BACKGROUND: The comprehensive evaluation of multimodal large language models (LLMs) in gastroenterological image-analysis remains limited. We aimed to assess the accuracy of the Eosinophilic Esophagitis Endoscopic Refere...BACKGROUND: The comprehensive evaluation of multimodal large language models (LLMs) in gastroenterological image-analysis remains limited. We aimed to assess the accuracy of the Eosinophilic Esophagitis Endoscopic Reference Score (EREFS) scoring system for Eosinophilic Esophagitis (EoE) across gastroenterology (GI) clinicians with varying levels of experience. METHODS: Fifty real-world endoscopic images of EoE patients were graded based on the original EREFS score by a gold standard anchor-rater. EREFS gradings of GI specialists, fellows and three multimodal LLMs (ChatGPT-4o, Claude Sonnet 3.5, Perplexity Sonar) were compared with the anchor-rater's scoring. LLMs were provided with both single-shot and few-shot prompting strategies to optimize performance. RESULTS: Overall assessment accuracy was significantly higher among GI fellows (72.4 %) compared to specialists (65.3 %, p = 0.004) and LLMs (58.9 %, p < 0.001). In the detection of edema, LLMs outperformed specialists (83.3 % vs 49.3 %, p < 0.001). However, LLMs showed significantly poor performance in rings assessment (30 %) compared to specialists (58 %) and fellows (58.7 %, both p < 0.001). After implementation of few-shot prompting, the overall performance of LLMs was comparable to GI specialists (62.7 % vs 65.3 %, p = 0.3). CONCLUSIONS: This study uncovers variability in EREFS scoring across human raters with different expertise and multimodal LLMs, and demonstrates that few-shot prompting can optimize LLM accuracy.
BACKGROUND AND AIMS: We performed a randomized, double-blind, placebo-controlled, trial to investigate the changes in microbiome composition induced by Butyrate-Lsc-Microincapsulated (BLM) supplementation in IBD patients...BACKGROUND AND AIMS: We performed a randomized, double-blind, placebo-controlled, trial to investigate the changes in microbiome composition induced by Butyrate-Lsc-Microincapsulated (BLM) supplementation in IBD patients and its impact on disease activity. METHODS: 140 IBD patients (n=60 Crohn's disease, CD and n=80 Ulcerative Colitis, UC) were randomized to oral administration of BLM, plus conventional therapy. Stool samples were assessed by 16S sequencing and fecal calprotectin (fCal) analysis. For the microbiota analysis, the Firmicutes/Bacteroidota (F/B) ratio was used. Clinical disease activity was assessed by using the Harvey-Bradshaw-Index (HBI) for CD and partial-Mayo-Score for UC, Quality-of-life (QoL) by using Inflammatory-Bowel-Disease-Questionnaire-32 (IBDQ) and adherence-dietary-recommendation was evaluated before and after supplementation RESULTS: microbiota analysis revealed two principal enterotypes, defined by the F/B ratio, in both CD and UC patients. BLM exerted a more pronounced effect on Enterotype 1 (low F/B ratio), resulting in greater clinical and biochemical improvements and potentially identifying a target population. After supplementation, clinical disease activity (p=0.013) and fCal (p=0.047) improved significantly in CD, while fCal showed a marginal reduction in UC (p=0.09). QoL increased significantly in both CD (p<0.001) and UC (p=0.003). CONCLUSIONS: Supplementation with BLM, by modulating the gut microbiota, significantly improved disease outcomes and QoL in patients with IBD. GOV REGISTRATION: NCT04879914.
Raso T, D'Arcangelo G, Renzo S
… +10 more, Strisciuglio C, Colucci A, Saccomani MD, Bramuzzo M, Bravin F, Sansotta N, Russo G, Lionetti P, Zullo A, Oliva S
BACKGROUND: Helicobacter pylori (H. pylori) infection remains prevalent in children, with significant clinical implications. While endoscopy with biopsy is the gold standard for diagnosis, non-invasive tests such as the...BACKGROUND: Helicobacter pylori (H. pylori) infection remains prevalent in children, with significant clinical implications. While endoscopy with biopsy is the gold standard for diagnosis, non-invasive tests such as the stool antigen test (SAT) and urea breath test (UBT) may offer alternatives. OBJECTIVES: To assess the diagnostic accuracy of SAT and UBT in children with suspected H. pylori infection and identify clinical predictors of infection. METHODS: This retrospective multicenter study included pediatric patients undergoing endoscopy for suspected H. pylori across six Italian centers. Histological analysis served as the reference standard. Diagnostic metrics of SAT and UBT were calculated. Demographic and clinical factors were analyzed to identify independent predictors. RESULTS: Of 256 patients, 150 (58.6 %) had confirmed infection. SAT showed higher sensitivity [94 % (95 % CI: 0.87-0.97)] than UBT [87 % (CI: 0.64-0.98)] but lower specificity [55 % vs 67 %], with lower PPV (64 % vs 78 %) and higher NPV (91 % vs 80 %). Independent predictors for H. pylori infection included family history [OR 4.4], positive SAT [OR 16.29], and non-Caucasian ethnicity [OR 4.3]. CONCLUSIONS: SAT demonstrates high sensitivity and NPV, supporting its role as a screening tool. In children without alarm symptoms, a negative SAT may safely exclude infection and help avoid unnecessary endoscopy.
Flagiello V, Gallo P, Terracciani F
… +10 more, Falcomatà A, De Vincentis A, Pasquale GD, Avagliano J, Tavaglione F, Sparapano S, Daniels SJ, Blau JE, Picardi A, Vespasiani-Gentilucci U
There is increasing evidence that visceral adipose tissue (VAT) plays a key role in the pathogenesis of metabolic dysfunction-associated steatotic liver disease (MASLD), yet its clinical assessment remains limited by the...There is increasing evidence that visceral adipose tissue (VAT) plays a key role in the pathogenesis of metabolic dysfunction-associated steatotic liver disease (MASLD), yet its clinical assessment remains limited by the need for advanced imaging. In this cross-sectional study of 747 patients at metabolic risk, we evaluated whether ultrasound (US)-measured VAT is more strongly associated with hepatic steatosis and metabolic dysfunction than conventional anthropometric indices. Hepatic fat was quantified via ultrasound-derived fat fraction (UDFF), while liver stiffness and FibroScan-AST (FAST) score assessed disease severity. Stratification by VAT tertiles showed a progressive increase in UDFF and the prevalence of diabetes, hypertension, and dyslipidaemia. VAT correlated more strongly than body mass index (BMI) and waist circumference (WC) with triglycerides and UDFF (p < 0.05), and was more strongly associated with HbA1c, ALT and FAST than BMI. In standardized multivariate analysis, VAT remained independently associated with UDFF (beta= 2.49, p < 0.001). Associations of VAT with UDFF and metabolic parameters were stronger in women than in men. These findings support the use of US-derived VAT as a non-invasive and low-cost biomarker to improve risk stratification beyond BMI and WC, serving as a practical tool to monitor response to therapies targeting visceral fat, in MASLD management.
Dell'Anna G, Crinò SF, Fuccio L
… +14 more, Maida MF, Cotsoglou C, Dinelli ME, Spadaccini M, Boskoski I, Machicado JD, Samanta J, Dhar J, Spampinato M, Stasi E, Dell'Anna A, Donatelli G, Almuhaidb A, Facciorusso A
BACKGROUND: It is unclear which is the best strategy for difficult biliary cannulation in endoscopic retrograde cholangiopancreatography (ERCP). AIMS: We compared the different techniques, through a network meta-analysis...BACKGROUND: It is unclear which is the best strategy for difficult biliary cannulation in endoscopic retrograde cholangiopancreatography (ERCP). AIMS: We compared the different techniques, through a network meta-analysis combining direct and indirect comparisons. METHODS: We identified 12 randomized controlled trials (1605 patients) comparing different techniques for difficult biliary cannulation (early and late needle knife techniques, pancreatic guidewire- and stent-assisted techniques, transpancreatic sphincterotomy, and endoscopic ultrasound rendez-vous [EUS-RV]) either with each other or with persistence with the standard cannulation techniques. The success rate of biliary cannulation and the incidence of post-ERCP pancreatitis (PEP) were the primary outcomes. RESULTS: Only transpancreatic sphincterotomy significantly outperformed pancreatic stent assisted cannulation (risk ratio [RR] 1.33, 1.00-1.55), whereas no difference was observed among the other techniques in terms of cannulation success. SUCRA ranking suggested EUS-RV and early needle knife techniques as the best performing approaches (SUCRA 0.78 and 0.68, respectively). In terms of PEP rate, only EUS-RV significantly outperformed pancreatic guidewire-assisted techniques (RR 0.21, 0.04-0.98). CONCLUSIONS: Based on low quality of evidence, the several techniques for difficult biliary cannulation show similar results although a trend in favour of needle knife techniques and EUS-RV was observed. EUS-RV seems to decrease the risk of PEP.
BACKGROUND AND AIMS: The increasing prevalence of Crohn's disease (CD) in South Korea has led to complications and reduced quality of life, emphasizing the need for early, non-invasive diagnosis. The International Organi...BACKGROUND AND AIMS: The increasing prevalence of Crohn's disease (CD) in South Korea has led to complications and reduced quality of life, emphasizing the need for early, non-invasive diagnosis. The International Organization for Inflammatory Bowel Disease (IO-IBD) developed the Red Flag Index (RFI) for early CD detection; however, it remains to be validated in the Korean population. This study aimed to develop and validate a Korea-specific RFI for early CD diagnosis using clinical and biomarker data. METHODS: Data from 121 patients with CD at Seoul National University Hospital were compared with those from two control groups: Group A (101 non-CD individuals) and Group B (198 high-risk individuals). The participants answered 15 IO-IBD-derived questions. Chi-square tests and logistic regression were used to identify key RFI predictors. Receiver operating characteristic (ROC) analysis was used to evaluate diagnostic performance, with subgroup analyses adjusted for RFI scores. The effect of fecal calprotectin on accuracy was also assessed. RESULTS: RFI identified key CD predictors: perianal lesions, weight loss, chronic anemia, chronic/nocturnal diarrhea, perianal pain, male sex, and fecal calprotectin >100 µg/g. The RFI's area under the curve (AUC) was 0.77 (Control A) and 0.88 (Control B), improving to 0.878 and 0.92, respectively, with fecal calprotectin. Subgroup analyses confirmed RFI's enhanced diagnostic accuracy, particularly in high-risk groups, except for the ankylosing spondylitis subgroup, which had lower AUCs. CONCLUSION: RFI demonstrated strong diagnostic performance, particularly when combined with fecal calprotectin, reinforcing its applicability in stratifying high-risk populations for early and accurate CD detection.
BACKGROUND: Advanced pancreatic ductal adenocarcinoma (PDAC) remains among the most lethal cancers, with >95 % of patients dying from the disease. Chemotherapy is the standard of care in advanced stages, with FOLFIRINOX...BACKGROUND: Advanced pancreatic ductal adenocarcinoma (PDAC) remains among the most lethal cancers, with >95 % of patients dying from the disease. Chemotherapy is the standard of care in advanced stages, with FOLFIRINOX and Gemcitabine-Nab-Paclitaxel (Gem-NabP) as the main first-line regimens. Both show moderate efficacy and significant toxicity. Except for the PASS-01 trial, no direct comparison exists, though observational studies suggest that specific subgroups may benefit differently. Given the modest outcomes and rapid clinical decline, most patients are ineligible for second-line therapy after progression. The PANThEON study evaluates whether switching from modified FOLFIRINOX (mFOLFIRINOX) to Gem-NabP after three months of induction with mFOLFIRINOX improves outcomes for PDAC. METHODS: PANThEON is a no-profit, phase III, randomized, open-label, multicenter trial. The primary endpoint is overall survival (OS). Secondary endpoints include progression-free survival (PFS), time to treatment failure (TTF), overall response rate (ORR), and quality of life (QoL). Exploratory analyses will assess tumor profiling, circulating tumor DNA (ctDNA), and radiomics to identify predictive markers. A total of 220 patients will be randomized 1:1 to Gem-NabP (arm B) or continued mFOLFIRINOX (arm A). DISCUSSION: The PANThEON trial addresses two challenges: improving efficacy while reducing toxicity. Switching to Gem-NabP may enhance tolerability without compromising benefit, prolonging survival and refining PDAC treatment strategies. TRIAL REGISTRATION: PANThEON is registered at ClinicalTrials.gov (NCT06897644) and EuCT (2024-515214-41-00).
Lefevre JH, Dubreuil O, Colle R
… +11 more, Tougeron D, Vienot A, Dupré A, Toullec C, Smith D, Guimbaud R, Chibaudel B, Garcia Larnicol ML, Vernerey D, Cohen R, PREMICES Study Group
BACKGROUND: Microsatellite instability (MSI) or mismatch‑repair deficiency (dMMR) defines a biologically distinct subgroup of colorectal cancers that are highly responsive to immune checkpoint inhibition. Major and compl...BACKGROUND: Microsatellite instability (MSI) or mismatch‑repair deficiency (dMMR) defines a biologically distinct subgroup of colorectal cancers that are highly responsive to immune checkpoint inhibition. Major and complete pathological responses have been observed with neoadjuvant immunotherapy, raising the possibility of avoiding upfront surgery in selected patients. METHODS: PREMICES is a multicenter, open‑label, randomized (1:1), non‑comparative phase II trial evaluating neoadjuvant pembrolizumab followed by a structured watch‑and‑wait strategy versus standard-of-care surgery ± adjuvant chemotherapy in adults with resectable, localized MSI/dMMR colon or upper rectal adenocarcinoma. Sixty patients will be randomized across nine French centers. The primary endpoint is 6-month strategy success after randomization (or after two successive colonoscopies), defined as the absence of death, disease progression, a decision to operate due to residual tumor on biopsies, or any primary‑tumor surgery. Secondary endpoints include the 24-month success, event‑free and overall survival, postoperative morbidity, health‑related quality of life (EORTC QLQ‑C30/CR29), tumor regression, and endoscopic complete response. Embedded translational studies include circulating tumor DNA and gut microbiota analyses. Randomization started in September 2025. DISCUSSION: PREMICES will determine whether a neoadjuvant PD‑1 blockade‑based non‑operative approach can be pursued safely and effectively in localized MSI/dMMR colon cancer and will generate prospective patient‑centered outcomes and biomarker data to inform future phase III trials. TRIAL REGISTRATION: EU‑CT (CTIS) number: 2023‑509322‑22‑00. Registered prior to first patient enrolment.