuals is pivotal for improving prognosis and alleviating the economic burden associated with this disease [3, 4].
Various scoring systems have been developed internationally to identify individuals at high risk for GC. Among these, the ABC method, which combines serum pepsinogen (PG), gastrin-17 (G-17), and Helicobacter pylori (H. pylori) antibody levels, is widely used [5]. However, these systems are based on foreign population standards and may not be fully applicable to the Chinese population. In response to this limitation, the National Clinical Research Center for Digestive Diseases in China introduced the Li’s scoring system (also known as the new gastric cancer screening risk scoring system) in 2017. This consensus proposed a novel gastric cancer screening risk scoring system tailored to the Chinese population, offering a stratified management approach for early detection [6]. This scoring system was further validated through a multicenter study in 2019, demonstrating good performance in identifying high-risk individuals in high-risk Chinese populations [7]. Additional investigation has demonstrated that this scoring system aligns closely with the ABC method in detecting gastric cancer and precancerous lesions, particularly in identifying low-risk individuals and defining populations requiring endoscopic follow-up [8].
In addition to Li’s scoring system, the Kyoto classification, introduced at the 85th Japanese Society of Gastroenterology in 2013, has gained attention. This classification, based on endoscopic findings, facilitates the assessment of H. pylori infection status and gastric cancer risk [9]. While widely adopted in clinical settings, its practical effectiveness in assessing gastric cancer risk, particularly in daily clinical practice, requires further evaluation.
However, despite their clinical utility and existing cross-sectional comparisons, there remains a critical knowledge gap regarding their comparative performance over time. Most existing studies have focused on their individual or comparative cross-sectional diagnostic accuracy, yet few have directly evaluated and compared their relative performance in terms of long-term longitudinal predictive value. The present study addresses this gap by conducting a large-scale, head-to-head evaluation of 5801 patients in a cross-sectional phase, and importantly, including a 4-year follow-up analysis of 2836 patients. By concurrently assessing the longitudinal divergence in clinical utility of these two distinct paradigms, this research seeks to provide a more robust rationale for their clinical integration and offer a novel perspective on optimizing long-term gastric cancer screening strategies.
Patients and methods
Study design and patients
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Wenzhou Central Hospital (Approval No. L2021-03-001x). The requirement for individual informed consent was waived by the committee due to the retrospective nature of the study and the use of fully anonymized data. This retrospective cohort study included patients who underwent esophagogastroduodenoscopy at Wenzhou Central Hospital between April 2017 and December 2023. Serological screening was conducted to measure serum PG, serum G-17, and H. pylori antibodies. Additionally, esophagogastroduodenoscopy and biopsy were performed to obtain pathological results. Inclusion criteria were as follows: (1) the interval between serological testing and the first esophagogastroduodenoscopy was less than two weeks, and (2) definitive pathological results were available. Exclusion criteria included: (1) use of acid inhibitors, such as H2 receptor blockers or proton pump inhibitors, within one month prior to esophagogastroduodenoscopy; (2) a history of gastric or duodenal surgery; and (3) inability to perform tissue biopsy due to severe cardiopulmonary dysfunction or long-term use of anticoagulant drugs. Ultimately, a total of 5801 patients (2996 males and 2805 females) were included in this study.
Li’s scoring system
Previous scoring systems for identifying high-risk populations for gastric cancer largely relied on qualitative assessments, such as the serum ABC method, which uses binary positive/negative classifications. However, the significance of different test results can vary, and assigning specific weights to various parameters would allow for a quantitative scoring approach. This refinement enables more accurate identification of individuals at truly high risk for gastric cancer. Based on data from the Chinese population, Li’s scoring system has been developed. This system incorporates five variables, including age, gender, H. pylori infection, pepsinogen ratio (PGR), and gastrin-17 (G-17) levels.
Age is scored from 0 to 10 based on defined age ranges, with older individuals receiving higher points. Male patients are assigned 4 points, while females receive 0. The presence of H. pylori infection adds 1 point, while its absence scores 0. For PGR, values below 3.89 are given 3 points, while higher values score 0. G-17 levels are stratified into three categories: <1.50 pmol/L scores 0, 1.50–5.70 pmol/L scores 3, and > 5.70 pmol/L scores 5. This detailed scoring framework facilitates stratified risk assessment and targeted screening, yielding a total score ranging from 0 to 23 points. Patients are categorized into three risk levels according to their scores: high-risk (17–23 points), indicating an extremely high risk of gastric cancer; intermediate-risk (12–16 points), suggesting a moderate risk; and low-risk (0–11 points), reflecting a generally low likelihood of gastric cancer.
All patients underwent serological testing within two weeks of their initial esophagogastroduodenoscopy to calculate their scores using Li’s scoring system.
Kyoto classification
The Kyoto classification evaluates gastritis and predicts gastric cancer risk based on five endoscopic findings: atrophy, intestinal metaplasia (IM), enlarged folds, nodularity, and diffuse redness. Each parameter is scored individually, with atrophy and IM ranging from 0 to 2 based on extent, enlarged folds and nodularity scored as 0 (absent) or 1 (present), and diffuse redness graded from 0 (none) to 2 (severe). Active H. pylori infection, often associated with these findings, may require further diagnostic confirmation. The total Kyoto score ranges from 0 to 8, with higher scores reflecting more severe mucosal damage and increased cancer risk. This system provides a standardized and efficient method for gastritis evaluation, aiding early detection and risk assessment for gastric cancer.
Esophagogastroduodenoscopy was performed by gastroenterologists using Olympus endoscopes (GIF-HQ290 and GIF-290Z, Tokyo, Japan). Each patient’s first esophagogastroduodenoscopy was independently evaluated by two experienced endoscopists (Lin JJ and Zhou QJ) based on the Kyoto classification. In cases of disagreement, a third endoscopist with over 20 years of experience (Pan J) resolved the discrepancies.
Follow-up
The longitudinal study period commenced from the date of each patient’s initial esophagogastroduodenoscopy and concurrent serological screening (baseline). The primary endpoint was defined as the incident diagnosis of GC, which was confirmed by histopathological biopsy results obtained during repeat endoscopic examinations. Time-at-risk (time-to-event) was calculated as the interval (in years) from the baseline examination to either the date of GC diagnosis (event) or the date of the last available endoscopic examination (censoring). Patients were censored if they: (1) reached the end of the study period (December 31, 2023) without a diagnosis of GC; (2) were lost to follow-up; or (3) died from causes other than gastric cancer during the follow-up period.
Statistical analysis
In the cross-sectional analysis, baseline characteristics of the GC and non-GC groups were compared using the Chi-square test for categorical variables and the Mann-Whitney U test for continuous variables. The diagnostic performance of the two systems was evaluated using receiver operating characteristic (ROC) curves, and the statistical significance of the differences between the areas under the curve (AUCs) was determined using the DeLong test to ensure methodological rigor. To evaluate the interobserver reproducibility of the Kyoto classification, Cohen’s kappa statistics were calculated for each endoscopic component.
For the longitudinal analysis, both univariate and multivariable Cox proportional hazards regression models were employed to evaluate the association of GC risk with the two systems. This retrospective cohort study used the forced-entry method to construct separate multivariable models for Li’s scoring system and the Kyoto classification, identifying independent predictors of GC development within their respective frameworks. The multivariable model for Li’s scoring system incorporated its specific constituent variables: age, sex, G-17, PGR, and H. pylori antibody status. Correspondingly, the multivariable model for the Kyoto classification included its endoscopic markers (atrophy, intestinal metaplasia, enlarged folds, and diffuse redness); however, nodularity was excluded from the multivariable analysis due to its extreme sparsity and lack of univariate significance (P > 0.05).
Kaplan-Meier curves were generated to visualize cumulative incidence distributions, and differences between risk groups were assessed using the log-rank test. The average annual incidence of GC was calculated by dividing the number of events by the total person-years of follow-up. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using the Cox proportional hazards model. The proportional hazards assumption was formally evaluated using scaled Schoenfeld residuals. For covariates showing borderline or statistically significant evidence of non-proportionality, stratified Cox sensitivity analyses were performed to assess the robustness of the remaining covariate estimates. Results of the proportional hazards diagnostics and sensitivity analyses are presented in Supplementary Tables S2 and S3. To evaluate potential selection bias, baseline characteristics were compared between patients included in the longitudinal analysis and those excluded (Supplementary Table S1) using the Chi-square test for categorical variables and the Mann-Whitney U test for continuous variables. All statistical analyses were performed using Python 3.9.6 with the scikit-learn (1.6.1), scipy (1.13.1), and lifelines (0.30.0) libraries. Data visualization was conducted using matplotlib (3.9.2). A P-value < 0.05 was considered statistically significant.
Results
Patient enrollment
After excluding 23 patients who had undergone gastric surgery, a total of 5801 individuals were included in the cross-sectional study. Among them, 2836 patients underwent at least two endoscopic examinations and were included in the longitudinal study (Fig. 1). The comparison of baseline characteristics between patients included (n = 2836) and excluded (n = 2965) from the longitudinal analysis is summarized in Table S1. Compared to the excluded group, the included group had a lower proportion of males (46.6% vs. 56.5%, P < 0.001), lower levels of G-17 and pepsinogen I (PGI) (P < 0.001), and a lower prevalence of severe atrophy (26.4% vs. 30.3%, P = 0.001) and diffuse redness (46.3% vs. 48.5%, P = 0.110). No significant difference was observed in age (P = 0.064) or H. pylori infection status (P = 0.057).
Fig. 1.
The flow diagram of patient enrollment
The baseline characteristics of the study population are presented in Table 1. In the cross-sectional study, the median age of the participants was 55 years (interquartile range [IQR]: 47–63 years), with males accounting for 51.7% of the population. The median score of Li’s scoring system was 10 (IQR: 6–13), with the proportions of low-, intermediate-, and high-risk groups being 66.0%, 28.7%, and 5.3%, respectively. The median score of the Kyoto classification was 3 (IQR: 0–4), with 35.5%, 28.5%, and 36.0% of patients classified into low-, medium-, and high-risk categories, respectively. In the longitudinal study, the median age was 55 years (IQR: 48–62 years), and males constituted 46.6% of the cohort. The median score of Li’s scoring system was 10 (IQR: 6–12), with 70.1%, 26.2%, and 3.6% classified into low-, intermediate-, and high-risk groups, respectively. The median score of the Kyoto classification was 3 (IQR: 0–4), with 36.0%, 34.6%, and 29.5% of patients categorized into low-, medium-, and high-risk groups, respectively. The two primary endoscopists demonstrated a high level of interobserver agreement for the Kyoto classification components. The Cohen’s kappa values were 0.90 for atrophy, 0.89 for intestinal metaplasia, 0.88 for enlarged folds, 0.85 for diffuse redness, and 0.78 for nodularity.
Table 1.
Baseline demographics of the study population
| Variable | Cross-sectional study (n = 5801) |
Longitudinal study (n = 2836) |
|---|---|---|
| Age (y) | 55 (47–63) | 55 (48–62) |
| Sex (male, n) | 2996 (51.7%) | 1322 (46.6%) |
| G-17 (pmol/L) | 3.3 (1.2–10.4) | 2.9 (1.1–9.4) |
| PGI (ug/L) | 103.8 (74.0-160.0) | 99.3 (72.1-148.4) |
| PGII (ug/L) | 9.9 (6.2–17.4) | 9.2 (5.9–16.3) |
| PGR | 10.7 (7.5–14.1) | 10.9 (7.8–14.3) |
| H. pylori Ab (positive, n) | 3451 (59.5%) | 1651 (58.2%) |
| Li’s scoring system | 10 (6–13) | 10 (6–12) |
| Low risk | 3829 (66.0%) | 1989 (70.1%) |
| Intermediate risk | 1662 (28.7%) | 744 (26.2%) |
| High risk | 310 (5.3%) | 103 (3.6%) |
| Atrophy | ||
| None | 1949 (33.6%) | 1122 (39.6%) |
| Mild to moderate | 2203 (38.0%) | 964 (34.0%) |
| Severe | 1649 (28.4%) | 750 (26.4%) |
| Intestinal metaplasia | ||
| None | 2322 (40.0%) | 1466 (51.7%) |
| Antrum-limited | 2989 (51.5%) | 1131 (39.9%) |
| Extending to corpus | 490 (8.5%) | 239 (8.4%) |
| Enlarged folds | ||
| Absent | 3921 (67.6%) | 1995 (70.3%) |
| Present | 1880(32.4%) | 841 (29.7%) |
| Nodularity | ||
| Absent | 5785 (99.7%) | 2827 (99.7%) |
| Present | 16 (0.28%) | 9 (0.3%) |
| Diffuse redness | ||
| None/Faint | 3050 (52.6%) | 1522 (53.7%) |
| Mild/Spotty | 2336 (40.3%) | 1136 (40.1%) |
| Severe/Uniform | 415 (7.2%) | 178 (6.3%) |
| Kyoto classification | 3 (0–4) | 3 (0–4) |
| Low risk | 2060 (35.5%) | 1020 (36.0%) |
| Intermediate risk | 1655 (28.5%) | 980 (34.6%) |
| High risk | 2086 (36.0%) | 836 (29.5%) |
Abbreviations G-17 gastrin-17, PGI pepsinogen I, PGII pepsinogen II, PGR pepsinogen ratio, H. pylori Helicobacter pylori, Ab antibody
Cross-sectional study
In the cross-sectional study, a total of 5801 individuals were categorized into the GC group (n = 157) and the non-GC group (n = 5644), as summarized in Table 2. The median age was higher in the GC group than in the non-GC group (65 years [IQR: 57–73] vs. 55 years [IQR: 47–63]; P < 0.001). Males accounted for 66.9% of the GC group compared to 51.2% in the non-GC group (P < 0.001). Serum biomarker levels, including G-17, PGI, and pepsinogen II (PGII), were significantly higher in the GC group compared to the non-GC group (P < 0.001), while the PGR value was lower (P < 0.001). H. pylori Ab-positivity was also significantly more frequent in the GC group (75.8% vs. 59.0%, P < 0.001). Significant differences were observed in endoscopic findings between the groups. The prevalence of atrophy (severe) and intestinal metaplasia (presence) was markedly higher in the GC group (93.6% and 98.1%, respectively) than in the non-GC group (26.6% and 53.9%, respectively; P < 0.001). Enlarged folds were more frequent in the GC group (68.2% vs. 31.4%, P < 0.001), while nodularity showed no significant difference (0.64% vs. 0.27%, P = 0.918). Diffuse redness (mild to severe) was significantly more prevalent in the GC group (81.5% vs. 46.5%, P < 0.001).
Table 2.
Comparison of characteristics between the two classification systems in the cross-sectional study
| Characteristics | GC Group (n = 157) |
Non-GC Group (n = 5644) |
P value |
|---|---|---|---|
| Age (y) + | 65 (57–73) | 55 (47–63) | < 0.001 |
| Sex (male, n) * | 105 (66.9%) | 2891 (51.2%) | < 0.001 |
| G-17 (pmol/L) + | 5.7(2.3–17.9) | 3.3(1.2–10.3) | < 0.001 |
| PGI (ug/L) + | 132.2(81.7-215.9) | 103.3(73.9-158.4) | 0.001 |
| PGII (ug/L) + | 15.0(9.2–27.5) | 9.8(6.2–17.2) | < 0.001 |
| PGR + | 8.1(6.0-11.6) | 10.7(7.6–14.1) | < 0.001 |
| H. pylori Ab (positive, n) * | 119 (75.8%) | 3332 (59.0%) | < 0.001 |
| Atrophy (severe, n) * | 147 (93.6%) | 1502 (26.6%) | < 0.001 |
| Intestinal metaplasia (present, n) * | 154 (98.1%) | 3325 (53.9%) | < 0.001 |
| Enlarged folds (present, n) * | 107 (68.2%) | 1773 (31.4%) | < 0.001 |
| Nodularity (present, n) * | 1 (0.64%) | 15 (0.27%) | 0.918 |
| Diffuse redness (mild to severe, n) * | 128 (81.5%) | 2623 (46.5%) | < 0.001 |
Abbreviations G-17 gastrin-17, PGI pepsinogen I, PGII pepsinogen II, PGR pepsinogen ratio, H. pylori Helicobacter pylori, Ab antibody, GC gastric cancer
* Variables were analyzed using the Chi-square test
+ Variables were analyzed using the Mann-Whitney U test
Figure 2 illustrates the ROC curves comparing the discriminative performance of Li’s scoring system and the Kyoto classification in identifying gastric cancer. The Kyoto classification achieved a robust AUC of 0.898 (95% CI: 0.859–0.938), which was significantly higher than that of Li’s scoring system (AUC = 0.760, 95% CI: 0.692–0.827). The DeLong test confirmed that this difference was statistically significant (P = 0.011), indicating that the Kyoto classification had higher diagnostic accuracy in the cross-sectional detection of gastric cancer.
Fig. 2.
ROC curves of two classification system in the cross-sectional study
Longitudinal study
In the longitudinal study, the longest follow-up duration was 7.15 years, with a median follow-up period of 4.05 years. Throughout the follow-up period, a total of 84 incident GC cases were identified among the 2836 participants. Cox regression analysis, as shown in Table 3, identified several significant risk factors associated with gastric cancer. Univariate analysis was first performed for all variables; nodularity was excluded from the subsequent multivariable model due to its extreme sparsity and lack of statistical significance (P > 0.05). In the univariate Cox regression analysis, several factors were significantly associated with GC risk, including age ≥ 69 years (HR: 2.18, P = 0.012), a lower PGR level (< 3.89) (HR: 3.09, P = 0.006), and the high-risk category of Li’s scoring system (HR: 4.27, P < 0.001). Additionally, the presence of endoscopic severe atrophy showed an increased risk in the univariate assessment (HR: 1.60, 95% CI: 1.03–2.48, P = 0.037). Nodularity was excluded from the multivariable Kyoto model due to its extreme sparsity and lack of statistical significance in the univariate analysis (P > 0.05). In the final multivariable model (Table 3), while Li’s scoring system remained a robust independent predictor (P < 0.001), the association for endoscopic severe atrophy was no longer statistically significant (P > 0.05) after adjusting for demographic and serological variables. These findings indicate that age, PGR level, and Li’s scoring system were the main predictors of long-term GC risk in this cohort.
Table 3.
Cox regression analysis of the two classification systems in the longitudinal analysis
| Variables | Crude HR | 95% CI | P value | Adjusted HR (95% CI) | P value | Incident GC cases |
|---|---|---|---|---|---|---|
| Age (y) | ||||||
| <50 | Reference | Reference | ||||
| 51–59 | 0.85 | 0.54–1.34 | 0.490 | 1.48 (0.79–2.77) | 0.220 | |
| 60–69 | 1.53 | 0.97–2.40 | 0.067 | 2.23 (1.19–4.20) | 0.013 | |
| >69 | 2.18 | 1.18–4.02 | 0.012 | 3.35 (1.54–7.28) | 0.002 | |
| Sex | ||||||
| Female | Reference | Reference | ||||
| Male | 1.13 | 0.73–1.73 | 0.589 | 1.20 (0.78–1.85) | 0.418 | |
| G-17 (pmol/L) | ||||||
| <1.5 | Reference | Reference | ||||
| 1.5–5.7 | 1.12 | 0.71–1.76 | 0.635 | 1.16 (0.68-2.00) | 0.584 | |
| >5.7 | 1.04 | 0.67–1.63 | 0.862 | 0.90 (0.52–1.58) | 0.726 | |
| PGR | ||||||
| ≥3.89 | Reference | Reference | ||||
| <3.89 | 3.28 | 1.51–7.11 | 0.003 | 3.09 (1.38–6.95) | 0.006 | |
| H. pylori Ab | ||||||
| Negative | Reference | Reference | ||||
| Positive | 1.01 | 0.66–1.56 | 0.965 | 0.94 (0.60–1.46) | 0.770 | |
| Li’s scoring system | ||||||
| Low risk | Reference | 50 (2.5%) | ||||
| Intermediate risk | 1.25 | 0.78–1.99 | 0.356 | 25 (3.4%) | ||
| High risk | 4.27 | 2.13–8.54 | < 0.001 | 9 (8.7%) | ||
| Atrophy | ||||||
| None | Reference | Reference | ||||
| Mild to moderate | 0.69 | 0.43–1.11 | 0.126 | 0.81 (0.29–2.25) | 0.691 | |
| Severe | 1.60 | 1.03–2.48 | 0.037 | 1.26 (0.37–4.24) | 0.710 | |
| Intestinal metaplasia | ||||||
| None | Reference | Reference | ||||
| Antrum-limited | 1.03 | 0.67–1.57 | 0.909 | 0.91 (0.34–2.42) | 0.846 | |
| Extending to corpus | 1.32 | 0.66–2.64 | 0.431 | 0.97 (0.30–3.17) | 0.959 | |
| Enlarged folds | ||||||
| Absent | Reference | Reference | ||||
| Present | 1.40 | 0.90–2.19 | 0.138 | 1.10 (0.59–2.04) | 0.766 | |
| Nodularity | ||||||
| Absent | Reference | |||||
| Present | 0.00 | 0.00-Inf | 0.994 | |||
| Diffuse redness | ||||||
| None/Faint | Reference | Reference | ||||
| Mild/Spotty | 1.09 | 0.71–1.68 | 0.694 | 1.12 (0.56–2.25) | 0.744 | |
| Severe/Uniform | 1.80 | 0.90–3.60 | 0.096 | 1.53 (0.56–4.18) | 0.406 | |
| Kyoto classification | ||||||
| Low risk | Reference | 28 (2.7%) | ||||
| Intermediate risk | 0.75 | 0.45–1.24 | 0.259 | 20 (2.4%) | ||
| High risk | 1.45 | 0.94–2.23 | 0.092 | 36 (3.7%) | ||
Abbreviations G-17 gastrin-17, PGI pepsinogen I, PGII pepsinogen II, PGR pepsinogen ratio, H. pylori Helicobacter pylori, Ab antibody, GC gastric cancer, HR hazard ratio
The proportional hazards assumption was assessed using scaled Schoenfeld residuals (Supplementary Table S2). In the Li’s scoring system component model, no covariate showed evidence of violation in the primary rank-based Schoenfeld test, although H. pylori antibody status showed a borderline diagnostic signal. In the Kyoto endoscopic feature model, mild/spotty diffuse redness showed evidence of non-proportionality. Stratified Cox sensitivity analyses by H. pylori antibody status and mild/spotty diffuse redness did not materially alter the interpretation of the remaining covariate estimates (Supplementary Table S3).
Kaplan-Meier analysis, as shown in Fig. 3, demonstrated distinct cumulative incidence rates among different risk groups. For Li’s scoring system (Fig. 3A), the absolute numbers of incident GC cases in the high-, intermediate-, and low-risk groups were 9, 25, and 50, respectively, yielding an incidence rate of 2.29% per year in the high-risk group, followed by the intermediate-risk (1.34% per year) and low-risk groups (0.58% per year). Similarly, for the Kyoto classification (Fig. 3B), incident GC occurred in 36, 20, and 28 cases across high-, intermediate-, and low-risk categories, with an incidence rate of 1.23% per year in the high-risk group, 0.75% per year in the intermediate-risk group, and 0.51% per year in the low-risk group. These findings indicate that both scoring systems provide effective stratification of gastric cancer risk over time.
Fig. 3.
Cumulative incidence of gastric cancer according to the two classification systems in the longitudinal study. A: Kaplan-Meier curves of gastric cancer development according to Li’s scoring system; B: Kaplan-Meier curves of gastric cancer development according to Kyoto classification
Discussion
The clinical imperative for precise GC risk stratification remains clinically important, particularly in East Asian populations where the disease burden remains disproportionately high [1, 2]. While previous literature has examined the cross-sectional comparison of serological and endoscopic markers, the present study significantly extends these findings by providing a rigorous, 4-year longitudinal assessment. Our dual-phase analysis reveals a critical functional divergence: the Kyoto classification functions as a high-precision tool for point-in-time assessment, whereas Li’s scoring system serves as a superior predictor of future malignant transformation. This longitudinal perspective is essential for shifting the screening paradigm from snapshot diagnosis to prognostic risk management.
The statistically significant superiority of the Kyoto classification in our cross-sectional analysis underscores the power of direct mucosal visualization. The Kyoto system provides a real-time assessment of gastric mucosal status that integrates five key markers of the gastric landscape: atrophy, IM, enlarged folds, nodularity, and diffuse redness. Our data specifically identified severe atrophy and diffuse redness as potent indicators of synchronous GC. From a pathobiological perspective, these features represent the advanced stages of Correa’s cascade, reflecting a gastric environment that has reached a high-risk threshold for malignant transformation [10, 11]. Diffuse redness and nodularity, in particular, are hallmarks of active H. pylori infection and the associated intense inflammatory response [12, 13]. When these visual cues are scored collectively, they provide a high-resolution snapshot of the malignant potential of the mucosa. The high AUC observed in our study suggests that for patients already presenting at tertiary centers, the Kyoto classification serves as an indispensable tool for immediate risk assessment, allowing endoscopists to prioritize high-risk areas for targeted biopsy [14–17]. However, the reliance on subjective visual interpretation remains a primary methodological limitation. Even with the expertise of senior endoscopists in our study, the potential for interobserver variability in identifying subtle findings highlights the need for standardized training or, ideally, the integration of artificial intelligence to ensure reproducibility [18].
A pivotal finding of our study is the longitudinal prognostic value of Li’s scoring system. Despite its lower AUC in the cross-sectional phase, Li’s scoring system demonstrated a more powerful association with future GC events. This difference may be attributed to the biological stability of its constituent parameters. Endoscopic features, such as diffuse redness, can be transient or masked by acid inhibitors [19–21]. In contrast, serological markers in Li’s scoring system, specifically the PGR and G-17, reflect the mucosal field cancerization of the gastric mucosa [22–24]. The PGR serves as a stable indicator of the functional mass of gastric fundic glands rather than a mere marker of current inflammation. A low PGR indicates mucosal exhaustion, a chronic state preceding carcinogenesis by several years. Furthermore, by incorporating irreversible demographic risk factors such as age and sex, Li’s scoring system captures the underlying biological risk and cumulative mucosal damage [25–28]. Our longitudinal analysis confirms the efficacy of this system in identifying high-risk individuals who may currently appear endoscopically stable but are prone to malignant transformation. Therefore, Li’s scoring system is highly suitable for population-based screening. It effectively identifies at-risk individuals from asymptomatic populations, thereby minimizing the logistical and invasive burdens associated with universal endoscopy.
The conceptual difference between the two systems is profound: the Kyoto classification identifies established mucosal structural changes associated with synchronous malignancy, while Li’s scoring system assesses the systemic and functional environment conducive to future neoplastic development. Our study demonstrates that these are not competing tools but complementary ones. The longitudinal analysis showed that atrophy (severe) in the Kyoto system was an independent risk factor (HR 1.60) in univariate analysis, but its predictive weight was lower than the comprehensive Li’s scoring score. This suggests that visual atrophy alone may underrepresent the total risk, which is further modulated by age, gender, and secretory function.
We propose a two-tiered precision screening model for high-burden regions such as China. In the first tier (population level), Li’s scoring system could be used to categorize individuals. Those in the high-risk and intermediate-risk categories should then proceed to the second tier (clinical level), where the Kyoto classification is applied during endoscopy to refine the surveillance interval. Such a model maximizes the cost-benefit ratio, reducing unnecessary procedures for low-risk individuals while ensuring that high-risk individuals are less likely to be missed during follow-up [29–31].
Regarding the longitudinal cohort, our analysis (Table S1) revealed certain baseline differences between included and excluded patients, suggesting a degree of selection bias. Specifically, patients who remained in the longitudinal study tended to have slightly lower baseline endoscopic scores compared to those who did not return for follow-up. This healthy screenee effect might lead to an underestimation of the absolute cumulative incidence of GC. However, since our primary aim was to compare the relative predictive performance of the two systems, this bias is likely to affect both paradigms similarly and is unlikely to invalidate our conclusions.
Despite the strengths of our study, several limitations remain. First, the single-center nature may reflect regional profiles that are not universal. The prevalence of H. pylori strains varies globally and may influence score weighting [6, 12]. Second, the median follow-up of 4.05 years, while robust, may not capture the full duration of gastric carcinogenesis; thus, longer observation is essential. Third, the retrospective design prevented consistent tracking of H. pylori eradication therapy, which can influence both serological and endoscopic findings. Fourth, other confounding factors such as smoking, alcohol, and family history were not incorporated. Future research should evaluate the differential regression rates of endoscopic and serological scores following eradication, which will be essential for the clinical management of the increasing post-eradication population.
Conclusion
In conclusion, this study clarifies the potentially complementary roles of the Kyoto classification and Li’s scoring system in gastric cancer risk assessment. The Kyoto classification, with its high immediate discriminative power, is optimized for point-in-time diagnosis and refined endoscopic management in clinical settings. Conversely, Li’s scoring system, by integrating stable serological biomarkers and demographic variables, demonstrates significant prognostic value for longitudinal risk stratification. We suggest a potential synergistic integration of these tools, utilizing Li’s system for non-invasive primary triage followed by the Kyoto classification for secondary endoscopic stratification, to achieve a balanced approach that may optimize resource allocation and personalize patient surveillance. However, given the current lack of prospective multicenter data and formal cost-effectiveness evaluations, this combined strategy requires further rigorous validation prior to clinical implementation in population-based screening programs.
Supplementary Information
Acknowledgements
We would like to thank all the patients who participated in this study. We are also grateful to the nursing staff and endoscopic technicians at the Endoscopy Center of Wenzhou Central Hospital for their invaluable support in clinical data collection and patient management.
Abbreviations
- AUC
Area under the curve
- CI
Confidence interval
- G-17
Gastrin-17
- GC
Gastric cancer
- H. pylori
Helicobacter pylori
- HR
Hazard ratio
- IM
Intestinal metaplasia
- IQR
Interquartile range
- PG
Pepsinogen
- PGI
Pepsinogen I
- PGII
Pepsinogen II
- PGR
Pepsinogen ratio
- ROC
Receiver operating characteristic
Authors’ contributions
JD was responsible for study conceptualization, supervision, and manuscript review and editing. JL and HS developed the methodology, while JL also provided supervision. HS performed the formal analysis, managed the software, and drafted the original manuscript. Investigation and data collection were conducted by JP, LZ, and QZ. All authors have read and agreed to the published version of the manuscript.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Data availability
The datasets generated and/or analyzed during the current study are not publicly available due to hospital privacy policies and patient confidentiality but are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
This study was approved by the Ethics Committee of Wenzhou Central Hospital (No. L2021-03-001x) and adhered to the principles of the Declaration of Helsinki.
Consent for publication
Not applicable. This manuscript does not contain any individual person’s identifiable data in any form (including individual details, images, or videos).
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Jiejun Lin and Huang Su contributed equally as first authors.
References
- 1.Inoue M. Epidemiology of gastric cancer—changing trends and global disparities. Cancers. 2024;16(17):17. 10.3390/cancers16172948. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394–424. 10.3322/caac.21492. [DOI] [PubMed] [Google Scholar]
- 3.Banks M, Graham D, Jansen M, et al. British society of gastroenterology guidelines on the diagnosis and management of patients at risk of gastric adenocarcinoma. Gut. 2019;68(9):1545–75. 10.1136/gutjnl-2018-318126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Quach DT, Hiyama T, Gotoda T. Identifying high-risk individuals for gastric cancer surveillance from western and eastern perspectives: Lessons to learn and possibility to develop an integrated approach for daily practice. World J Gastroenterol. 2019;25(27):3546–62. 10.3748/wjg.v25.i27.3546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Watanabe M, Kato J, Inoue I, et al. Development of gastric cancer in nonatrophic stomach with highly active inflammation identified by serum levels of pepsinogen and helicobacter pylori antibody together with endoscopic rugal hyperplastic gastritis. Int J Cancer. 2012;131(11):2632–42. 10.1002/ijc.27514. [DOI] [PubMed] [Google Scholar]
- 6.National Clinical Research Center for Digestive Diseases, Chinese Society of Digestive Endoscopy, Chinese Society of Health Management, et al. Expert consensus opinion on screening procedures for early gastric cancer in China (draft 2017, shanghai). Chin J Dig Endosc et al. 2018;35(2):77–83.
- 7.Cai Q, Zhu C, Yuan Y, et al. Development and validation of a prediction rule for estimating gastric cancer risk in the Chinese high-risk population: A nationwide multicentre study. Gut. 2019;68(9):1576–87. 10.1136/gutjnl-2018-317556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wang XT, Ji ZZ, Han F, Lyu B. A comparative study of new gastric cancer screening scoring system and new ABC method for screening gastric cancer and precancerous lesions. Zhonghua nei ke za zhi. 2021;60(3):227–32. 10.3760/cma.j.cn112138-20200512-00476. [DOI] [PubMed] [Google Scholar]
- 9.Kamada T, Haruma K, Inoue K, Shiotani A. Helicobacter pylori infection and endoscopic gastritis -kyoto classification of gastritis. Nihon Shokakibyo Gakkai Zasshi Jpn J Gastro-Enterol. 2015;112(6):982–93. 10.11405/nisshoshi.112.982. [DOI] [PubMed] [Google Scholar]
- 10.Toyoshima O, Nishizawa T. Kyoto classification of gastritis: Advances and future perspectives in endoscopic diagnosis of gastritis. World J Gastroenterol. 2022;28(43):6078–89. 10.3748/wjg.v28.i43.6078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Toyoshima O, Nishizawa T, Koike K. Endoscopic kyoto classification of helicobacter pylori infection and gastric cancer risk diagnosis. World J Gastroenterol. 2020;26(5):466–77. 10.3748/wjg.v26.i5.466. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kumar S, Metz DC, Ellenberg S, Kaplan DE, Goldberg DS. Risk factors and incidence of gastric cancer after detection of helicobacter pylori infection: A large cohort study. Gastroenterology. 2020;158(3):527–e5367. 10.1053/j.gastro.2019.10.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Collatuzzo G, Pelucchi C, Negri E, et al. Exploring the interactions between helicobacter pylori (hp) infection and other risk factors of gastric cancer: A pooled analysis in the stomach cancer pooling (StoP) project. Int J Cancer. 2021;149(6):1228–38. 10.1002/ijc.33678. [DOI] [PubMed] [Google Scholar]
- 14.Akbari M, Tabrizi R, Kardeh S, Lankarani KB. Gastric cancer in patients with gastric atrophy and intestinal metaplasia: A systematic review and meta-analysis. PLoS ONE. 2019;14(7):e0219865. 10.1371/journal.pone.0219865. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Spence AD, Cardwell CR, McMenamin ÚC, et al. Adenocarcinoma risk in gastric atrophy and intestinal metaplasia: A systematic review. BMC Gastroenterol. 2017;17(1):157. 10.1186/s12876-017-0708-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dinis-Ribeiro M, Areia M, de Vries AC, et al. Management of precancerous conditions and lesions in the stomach (MAPS): Guideline from the european society of gastrointestinal endoscopy (ESGE), european helicobacter study group (EHSG), european society of pathology (ESP), and the sociedade portuguesa de endoscopia digestiva (SPED). Endoscopy. 2012;44(01):74–94. 10.1055/s-0031-1291491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Burke E, Harkins P, Arumugasamy M, Burke E, Harkins P, Arumugasamy M. Incidence of gastric adenocarcinoma in those with gastric atrophy: A systematic review. Cureus. 2024;16(10). 10.7759/cureus.71768. [DOI] [PMC free article] [PubMed]
- 18.Nishikawa I, Kato J, Terasoma S, et al. Nodular gastritis in association with gastric cancer development before and after helicobacter pylori eradication. JGH Open Open Access J Gastroenterol Hepatol. 2018;2(3):80–6. 10.1002/jgh3.12049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sugimoto M, Ban H, Ichikawa H, et al. Efficacy of the kyoto classification of gastritis in identifying patients at high risk for gastric cancer. Intern Med. 2017;56(6):579–86. 10.2169/internalmedicine.56.7775. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kutsuma A, Oshida H, Suwa K, Nakajima K. A possible association of low pepsinogen I and pepsinogen I/II with low and high body weight in japanese men. Clin Biochem. 2014;47(1):126–8. 10.1016/j.clinbiochem.2013.10.003. [DOI] [PubMed] [Google Scholar]
- 21.Wang Y. Pepsinogen II and a no-pickled food diet are risk factors for female patients with anxiety: A cross-sectional study. Am J Transl Res. 2024;16(2):617–24. 10.62347/TZRM6783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ren JS, Kamangar F, Qiao YL, et al. Serum pepsinogens and risk of gastric and oesophageal cancers in the general population nutrition intervention trial cohort. Gut. 2009;58(5):636–42. 10.1136/gut.2008.168641. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Yoshihara M, Sumii K, Haruma K, et al. Correlation of ratio of serum pepsinogen I and II with prevalence of gastric cancer and adenoma in japanese subjects. Off J Am Coll Gastroenterol ACG. 1998;93(7):1090. 10.1111/j.1572-0241.1998.00335.x. [DOI] [PubMed] [Google Scholar]
- 24.Tu H, Sun L, Gong Y, et al. Abstract C13: Population-based evaluation of serum pepsinogen, anti-H. pylori IgG antibody and gastrin-17 tests for gastric cancer screening. Cancer Prev Res (Phila Pa). 2014;6(11Supplement):C13. 10.1158/1940-6215.PREV-13-C13. [Google Scholar]
- 25.Xie X, Lai Y, Xu P, Xu K, Wang J, Wang L. Comparison of three prediction rules for assessing the risk of gastric cancer in Chinese health examination populations. Preprint posted online Febr. 2021;25. 10.21203/rs.3.rs-257833/v1.
- 26.Hu Y, Bao H, Jin H, et al. Performance evaluation of four prediction models for risk stratification in gastric cancer screening among a high-risk population in China. Gastric Cancer. 2021;24(6):1194–202. 10.1007/s10120-021-01204-6. [DOI] [PubMed] [Google Scholar]
- 27.Zhang W, Li Z, Akram MS, et al. Gastric cancer screening methods: A comparative study of two scoring methods. Cancer Manag Res. 2021;13:5785–91. 10.2147/CMAR.S308395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Liu Xming, Ma Xyu, Liu F et al. Gastric cancer screening methods: A comparative study of the Chinese new gastric cancer screening score and kyoto classification of gastritis. Moreira L, ed. Gastroenterol Res Pract. 2022;2022:1–7. 10.1155/2022/7639968 [DOI] [PMC free article] [PubMed]
- 29.Tu H, Sun L, Dong X, et al. Abstract 5304: A serological biopsy using five stomach-specific circulating biomarkers for gastric cancer risk assessment: a multi-phase study. Cancer Res. 2017;77(13Supplement):5304. 10.1158/1538-7445.AM2017-5304. [DOI] [PubMed] [Google Scholar]
- 30.Ikeda F, Shikata K, Hata J, et al. Combination of helicobacter pylori antibody and serum pepsinogen as a good predictive tool of gastric cancer incidence: 20-year prospective data from the hisayama study. J Epidemiol. 2016;26(12):629–36. 10.2188/jea.JE20150258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Karra S, Sinduja R, Gurushankari B, et al. Development of panel of three-dimensional biomarkers to identify gastric carcinoma and precancerous lesions of the stomach - an analytical cross-sectional study. Indian J Clin Biochem Published online August. 2024;10. 10.1007/s12291-024-01257-1. [DOI] [PMC free article] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analyzed during the current study are not publicly available due to hospital privacy policies and patient confidentiality but are available from the corresponding author on reasonable request.