Introduction

Despite the usefulness of white light endoscopy (WLE) and nonmagnified narrow band imaging (NBI) for screening superficial esophageal squamous cell carcinoma (SESCC) and precancerous lesions, these lesions may be missed due to their subtle features and interpretation variations among endoscopists.

Aims & Methods

Our team has developed an artificial intelligence (AI) system to detect SESCC and precancerous lesions using WLE and nonmagnified NBI. This study aims to evaluate the impact of this AI system on clinical diagnosis.
A prospective tandem study was conducted to evaluate the auxiliary diagnostic performance of the AI system in a real clinical setting. Consecutive patients undergoing sedated upper gastrointestinal endoscopy at 12 hospitals in China were recruited. Patients were randomly assigned to either the AI-assisted first group or the routine first group at a 1:1 ratio. The same endoscopist at each center performed tandem upper gastrointestinal endoscopy for each eligible patient on the same day. In the AI-assisted group, the endoscopist performed the first examination with the assistance of the AI system and the second examination without it. In the routine first group, the order of examinations was reversed. The primary outcome was the miss rate of SESCC and precancerous lesions. The secondary outcome including the detection rate of SESCC and precancerous lesions.

Results

Between October 2021 and June 2022, a total of 5,856 eligible patients were included and randomly assigned to the AI-assisted first group and 5,850 were assigned to the routine first group. The per-lesion miss rate was significantly lower in the AI-assisted first group than in the routine first group (1.7% vs 6.7%; Risk ratio (RR)=0.25; p=0.08). The detection rate in the AI-assisted first group was significantly higher than that in the routine first group (1.8% vs 1.3%; RR=1.38; p=0.03).

Conclusion

The proposed AI system can reduce the miss rate and improve the detection rate of SESCC and precancerous lesions during upper gastrointestinal endoscopy under WLE and non-magnified NBI.

Introduction

Recent research from the Netherlands, the United States, and Korea suggests that individuals who receive a positive result on the fecal immunochemical test (FIT) may be at a higher risk of developing gastroesophageal cancer (GEC) within the following 36 months. However, the associations reported in these studies may be biased due to differences in age and sex between the FIT-positive and FIT-negative groups and other confounding factors.

Aims & Methods

We aimed to evaluate the association between FIT results and the risk of GEC in a matched cohort by age and sex. Additionally, we assessed the cost-effectiveness of a one-time Esophago-Gastro-Duodenoscopy (EGD) for individuals with positive FIT results.. We created a matched cohort of individuals aged 50 to 75 who underwent FIT testing at Clalit Health in Israel between June 2016 and June 2019. For each FIT-positive individual, we matched three FIT-negative individuals based on age, sex, and H. pylori status. Adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) were utilized to evaluate the association between a positive FIT result and the risk of GEC within 36 months. We calculated the incremental cost-effectiveness ratio (ICER) for a one-time EGD at $400 in individuals who tested positive for FIT. Screening is considered cost-effective if the ICER is below $50,000.

Results

The primary cohort included 483,710 individuals, while the matched cohort comprised 151,500 individuals (47.6% female, median age 62.5 years). During the follow-up period, 202 cases of GEC occurred: 0.2% (64/38,575) among FIT-positive individuals and 0.1% (138/112,925) among FIT-negative individuals, adjusted HR 1.39 (95% CI 1.03-1.87). Age, male sex, H. pylori exposure, country of birth, proton-pump inhibitor usage, smoking status, and anemia were all significantly associated with an increased risk of gastric cancer. A one-time EGD has been determined to be a cost-effective option in a range of ages and rates of localized disease at the time of EGD.

Conclusion

This matched cohort study suggests that individuals with a positive FIT may have an increased risk of GEC, which occurs at an overall rate of 0.2% within 36 months. A one-time EGD priced at $400 can be cost-effective in many scenarios.

Disclosure

NONE

Introduction

Several risk stratification scores have been developed for predicting endoscopic treatment or patient survival in non-variceal upper gastrointestinal bleeding (NVUGIB).

Aims & Methods

Our aim was to compare the performance of some previously described scores such as Canada-United Kingdom-Adelaide score (CANUKA), Glasgow-Blatchford score (GBS), Harbinger score, neutrophil-to-lymphocyte ratio (NLR), red cell distribution width (RDW) and Rockall (pre-endoscopy and complete) scores in predicting with the following outcomes in patients with variceal upper gastrointestinal bleeding (VUGIB): 30-day mortality, in-hospital mortality, need for interventional radiological procedure, need for RBC transfusion and rebleeding. Patients undergoing emergency endoscopy for upper GI bleeding at the Department of Emergency Medicine at Semmelweis University between January 1, 2022 and December 31, 2023 were included in a retrospective cohort analysis. Cases with VUGIB were managed according to the ESGE guideline. Statistical methods included ROC analysis, through which we obtained the AUROC values.

Results

There were 126 patients with VUGIB (79 males, 47 females, mean age 58). In-hospital rebleeding rate was 7.9% (n=10/126) for VUGIB. Second look endoscopy was performed in 30 cases (23.8%). The indication for second look endoscopy was inadequate index endoscopy in 10 cases, suspicion of re-bleeding in 6 cases, control of endoscopic intervention in 12 cases. The 30-day mortality rate was 23.8% (n=30/126). Due to missing data in 5 patients scoring systems were analyzed in 121 patients. Area under receiver operator characteristic curves (AUROC) for predicting various outcomes were as follow: 30-day mortality (HARBINGER: 0.63, CANUKA: 0.75, Rockall (pre-endoscopy): 0.73, Rockall (complete): 0.73, GBS: 0.71, NLR: 0.65, RDW: 0,54); in-hospital mortality (HARBINGER: 0.59, CANUKA: 0.7, Rockall (pre-endoscopy): 0.67, Rockall (complete): 0.67, GBS: 0.65, NLR: 0.63, RDW: 0.56); need for interventional radiological procedure (HARBINGER: 0.62, CANUKA: 0.49, Rockall (pre-endoscopy): 0.56, Rockall (complete): 0.56, GBS: 0.54, NLR: 0.54, RDW: 0.67); need for RBC transfusion (HARBINGER: 0.66, CANUKA: 0.83, Rockall (pre-endoscopy): 0.7, Rockall (complete): 0.7, GBS: 0.78, NLR: 0.65, RDW: 0.63); rebleeding (HARBINGER: 0.45, CANUKA: 0.44, Rockall (pre-endoscopy): 0.42, Rockall (complete): 0.42, GBS: 0.5, NLR: 0.53, RDW: 0.56).

Conclusion

Currently available risk stratification scores developed mainly for NVUGIB, when used for VUGIB have a suboptimal performance (AUROC ≤0.8) in predicting important clinical outcomes as mortality, rebleeding, need for endoscopic or interventional radiological procedure. CANUKA was clinically effective and superior to other scoring systems in estimating the need for RBC transfusion. Further prospective studies on larger patient cohorts are needed to further analyze and improve risk stratification scores for VUGIB.

Disclosure

No Conflict of Interest

Introduction

Acute non-variceal upper gastrointestinal bleeding (NVUGIB) is a common emergency with high morbidity and mortality. Lactate levels and comorbidity indices, such as the Charlson Comorbidity Index (CCI), have shown promise in predicting patient outcomes. While these parameters individually provide valuable prognostic insights, their collective utility in predicting mortality among NVUGIB patients has not been explored.

Aims & Methods

The data for this study were collected from hospital records of patients admitted to Hitit University Hospital with NVUGIB between June 2021 and June 2024.
Demographic information, including age, gender, chronic illnesses, current medications, and history of previous gastrointestinal bleeding, was retrieved from computer database. Clinical data, such as outcomes (discharge or death), reasons for mortality, length of hospital stay were documented. Laboratory parameters, including lactate levels were recorded. Additionally, CCI and Glasgow Blatchford Score (GBS) were calculated for each patient.
Patients were stratified based on mortality status, which was considered the primary outcome. The associations between mortality and potential predictive factors were statistically analyzed. To improve mortality prediction, parameters that were significant in the logistic regression analysis were combined and assessed independently.

Results

A total of 390 patients with acute upper gastrointestinal bleeding were initially assessed. Of these, 180 patients were excluded due to missing data, leaving 210 patients included in the final study cohort. Key variables such as age, CCI, GBS, length of hospital stay, and mortality status were presented. Laboratory findings upon admission and the etiological distribution of gastrointestinal bleeding sources were also detailed.
Variables significantly associated with mortality in the univariate analysis include age, CCI, GBS, urea, aspartate aminotransferase (AST), alanine aminotransferase (ALT) and lactate levels. In the logistic regression analysis, significant predictors of mortality were CCI (OR: 1.318, p=0.001) and lactate levels (OR: 1.575, p=0.003). GBS was not a predictor for mortality (p=0.468). CCI plus lactate levels were significantly higher in deceased patients compared to survivors (12.96 vs.7.31, p<0.001).
The correlation coefficients (r) showed significant positive correlations between the combination of lactate levels and CCI with GBS (r=0.498, p<0.001), length of stay (r=0.213, p=0.014), age (r=0.634, p<0.001) and urea levels (r=0.417, p<0.001). No significant correlation was found with AST or ALT.
ROC curves for lactate levels, CCI and CCI plus lactate levels predicting mortality showed that the area under the curve (AUC) for lactate was 0.704 (95% CI: 0.577–0.830, p=0.002) with a cut-off value of 1.83, yielding a sensitivity of 76% and a specificity of 52%. For CCI, the AUC was 0.769 (95% CI: 0.670–0.869, p<0.001) with a cut-off value of 6.5, showing a sensitivity of 76% and a specificity of 70%. And for the ‘CCI plus lactate’ level, the AUC was higher as 0.799 (95% CI: 0.703–0.895, p<0.001) with a cut-off value of 8.3, showing a sensitivity of 88% and a specificity of 62%.

Conclusion

Both lactate levels and the CCI are valuable tools for predicting mortality in NVUGIB patients. Our findings support the use of these parameters collectively to improve risk stratification. Combining lactate levels and CCI offers the advantage of better predicting mortality and improving sensitivity compared to using lactate levels or CCI alone.