We retrospectively investigated 1,007 upper gastrointestinal endoscopy patients and obtained gastric mucosal tissues for H. pylori culture between January 2019 and December 2021 from Asan Medical Center, Seoul, Korea. Patients with previous H. pylori eradication treatment (n = 216), gastric surgery history (n = 12), familial adenomatous polyposis (n = 1), and advanced gastric cancer (n = 37) or lymphoma (n = 29) malignancies, which interfere with intact gastric mucosal observation, were excluded. We excluded patients who did not undergo further UBT when the results of both the culture and RUT were negative. (n = 222) (Fig. 1). We defined H. pylori-positive patients as those with positive results for the culture, or both positive in RUT, and UBT and H. pylori-negative patients as those with all negative results for the culture, RUT, and UBT. The Institutional Review Board of Asan Medical Center approved this study (approval number: 2022-0605). Informed consent was waived by Institutional Review Board.

The RUT (Chongkeundang Pharma, Seoul, Korea) assessed the two gastric specimens obtained from the antrum and body. Two samples of gastric tissues were separately placed into RUT kits A and B, and a negative result was obtained when there was no color change for 24 hours. After at least 4 hours of fasting, a breath sample was collected, and a testing tablet containing 100 mg ¹³C-urea (UBIT^® tablet; Otsuka Pharmaceutical Co. Ltd., Tokyo, Japan) was given to the patient. A second breath sample was procured 20 min later and was tested for H. pylori infection by the ¹³C-UBT (Analyzer POCone; Otsuka Electronics Co. Ltd., Osaka, Japan). The cutoff value was 2.5 ppm. The UBT could detect H. pylori infection before eradication treatment and indicate if the treatment was successful.

During endoscopy, the two H. pylori culture specimens were obtained from the antrum and body. For bacterial isolation, the medium used in H. pylori culture plates was Brucella broth agar, which was coated with 5% sheep blood (Becton Dickinson, Franklin Lakes, NJ, USA), trimethoprim (5 μg/mL), vancomycin (10 μg/mL), amphotericin B (5 μg/mL), and polymyxin B (2.5 IU) (Sigma-Aldrich, Darmstadt, Germany). These specimens were incubated at 37°C under micro-ventilation conditions (5% O₂, 10% CO₂, and 85% N₂) for 7 days. The H. pylori colonies collected from each culture were confirmed through gram staining and biochemical methods.

We used the GI-H290 scope (Olympus Medical, Inc., Tokyo, Japan) to examine mucosal lesions, and a senior endoscopist (AJY, more than 15 years of experience) assessed endoscopic findings according to the Kyoto gastritis classification [7]. Atrophy (closed and open), intestinal metaplasia, diffuse redness, spotty redness, mucosal edema, sticky mucus, enlarged folds, nodularity, xanthoma, hyperplastic polyp, regular arrangement of collecting venules (RAC), fundic gland polyp, linear streak, raised erosion, hematin, patchy redness, map-like redness, and multiple whitish flat lesions were assessed based on previous endoscopic images. RAC presence was evaluated in the lesser curvature. We divided the endoscopic findings into non-, current, and past infection. Endoscopic findings associated with non-infection were RAC, raised erosion, fundic gland polyp, linear streak, and hematin. Current infection indicators were open and closed atrophies, spotty redness, diffuse redness, hyperplastic polyp, xanthoma, intestinal metaplasia, edema, nodularity, enlarged folds, and sticky mucus. Past infection was determined by the presence of patchy redness, map-like redness, and multiple whitish flat lesions.

Initial eradication treatment included standard triple therapy (n = 186), bismuth quadruple therapy (n = 22), HDDT (high-dose dual therapy) (n = 17), combination therapy (n = 2), or PBTL (proton pump inhibitor, bismuth, clarithromycin, and levofloxacin) (n = 1) depending on the available antibiotic resistance results. Next, patients took a UBT at least 6 weeks post-treatment to confirm eradication. If the initial eradication therapy failed, a second treatment was recommended with a bismuth-based quadruple therapy when culture results were unavailable or a tailored therapy if culture results were available. If they agreed, patients who failed second-line treatment were treated with a different regimen.

The scoring system assessed three endoscopic parameters with scores ranging from 0 to 3 as follows: diffuse redness or spotty redness, nodularity, and RAC (Supplementary Table 1). To develop the scoring system for predicting H. pylori infection, we divided factors into development (January 2019 to June 2021) and validation sets (July 2021 to December 2021) based on the test date. Initially, since there was a gray area between spotty and diffuse redness when viewed endoscopically, we defined diffuse redness and spotty redness as one factor. For the development set, we performed logistic regression analysis with backward elimination to select factors and predict the H. pylori infection status. In multivariable analysis, we found that diffuse redness or spotty redness and nodularity significantly increased the risk of H. pylori infection, and RAC significantly decreased the risk of H. pylori infection. Previous studies have shown that the presence of RAC, diffuse redness or spotty redness, and nodularity has low inter-observer variability, and they may be easily observed under white light endoscopy [11–14]. Therefore, we selected nodularity, RAC, and spotty redness or diffuse redness as the key indicators (Fig. 2). The risk score was the weighted sum of these predictors, for which weights were the rounded integer value of the quotient of regression coefficients divided by the coefficient of the reference predictor (nodularity). The predicted probability was estimated by a logistic regression equation with the risk score in the validation set. The event probability was one divided by the sum of one and the exponential function of negative one times the sum of negative 0.9884 and 4.4197 times the score squared:

The risk score’s discrimination capability was assessed using the C statistic, and the calibration capability was assessed using a calibration plot and the Hosmer–Lemeshow test in both the development and validation sets.

Statistical analysis was performed with SPSS ver. 22.0 for Windows (IBM Corp., Armonk, NY, USA) and R (version 3.6.1; https://www.R-project.org). All descriptive variables are expressed as the mean with standard deviation representative of the distribution. In addition, each group characteristic was compared using the chi-squared or Fisher’s exact tests for categorical variables and two-sample t-tests for continuous variables. A p value less than 0.05 (represented by “< 0.05”) indicated statistical significance.

This study examined 358 H. pylori-positive and 132 H. pylori-negative patients. Based on endoscopic findings, compared with the H. pylori-negative group, the H. pylori-positive group exhibited a significantly higher prevalence of atrophy (98.6% vs. 81.1%, p < 0.001), intestinal metaplasia (64.0% vs. 45.5%, p < 0.001), diffuse redness (47.5% vs. 3.0%, p < 0.001), spotty redness (77.1% vs. 6.8%, p < 0.001), edema (14.2% vs. 5.3%, p = 0.010), sticky mucus (28.5% vs. 1.5%, p < 0.001), enlarged folds (6.1% vs. 0.0%, p = 0.008), and nodularity (7.5% vs. 0.0%, p = 0.003). On the other hand, compared with the H. pylori-positive group, the H. pylori-negative group had a significantly higher occurrence of RAC (56.1% vs. 1.1%, p < 0.001), fundic gland polyp (12.9% vs. 0.6%, p < 0.001), linear streak (37.9% vs. 0.8%, p < 0.001), and raised erosion (28.8% vs. 2.0%, p < 0.001) (Table 1).

The H. pylori-negative group was further divided into no atrophy (n = 25) and atrophy (n = 107). The mean age of the no atrophy group was lower than that of the atrophy group (48.4 ± 17.3 yr vs. 63.4 ± 11.3 yr). Endoscopic features demonstrated that all 25 patients in the no atrophy group had notably more RAC (100.0% vs. 45.8%, p < 0.001) and linear streaks (84.0% vs. 27.1%, p < 0.001) than those in the atrophy group. The no atrophy group did not exhibit intestinal metaplasia, diffuse redness, spotty redness, edema, sticky mucus, enlarged folds, nodularity, or hyperplastic polyp. On the other hand, 56.1% of patients in the atrophy group had intestinal metaplasia (p < 0.001), and 24.3% of them had xanthoma (p = 0.025) (Supplementary Table 2).

The development set consisted of 349 patients, and the validation set included 141 patients. In our scoring system (0–3), diffuse or spotty redness, nodularity, and RAC loss were worth 1 point each. At a score of 2, 97% of patients in the development and validation sets were H. pylori-positive, and at a score of 3, 100% of patients were H. pylori-positive (Table 2). The scoring system indicated 2%, 45%, 96%, and 99% infection risk at scores of 0, 1, 2, and 3, respectively (Fig. 3).

The performance of the scoring system was assessed using discrimination and calibration metrics (Table 3). Discrimination capability was evaluated using the C statistic, and calibration was assessed using the Hosmer–Lemeshow test. For the development set, the scoring system demonstrated a high level of discrimination, with a C statistic of 0.929 (standard error [SE] = 0.018) and a 95% confidence interval (CI) ranging from 0.894 to 0.963. The calibration of the model showed acceptable agreement with the observed outcomes, as indicated by a Hosmer–Lemeshow χ² value of 2.000 (degrees of freedom [DF] = 2) and a non-significant p value of 0.368, suggesting no evidence of poor fit. For the temporal validation set, the model maintained strong discrimination with a C statistic of 0.951 (SE = 0.017) and a 95% CI of 0.918–0.984. The calibration test yielded a χ² value of 0.084 (DF = 2) with a p value of 0.959, indicating a good fit between the predicted and observed event rates over the deciles of risk.

The endoscopic findings of patients who underwent the initial eradication therapy were categorized into the success (n = 198) or failure (n = 30) groups, with an 86% success rate observed with initial eradication treatment. The success group was treated with various eradication regimens, including standard triple therapy (163/186, 87%), bismuth-based quadruple therapy (22/22, 100%), combination therapy (1/2, 50%), HDDT (14/17, 82%), and PBTL (1/1, 100%). Patients with an unsuccessful initial treatment were re-treated with the following regimens: standard triple therapy (0/1, 0%), bismuth-based quadruple therapy (12/14, 86%), HDDT (0/2, 0%), and PAL (proton pump inhibitor, amoxicillin, and levofloxacin) (1/1, 100%). All patients with an unsuccessful second eradication treatment with standard triple and high-dose dual therapies were successfully treated with bismuth-based quadruple therapy. The remaining patients were lost to follow-up or refused further treatment. The proportions of patients with diffuse redness and sticky mucus were significantly higher in the failure group than in the success group (43.9% vs. 66.7%, p = 0.033 and 26.8% vs. 46.7%, p = 0.044, respectively) (Supplementary Table 3).