Introduction

Ulcerative colitis (UC) is a chronic, relapsing-remitting inflammatory bowel disease (IBD) with increasing global prevalence. The FDA has emphasized the importance of histologic endpoints in clinical trials, noting mucosal healing claims should be supported by more than just endoscopic findings. However, there are many challenges to achieving a standardized assessment of histologic disease activity in a UC clinical trial, including: the heterogeneity in the collection of mucosal biopsies (number, location) from study to study, a lack of fully validated histology scoring systems and an absence of community standards for definitions of histologic improvement or histologic remission.

Aims & Methods

This work was conducted by the FNIH Biomarkers Consortium Mucosal Healing in UC Project, a collaborative initiative involving academic institutions, pharmaceutical companies, patient advocacy organizations, and regulatory agencies to foster consensus, encourage adoption of findings, and support regulatory decision-making.
Aim 1: McBride et al. analyzed 994 biopsies from UC colectomy specimens to determine optimal biopsy sampling strategies for histologic grading.
Aim 2: A prospective study enrolled 153 UC patients to compare targeted versus non-targeted biopsy strategies across four colonic segments, assessing both endoscopic and histologic remission.
Aim 3: Data from three clinical trials (NCT02840721, NCT02958865, NCT02065622) and four independent datasets were compiled, yielding 9,652 whole slide images (WSIs) were used to train two AI tools:
• AIM-HI UC: Multiple instance learning models for predicting Geboes Score (GS) subgrades directly from H&E-stained WSIs.
• IBD Explore: A suite of convolutional neural networks (CNNs) for segmentation and classification of histologic features, including tissue types, crypt architecture, and cells. Performance was evaluated using a held-out test set (N=578 WSIs) and inter-pathologist agreement metrics.

Results

Aim 1: Sampling 2–3 biopsies per colonic region most effectively captured histologic heterogeneity and improved grading reliability. Agreement across histologic indices increased with biopsy number, particularly after the second and third biopsies.
Aim 2: Rectal histologic remission was highly predictive of proximal colonic remission (93% endoscopic, 88% histologic concordance). Histologic scores showed high correlation between targeted and non-targeted biopsies (ρ = 0.80–0.86), especially in remission.
Aim 3: Most AIM-HI UC models outperformed individual pathologists in predicting GS subgrades. IBD Explore generated interpretable histologic features—such as epithelial-to-lamina propria ratio, crypt distortion, and lymphocyte density—supporting advanced spatial analysis beyond conventional scoring systems.

Conclusion

This consortium-based endeavor generated several key findings: 1) Up to 3 biopsy samples per region of interest are required to overcome microscopic heterogeneity and ensure accurate histological grading in active colitis; 2) Targeted and random biopsies provide comparable histologic information, supporting the use of random biopsies in future clinical trials and practice; 3) Rectal endoscopic and histologic remission are highly predictive of pancolonic remission in extensive UC, supporting assessment limited to rectum in stable disease. Development of a ML-based approach for predicting GS subgrades, combined with these sampling insights, has the potential to improve standardization and automation of GS histological assessment, enhancing measurement of histological treatment response and endpoints in clinical trials.

References

IBD Explore and AIM-HI UC are for Research Use Only. Not for use in diagnostic procedures
McBride, R. B., Suarez-Farinas, M., Ko, H. M., Chen, X., Liu, Q., & Harpaz, N. (2023). Density of Biopsy Sampling Required to Ensure Accurate Histological Assessment of Inflammation in Active Ulcerative Colitis. Inflammatory bowel diseases, 29(11), 1706–1712. https://doi.org/10.1093/ibd/izad063
Danese S, Klopocka M, Scherl EJ, et al. Anti-TL1A Antibody PF-06480605 Safety and Efficacy for Ulcerative Colitis: A Phase 2a Single-Arm Study. Clin Gastroenterol Hepatol. 2021;19(11):2324-2332.e6. doi:10.1016/j.cgh.2021.06.011
Sandborn WJ, Danese S, Leszczyszyn J, et al. Oral Ritlecitinib and Brepocitinib for Moderate-to-Severe Ulcerative Colitis: Results From a Randomized, Phase 2b Study. Clin Gastroenterol Hepatol. 2023;21(10):2616-2628.e7. doi:10.1016/j.cgh.2022.12.029
Panés J, Colombel JF, D'Haens GR, et al. Higher vs Standard Adalimumab Induction and Maintenance Dosing Regimens for Treatment of Ulcerative Colitis: SERENE UC Trial Results. Gastroenterology. 2022;162(7):1891-1910. doi:10.1053/j.gastro.2022.02.033

Disclosure

Disclosures of Primary Authors
Autumn Jones: None
Russell McBride: None
Stephen Laroux: holds shares of stock in AbbVie
Christina Jayson: PathAI stock holder
Noam Harpaz: has served as an advisory board member or consultant for Lilly, Inc., Abbvie Inc., Bristol Meyer Squibb, Inc., and PathAI, Inc.

All listed organizations are financially contributing partners and/or consulting partners of the Foundation for the National Institutes of Health Biomarkers Consortium project team.
PathAI employees, previous and current are PathAI stock holders.

Introduction

Colectomy is the primary treatment for dysplastic lesions in inflammatory bowel disease (IBD). Since the development of advanced endoscopic techniques, the management of dysplasia is evolving. Endoscopic submucosal resection (ESD) could be used in selected cases at referral centres as an alternative to colectomy. However, the feasibility and safety in IBD remains limited.

Aims & Methods

To evaluate en-bloc resection rate and the need of colectomy due to ESD failure in IBD.
Secondary aims: to assess the complications of ESD, the local recurrence of dysplastic lesions, the presence of synchronous and metachronous lesions and the risk of IBD relapse following the procedure.
Multicentre, retrospective study of IBD patients who underwent ESD for suspected colonic dysplastic lesions at least one month before the inclusion date. Patients with total colectomy before ESD, invisible dysplasia at the time of ESD, or pregnancy were excluded. En-bloc resection rate and colectomy due to ESD failure at the short term were evaluated by logistic regression analysis. The risk of IBD relapse was evaluated by Cox regression analysis.

Results

Sixty patients from 22 centres were included: 47 with ulcerative colitis, 11 with Crohn’s disease and 2 with IBD-unclassified. The mean age at ESD was 64 years (SD 13). The majority of the lesions (n=33, 55%) were located in rectum. Fifty lesions (83%) were located in areas previously inflamed. En-bloc and R0 were achieved in 58 (97%) and 50 patients (83%), respectively. Colectomy due to ESD failure was required in 5 patients (8.3%), all of them occurred within 6 months after ESD. No risk factors were associated with colectomy in multivariable analysis, including inflammatory activity, IBD treatments or characteristics of the lesions. Complications occurred in 17 patients (28%): 3 did not required any treatment, 6 required pharmacologic therapy, 6 needed endoscopic or radiologic intervention, and 2 underwent surgery (Table 1). No deaths were reported.
At least one surveillance colonoscopy after ESD was performed in 42 patients (70%). Local recurrence was observed in 2 patients (4.7%) while other dysplastic lesions were detected in 11 (26%) during the follow-up (median 1.5 years, IQR 0.5-3.4). All these lesions were removed endoscopically. Synchronous and metachronous lesions were identified in 4 (20%) and 9 patients (19%), respectively. IBD relapse occurred in 8 patients during the follow-up, with an incidence rate of 7% person-year (CI 95%:3-13). The occurrence of relapses was gradually throughout the follow-up period.

Table 1. Complications of endoscopic submucosal dissection

Conclusion

ESD avoids colectomy in 90% of IBD patients with dysplastic lesions. One-third of the patients experienced complications, all of them resolved with no sequelae. Given the risk of new dysplastic lesions, ongoing monitoring with surveillance colonoscopies is recommended.

Disclosure

María José García has received financial support for travelling and educational activities from Janssen, Pfizer, AbbVie, Takeda, Kern Pharma, Faes Farma and Ferring.
Edoardo Vincenzo Savarino has served as speaker for Abbvie, Abivax, Agave, AGPharma, Alfasigma, Apoteca, Biosline, CaDiGroup, Celltrion, Dr Falk, EG Stada Group, Fenix Pharma, Galapagos, Johnson&Johnson,consultant for Abbvie, Agave, Alfasigma, Biogen, Bristol-Myers Squibb, Celltrion, Dr. Falk, Eli Lilly, Fenix Pharma, Ferring, Giuliani, Grunenthal, Johnson&Johnson, JB Pharmaceuticals, Merck & Co, Nestlè, Pfizer, Reckitt Benckiser, Sanofi/Regeneron, SILA, Sofar, Takeda, Unifarco; he received research support from Bonollo, Difass, Pfizer, Reckitt Benckiser, Sanofi/Regeneron, SILA, Sofar, Unifarco, Zeta Farmaceutici.
Irene González-Partida has received support for conference attendance and speaker fees from AbbVie, Dr. Falk Pharma, Ferring Pharmaceuticals, Janssen, MSD, Kern, Pfizer, Shire, Takeda, Teva and Tillotts Pharma AG.
Alicia López has served as speaker for Chiesi, Takeda, ABBVIE, Lilly. I have received support for conference attendance from Pfizer, Lilly, Kern Pharma, Falk, Ferring, and Janssen
Jordina Llao has received educacional suport from Abbvie, Tillots, Ferring and Lilly.

Introduction

Tumor necrosis factor-like ligand 1A (TL1A) is a member of the Tumor necrosis factor (TNF) superfamily implicated in the pathogenesis of inflammatory bowel disease (IBD) as a key amplifier of the effector function of several immune cells. TL1A can be expressed by various immune cells, including macrophages, dendritic cells, and some T cell subsets, and exists either as a membrane-bound or as a cleaved soluble trimer. TL1A signals predominantly by activating NF-kB pathway via the only known functional receptor, death receptor 3 (DR3), which has restricted expression mostly on effector T cells and innate lymphoid cells (ILCs). Afimkibart is a fully human anti-TL1A antibody that has shown clinical benefit in Phase 2 clinical studies and specifically blocks the activity of trimeric forms of TL1A. Monomeric TL1A has also been described; however, the relevance of monomeric TL1A in DR3 signaling and disease is unclear. This study aimed to understand the functional characteristics of both monomer and trimer TL1A species in primary cells and cell lines.

Aims & Methods

Immunoassays were used to measure the detection of TL1A species in IBD serum. Recombinant human monomeric and trimeric TL1A proteins were assessed for binding to afimkibart using SPR and in cellular assays using primary immune cells and cell lines for nuclear factor-kappa B (NF-κB) signaling and interferon-gamma (IFNγ) production.

Results

In functional cellular assays with primary immune cells, we demonstrate that trimeric recombinant human TL1A (rhTL1A) protein causes significant NF-κB signaling in human peripheral T lymphocytes, and in the presence of interleukin 12 (IL12) and interleukin 18 (IL18) co-stimulation, a significant increase of IFNγ production by human primary peripheral blood mononuclear cells (PBMCs). In line with the mechanism of action, both the NF-κB signaling and IFNγ production are blocked by afimkibart. In contrast, only high concentrations of monomeric rhTL1A, >100 ng/ml, induced moderate NF-κB signal and demonstrated a minimal increase in IFNγ production above IL12 + IL18 stimulation alone. However, this activity was blocked by afimkibart. Lastly, afimkibart blocks endogenous soluble and membrane-bound TL1A.

Conclusion

Herein, we demonstrate that afimkibart blocks NF-kB signaling and IFNγ production induced by TL1A stimulation. While some signaling activity was observed at high concentrations of monomeric TL1A protein, we hypothesize that any signaling observed with monomeric form is likely due to the formation of trimers, which are efficiently blocked by afimkibart. Overall, any monomeric TL1A present should not be functionally active nor be available as a sink for afimkibart, thus preserving the optimal pharmacokinetic (PK) and pharmacodynamic (PD) properties needed for effective target inhibition.

Disclosure

All the authors are employees of Genentech-Roche.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) has a global prevalence of 30%, with 10-30% progressing to advanced liver fibrosis in 5 years. It is essential to assess advanced liver fibrosis – a predictor of liver-related outcomes. FIB-4 is a good screening test, but its low specificity leads to unnecessary referrals to hospital, especially in individuals over 65 years and with type 2 diabetes (T2D). To address these limitations, machine-learning techniques are developing scores like FIB-6 (validated for Hepatitis B and C).

Aims & Methods

We aimed to assess the performance of FIB-6 score in screening for advanced fibrosis in a Portuguese cohort of MASLD patients.
A retrospective-cohort study was conducted with MASLD patients who had undergone liver vibration-controlled transient elastography (VCTE) between January-October 2024. inclusion criteria: adult patients (age >18years) diagnosed with MASLD; undergone liver VCTE; complete biochemical and hematological data. Exclusion criteria: coexistence of other liver diseases; decompensated cirrhosis; hepatocellular carcinoma; hematological/bone‑related diseases. We used liver stiffness measurement by VCTE as a reference standard for liver fibrosis. Fibrosis was staged on a 0–4 scale: F0, F1 and F2, patients without advanced fibrosis; F3 and F4, patients with advanced fibrosis/cirrhosis.
Non invasive tests (FIB-6, FIB‑4, APRI, NFS) were calculated using blood results available within the month of the VCTE. Statistical analysis was performed using R version 4.2.3. AUC-ROC curve analysis with cutoff points were calculated using the Youden index.

Results

115 MASLD patients (58 males (50.43%),57 females (49.57%), median age of 62years) where included. 41 (35.65%) had diabetes mellitus type 2 (T2D) (10 of those insulin-treated), 74 (64.35%) arterial hypertension, 76 (66.09%) dyslipidemia and with a median body mass index of 29.9mg/Kg/m². Regarding hepatic fibrosis, 79 patients had F0-2 and 17 patients had F3-4.
In the general sample, AUC-ROC curve analysis showed an AUC 0.771 (CI 95% :0.662, 0.88) for FIB-4, AUC 0.788 (CI 95% :0.686, 0.889) for APRI, AUC 0.709 (CI 95% :0.595, 0.823) for NFS and AUC 0.695 (CI 95% :0.571, 0.819) for FIB-6.
A subanalysis of the sample over 65 years old revealed AUC 0.807 (CI 95% : 0.665, 0.949) for FIB4; AUC 0.788 (CI 95% : 0.686, 0.889) for APRI; AUC 0.788 (CI 95% : 0.65, 0.926) for NFS; AUC 0.663 (CI 95% : 0.475, 0.852) for FIB-6. A subanalysis of the sample of T2D patients revealed AUC 0.747 (CI 95% :0.584, 0.911) for FIB4; AUC 0.811 (CI 95% :0.676, 0.946) for APRI; and AUC 0.765 (CI 95% : 0.612, 0.918) for FIB6.
Using the optimal cut‑offs of FIB6 (cutoff= 1.8992) for the diagnosis of F3+F4 indicated a reliable performance of FIB‑6 score (sensitivity was 70%, specificity was 52.6%).
The cutoff values of FIB-6 for the general sample was 2.4420 (sensitivity 53%, specificity 87%); for the subset of patients over 65 years was 2.7668 (sensitivity 40%, specificity 96%); for the subset of T2D patients was 2.2230 (sensitivity 55%, specificity 90%).

Conclusion

FIB-6 demonstrated good performance in screening advanced fibrosis in MASLD patients, when compared to those of FIB‑4, APRI, and NFS, surpassing FIB-4, particularly in individuals with T2D. This study is the first to evaluate the applicability of FIB-6 in a Portuguese MASLD cohort. Although the findings warrant validation in larger and more diverse populations, they suggest that FIB-6 may be more efficient to evaluate liver fibrosis in MASLD and offer a more efficient tool to identify patients that need specialized care.

References

Anushiravani A, et al. 2023.
Alswat K, et al. 2024.
Shiha G, et al. 2022
Tacke, Frank et al. 2023
Boursier, Jerome et al. 2022
Papatheodoridi, Margarita et al. 2021
Hannes Hagström et al. 2024.
Rinella, Mary E. et al.2023