National Gastroenterology and Hepatology Conference Features UCSF Health Experts

Gastroenterology and hepatology experts from UCSF Health participated in Digestive Disease Week (DDW) 2026, held May 2 to 5, 2025 in Chicago. The annual conference is the premier meeting for professionals working in gastroenterology, hepatology, GI endoscopy, gastrointestinal surgery and related fields. 

UCSF clinicians and researchers from the UCSF Department of Gastroenterology and Hepatology shared presentations in such areas as clinical trials in inflammatory bowel disease and pregnancy, clinical research in CAR T-mediated enteritis, and the use of AI in diagnosis and documentation for gastroenterology and hepatology practice and administration.

UCSF Invited Lectures and Oral Presentations:

Uma Mahadevan, MD, the Marc and Lynne Benioff Professor of Gastroenterology and director of the UCSF Colitis and Crohn's Disease Center, presented “Recent Advances in The Management of Acute Severe Ulcerative Colitis” (Sp387) during the invited lecture “Dr. Ellen Scherl — Jill Roberts Lecture: Advances in the Management of Complicated IBD.” Mahadevan reviewed current therapies to treat acute severe ulcerative colitis (ASUC), comparing the efficacy of several treatments by reviewing recent clinical trial results including the “PREDICT-UC” trial; the TACOS trial; the DUET-Uc trial and the use of autologous CD19-directed CAR T-cell therapy.

Mahadevan also presented the invited lecture “Care of the IBD Patient during Pregnancy” (Sp1166) during the session “GI in Gestation: Management of Gastrointestinal Diseases During Pregnancy.” Last year, Mahadevan led the Helmsley PIANO Expert Global Consensus which released standardized, evidence-based recommendations to providers caring for pregnant women with IBD. The PIANO (Pregnancy Inflammatory Bowel Disease And Neonatal Outcomes) study upon which many of the recommendations were based, looked at the safety of IBD medications in pregnancy and short and long-term outcomes of the children. Mahadevan discussed these findings as well as recommendations that women with IBD receive preconception counseling, ideally be in remission for three to six months prior to considering conception and that all women with IBD are followed as high-risk pregnancies.

In addition, Mahadevan presented “Pregnancy Outcomes In Maternal Exposure To Guselkumab: Review Of Cases Reported To The Company Global Safety Database” (#95) and “Pregnancy and Early Childhood Outcomes After In Utero Exposure To JAK Inhibitors In Women With IBF: A Global Multicenter Cohort Study” (#96) during the abstract lecture session “Management of IBD in Children, Pregnancy and Older Adults.”

In the first presentation (#95), Mahadevan reported on the Global Consensus Consortium statement on the management of pregnancy in women with IBD, suggesting women can continue treatment with IL-23 inhibitors throughout pregnancy. The findings of this study suggest no apparent impact of Guselkumab on pregnancy outcomes. Live births, elective terminations, spontaneous abortions, and congenital anomalies were consistent with the US population. More studies are warranted to confirm these observations and to further characterize the safety profile of Guselkumab exposure during pregnancy. In the latter presentation (#96), Mahadevan reported that ongoing disease remission in pregnancy results in optimal pregnancy outcome, but pregnant women can be treated with JAK inhibitors only if there is no other viable option for maternal health. Since JAK inhibitors cross the placenta via passive diffusion and safety data in pregnancy remain very limited, JAK inhibitors should be used with caution and shared decision-making with the patient is key.

Jin Ge, MD, MBA, gastroenterologist, transplant hepatologist and director of Clinical AI at the UCSF Divisions of Gastroenterology and Hepatology, presented the invited lecture “Practical and Chat-Based Integration of LLMs in Hepatology” during the AGA session “Artificial Intelligence in Hepatology: From Diagnosis to Management.” Ge’s lecture made the practical case for large language models (LLMs) in hepatology, a field where the clinical burden is steep (cirrhosis drives more than 200,000 hospitalizations and roughly $20 billion in spending each year, with nearly 37% of patients readmitted within 30 days) and where about 80% of clinical data is locked in unstructured free-text notes. He discussed the spectrum of chat-based LLM deployment, from off-the-shelf tools to custom GPTs to retrieval-augmented generation (RAG) to EHR-embedded copilots, illustrating each with UCSF's own work: LiVersa, a liver-disease-specific RAG chatbot that drafts hepatology e-consults, and a Transplant Hepatology Note Co-Pilot built on PHI-protected ChatGPT Enterprise that cuts pre-charting time for new-patient visits from about 30 minutes to 10. A recurring theme was that the most valuable AI tool is the one that clinicians will actually use, and that rigorously evaluating generative outputs for safety, bias, and equity remains the field's central unsolved challenge.

Ge also gave the invited lecture “Digital Listening Technology in GI and Hepatology Clinical Practice: AI Scribes and Beyond” (Sp1025) during the AASLD session “Beyond the Bot: Provider and Patient Perspectives of AI Applications in Gastroenterology and Hepatology Clinical Practice.” This talk examined ambient AI "scribes" (tools that listen to clinical conversations and draft notes) against a documentation crisis in which hepatology and IBD providers spend roughly 131 minutes a day in the EHR and about half of gastroenterologists report burnout. He then advocated a more ambitious idea: that the same passively captured audio could move from documentation to diagnosis. Using covert hepatic encephalopathy (CHE) as the model (a common, costly, and frequently missed complication of cirrhosis), Ge and his colleagues showed how spontaneous speech carries acoustic and linguistic signals that scripted reading tests miss, and previewed UCSF's Ambient-CHE study (supported by an AGA Research Foundation Pilot Award), which analyzes multimodal speech features from routine clinic visits to detect CHE. He closed with a vision of ambient recordings as a passive, multi-disease screening exam, with potential applications spanning IBD, heart failure, cognitive decline, and hospital delirium.

Ge also gave an oral presentation “A Large Language Model (LLM)-Based Chatbot Approach for Detecting Covert Hepatic Encephalopathy” (#844) during the session “AI Applications in Liver Diseases and Transplantation.” This oral presentation reported on a UCSF pilot testing whether a text-based chatbot could detect covert hepatic encephalopathy (CHE). Twenty outpatients with chronic liver disease (10 with CHE and 10 without, defined by formal psychometric testing) held supervised conversations with LiVersa, UCSF's liver-disease-specific chatbot; the transcripts were converted into numerical embeddings and analyzed with several machine-learning models. K-means clustering performed best (modestly but consistently better than chance) while more complex models overfit the small sample. The team framed this as an encouraging proof-of-concept that the content of patient–chatbot conversations may carry detectable signals of CHE, while emphasizing the need for validation in larger, real-world cohorts and noting the practical challenges of deployment in this older patient population.

Rishika Chugh, MD, a gastroenterologist and assistant professor of Medicine in the UCSF division of Gastroenterology, presented the invited lecture “Disease Monitoring in IBD - Tips and Tricks” (Sp843). Her discussion focused on how to monitor disease progression in inflammatory bowel disease. She also discussed the utility of MRI, CT, and intestinal ultrasound and provided the latest guidance on combining imaging and biomarkers for more accurate restaging. In addition, Chugh also led two intestinal ultrasound “iUSCAN” workshops at DDW: the “iUCSCAN Clinical Trial Certification” workshop and “iUSCAN Fellows Certification Course Phase 1,” which was a kickoff to a yearlong training for IBD fellows interesting in learning bowel ultrasound.

Donna Leet, MD, a UCSF gastroenterology fellow, gave the oral presentation “Multi-Omic Profiling of CAR-T Enteritis Reveals Distinct Pathogenic Mechanisms of Car+ T Cell–Mediated Tissue Injury” (#25) during the session “Novel Mechanisms of Inflammatory Bowel Disease Therapeutics.” Chimeric antigen receptor T-cell (CAR-T)-mediated enteritis is a newly described and poorly understood disease. Defining the pathogenesis of CAR-T enteritis is urgently needed, as CAR-T therapies are being increasingly investigated for use in inflammatory disorders like ulcerative colitis (UC). Leet and her colleagues studied the BCMA-directed CAR-T therapy ciltacabtagene autoleucel (cilta-cel), which can induce durable remission in relapsed/refractory multiple myeloma, yet up to 5% of patients develop severe CAR-T enteritis. Leet and her colleagues performed integrated multi-omic profiling of cryopreserved duodenal biopsies from two cilta-cel–treated patients with CAR-T enteritis, one cilta-cel–treated patient without enteritis, and controls with immune-checkpoint inhibitor enteritis, gastrointestinal graft-versus-host disease, and healthy tissue. Their findings revealed two distinct pathogenic patterns within the duodenal niche: a monoclonal, highly cytotoxic CAR+ expansion in one patient, and an oligoclonal process with possible endogenous T-cell involvement in the other. Ongoing analyses include lentiviral integration mapping and whole-genome sequencing for tumor-driver mutations to distinguish inflammatory from lymphoproliferative mechanisms and guide diagnostic and therapeutic strategies. This study represents the first comprehensive multi-omic single-cell and spatial characterization of CAR-T enteritis.

Goktug Onal, MD, a UCSF research scholar, gave the oral presentation “Early Identification of Inflammatory Bowel Disease Using Longitudinal Electronic Health Record Data: A Multi-Site Machine Learning Study” (#100) during the session “Trends in Disease Burden and Natural History of IBD.” Onal reported on his study about reducing diagnostic delay in inflammatory bowel disease using longitudinal EHR data. By using sequential deep learning models trained on routinely collected clinical data, Onal and his colleagues were able to identify patients at high risk of IBD months to years before clinical diagnosis, with consistent performance across both academic and safety-net hospital cohorts. The research team is now expanding this work through external validation and exploring prospective deployment strategies.

UCSF Posters:

Meera Garriga, MD, Pregnancy Outcomes in an International Cohort of Autoimmune Hepatitis, (Monika Sarkar, MD, senior author) Abstract: Mo1620

Holly Huang, MD, Performance of LLM-Drafted Hepatology E-Consult Responses as Evaluated by Human and LLM Reviewers (Jin Ge, MD, senior author) Abstract Su2014

William Ge, MD, How Do Primary Care Providers Respond to Radiographic Findings of Hepatic Steatosis? (Jin Ge, MD, senior author), Abstract Su2080

William Ge, MD, A Practical and Deployable Solution for Automated Steatosis Identification in Radiology Reports (Jin Ge, MD, senior author), Abstract Su2021

Mao-Yuan Chen, MD, Causes of Diagnostic Delays in Patients with IBD: A Tertiary Center Study Using Electronic Health Records and Large Language Models (Vivek Rudrapatna, MD, senior author), Abstract Su1617

Mao-Yuan Chen, MD, Environmental Exposures in Surface Water Predict Inflammatory Bowel Disease: A Multicenter Cohort Study with Geospatial Methods (Vivek Rudrapatna, MD, senior author), Abstract Su1553

Aryan Ayati, MD, MPH, Closing Gaps in ASUC Management: LLMs for Automated Guideline Application and Adherence Monitoring (Vivek Rudrapatna, MD, senior author), Abstract Su1610

Yuntao Zou, MD, Hiatal Hernia Repair and Atrial Fibrillation: Evidence for a Protective Association, Abstract Sa1147

Linda Huang, MD, Defining a Unique Overlap Population: Clinical Characteristics and Treatment Outcomes of Patients with Eosinophilic Esophagitis and Inflammatory Bowel Disease, (Priya Kathpalia, MD, senior author) Abstract Tu1095

Vu Dinh, MD, PhD, Magnamosis Can Create a Neo Biliary Anastomosis to Bybass a Complete Biliary Obstruction (Michael Larsen, MD, senior author), Abstract Tu1963

Justin Field, MD, Large Language Models Markedly Improve Identification of Perianal Crohn’s Disease Compared with ICD-Based Methods, Abstract Tu1499

Rohit Khanna, DO, Impact of In Vitro Fertilization on Disease Activity and Pregnancy Outcomes in Women with Inflammatory Bowel Disease (Uma Mahadevan, MD, senior author), Abstract Tu1570

Rohit Khanna, DO, The Use of Intestinal Ultrasound Reduced the Need for Contrast-Enhanced Studies and Endoscopic Procedures in Adults Over the Age of 60, (Rishika Chugh, MD, senior author), Abstract Sa1511

Edgar Corona, MD, MPH, Impact of Patient Navigation to Improve Follow-Up After Colonoscopy No-Show in A Safety-Net Population: Preliminary Results from an Ongoing Study, Abstract Mo1021

Please visit DDW2026 for complete meeting abstract and session information.

About UCSF Health: UCSF Health is recognized worldwide for its innovative patient care, reflecting the latest medical knowledge, advanced technologies and pioneering research. It includes the flagship UCSF Medical Center, which is a highly-ranked hospital, as well as UCSF Benioff Children’s Hospitals, with campuses in San Francisco and Oakland; two community hospitals, UCSF Health Stanyan Hospital and UCSF Health Hyde Hospital; Langley Porter Psychiatric Hospital; UCSF Benioff Children’s Physicians; and the UCSF Faculty Practice. These hospitals serve as the academic medical center of the University of California, San Francisco, which is world-renowned for its graduate-level health sciences education and biomedical research. UCSF Health has affiliations with hospitals and health organizations throughout the Bay Area. Visit http://www.ucsfhealth.org/. Follow UCSF Health on Facebook, Threads or LinkedIn.