BMJ Open Gastroenterology最新文献

The integration of machine learning (ML) into medical diagnostics has significantly advanced endoscopic examinations for gastrointestinal diseases. By leveraging extensive datasets and sophisticated algorithms, ML technologies enhance diagnostic precision, detect subtle abnormalities, classify diverse pathologies and predict disease progression. However, their widespread adoption is hindered by the inherent heterogeneity of gastrointestinal diseases, technical limitations, limited generalisability across different populations and ethical challenges related to patient privacy, data security and algorithmic bias.This review provides a comprehensive structural analysis of ML approaches in endoscopy, starting with an overview of the classical endoscopic methodology that relies on direct visualisation of the gastrointestinal tract for diagnosis and therapeutic interventions. Then, current ML applications that hold promise for reducing physician-dependent variability, improving diagnostic accuracy and streamlining procedural workflows were explored. Despite these advances, the effectiveness of ML models often remains constrained by the quality and diversity of training data, which can undermine both reliability and generalisability.Ethical considerations - such as safeguarding patient information, upholding data security and mitigating biases embedded in algorithms - are integral to responsibly deploying ML in clinical settings. By examining these technical and ethical barriers, this work contributes to the evolving discourse on integrating advanced ML techniques into gastroenterology. Ultimately, our goal is to pave the way for more effective and reliable ML-driven endoscopic practices that will enhance disease detection, optimise patient care and benefit healthcare providers worldwide.

Objective: Food-related quality of life (FR-QoL) is the psychosocial impact of food, nutrition, eating and drinking on QoL and can be profoundly affected by inflammatory bowel disease (IBD). We aimed to translate and validate the FR-QoL-29 Questionnaire in a Belgian IBD population and investigate associations with relevant clinical variables.

Methods: The English FR-QoL-29 was translated to Dutch using the forward-backward method. Consecutive patients with IBD attending the outpatient clinic in a university hospital in Belgium completed the FR-QoL-29 Dutch (Flemish), alongside questionnaires assessing disease severity and IBD-related disability. Clinical and biochemical data were collected. Exploratory factor analysis (EFA) with promax rotation assessed the underlying factor structure. Reliability measures (internal consistency, test-retest reliability) were evaluated. Pearson and Spearman correlations assessed relationships between FR-QoL-29 Score and continuous demographic and clinical variables, while categorical parameters were analysed using independent t-tests and one-way analysis of variance with Tukey post hoc tests.

Results: 301 patients were included, with 31 (10.3%) completing the retest. EFA revealed a one-factor structure explaining 55% of the variance. The FR-QoL-29-Dutch (Flemish) showed excellent internal consistency (Cronbach's α=0.97) and very good test-retest reliability (intraclass correlation=0.94). Lower FR-QoL-29 Score was associated with higher serum albumin levels, younger age, disease activity and IBD-related disability. FR-QoL-29 Score was lower in females, those with Crohn's disease (CD), CD patients with a stricturing phenotype and those previously receiving IBD-related surgery.

Conclusion: The FR-QoL-29-Dutch (Flemish) is valid and reliable and correlates with disease activity and IBD-related disability. Patients with CD, stricturing disease and who previously underwent IBD-related surgery have a significantly lower FR-QoL and should be targeted for support.

The incidence of incidental pancreatic cystic lesions (PCLs) has risen in recent years, largely due to advances in and increased use of imaging techniques. Endoscopic ultrasound (EUS) has become a crucial tool for evaluating and characterising PCLs, allowing for detailed morphological assessment and aiding in the identification of lesions with a higher risk of progression to high-grade dysplasia or invasive pancreatic carcinoma. This review aims to outline the key aspects of EUS in the evaluation of PCLs, covering a range of modalities from morphological assessment and contrast-enhanced imaging to elastography, fine-needle aspiration for biomarker analysis, cytology, DNA sequencing, histological evaluation and the emerging role of confocal laser endomicroscopy or artificial intelligence. Additionally, we address therapeutic EUS modalities for PCLs, the current limitations of EUS, anticipated technological advancements and the diverse management strategies recommended by leading scientific societies for the clinical handling of PCLs.

Objective: Accurate data resources are essential for impactful medical research, but available structured datasets are often incomplete or inaccurate. Recent advances in open-weight large language models (LLMs) enable more accurate data extraction from unstructured text in electronic health records (EHRs), however, thorough validation of such approaches is lacking. Our objective was to create a validated approach using LLMs for identifying histopathologic diagnoses in pathology reports from the nationwide Veterans Health Administration (VHA) database, including patients with genotype data within the Million Veteran Program (MVP) biobank.

Methods: Our approach utilises search term filtering followed by simple 'yes/no' question prompts for the following phenotypes of interest: any colorectal dysplasia, high-grade dysplasia and/or colorectal adenocarcinoma (HGD/CRC) and invasive CRC. We first developed the LLM prompts using example reports from patients with inflammatory bowel disease (IBD). We then validated the approach in IBD and non-IBD by applying the fixed prompts to a separate corpus of 116 373 pathology reports generated in the VHA between 1999 and 2024. We compared model outputs to blinded manual chart review of 200-300 pathology reports for each patient cohort and diagnostic task, totalling 3816 reviewed reports, and calculated F1 scores as a balanced accuracy measure.

Results: In patients with IBD in MVP, we achieved F1-scores of 96.9% (95% CI 94.0% to 99.6%) for identifying dysplasia, 93.7% (88.2%-98.4%) for identifying HGD/CRC and 98% (96.3%-99.4%) for identifying CRC. In patients without IBD in MVP, we achieved F1-scores of 99.2% (98.2%-100%) for identifying any colorectal dysplasia, 96.5% (93.0%-99.2%) for identifying HGD/CRC and 95% (92.8%-97.2%) for identifying CRC using LLM Gemma-2.

Conclusion: LLMs provided excellent accuracy in extracting the diagnoses of interest from EHRs. Our validated methods generalised to unstructured pathology notes, even withstanding challenges of resource-limited computing environments. This may, therefore, be a promising approach for other clinical phenotypes given the minimal human-led development required.

Objective: Gallstone disease (GD) is a common gastrointestinal disorder with a significant genetic component. Despite known risk factors, the genetic basis of GD remains incompletely understood. We aimed to identify novel genetic loci associated with GD, explore their clinical implications and investigate their therapeutic potential.

Methods: We conducted a genome-wide association study from the UK Biobank followed by a meta-analysis, integrating summary statistics from the FinnGen R11, with further replication from Biobank Japan. Using systematic bioinformatic approaches, we performed gene prioritisation, colocalisation analysis, transcriptome-wide association study, Mendelian randomisations, cross-trait genetic correlations, phenome-wide association study, clinical investigations and gene-environment interactions by leveraging data from the FinnGen, Genotype-Tissue Expression project and Liver Cell Atlas single-cell transcriptomics data set.

Results: Our study highlighted novel susceptibility loci near candidate genes (ie, UGT1A4, FADS1/3) associated with GD, expanding the known genetic landscape. Functional annotation and colocalisation analysis implicated that the independent variants are involved in various hepatocyte functions, including bile secretion, cellular glucuronidation and cholesterol gallstone pathway. Mendelian randomisation established causal relationships between the level of unsaturated fatty acids and GD risk. We also demonstrated the implications of indirect bilirubin level in GD risk stratification and the protective effect of oily fish intake in genetically susceptible individuals.

Conclusions: This study provides new insights into the genetic basis of GD and highlights the role of hepatocytes in GD pathogenesis. These findings have implications for the personalised prevention strategies and new therapeutic interventions in individuals predisposed to GD.

Intestinal failure (IF) is defined as 'the reduction of gut function below the minimum necessary for the absorption of macronutrients and/or water and electrolytes, such that intravenous supplementation is required to maintain health and/or growth'. Type 1 IF is short-term and often perioperative in nature. Patients are managed in a multitude of healthcare settings, particularly surgical and critical care areas. Type 1 IF can lead to malnutrition, which is prevalent in hospitals and has significant consequences for patient outcomes. Type 1 IF patients require short-term parenteral nutrition (PN) support; the provision of PN in hospitals should be managed by a multidisciplinary nutrition support team (NST).Nutritional assessment should include estimating requirements for energy, protein, fluid and electrolytes, based on basal requirements with adjustments for higher demands. All patients should be assessed for risk of refeeding syndrome and managed appropriately. Most patients can be managed using multichamber PN bags. A small minority may require bespoke PN. PN should always be provided with micronutrients and electrolytes.A central venous catheter is the preferred choice for venous access. All intravenous devices used for PN should be handled using an aseptic non-touch technique. If a catheter-related bloodstream infection is suspected, the diagnosis should be made using paired blood culture sampling.All patients should undergo ward-based and blood monitoring. The frequency of monitoring may be reduced if PN continues for a longer duration, under NST advice. Daily assessment should also include monitoring for signs of resolution of IF, and the introduction of oral/enteral nutrition.

Objective: Capsule endoscopy (CE) provides a minimally invasive exam modality for panendoscopic evaluation of the entire gastrointestinal (GI) tract. However, conventional reading methods can be time-consuming and error-prone. Protruding lesions are a relatively common entity that can be found with a variable incidence and different pathological significance throughout the GI tract. The aim of this study was to develop and test a convolutional neural network (CNN)-based algorithm for panendoscopic automatic detection of protruding lesions on CE exams.

Methods: A multicentric retrospective study was conducted, based on 1245 CE exams. We used a total of 191 455 frames, from six types of CE devices, of which 52 717 had protruding lesions (polyps, epithelial tumours or subepithelial lesions) after triple validation. Data were divided into a training and test set (90% vs 10%), in an exam-split design. During the training stage, we performed a fivefold cross-validation. Our outcome measures were sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and areas under the conventional receiver operating characteristic curve (AUC-ROC) and the precision-recall curve (AUC-PR).

Results: In the test set, the sensitivity was 79.7% and the specificity was 96.5%. The PPV and NPV were 81.5% and 96.0%, respectively. The global accuracy was 93.7%.

Conclusion: This study aims to address a gap in artificial intelligence (AI)-enhanced capsule panendoscopy by reporting the development of the first CNN for the detection of protruding lesions across the GI tract. AI's improvement of CE's diagnostic accuracy, along with the growing interest in minimally invasive procedures, may contribute to increasing access to this diagnostic tool. Further multicentric and prospective studies are needed to validate our preliminary results to ultimately introduce deep learning models into clinical practice.

Objective: People with cystic fibrosis (pwCF) are at significantly increased risk of colorectal cancer (CRC), prompting international recommendations for earlier screening with colonoscopy. The utility of faecal immunochemical testing (FIT) as a screening adjunct in pwCF remains unclear. This study evaluates FIT's diagnostic performance and uptake within a CRC screening programme in a UK CF centre.

Methods: PwCF aged ≥40 years were invited in person to participate in a screening protocol including FIT and colonoscopy. FIT results were interpreted using three thresholds: ≥10 µg Hb/g (primary), ≥80 µg Hb/g and ≥120 µg Hb/g. Colonoscopy findings and polyp histology were recorded.

Results: Of 113 eligible patients, 66 (58.4%) returned FIT, 49 (43.4%) had FIT and colonoscopy, and 27 (23.9%) underwent colonoscopy only. Colonic polyps were detected in 27.6% (21/76), which were predominantly adenomatous, and no CRCs were detected. For polyp detection, FIT demonstrated poor sensitivity (14%) and modest specificity (86%) at the 10 µg Hb/g threshold. Quantitative FIT values did not correlate with polyp presence (area under the receiver operating characteristic curve 0.48). Bowel preparation was generally adequate (mean Boston Bowel Preparation Scale 6.8), with a low repeat colonoscopy rate (14.5%).

Conclusion: In this study, no cancer was detected in 76 consecutive eligible individuals who underwent colonoscopy. In 49 patients who had both FIT and colonoscopy, FIT did not aid colonic polyp detection, showing low sensitivity and no correlation between faecal haemoglobin and adenoma detection. Given the high adenoma prevalence, the limitations of FIT and the poor FIT return rate, colonoscopy should remain the preferred modality for CRC screening in pwCF.

Objective: Patients with acute pancreatitis show reduced gut microbiome diversity and high abundance of pathogenic bacteria compared with healthy subjects. Admission microbiome profiles are increasingly linked to severity, but methodology and study quality hamper interpretation. Our aim was to investigate whether admission microbiome analysis provides robust and reproducible associations with severity and complications of acute pancreatitis.

Methods: Patients with acute pancreatitis were prospectively enrolled from 20 Dutch hospitals (2019-2022). Admission saliva and rectal samples from 276 patients underwent 16S rDNA sequencing for microbiome profiling. Subgroups were defined based on a literature search. The microbiota endpoints (alpha- and beta-diversity, and genus abundance) were compared across subgroups and with previous studies. Robustness of the significant associations was classified as 'moderate' or 'high' in case of statistical significance in, respectively, 2 or ≥3 differential abundance models.

Results: Rectal alpha diversity (Shannon Index 3.55 vs 3.63, p=0.026) was decreased in necrotising (n=49) versus oedematous pancreatitis (n=218). Microbiota communities of either saliva or rectal samples differed in all the subgroups. In total, 270 (rectal) and 138 (saliva) genera were associated with severity or complications, of which 35 and 3 (Anaeroglobus and Finegoldia in saliva; Lachnospiraceae_FE2018_group in rectal) were classified as, respectively, moderately and highly robust. Fourteen associations were previously reported, of which 10 were in the opposite direction compared with this study.

Conclusion: Three admission microbiome taxa associated with severity and complications were highly robust, although their biological relevance remains unclear. This study also shows the lack of replicable findings of admission microbiome associations, highlighting the need for longitudinal studies to establish temporal relationships between microbiome changes and disease progression.

Introduction: People with inflammatory bowel disease (IBD) commonly experience pain, whether during active disease or remission, which interferes with daily life and major goals and causes distress. Current psychological methods of pain management draw from musculoskeletal pain interventions, but it has not been established that the musculoskeletal model is a good fit. We aimed to outline a psychological model of IBD pain.

Methods: We used qualitative methods: a very open interview (Grid Elaboration Method), conducted online and transcripts analysed for themes and subthemes. 15 men and 15 women with IBD pain, recruited from a national charity, took part in 4 months to February 2024. Participants scored their average pain 5/10 and interference by pain with activity 6/10, where 10 is maximum pain or interference.

Results: We extracted five inter-related themes: on the emotional impact of pain and symptoms; the challenge of pain; restrictions due to pain and other IBD symptoms; shortcomings in healthcare, particularly for pain; and poor public understanding of IBD. Although the first theme, universally endorsed, covered anxiety about the meaning of pain, we did not find the fears about physical integrity that characterise much musculoskeletal pain, nor the avoidance of physical activities based on those fears.

Conclusion: We propose that further exploration is warranted of the experience of IBD-related pain and how people adjust to it. This will inform the design of better psychologically-informed interventions to help people with IBD manage their pain, independently and in partnership with healthcare.