Introduction

Endoscopic submucosal dissection (ESD) is the standard treatment for superficial esophageal cancer. However, esophageal stricture following endoscopic ESD is a common and severe complication, leading to malnutrition, increased financial burden, and reduced quality of life.

Aims & Methods

This study aimed to develop a risk prediction model for stricture after non-circumferential ESD to improve prevention, treatment, and personalized management of post-ESD esophageal stricture. Clinical data were retrospectively collected from patients with suspected early esophageal squamous cell carcinoma undergoing ESD between January 2014 and December 2022 at four tertiary hospitals in Sichuan Province, China.
Cases were divided into non-circumferential and circumferential ESD groups. Non-circumferential ESD cases were split into a development cohort (70%) and a validation cohort (30%). Independent risk factors for stricture were identified using univariate and multivariate logistic regression analyses. A risk scoring model was established and internally validated.
For patients with circumferential lesions, a case-control study was first conducted to validate the efficacy of steroid-based interventions in reducing ​stricture​ rates, followed by subgroup analyses comparing the therapeutic differences between oral administration alone and oral combined with local injection.

Results

Among 2009 enrolled patients, the stricture rate was 12.9% (259/2009). Data from 1854 non-circumferential ESD cases were used for model construction. Independent risk factors included wound circumferential extent, lesion location, intraoperative titanium clip usage, and non-R0 resection. The risk stratification model (score range: 0–8) categorized patients into low- (0–2 points, 1.7% stricture), intermediate- (3–5 points, 25%), and high-risk (6–8 points, 80%) groups. In the validation cohort, the model demonstrated an AUC of 0.864 (95% CI: 0.814–0.914, P<0.001), Hosmer-Lemeshow test (P=0.187), and Brier score of 0.0456, indicating excellent discrimination and calibration (Table 1）.
Among 155 patients undergoing circumferential esophageal ESD, the stricture rate was 72.9% (113/155). The stricture rate was 100% (14/14) in the non-steroid group versus 70.2% (99/141) in the steroid group (P=0.017). Patients receiving steroids required an average of 7.4 ± 8.9 dilation sessions, which was significantly lower than the non-steroid group (P=0.031). Nevertheless, no significant differences were observed between oral steroids alone and combined oral with local injection steroids in preventing strictures (P=0.908) or reducing dilation sessions (P=0.652).
Table 1. Predictive Scoring System for Post-ESD Stricture (Non-Circumferential Lesions）

Conclusion

This multicenter study identified ​​wound circumferential extent, lesion location, titanium clip use, and non-R0 resection​​ as independent risk factors for stricture after non-circumferential ESD. A validated ​​risk stratification model (low/intermediate/high-risk)​​ enables tailored postoperative management. While steroids reduced strictures in circumferential ESD, ​​oral versus combined oral/local injection regimens​​ showed no difference in efficacy or dilation frequency.

References

1. Pimentel-Nunes P, Libânio D, Bastiaansen B A J, et al. Endoscopic submucosal dissection for superficial gastrointestinal lesions: European society of gastrointestinal endoscopy (ESGE) guideline - update 2022[J]. Endoscopy, 2022, 54(6): 591-622.
2. Park C H, Yang D H, Kim J W, et al. Clinical practice guideline for endoscopic resection of early gastrointestinal cancer[J]. Intestinal Research, 2021, 19(2): 127-157.
3. Ishihara R, Arima M, Iizuka T, et al. Endoscopic submucosal dissection/endoscopic mucosal resection guidelines for esophageal cancer[J]. Digestive Endoscopy, 2020, 32(4): 452-493.
4. Kitagawa Y, Uno T, Oyama T, et al. Esophageal cancer practice guidelines 2017 edited by the japan esophageal society: Part 2[J]. Esophagus, 2019, 16(1): 25-43.

Introduction

Gastric cancer remains a major global cause of cancer mortality. Early detection and accurate risk stratification are crucial for improving prognosis. Noninvasive, affordable serum tests are essential for mass screening. We compared the updated ABC method (pepsinogen I/II and gastrin-17) and a novel gastric cancer scoring system (“new scoring system”) regarding their effectiveness in screening early gastric cancer and precancerous lesions, risk stratification, and missed diagnosis prevention.

Aims & Methods

From April to November 2024, 7,304 asymptomatic individuals from Yuhang District, Hangzhou, China, participated in community-based screening. Serum pepsinogen I (PGI), PGII, PGI/II ratio (PGR), gastrin-17 (G-17), and H. pylori antibody were tested. Participants were stratified into low-, medium-, or high-risk groups using the ABC method and the new scoring system. A total of 654 medium- and high-risk participants underwent endoscopy and histopathology. Detection rates of gastric cancer, intraepithelial neoplasia, and atrophic gastritis were compared. Missed diagnosis rates in low-risk groups were analyzed assuming no endoscopy.

Results

The ABC method classified 90.55% as low-risk, 7.69% medium-risk, and 1.75% high-risk. The new scoring system classified 59.42% as low-risk, 32.67% medium-risk, and 7.91% high-risk. Endoscopy detected 11 gastric cancers (1.68%, 10 early-stage), 21 intraepithelial neoplasias (3.21%), and 98 atrophic gastritis cases (14.98%).
Gastric cancer detection rates in medium/high-risk groups were similar: ABC 1.58% (3/190), new scoring 1.77% (10/565) (p=1.000). Atrophic gastritis rates were comparable (ABC 14.21%, new scoring 15.04%, p=0.872). In the high-risk subgroup, the new scoring system detected significantly more atrophic gastritis cases (29.87% vs ABC 16.33%).
In low-risk groups, the ABC method would miss 8 of 11 gastric cancers (72.73%) without endoscopy; the new scoring system would miss only 1 case (9.09%, p=0.009). ROC analysis showed higher AUC for the new scoring system (0.60 vs 0.51). Decision curve analysis indicated superior net clinical benefit for the new scoring system at threshold probabilities above 0.2. The new scoring system showed higher sensitivity for early-stage disease detection, whereas ABC had slightly higher specificity.

Conclusion

Both systems showed similar performance in detecting gastric cancer and precancerous lesions among medium- and high-risk populations. The new scoring system significantly reduced missed diagnoses in low-risk groups and improved identification of high-risk precancerous conditions. Incorporating ROC, DCA, and accuracy metrics, the new scoring system offers enhanced flexibility for risk stratification, making it potentially more effective for population-based gastric cancer screening, especially in resource-limited settings. Further validation in larger cohorts is warranted.

Disclosure

Nothing to disclose.

Introduction

Erosive esophagitis (EE) is a clinically significant manifestation of gastroesophageal reflux disease (GERD), characterized by endoscopically visible mucosal breaks due to chronic gastric acid exposure. EE increases the risk of complications such as Barrett’s esophagus and esophageal adenocarcinoma. Although esophagogastroduodenoscopy (EGD) remains the diagnostic gold standard, it is invasive and often requires sedation. Wired magnetic-assisted capsule endoscopy (MACE) is a novel, minimally invasive technique using a magnetically navigable capsule tethered via a soft wire, enabling upper gastrointestinal tract examination. However, its diagnostic value for EE remains underexplored.

Aims & Methods

From February 2021 to November 2024, 85 patients underwent wired MACE followed by EGD within six hours at National Taiwan University Hospital. Indications included acute upper gastrointestinal bleeding (n=30), esophageal varices screening (n=33), and pre-bariatric surgery assessment (n=22). Diagnostic accuracy of wired MACE for EE was assessed using EGD as the reference. Patient-reported discomfort (0–10 scale) and acceptance (0–5 scale) were compared between the two modalities.

Results

The mean age was 57.8 years (SD 15.2; range 23–83), with 62.4% male. EGD was performed without sedation in 92.9% of cases. Wired MACE showed a sensitivity of 86.4% (95% CI: 76.2–97.0%), specificity of 82.9% (95% CI: 71.4–94.0%), PPV of 84.4%, NPV of 85.0%, and overall accuracy of 84.7%. Agreement with EGD was substantial (κ = 0.69). MACE was associated with lower mean discomfort scores (1.9 vs. 5.5, p < 0.0001) and higher acceptance scores (4.2 vs. 2.3, p < 0.01), indicating superior tolerability.

Conclusion

Wired MACE demonstrated high diagnostic accuracy and substantial agreement with EGD in detecting EE. Its significantly lower discomfort and higher patient acceptance support its potential as a less invasive, better-tolerated diagnostic alternative to conventional endoscopy.

References

Sharma, P., et al., The Diagnostic Accuracy of Esophageal Capsule Endoscopy in Patients With Gastroesophageal Reflux Disease and Barrett's Esophagus: A Blinded, Prospective Study. Official journal of the American College of Gastroenterology | ACG, 2008. 103(3): p. 525-532.
Lien, G.S., et al., Feasibility and safety of a novel magnetic-assisted capsule endoscope system in a preliminary examination for upper gastrointestinal tract. Surg Endosc, 2018. 32(4): p. 1937-1944.

Introduction

Risk stratification is needed to determine which Barrett’s esophagus (BE) patients benefit from regular endoscopic surveillance given its burden for patients and society. Several risk classification models have been proposed based on clinical and histopathological features, such as the Parasa¹ model which achieves a modest AUC of 0.76. Machine learning offers a new approach to developing risk prediction tools, potentially enhancing predictive performance and model flexibility compared to traditional methods.

Aims & Methods

We developed a machine learning model for predicting neoplastic progression in BE based on a large prospective cohort. Demographic, endoscopic and p53 immunohistochemistry (IHC) data were collected from patients in the ProBar study, an ongoing multicenter prospective cohort study in the Netherlands (2003- onwards). Patients received regular endoscopic surveillance according to the guidelines and were excluded if they had less than 6 months of follow-up, high grade dysplasia (HGD) or esophageal adenocarcinoma (EAC) at baseline or within 6 months of inclusion, or a BE segment < 2cm. The model we developed used 17 variables, including demographic factors (sex, age at inclusion, ethnicity, education level), medical history (family history of BE or EAC, previous low grade dysplasia), lifestyle factors (smoking, alcohol use), clinical features (weight, height, BMI, gastroesophageal reflux disease, presence of esophagitis, proton pump inhibitor (PPI) use, BE segment length), and p53 IHC status. An XGBoost model was developed to predict progression to HGD or EAC using 10-fold stratified cross-validation for training and validation. To enhance model performance and reduce complexity, SelectFromModel was applied to identify the most relevant features. The top seven predictors were sex, smoking history, alcohol use, education level, PPI use, presence of esophagitis and p53 IHC status — the strongest predictor. Model performance was evaluated using accuracy, balanced accuracy, sensitivity, specificity and AUC.

Results

Of 2052 patients in the ProBar cohort, 1246 were included for analysis (73% male; 81% Caucasian; median age 61.0 ±11.2 years). The average BE segment length was 4.9 ± 2.9 cm and p53 was aberrant in 20.5%. During follow-up, 121 patients (9.7%) showed neoplastic progression, while 1125 (90.3%) did not. Using stratified 10-fold cross-validation, the model demonstrated strong predictive performance for neoplastic progression. It achieved a mean accuracy of 84.4% (95% CI: 82.7%–86.2%) and an AUC of 0.837 (95% CI: 0.810–0.863), indicating good discriminative ability. The balanced accuracy was 80.0% (95% CI: 77.3%–82.8%), sensitivity was 74.6% (95% CI: 69.0%–80.1%), and specificity was 85.5% (95% CI: 83.5%–87.5%). Although class weighting was used to address the class imbalance (around 1:9 ratio), the model performed better on the majority class. The improvements in sensitivity and balanced accuracy suggest the model is becoming more reliable at identifying progression cases. Adding more samples to the minority class could further balance the data and enhance model performance.

Conclusion

We developed a machine learning model based on clinical and histopathological factors with high diagnostic accuracy to predict neoplastic progression in BE patients. Clinical risk stratification could be an accessible route for more personalized surveillance strategies.

References

1. Parasa S, Vennalaganti S, Gaddam S, et al. Development and Validation of a Model to Determine Risk of Progression of Barrett's Esophagus to Neoplasia. Gastroenterology 2018;154:1282-1289 e2.

Disclosure

M.C.W. Spaander received research support from Capsulomics, Inc. and Lucid Diagnostics. The remaining authors disclose no conflicts of interest.

Introduction

Patients with Barrett’s Esophagus (BE) undergo regular endoscopic surveillance to timely intervene in the development of esophageal adenocarcinoma (EAC). The current European Society of Gastrointestinal Endoscopy (ESGE) guideline recommends stopping BE surveillance at the age of 75 years, based on average life expectancy. However, clinical studies to support this recommendation are lacking.

Aims & Methods

Therefore, this study aims to determine the optimal age to discontinue surveillance by analyzing incidence and mortality rates in a real-life cohort study. We analyzed data from a multicenter, prospective cohort of BE patients in the Netherlands with a segment length of ≥2 cm, all of whom underwent endoscopic surveillance between 2004 and 2023, in accordance with American College of Gastroenterology guidelines. Patients were linked with the Netherlands Cancer Registry (NCR) to optimize data completeness. Vital status and cause of death of patients were verified with their general practitioner and through the Central Statistical Office (CBS). The cumulative incidence of EAC was calculated, and standardized incidence (SIR) and standardized mortality rate (SMR) were analyzed using a Poisson regression model.

Results

A total of 1,378 BE patients (median age at inclusion 62 years; 73% male) were included for analysis. Median follow-up time was 6.3 years, and median age at cessation of surveillance was 75 years. In total, 57 EAC cases were identified at a median age of 69 years, of which 89% under active surveillance. The majority of cancers was diagnosed at an early stage (47% Stage I, 12% stage II, 3.5% stage III). The cumulative incidence after 10 years of follow-up was 4.4.% in the complete cohort, which was lower in the age group >75 (1.7%). The incidence rate per 1,000 person years was highest amongst patients aged 65-69 years. The SIR was 19.2 in the complete cohort, decreasing to 12.1 in patients >75 years. A total of 350 patients died during follow-up, of which 15 (4.3%) due to EAC. Median age at death was 79 years. The most frequent causes of death were cardiovascular disease (25%), and other malignancies (27%). The mortality rate per 10,000 person years was 10 for EAC-related death and 240 for other causes. The risk of other cause mortality was 24 times higher than EAC-related mortality. The overall SMR for EAC-related death was 3.9, decreasing to 2.5 in patients > 75 years of age.

Conclusion

Our findings support the ESGE guideline recommendation to discontinue surveillance in BE patients older than 75 years, as the cumulative incidence, SIR, and SMR all declined notably in this age group. Although the SIR was still high in patients aged 75 and above, patients in this cohort were 24 times more likely to die from other causes than from EAC.

Introduction

Risk stratification is vital in upper gastrointestinal bleeding (UGIB) to guide management decisions. Several risk scores to predict adverse outcomes of UGIB have been developed, however, real-world comparisons across diverse patient outcomes are limited. We aimed to validate and compare the discriminatory performance of five established UGIB risk scores for inpatient mortality, including subgroup analysis for variceal and non-variceal bleeding aetiologies.

Aims & Methods

We retrospectively analysed 1,003 patients who received inpatient esophagogastroduodenoscopy (OGD) on admission between January 2025 and August 2023 at a tertiary centre. Each patient’s ABC, CHAMPS, MAP(ASH), Glasgow-Blatchford (GBS), and Japanese scores were calculated. The outcome measured was 30-day inpatient mortality. Receiver operating characteristic (ROC) and area under the curve (AUC) were calculated to assess each score’s discriminative ability. Further subgroup analyses were performed depending on the aetiology of UGIB (variceal vs non-variceal).

Results

Between the five analysed scores, the ABC score (AUC: 0.817) had the highest overall prediction ability for 30-day inpatient mortality, followed by CHAMPS (AUC: 0.791) and MAP(ASH) (AUC: 0.774). GBS and the Japanese scores demonstrated moderate to poor accuracy (AUC: 0.713 and 0.614, respectively). A similar pattern was displayed for variceal UGIB subgroup, with the ABC score being the best performing (AUC: 0.826), followed by CHAMPS (AUC: 0.774). MAP(ASH), GBS, and the Japanese score had poor predictive values (AUCs: 0.694, 0.660, and 0.540, respectively). ABC, CHAMPs, and MAP(ASH) demonstrated strong predictive accuracy in the non-variceal UGIB subgroup, with AUCs: 0.818, 0.792, and 0.787, respectively. Whilst GBS (AUC = 0.728) and the Japanese score (AUC = 0.627) had less reliability.

Conclusion

In this real-world UK cohort, the ABC score outperformed four other established scoring systems in predicting inpatient mortality in UGIB. Thus, our findings may indicate ABC score to be the most dependable score for early risk assessment in acute UGIB patients, independent of bleeding aetiology, followed by CHAMPS and MAP(ASH).

Introduction

Endoscopy services in many healthcare systems face increasing demand, and novel approaches to patient triage are needed. Recent research suggests that capsule sponge testing combined with immunohistochemical biomarkers can prioritise patients with Barrett’s oesophagus (BO) and gastro-oesophageal reflux for oesophagogastroduodenoscopy (OGD).^1,2 However, it is not yet widely integrated into clinical practice globally. The capsule sponge is not currently used in South Wales and Northern Ireland, and further investigation into its implementation in these regions is required.

Aims & Methods

This multicentre feasibility study aimed to evaluate whether the capsule sponge could be used in Wales and Northern Ireland to triage patients with gastro-oesophageal reflux or BO for OGD, and to assess feedback from patients and staff. Patients referred to gastroenterology due to gastro-oesophageal reflux and patients with known BO under surveillance were eligible. Patients were recruited from 3 hospital sites in Wales and 2 in Northern Ireland, undergoing the test in nurse-led clinics between December 2024 and March 2025. Pathology samples were processed centrally in an accredited laboratory (Cyted), examining for TFF3 (suggestive of intestinal metaplasia), p53 and atypia (high-risk findings associated with dysplasia). Patient feedback was collected at the clinics, and feedback forms were also sent to patients and staff.

Results

201 patients were recruited, with 196 (97.5%) completing the capsule sponge test: 116 with BO (median age 59, 84% male) and 80 with reflux (median age 49, 56% female). 101 BO samples (87%) were adequate for analysis. Of these, 21 (21%) were positive/equivocal for atypia/p53 and deemed high-risk, and 80 (79%) were negative for both atypia and p53. High-risk patients were prioritised for urgent OGD, with the remainder followed-up as per the British Society of Gastroenterology guidelines. 61 samples (76%) from patients with reflux were adequate for analysis. Of these, 3 (5%) were positive/equivocal for atypia/p53 and prioritised for urgent OGD, 11 (18%) were TFF3 positive but negative for both atypia and p53 and were booked for routine OGD, and 47 (77%) were negative for all 3 biomarkers and were managed as per usual local practice.
148 patients (76%) who completed a capsule sponge test had previously had an OGD, with 78% of these preferring the capsule sponge, 4% preferring the OGD and 9% being unsure. Feedback forms were completed by 52 patients, with response rates varying by question. 92% rated the capsule sponge procedure good/very good, 3% fair and 5% poor/very poor. When asked how likely they would be to recommend the test, on a scale from 1 (least likely) to 10, patients with reflux gave a mean score of 8.7 and patients with BO 9.6. For whether they would be willing to have the test done again, 83% said yes, 14% not sure and 3% no. Patients’ free text comments indicated they found the capsule sponge to be quicker and easier than an OGD, however some experienced discomfort on sponge withdrawal. 14 staff completed feedback forms. 100% agreed the capsule sponge test was easy to deliver, and 100% rated the implementation of the test pathway as good/very good.

Conclusion

This feasibility study showed that the capsule sponge could successfully triage patients under BO surveillance or with reflux for OGD in Wales and Northern Ireland, with positive feedback from patients and staff. These findings support future integration of the capsule sponge into clinical pathways in these regions, with potential transferability to other healthcare systems internationally.

References

1. Chien S, Glen P, Penman I, Cruickshank N, Bryce G, Crumley A, et al. Oesophageal cell collection device and biomarker testing to identify high-risk Barrett’s patients requiring endoscopic investigation. Br J Surg. 2024 May 13;111(5):znae117.
2. Chien S, Glen P, Bryce G, Cruickshank N, Penman I, Robertson K, et al. Clinical application of capsule sponge testing in symptomatic reflux disease: a national prospective cohort study. BMC Gastroenterol. 2024 Nov 26;24:431.

Disclosure

R McNally, B Afinowi, I Aldred and R Hardwick are employed by Cyted Health, who manufacture the capsule sponge device used in this study. The remaining authors have nothing to disclose.