Introduction

Liver cirrhosis is a progressive disease characterized by a continuum from compensated to decompensated stages. The development of portal hypertension is the main pathophysiological process in this clinical course. Clinically significant portal hypertension (CSPH) is one of the leading causes of complications, with the portal venous pressure gradient (PPGG) being the most powerful predictive factor for decompensation, showing a negative predictive value of 90% for PPG < 10 mmHg (1). In this context, nonselective beta-blockers (NSBBs) have proven effective in reducing the portal hypertension gradient and preventing variceal bleeding, with a reduction of >10% in HVPG considered clinically significant (2).
In our study, we aim to:
- Assess the safety and efficacy of portal pressure gradient measurement obtained via endoscopic ultrasound in patients diagnosed with liver cirrhosis who are indicated to start primary prophylaxis with non-cardioselective beta-blockers.
- Evaluate the acute hemodynamic response to intravenous propranolol through EUS-PPG measurement in patients diagnosed with liver cirrhosis.
- Identify patients who may benefit from NSBB treatment for primary prophylaxis to prevent decompensation using a strategy guided by EUS-PPG measurement.

Aims & Methods

Non-randomized, prospective, observational, single-center clinical trial including all patients at a tertiary care hospital (Hospital Virgen Macarena) who attend outpatient clinics or are hospitalized with a diagnosis of liver cirrhosis and are candidates for primary prophylaxis with NSBBs according to the Baveno VII guidelines. The following will be performed on all patients:
- Comprehensive laboratory tests: complete blood count, coagulation panel, liver function tests
- Complete Etiological study of liver disease.
- Abdominal ultrasound.
- FibroScan.

Results

Patient	Age	Sex	Etiology	Child-Pugh	MELD	Fibroscan	Platelets	HVPG pre	HVPG post
1	60	Male	Alcohol	B8	12	75	164.000	3 mmHg
2	67	Female	Alcohol	A5	8	19	108.000	3.75 mmHg
3	60	Female	Metabolic	A5	7	29	170.000	10.5 mmHg	4.75 mmHg
4	65	Male	Alcohol	B9	11	No	147.000	10.5 mmHg	2 mmHg

Four patients were included, whose characteristics are summarized in Table 1. Fifty percent were male (n=2), with a mean age of 63 ± 3.96 years. The etiology was alcoholic in 75% (n=3) and metabolic in 25% (n=1), with a mean FibroScan value (available in n=3) of 41 ± 29.87 kPa. Fifty percent (n=2) had decompensated liver function according to the Child-Pugh classification. The mean MELD score was 9.35 ± 2.38. The mean platelet count was 147,500 ± 27,921.

Fifty percent (n=2) had a normal pressure gradient, while the other 50% (n=2) had a gradient compatible with clinically significant portal hypertension (10.5 mmHg). These latter patients received the standard IV propranolol protocol (0.15 mg/kg body weight bolus administered over 10 minutes, with measurement taken 5 minutes later), showing a significant reduction >10% in both cases. No patient experienced complications during or immediately after the procedure.

Conclusion

The clinical efficacy of NSBBs is based on a significant reduction in portal pressure, estimated by measuring the PPG. However, the main limitation of NSBBs is the high interindividual variability in hemodynamic response. Therefore, the development of tools to identify patients most likely to benefit from beta-blocker therapy remains an unmet clinical need.

In this regard, measuring the portal pressure gradient before and after intravenous administration of an NSBB may help overcome these limitations and potentially predict the chronic hemodynamic response. This could serve as a strategy to guide therapy and identify patients who may benefit from it in a single hemodynamic study, substantially reducing the need for repeated measurements.

References

1. T. Reiberger, A. Puspok, M. Schoder, et al. Austrian consensus guidelines on the management and treatment of portal hypertension (Billroth III). Wien Klin Wochenschr, 129 (2017), pp. 135–145.
2. L. Turco, C. Villanueva, V. La Mura, J.C. García-Pagán, T. Reiberger, J. Genescà, et al. Portal pressure reduction improves outcomes in patients with cirrhosis, with or without ascites: a meta-analysis. Clin Gastroenterol Hepatol, 18 (2020), pp. 313–327.

Introduction

Interest in large language models (LLMs) across multiple medical specialities has grown rapidly as a source of patient‑facing advice., No systematic review has appraised the accuracy and readability of LLM‑generated responses for gastrointestinal (GI) procedures, resulting in a lack of current evidence on the reliability of these tools.

Aims & Methods

We aimed to quantify the accuracy and readability of LLM responses to both patient‑ and clinician‑facing questions on GI investigations and interventions, and to identify limitations hindering clinical use. We searched PubMed, Ovid MEDLINE, Ovid Embase, and Google Scholar for peer‑reviewed studies published from 2023 to 2025 that evaluated text-based answers from any LLM on at least one GI diagnostic or therapeutic procedure (colonoscopy, endoscopy, EGD, ERCP, EUS, PEG, GI surgery and others). Inclusion criteria were English language, full‑text primary studies reporting at least one accuracy or readability metric; exclusion criteria included studies with a non-GI focus, image-only tasks, reviews, abstracts, or missing data.. Multiple reviewers independently screened records and extracted data, resolving disagreements by consensus. Risk of bias was appraised using the NIH Quality Assessment Tool (NIH‑QAT) for observational cross‑sectional studies [1]. Accuracy was quantified using a binary (accurate/inaccurate), three-point (accurate, partially accurate, inaccurate), or numerical 1-5 point scales. Readability was assessed using a five‑point Likert scale.

Results

Of the 409 records screened, 45 full‑text articles were reviewed, and 14 met the inclusion criteria, covering seven LLMs. Accuracy varied across models: on the binary scale, Google Bard performed best (82 %, n = 11), followed by Microsoft Copilot (60 %, n = 10). On the three‑point scale, Claude performed best (2.44 / 3, n = 124), with ChatGPT second (2.15 / 3, n = 124). On the five‑point scale, ChatGPT ranked first (3.6 / 5, n = 308), and Claude next (3.04 / 5, n = 16).

By procedure, colonoscopy scored highest on binary (73 %, n = 30) and three‑point (2.73 / 3, n = 14) scales, whereas endoscopy topped the five‑point scale (4.01 / 5, n = 32). Surgery showed mixed performance — binary 59 % (n = 141), three‑point 1.93 / 3 (n = 105), and five‑point 3.54 / 5 (n = 244). ERCP, EUS, PEG, and EGD were represented by smaller samples, giving variable three‑point scores (ERCP 2.66, EUS 2.56, EGD 2.22, PEG 1.83).

Conclusion

LLMs can provide readable answers on GI procedures, but none reach the ≥ 95 % accuracy typically expected for clinical use, with performance still inconsistent across procedures. Standardised benchmarking, balanced prompt libraries, and updated model evaluations are needed before LLMs can be safely used in clinical gastroenterology.

References

[1] National Heart, Lung, and Blood Institute. Study Quality Assessment Tools [https://www. nhlbi. nih. gov/health-topics/study-quality-assessment-tools].

Introduction

Gallstone disease affects approximately 6% of the global population. Adequate knowledge about gallstone disease among the public can facilitate early medical consultation and enable them to navigate healthcare pathways effectively. In the digital age, patients increasingly turn to the internet for health-related information, with over half of the global population relying on it for access to knowledge.

Aims & Methods

This study aimed to evaluate the quality, readability, usability, and reliability of patient-oriented websites providing information on gallstone disease, using validated assessment tools. The three most widely used search engines, Google, Yahoo, and Bing were searched employing keywords related to gallstone disease. The first 50 search results per keyword per engine were screened. Applying the inclusion and exclusion criteria, 52 websites were selected for detailed analysis. The information was assessed for readability using the Flesch Reading Ease Score (FRES), for usability and reliability using the LIDA tool, and for overall quality using the DISCERN instrument. Statistical analysis was performed using SPSS (v27), with significance set at p < 0.05.

Results

Among the 52 websites analyzed, only sixteen (30.8%) sites had obtained Health on the Net (HON) code certification. The median FRES score of the included websites was 60.5 out of 100 (range: 33-79.6), indicating readability appropriate for readers aged 13–15 years. The median LIDA usability and reliability scores were 40 out of 54 (range: 31–54) and 24.5 out of 51 (range: 10–51), respectively. The median DISCERN score was 69.8 out of 80 (range: 30-80) with 5.8% (n=3) of websites scoring below 50%. A low DISCERN score (≤ 50%) was found in three (5.8%) websites. The DISCERN score was found to be significantly associated with LIDA reliability and LIDA overall scores (p < 0.001), government run websites (p= 0.028) and HON code certification (p = 0.01). However, no significant association was found between DISCERN scores and LIDA usability or FRES. Most websites (86.5%) provided accurate clinical information, but only 76.9% discussed treatment complications.

Conclusion

The quality of online patient information on gallstone disease varied considerably, with only a minority meeting optimal standards for readability and certification. While clinical accuracy was generally satisfactory, gaps remain in presenting treatment risks and maintaining content currency. These findings emphasize the need for improved oversight and standardization of online health content to support patient education and informed decision-making.

References

1. Internet World Stats. Internet world stats: Usage and population statistics, http://www.internetworldstats.com/stats.htm (2013, accessed 16 June 2023).
2. Lee JS, Seo HS, Hong TH. YouTube as a source of patient information on gallstone disease. World J Gastroenterol 2014; 20: 4066–4070.
3. Raptis DA, Sinanyan M, Ghani S, et al. Quality assessment of patient information on the management of gallstone disease in the internet. HPB 2019; 21: 1632–1640.
4. European Association for the Study of the Liver. EASL Clinical Practice Guidelines on the prevention, diagnosis and treatment of gallstones. J Hepatol 2016; 65: 146–181.

Disclosure

There is no conflict of interest

Introduction

Artificial intelligence (AI) is increasingly integrated into medical decision-making and education, particularly through large language models such as ChatGPT. In gastroenterology, where clinical reasoning relies heavily on imaging, biological interpretation, and complex decision processes, the effective implementation of AI tools remains poorly documented in real-world settings.

Aims & Methods

This study aimed to compare how gastroenterologists and physicians from other specialties perceive the usefulness and practical applications of AI in gastroenterology, with a focus on diagnostic, therapeutic, and educational support.
A cross-sectional survey was distributed to physicians practicing in various hospital settings in March 2025. Participants were grouped as gastroenterologists or non-gastroenterologists. The questionnaire assessed demographics, clinical exposure to digestive diseases, prior AI usage, and perceived utility of AI in gastroenterology. Statistical comparisons were conducted using the Chi-squared test, with significance set at p<0.05.

Results

Gastroenterologists were significantly more likely to have used AI for gastroenterology-specific clinical questions (64% vs. 41% among other physicians). While 96% of all respondents acknowledged the potential usefulness of AI in gastroenterology, gastroenterologists expressed greater caution: 60% described AI as useful but with limitations, 36% viewed it as a promising tool, and 4% saw no particular benefit. Similar distributions were observed among other physicians (56%, 42%, and 2%, respectively), with no statistically significant difference in overall perception (p = 0.77).
However, the perceived domains of utility differed significantly between the groups (p = 0.047). Gastroenterologists most often highlighted the role of AI in continuing medical education (24.5%), diagnostic support (20.8%), and public education (15.1%). In contrast, other physicians emphasized diagnostic assistance (21.9%), pedagogical support (23.5%), and notably, therapeutic decision-making (13.8%), which was less frequently cited by gastroenterologists (7.5%).

Conclusion

Gastroenterologists—although more experienced with AI in their field—tend to adopt a more cautious and targeted perspective on its clinical integration. They see it as a complementary tool for interpretation and professional development, rather than a primary decision-making assistant. Meanwhile, non-specialist physicians show a slightly broader enthusiasm, seeing AI as helpful both in clinical decisions and educational support. While both groups recognize the potential of AI in gastroenterology, its successful implementation will require tailoring to the specific expectations, workflows, and trust levels of each medical discipline.

Introduction

Endoscopic scores are crucial for assessing Crohn's disease (CD) severity, guiding therapy, and monitoring progression. The Simplified Endoscopic Score for Crohn's Disease (SES-CD) is widely used for this purpose [1]. Accurate scoring is vital for patient care. Recently, artificial intelligence has demonstrated potential in medical image interpretation [2]. This study aims to evaluate ChatGPT's effectiveness in rating the SES-CD.

Aims & Methods

This retrospective study evaluated ChatGPT's accuracy in assessing CD endoscopic activity using the SES-CD. One representative still image was selected for each bowel segment, and stricture degrees were provided separately if present. ChatGPT scored the SES-CD based on these images. The IBD expert's ratings were validated by calculating the intra-class correlation coefficient (ICC) with clinical trial central readers. The ICC between the IBD expert and ChatGPT scores was calculated to determine ChatGPT's reliability in SES-CD scoring.

Results

Nine (mean age 32.0 ± 11.8 years, 77.8% male) were included, with 45 bowel segments. Three (33.4%), 5 (55.6%), and 1 (11.1%) had nonstricturing nonpenetrating, structuring, and penetrating disease, respectively. Four (44.4%) had ileal disease and 5 (55.6%) had ileocolonic disease. Five (55.6%) had previous IBD surgery.
The SES-CD scores of the participants, rated by IBD expert in the hospital, revealed the following mean and standard deviation for each segment of the bowel: 6.9 ± 2.3 in the ileum, 1.3 ± 2.6 in the right colon, 1.6 ± 2.4 in the transverse colon, 1.3 ± 2.6 in the left colon, and 1.3 ± 2.6 in the rectum. The total SES-CD score was 12.4 ± 10.9. The score reliability of the institute IBD expert was tested compared to the clinical trial central reader and an intraclass correlation coefficient was calculated and found to be 0.96 (95% CI, 0.84-0.99) which is considered extremely correlated.
The mean and standard deviation of ChatGPT score in each segment were: 6.4 ± 2.4 in the ileum, 2.4 ± 3.8 in the right colon, 0.9 ± 2.7 in the transverse colon, 0.3 ± 1.0 in the left colon, 0.6 ± 1.3 in the rectum, with total SES-CD score of 10.6 ± 6.8. The calculated ICC between the expert and ChatGPT obtained SES-CD scores was 0.80 (95% IC, 0.64- 0.89) which is considered highly correlated.

Conclusion

Despite being based on only one photo per segment, ChatGPT demonstrated satisfactory performance in estimating endoscopic activity in CD. The differences between the IBD expert obtained scores and ChatGPT scores were acceptable, suggesting that artificial intelligence has the potential to be a useful tool in clinical and research settings, providing an accurate estimation of disease severity based on limited visual data.

References

1- Daperno M, D'Haens G, Van Assche G, et al. Development and validation of a new, simplified endoscopic activity score for Crohn's disease: the SES-CD. Gastrointest Endosc. 2004;60(4):505-512.
2- Levartovsky A, Albshesh A, Grinman A, et al. Enhancing diagnostics: ChatGPT-4 performance in ulcerative colitis endoscopic assessment. Endosc Int Open. 2025;13:a25420943. Published 2025 Mar 14.

Introduction

Current therapeutic strategies for inflammatory bowel disease (IBD) remain limited, with clinical efficacy often falling short of expectations. Post-translational modifications play a pivotal role in IBD pathogenesis. This study aims to elucidate the functional significance of ubiquitination in IBD development and progression.

Aims & Methods

By analyzing colon RNA sequencing (RNA-seq) data from healthy and dextran sulfate sodium (DSS)-induced colitis mice, combined with RNA-seq datasets from healthy individuals and ulcerative colitis (UC) patients, we identified differential RNA expression levels of ATDC between healthy and inflammatory groups. Single-cell RNA sequencing (scRNA-seq) analysis further pinpointed the primary cell types exhibiting ATDC expression discrepancies. These findings were validated through immunohistochemistry, immunofluorescence, and western blot. Co-immunoprecipitation (Co-IP) assays and molecular docking simulations were employed to confirm and model the interaction between ATDC and PRMT5 in vitro. The role of ATDC in IBD pathogenesis was investigated using an ATDC-knockout (ATDC-KO) murine DSS model, with flow cytometry analyses delineating its impact on immune status.

Results

RNA-seq analysis revealed elevated ATDC mRNA levels in DSS-treated mice and UC patient intestinal tissues. scRNA-seq data indicated ATDC upregulation predominantly in the intestinal epithelial cells of UC patients. Protein-level upregulation of ATDC was further confirmed in both DSS mice and UC patient specimens. Lipopolysaccharide stimulation increased ATDC expression in Caco2 cells. Overexpression of ATDC in Caco2 cells, coupled with mass spectrometry, identified ATDC-PRMT5 binding, which was corroborated by Co-IP assays and molecular docking simulations. ATDC overexpression downregulated PRMT5 protein (but not mRNA) levels. In DSS models, PRMT5 expression was reduced compared to healthy controls; however, ATDC-KO mice exhibited PRMT5 upregulation and exacerbated colitis severity. Flow cytometry demonstrated an elevated Th17 cell proportion in ATDC-KO DSS mice relative to wild-type counterparts.

Conclusion

This study is the first to demonstrate the protective role of ATDC in colitis. ATDC attenuates colonic inflammation by promoting the degradation of PRMT5, thereby suppressing Th17 cell-mediated immune responses. Enhancing ATDC activity may represent a novel therapeutic strategy for mitigating colitis progression.

References

-

Introduction

In recent years, both the incidence and early-stage diagnosis of pancreatic neuroendocrine tumors (pNETs) have increased, likely due to advancements in cross-sectional imaging techniques (1). Management strategies are size-dependent, with observation generally preferred for lesions <10 mm and surgical resection for those >20 mm. However, there is no consensus regarding the optimal approach for tumors between 10 and 20 mm, highlighting the need for multidisciplinary decision-making (2). In this setting, endoscopic ultrasound (EUS) with fine-needle aspiration (FNA) or fine-needle biopsy (FNB) plays a crucial role not only in diagnosis but also in tumor grading (3). However, samples obtained through EUS-FNA or FNB may not provide sufficient histological material to determine tumor grade in pNETs smaller than 20 mm. This study aims to summarize the current evidence regarding the adequacy of histological samples obtained via EUS-FNA/FNB for tumor grading.

Aims & Methods

This study followed the PRISMA guidelines for Systematic Reviews and Meta-Analyses. MEDLINE, EMBASE, and SCOPUS databases were searched up to December 2024 for studies reporting on the sample adequacy of EUS-FNA or FNB for grading pNETs ≤ 20 mm. Sample adequacy was defined as the proportion of samples in which the Ki-67 index was successfully obtained for tumor grading among all samples collected. Pooled estimates were calculated using random-effects models. Heterogeneity was assessed using the I² statistic. Subgroup analyses were conducted based on needle type (FNA vs. FNB).

Results

17 studies were included, analyzing a total of 1414 patients. The pooled adequacy rate was 71.6% (95% CI: 0.592–0.841), with substantial heterogeneity across studies (I² = 98.69%). The risk of publication bias was considered low based on Egger’s test (p = 0.162). When considering only studies that used FNA, adequacy rates were lower, with a pooled estimate of 52.5% (95% CI: 0.302–0.748; I² = 97.99%). In contrast, studies employing FNB showed higher adequacy rates, with a pooled estimate of 79.8% (95% CI: 0.605–0.991), although heterogeneity remained high (I² = 87.32%). In this subgroup, there was also evidence of publication bias, as suggested by Egger’s test (p = 0.017) and visual asymmetry in the funnel plot.

Conclusion

Our study found that EUS-FNA/FNB provides adequate histological specimens for tumor grading in pNETs < 20 mm, although significant heterogeneity was observed. As expected, FNB demonstrated considerably better performance than FNA in establishing tumor grade, suggesting it should be preferred for pNET sampling. However, even with FNB, tumor grade could not be determined in up to 20% of cases, indicating that it remains a suboptimal tool for grading and may offer limited additional value beyond diagnostic confirmation in some patients. Notably, our study did not distinguish between first-, second-, and third-generation FNB needles; therefore, results from newer-generation FNBs may show improved sample adequacy rates. Furthermore, the observed heterogeneity highlights the need for studies specifically evaluating the ability of EUS-FNB to obtain high-quality tissue for grading, thereby supporting the role of EUS in clinical decision-making for these small tumors.

References

1. Souche R, Hobeika C, Hain E, et al. Surgical Management of Neuroendocrine Tumours of the Pancreas. J Clin Med 2020;9.
2. Mansour JC, Chavin K, Morris-Stiff G, et al. Management of asymptomatic, well-differentiated PNETs: results of the Delphi consensus process of the Americas Hepato-Pancreato-Biliary Association. HPB (Oxford) 2019;21:515-523.
3. Franchellucci G, Andreozzi M, Carrara S, et al. Contrast Enhanced EUS for Predicting Solid Pancreatic Neuroendocrine Tumor Grade and Aggressiveness. Diagnostics (Basel) 2023;13.