Introduction

Confocal laser endomicroscopy (CLE) enables real-time microscopic visualization of the duodenal mucosa during upper endoscopy and has shown acute food-triggered disruption of the epithelial barrier in the duodenum of irritable bowel syndrome (IBS) patients. While this technique offers novel scientific insights, image interpretation is subjective and agreement rates between observers are unknown.

Aims & Methods

We aimed to assess the agreement rate between assessors regarding acute food induced mucosal changes visualized by CLE in IBS. To this end we performed a multicentric, international post-hoc evaluation of CLE recordings. CLE sequences from three centers were blindly assessed. Five investigators from the three participating centers blindly assessed all videos both in real-time and in slow-motion in a random order, including some duplicates.
Mucosa was considered altered if fluorescein leakage AND “cell shedding” were simultaneously present. Outcomes of the global assessment were assessed using the binary outcome “altered” vs. “unaltered”. Agreement was quantified using Fleiss’ kappa for >2 assessors and Cohens’ kappa for 2 assessors – both are reported as κ.

Results

A total of 124 video sequences (94572 frames, 197 minutes of real-time recording) were reviewed. Intraindividual agreement rate was between 90 and 100% (κ=0.74-1.00). Overall interobserver agreement rate for all assessors and all videos was 86% (κ=0.69). Agreement rates were not higher when videos were assessed in slow-motion (81%, κ=0.58). Among the 3 centers, agreement between all blinded assessors and real-time assessment by the endoscopist ranged from 78% (κ=0.55) to 86% (κ=0.67). A total of 63 videos were judged the same by all assessors (100% agreement) – here agreement rates with the endoscopist were 97% (κ=0.91). Using these videos as reference, a cutoff of 2 altered spots in a recording of 5 spots showed a sensitivity and specificity of 100% for the presence or absence of alterations.

Conclusion

We report here for the first time the interobserver agreement rates of acute food induced mucosal alterations as visualized on CLE. The agreement rates were substantial overall – both among blinded assessors and between blinded assessors and real-time judgement by the endoscopist. Slow-motion revision did not increase agreement rates. A cutoff of 2/5 spots was ideal for the detection of acute mucosal alterations.

Introduction

Accurate estimation of cancer risk in patients with Barrett’s esophagus (BE) and low-grade dysplasia (LGD) or indefinite for dysplasia (ID) is difficult, leading to potential under- and over-treatment. We aimed to evaluate the utility of machine learning (ML) and copy number alterations (CNAs) assessed by shallow whole genome sequencing (sWGS) on standard formalin-fixed paraffin-embedded (FFPE) samples for precise risk stratification in this group of BE patients.

Aims & Methods

We conducted a retrospective analysis on a cohort of 164 BE patients encompassing 990 tissue samples with sWGS (~0.4X) data from 2001-2018. These included 215 samples with LGD or ID from 64 patients, of which 38 progressed to high-grade dysplasia (HGD) or cancer at later timepoints (P), and 26 who did not progress (NP). CNAs were quantified using QDNASeq, and outlier bins were determined by Tukey's fences. Subsequently, the CNAs of 64 esophageal cancer driver genes were compared between P and NP groups using t-tests and ranked by p-values. A random forest (RF) ML model was developed on the CNAs of the 64 genes, employing nested 5-fold cross-validation split by patients and iterated 100 times to mitigate bias. The model's performance was compared against routine p53 immunohistochemistry (IHC), a decision tree that incorporates p53 followed by the ML model, and the current ESGE clinical guideline recommending treatment after LGD is identified in two endoscopies. Both the model and model+p53 were assessed across 100 test sets derived from the cross-validation. The guideline's performance was evaluated using a historical extended cohort comprising 253 LGD patients, of which 32 progressed, and underwent at least one surveillance endoscopy after LGD diagnosis and before HGD/cancer diagnosis. According to the guideline, two occurrences of LGD result in a positive prediction.

Results

Significant variance was observed in the distribution of relative CNAs across genomic bins between P- and NP-LGD/ID samples, with the proportions of outliers at 2.15% (95% CI: 1.88%, 2.42%) vs. 1.35% (95% CI: 1.26%, 1.44%). The most divergent genes included TP53, ABCB1, MET, BRAF, and SIN3A, exhibiting higher CNAs in progressed samples. Among the 215 samples, the RF model achieved an average sensitivity of 73.4% and specificity of 80.2%. Of these, 160 samples were tested for p53 IHC staining, where p53 alone showed a sensitivity of 22.1% and specificity of 100%, and a combined analysis of p53 and model increased the sensitivity of model alone (81.5% vs. 78.1%, p<0.001) while maintaining the same specificity at 78.9%. As a benchmark, the performance of the current ESGE guideline on biopsy diagnosed LGD patients demonstrated a sensitivity of 75.0% and specificity of 59.6%.

Conclusion

Our findings affirm the utility of ML and CNAs in predicting progression in BE patients with LGD or ID. Incorporating p53 followed by ML can further improve the performance. Molecular biomarkers would allow anticipation of intervention in high-risk cases and reduction of overtreatment in low-risk cases.

Disclosure

Nothing to disclosure.

Introduction

Oesophageal ulceration (OU) is a frequent finding at oesophago-gastro-duodenoscopy (OGD). Despite limited data on the benefit of a repeat endoscopy, current guidelines recommend a follow-up OGD after six weeks to exclude underlying malignancy¹.

Aims & Methods

Aims: We aimed to explore the utility of a follow-up endoscopy after a diagnosis of oesophageal ulcer and identify predictors of oesophageal malignancy and dysplasia. Secondary aims were to estimate the environmental and financial impact of a follow-up endoscopy.
Methods: We collected demographic, endoscopy, and histology data from patients diagnosed with OU over a 5-year period (from 1^st January 2016 to 31^st December 2020 inclusive) at Leeds Teaching Hospitals NHS Trust. We collected data from the index OGD, follow-up OGD, and from their medical records up to 3 years after a diagnosis of OU at their index OGD. We examined the proportion of patients who were diagnosed with dysplasia or malignancy at follow-up or up to 3 years after their index OGD. Binary logistic regression was performed to identify predictors of oesophageal malignancy or dysplasia at index OGD, follow-up OGD, or 3 years after the index OGD. NHS national cost data for OGD, and carbon dioxide (CO2) emissions were estimated for all unnecessary follow-up OGDs^2,3.

Results

In total, 409 patients were diagnosed with OU. Of these, 33 patients were diagnosed with oesophageal cancer or dysplasia, with 30 diagnoses (90.9%) made at the index OGD and the remaining 3 (9.1%) at either the follow-up OGD or within 3 years of the index OGD. Of those 3 patients, one was diagnosed with oesophageal carcinoma, one with low grade dysplasia, and one with high grade dysplasia, all within a segment of Barrett’s oesophagus already diagnosed at the index OGD. Those who were diagnosed with oesophageal malignancy or dysplasia, compared to those who were not, were more likely to have larger ulcers (mean size (SD), mm, 15.9 (17.2) vs 9.50 (11.1), P = 0.003), endoscopic features suspicious for malignancy (69.7% vs 8.0%, P < 0.001), Barrett’s oesophagus (54.5% vs 14.4%, P < 0.001), and biopsies taken at index OGD (100% vs 64.1%, P < 0.001), and less likely to have a proton pump inhibitor prescribed (57.6% vs 86.4%, P < 0.001) after the index OGD. Following logistic regression controlling for all data, those with endoscopic features suspicious for malignancy (OR = 45.9; 95% CI 10.3 to 204) and those with Barrett’s oesophagus at index OGD (OR = 5.29; 95% CI 1.56 to 18.0) were more likely to be diagnosed with oesophageal cancer or dysplasia up to 3 years after the index OGD. A total of 376 unnecessary OGDs were undertaken over a 5-year period, the costs of which were estimated at £140, 248 and 10.6 tonnes of CO2 emissions.

Conclusion

Repeat OGD to exclude underlying malignancy following a diagnosis of oesophageal ulcer may only be required in those with endoscopic features suspicious for malignancy and those with Barrett’s oesophagus. Larger prospective multi-centre studies, with cost and carbon footprint analyses, are required to confirm our results and to inform guidelines.

References

1. Beg S. et al. (2017). Quality standards in upper gastrointestinal endoscopy: a position statement of the British Society of Gastroenterology (BSG) and Association of Upper Gastrointestinal Surgeons of Great Britain and Ireland (AUGIS), Gut, 66(11):1886-1899. DOI: 10.1136/gutjnl-2017-314109
2. NHS (2023): National tariff payment system documents, annexes and supporting documents. Available at: https://www.england.nhs.uk/publication/national-tariff-payment-system-documents-annexes-and-supporting-documents/
3. Lacroute, J. et al. (2023), The carbon footprint of ambulatory gastrointestinal endoscopy, Endoscopy, 55(10): 918–926. DOI: 10.1055/a-2088-4062

Introduction

Psychological therapies have been shown to be effective for treatment of patients with disorders of gut-brain interaction like irritable bowel syndrome (IBS). However, these therapies are time and personnel demanding which limits the access to this therapeutic option for many patients. Virtual reality (VR) creates an immersive environment where psychological interventions can be self-administered by the patient at home, thereby facilitating the access of these interventions to a large number of patients with minimal costs.

Aims & Methods

Our aim was to evaluate the effect of a psychological intervention delivered using VR on digestive symptoms in patients with IBS. Thirty-six patients with IBS (mean age 42±2 years, 89 % female) were randomized to receive a VR-based psychological intervention or placebo, using a randomized, single blind, parallel study design. The psychological intervention consisted in a 6-week program that started with a session of psychoeducation using a specifically created VR-video, followed by weekly sessions of group psychotherapy applied on-line by a psychologist. During the 6 treatment weeks, patients were asked to perform 15-30-min VR-sessions at home every day following a pre-designed protocol that incorporated progressive psychological interventions using VR-videos developed by a specialized company (XRHealth, Massachusetts, USA). The placebo group took a pill containing 0.3 g of glucose once daily. The primary outcome was the improvement of gastrointestinal symptoms (assessed by IBS-SSS) at the end of the intervention (6 weeks). Secondary outcomes included psychological comorbidities using the HADS (Hospital Anxiety and Depression Scale) and PSS (Perceived Stress Scale) questionnaires.

Results

Twenty-one patients were randomized to VR and fifteen patients to placebo. After 6 weeks of treatment, patients receiving the psychological intervention had a significantly greater reduction in IBS-SSS scores (score change -39±12) compared to patients on placebo (score change 1±19; p= 0.009). When analyzing the specific symptoms, abdominal pain had a significant reduction after psychological intervention (score reduction -8±3; p=0.015) and a tendency was identified for abdominal bloating (score reduction -6±5; p=0.097). Both symptoms remained unchanged after placebo (score change 1±6; p=0.719 and -0±6; p=0.695, pain and distension, respectively).
In addition, psychological intervention was followed by a greater improvement than placebo in HADS scores for anxiety (score change -2.35 ± 0.56; p=0.041 vs score change 0.00 ± 0.85 after placebo) and depression (score change -1.40± 0.71; p=0.005 vs score change 1.36 ± 0.50 after placebo). Stress evaluation using PSS also showed a greater reduction in stress perception in patients on psychological intervention (score change -5.15 ± 1.50) compared with patients on placebo 0.07 ± 1.35 (p=0.023).

Conclusion

A psychological intervention delivered at home using virtual reality is superior to placebo in relieving IBS symptoms, an effect that is associated to a reduction in anxiety, depression and stress in patients with IBS. These findings open the door to development of treatments in patients own home, and should facilitate the access to these treatments in large populations of patients. NTC06318572

Disclosure

No potential conflict of interest of any author

Introduction

Aims & Methods

We conducted a retrospective analysis of 402 medical data from patients who were hospitalized with AP at our gastroenterology department of a university hospital from January 2018 to December 2023. The study included patients who underwent abdominal and pelvic computed tomography (CT) and our objective was to determine the prevalence of DVT in hospitalized AP patients and evaluate its influence on morbidity and mortality.

Results

Among 402 patients with acute pancreatitis (AP), 5.5% (n=22) developed DVT, with Portal vein thrombosis in 54.5%, Splenic vein thrombosis in 32%, and Superior mesenteric vein thrombosis in 13.6%. The mean age among patients with DVT was 58 ± 22 years compared to 56 ± 18 years (P=0.03). DVT incidence was significantly higher among patients with a Computed Tomography Severity Index score of ≥ 3 (63.6%, P=0.02). Among patients with pleural effusion, 13.7% without thrombosis and 31.8% with thrombosis were observed (P=0.02). An association was found between Systemic Inflammatory Response Syndrome (SIRS) and thrombosis, with 16.1% of patients with persistent SIRS not developing thrombosis compared to 40.9% with thrombosis (P=0.006). Similarly, among patients with a Bedside Index for Severity score > 2, 15.0% without thrombosis and 31.8% with thrombosis were observed (P=0.03). Among patients with WON, 6.9% without thrombosis and 35.0% with thrombosis were observed (P<0.001). The mean CRP level was significantly higher in patients with thrombosis (187 ± 100 vs 130 ± 107 mg/dl, P=0.02). Hypoalbuminemia was significantly more frequent in cases of thrombosis (P=0.002). During follow-up, among patients with pancreatic insufficiency, 0.9% without thrombosis and 10.0% with thrombosis were observed (P=0.001). However, no significant association was observed between thrombosis and mortality rate.

Conclusion

In summary, our study reveals the prevalence of DVT in AP patients, emphasizing its clinical significance. Factors such as higher severity scores and association with systemic inflammation underscore the importance of proactive management to mitigate DVT risk and enhance patient outcomes.