Introduction

Immune checkpoint inhibitors (ICIs) have marked a breakthrough in the management of a wide range of cancer types, leading to increased overall survival. However, this treatment is frequently associated with the onset of various immune-related adverse events (irAEs), including ICI-related diarrhea and colitis.¹ Current recommendations for patients with diarrhea or colitis grade 2 or higher (graded via CTCAE)² include the discontinuation of ICI and administration of immune suppressants (e.g. steroids or biologics), which may lead to worse outcomes in these patients.³ Therefore, alternative therapeutic strategies to manage irAEs are needed. The gut microbiota plays a pivotal role in the onset of gastrointestinal irAEs, but the role of its therapeutic modulation on these disorders is mostly unknown. ^4,5

Aims & Methods

Our aim was to evaluate the efficacy and safety of therapeutic microbiome modulation for gastrointestinal irAEs in cancer patients. We enrolled cancer patients affected by ICI-related diarrhea/colitis referred to our Microbiome Clinic from January 2023 to March 2025. After clinical evaluation, patients were stratified according to severity of diarrhea (graded via CTCAE)² and then received a microbiome-modulating therapy. Patients were followed-up for at least 8 week after the initiation of therapy. We collected data about patient demographics (age and gender), patient’s overall health status as assessed by the the Charlson Comorbidity Index (CCI)⁶, type of cancer, oncological therapy and its suspension, time in weeks from initiation of cancer therapy to onset of diarrhea, severity of diarrhea (graded via CTCAE)², time in weeks until resolution of diarrhea and resumption of oncological therapy

Results

We enrolled 42 patients in the study period (8 males, 34 females, mean age 58,3 years-old, mean CCI-index 5.8). All patients were treated with ICI (Pembrolizumab: n=25 , Dostarlimab: n=4, Nivolumab: n=4; Atezolizumab: n=3; Novulumab and Ipilumab: n=2, Cemiplimab: n=2; Mogamulizumab: n=1; Pertuzumab: n=1). Ten patients experienced G1 diarrhea (23,8%), 16 (38,1) G2, 15 (35,7%) G3 and one G4 (2,4%). As indicated by current European Guidelines³, 30 patients (71,4%) had to interrupt ICI due to diarrhea. Multi-strain probiotics were administered to 27 (64.3% ) patients, rifaximin to 22 (52,4%). Overall, 21 patients (50%) needed therapy with systemic steroids. Seven patients who experienced ICI-related colitis refractory to steroids and biologics were successfully treated with FMT. Twenty-eight patients (66.7% ) had a resolution of the diarrhea (G0) and 11 patients (26.2%) had diarrhea G1 at a mean time of 8 weeks. Upon resolution of diarrhea, nine of the 30 patients who had interrupted ICI (30%) were able to resume the treatment. None of the patients experienced any adverse event related to gut microbiota modulation as defined by CTCAE².

Conclusion

Therapeutic modulation of gut microbiota appears to be a promising safe and effective approach for treating ICI-related diarrhea and colitis. However, these preliminary findings need confirmation through more consolidated and well-designed studies.

References

1. Dougan M, Wang Y, Rubio-Tapia A, Lim JK. AGA Clinical Practice Update on Diagnosis and Management of Immune Checkpoint Inhibitor Colitis and Hepatitis: Expert Review. Gastroenterology. 2021;160(4):1384-1393. doi:10.1053/j.gastro.2020.08.063
2. Freites-Martinez A, Santana N, Arias-Santiago S, Viera A. Using the Common Terminology Criteria for Adverse Events (CTCAE - Version 5.0) to Evaluate the Severity of Adverse Events of Anticancer Therapies. Actas Dermosifiliogr. 2021;112(1):90-92. doi:10.1016/j.ad.2019.05.009
3. Haanen J, Obeid M, Spain L, et al. Management of toxicities from immunotherapy: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up. Ann Oncol. 2022;33(12):1217-1238. doi:10.1016/j.annonc.2022.10.001
4. Wang Y, Wiesnoski DH, Helmink BA, et al. Fecal microbiota transplantation for refractory immune checkpoint inhibitor-associated colitis. Nat Med. 2018;24(12):1804-1808. doi:10.1038/s41591-018-0238-9
5. Wang T, Zheng N, Luo Q, et al. Probiotics Lactobacillus reuteri Abrogates Immune Checkpoint Blockade-Associated Colitis by Inhibiting Group 3 Innate Lymphoid Cells. Front Immunol. 2019;10. doi:10.3389/fimmu.2019.01235
6. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis. 1987;40(5):373-383. doi:10.1016/0021-9681(87)90171-8

Disclosure

G.I. has received personal fees for acting as speaker for Biocodex, Danone, Sofar, Malesci, Metagenics and Tillotts Pharma, and for acting as consultant and/or advisor for Ferring Therapeutics, Giuliani, Malesci and Tillotts Pharma. A.G. reports personal fees for consultancy from Eisai Srl, 3PSolutions, Real Time Meeting, Fondazione Istituto Danone, SinergieSrl, Board MRGE and Sanofi SpA personal fees for acting as a speaker for Takeda SpA, AbbVie and Sandoz SpA and personal fees for acting on advisory boards for VSL3 and Eisai. G.C. has received personal fees for acting as advisor for Ferring Therapeutics. All other authors have no conflicts of interest to disclose.

Introduction

Artificial Intelligence (AI)- enabled endoscopy and histology offer accurate, objective, and rapid assessment of disease activity in Ulcerative Colitis (UC). Emerging multi-source AI models integrating diverse datasets may enhance standardised disease evaluation and outcome prediction.

Aims & Methods

This study aimed to develop a novel AI model fusing endoscopic and histological findings to improve the assessment of disease remission and predict early response to therapy in UC clinical trials. A novel multimodal AI fusion algorithm was developed by integrating paired endoscopic videos and histological whole-slide images (WSIs) from the phase 2 clinical trial of Mirikizumab in UC (NCT02589665). The endoscopy branch of the model was trained with 291 white-light videos, using a convolutional neural network to select informative frames and the BioMedCLIP foundational model to extract features. The histology branch utilised the CONCH foundational model and was fine-tuned on 291 WSIs to obtain patch-level features. Features from both modalities were aggregated and fused using multi-head self-attention. The model’s ability to assess histological remission and response to therapy at weeks 12 and 52 was evaluated. Histological remission was defined as Geboes ≤2B.0, while a response to therapy at different time points was based on histological remission or improvement (Geboes <3.1).

Results

The fusion model outperformed single-modality assessments for histological remission, achieving a sensitivity of 89.72% (95% CI: 82.35–94.76), specificity of 89.67% (95% CI: 84.34–93.67), and accuracy of 89.69% (95% CI: 85.61–92.94). It demonstrated remarkable performance in assessing response to therapy at 12 and 52 weeks, with sensitivity of 97.96% (95% CI: 89.15–99.95), specificity of 86.84% (95% CI: 71.91–95.59) and accuracy of 93.10% (95% CI 85.59 – 97.43) for histological remission at week 52. Substantial agreement was observed between the AI fusion model and central readout.

Conclusion

This innovative multimodal fusion AI model enhances the assessment of histological remission and accurately predicts response to therapy. By potentially standardising central readouts and enabling automated disease assessment, this novel tool marks a significant advancement towards precision medicine in clinical trials.

Disclosure

Marietta Iacucci received research grants and consultant fees from Olympus, Pentax, Eli Lilly, and Janssen. All other authors declare no competing interests.

Introduction

Gastroenterologists face dual demands from clinical practice and academic research, necessitating efficient literature processing. While large language models (LLMs) show promise as healthcare tools, their variable performance in literature interpretation remains understudied, particularly regarding content completeness and reliability. This study systematically evaluates four prominent LLMs (ChatGPT-4o, Claude-3.5sonnet, Doubao-pro, Deepseek-V3) in explaining five gastroenterology literature types (original articles, case reports, reviews, editorials, guidelines), providing evidence-based recommendations for practical applications.

Aims & Methods

We tasked four LLMs with interpreting 125 gastroenterology publications (25 per literature type, 5 interpretations per article) spanning artificial intelligence, endoscopy, gastrointestinal cancers, and inflammatory bowel disease. Five gastroenterologists assessed interpretations using five criteria:
1. Accuracy: Correct identification of study design.
2. Format compliance: Adherence to predefined structure.
3. Content completeness: Key concept coverage (quantified via standardized rating scales).
4. Conciseness: Absence of extraneous content.
5. Reliability: Absence of hallucinated content.
Domain experts developed literature-specific completeness scales. Statistical analysis employed Pearson's chi-square tests for categorical variables (accuracy, format compliance, conciseness, reliability) and Kruskal-Wallis tests with Dunn's post-hoc comparisons for completeness scores. We applied Bonferroni correction for multiple comparisons (R version 4.4.2), with α=0.05 defining statistical significance.

Results

The 625 evaluations revealed significant inter-model differences across all metrics (p<0.001). ChatGPT-4o and Deepseek-V3 demonstrated superior accuracy (99.2% vs 100.0%) and format compliance (97.6% vs 99.2%) compared to Claude-3.5sonnet and Doubao-pro. Claude-3.5sonnet produced the most concise outputs (99.4%), while Deepseek showed the lowest concision (92.8%). Doubao exhibited perfect reliability (100.0%) versus 93.6%-97.0% in other models. Median completeness scores ranked: ChatGPT-4o (84.5%) > Claude-3.5sonnet (80.0%) = Doubao-pro (80.0%) > Deepseek-V3 (79.3%), with ChatGPT-4o significantly worse than others (Dunn's test, p<0.05).

Conclusion

Gastroenterologists can select LLMs based on the application priority of literature processing: ChatGPT-4o and Deepseek-V3 excel in accuracy and format compliance; Claude-3.5sonnet offers superior conciseness; while Doubao-pro demonstrates optimal reliability.

Introduction

Irritable bowel syndrome (IBS), the most prevalent gastrointestinal disorder globally, has a significant impact on the quality of life of 10 - 15% of the population, with a higher prevalence among women than men. Despite the identification of certain contributing factors, the pathophysiology of IBS remains largely unknown.

Aims & Methods

We firstly performed Genome-wide association study (GWAS) on the IBS cohort from the All of Us Research Program and adjusted for sex, age, and top 10 principal components. We conducted GWAS meta-analyses of IBS and its subtypes in 80,737 IBS cases and 1,019,143 asymptomatic control subjects from five independent cohorts (All of Us Research Program, UK Biobank, Bellygenes, Genetic Epidemiology Research on Aging, FinnGen). IBS risk loci were functionally annotated to identify candidate genes. Systematic variant-level analysis, including variants annotation, functional genomics and expression quantitative trait loci, prioritized potential causal variants. Gene-level analyses, including transcriptome-wide association study, colocalization analysis and Mendelian randomization, nominated the potential causal genes.

Results

We identified two independent IBS susceptibility loci at genome-wide significance (P < 5 × 10⁻⁸) in All of Us Research Program cohort totaling 14,897 cases and 245,388 controls. GWAS-meta analysis identified 12 novel independent (P < 5×10^-8) signals and 5 unreported loci (rs143348218, rs3748618, rs6432674, rs9536395, rs976714). Subtype-specific (P < 5×10^-8) findings based on subtype-specific GWAS meta-datasets included 3 loci for subtype IBS-C (3 novel independent loci) , 44 loci for IBS-M (2 novel independent loci). By integrating evidence from layers of different analyses, we prioritized the most plausible causal genes for IBS including CADM2, PCLO, PHF2 and SHISA6. Notably, PCLO (associated with bipolar disorder and major depressive disorder) and SHISA6 (overexpression in Platelet and T-lymphocyte) have been previously unreported in association with IBS.

Conclusion

Our study not only showed the power of the last large-scale GWAS in deciphering the genetic aetiology of IBS, but also uncovered 5 new genetic risk variants, potential unreported causal variants, genes and therapeutic targets for IBS.

References

1 Chey WD, Kurlander J, Eswaran S. Irritable bowel syndrome: a clinical review. JAMA. 2015;313:949–58. doi: 10.1001/jama.2015.0954
2 Soubieres A, Wilson P, Poullis A, et al. Burden of irritable bowel syndrome in an increasingly cost-aware National Health Service. Frontline Gastroenterology. 2015;6:246–51. doi: 10.1136/flgastro-2014-100542
3 Eijsbouts C, Zheng T, Kennedy NA, et al. Genome-wide analysis of 53,400 people with irritable bowel syndrome highlights shared genetic pathways with mood and anxiety disorders. Nat Genet. 2021;53:1543–52. doi: 10.1038/s41588-021-00950-8
4 Bonfiglio F, Zheng T, Garcia-Etxebarria K, et al. Female-Specific Association Between Variants on Chromosome 9 and Self-Reported Diagnosis of Irritable Bowel Syndrome. Gastroenterology. 2018;155:168–79. doi: 10.1053/j.gastro.2018.03.064
5 DeBoever C, Tanigawa Y, Aguirre M, et al. Assessing Digital Phenotyping to Enhance Genetic Studies of Human Diseases. Am J Hum Genet. 2020;106:611–22. doi: 10.1016/j.ajhg.2020.03.007
6 Venner E, Muzny D, Smith JD, et al. Whole-genome sequencing as an investigational device for return of hereditary disease risk and pharmacogenomic results as part of the All of Us Research Program. Genome Med. 2022;14:34. doi: 10.1186/s13073-022-01031-z
7 Weeks EM, Ulirsch JC, Cheng NY, et al. Leveraging polygenic enrichments of gene features to predict genes underlying complex traits and diseases. Nat Genet. 2023;55:1267–76. doi: 10.1038/s41588-023-01443-6