Introduction

Histopathological analysis of mucosal biopsies in ulcerative colitis (UC) is currently performed by pathologists on digital whole slide images (WSIs) or under the microscope; but established grading scales are not routinely employed and are subject to well-described intra- and inter-observer variations. Previously, we demonstrated that our Artificial Intelligence (AI) Tool can assign a Nancy Index (NI) [1] to a biopsy case with accuracy comparable to an expert pathologist [2]. Here we introduce the novel concepts of “smart tiling” and HeatMaps for comprehensive AI-powered assessment of WSIs and deploy these to describe a sample of WSIs.

Aims & Methods

We refer to a high-power location in a WSI as a ‘tile’, a fixed size region of interest of 1024 by 1024 pixels, with 0.5 microns per pixel in WSI with 20-fold magnification. Our previously validated AI Tool identifies tissue types, cell types (neutrophils, lymphocytes, plasmacytes), cell density, crypts, and ulcerations, and calculates an NI with high degree of agreement [2]. In this analysis, we ‘mapped’ the WSIs on a tile-by-tile basis and using the NIs, created a HeatMap of grades from NI 0 to 4. The NI for the WSI was selected on the basis of the most frequently appearing NI. We included fixed size tiles that contained more than 15% of tissue in them and excluded the rest. In this feasibility study, we processed 11 WSIs received from 2 different scanners, which resulted in a 1175 tile database.

Results

The average number of tiles per WSI was 107, with a range from 30 tiles (WSI with NI=0) to 203 tiles (WSI with NI=4). Every WSI contained multiple tiles with a variety of NIs. WSIs scored as NI= 3 or 4 contained on average 53 tiles of NI=4, 36 with NI=3, 16 with NI=2, and 8 with NI =1 and 8 with NI=0. WSIs scored as NI= 0-2 contained on average 16 tiles of NI=4, 5 with NI=3, 34 with NI=2, and 10 with NI =1 and 18 with NI=0. The average neutrophil or eosinophil / lymphocyte / plasmacyte counts per tile were: in WSIs with NI=0 15/22/52, in WSIs with NI=2 22/35/88, in WSI with NI=3 57/53/125, in WSI with NI=4 73/31/85. The average sum of neutrophils divided by the number of tiles in WSIs with NI=0 0.83, in WSIs with NI=2 1.92, in WSI with NI=3 20.3, in WSI with NI=4 25.6.

Conclusion

This is the first description of novel concepts for histopathological analysis of mucosal biopsies in UC with smart tiling and HeatMaps. We identified significant heterogeneity and variation among tiles and within WSIs in both degree of inflammation by the AI-determined NI, but also in cell types. These novel findings and automatic assessments provide new insight into the pathophysiology of UC and may allow for more reliable clinical assessments in our patients.

References

1. Marchal-Bressenot, A et al. Development and validation of the Nancy histological index for UC. Inflammatory Bowel Disease (2014). https://doi.org/10.1136/gutjnl-2015-310187
2. Peyrin-Biroulet, L et al. An artificial intelligence‐driven scoring system to measure histological disease activity in ulcerative colitis. United European Gastroenterology Journal (2024) https://doi.org/10.1002/ueg2.12562
3. Tanevski et al. Learning tissue representation by identification of persistent local patterns in spatial omics data. bioRxiv (2024). https://doi.org/10.1101/2024.03.06.58369

Footnote: David T Rubin and Viktor H Koelzer share joint first authorship. Laurent Peyrin-Biroulet and Olga Kubassova share joint last authorship.

Disclosure

DTR has received grant support from Takeda; and has served as a consultant for Abbvie, Altrubio, Apex, Avalo Therapeutics, Bristol-Myers Squibb, Buhlmann Diagnostics Corp, Celgene, Connect BioPharma, Intouch Group, Iterative Health, Janssen Pharmaceuticals, Lilly, Pfizer, Samsung Neurologica, and Takeda. He serves on the Board of Trustees for the Crohn’s & Colitis Foundation and is on the Board of Directors for Cornerstones Health.
VHK has been a speaker for Indica Labs; has received research support/grants from Image Analysis Group and Roche; and other support from Sharing Progress in Cancer Care (invited chair).
WX has no disclosures.
BH, SA, and MF are employees of Takeda. CPdC, JH, and OK are employees of Image Analysis Group.
LPB has received consulting fees from AbbVie, Abivax, Alimentiv, Alma Bio Therapeutics, Amgen, Applied Molecular Transport, Arena, Biogen, Bristol Myers Squibb, Celltrion, Connect Biopharm, Cytoki, Eli Lilly, Enthera, Ferring, Fresenius Kabi, Galapagos, Genentech, Gilead, GlaxoSmithKline, Gossamer Bio, H.A.C. Pharma, Image Analysis Group, InDex, Inotrem, Janssen, Medac, Merck Sharp and Dohme, Mopac, Morphic, Norgine, Novartis, OM Pharma, Ono Pharmaceutical, OSE Immunotherapeutics, Pandion, Par’Immune, Pfizer, Prometheus, Protagonist, Roche, Sanofi , Sandoz, Takeda, Theravance, Thermo Fisher, TiGenix, Tillotts, Viatris, Vifor, and Ysopia; grants from Celltrion, Fresenius Kabi, and Takeda; lecture fees from AbbVie, Amgen, Arena, Biogen, Celltrion, Eli Lilly, Ferring, Galapagos, Genentech, Gilead, Janssen, Medac, Merck Sharp and Dohme, Pfizer, Sandoz, Takeda, Tillotts, Viatris, Vifor, and Ysopia; travel support from AbbVie, Amgen, Celltrion, Connect Biopharm, Eli Lilly, Ferring, Galapagos, Genentech, Gilead, Gossamer Bio, Janssen, Medac, Merck Sharp and Dohme, Pfizer, Sandoz, Takeda, Thermo Fisher, and Tillotts.
Funding: This study was sponsored by Takeda.
Acknowledgements: Medical writing support was provided by Paul Hassan, PhD, of Excel Scientific Solutions, and was funded by Takeda.

Introduction

Precision medicine aims to expand upon the summary average treatment effects reported in clinical trials by providing more refined, individualised estimates of treatment effects. The phase IIb TUSCANY-2 study (NCT04090411), which examined the efficacy and safety of afimkibart (RO7790121/RG6631; previously PF-06480605/RVT-3101), an anti-TL1A antibody, in patients (pts) with moderate to severe ulcerative colitis (UC), previously reported clinically meaningful treatment effects at UEGW 2024.¹ Here, we report post hoc analyses of the relationship between baseline clinical characteristics and Week 14 efficacy in pts enrolled in TUSCANY-2.

Aims & Methods

Pts were randomised to receive subcutaneous afimkibart 50mg, 150mg, 450mg or matched placebo (PBO) monthly during induction. In this post hoc analysis, all afimkibart doses were pooled, and the endpoint of interest was clinical remission by modified Mayo score (mMS; defined as stool frequency subscore=0 or 1; rectal bleeding subscore=0; and endoscopic subscore [ES]=0 or 1) at Week 14. Our investigation of more patient-specific treatment effects was guided by the Predictive Approaches to Treatment effect Heterogeneity (PATH) statement.² We developed a regularised logistic regression model to predict the probability of Week 14 remission and to help examine heterogeneity of treatment effects based on the following baseline characteristics: mMS, C-reactive protein (CRP), serum albumin, Robart’s Histopathology Index (RHI), number of failed advanced therapy classes, corticosteroid use, disease extent, disease duration, age and sex. Baseline faecal calprotectin (FCP) was investigated separately due to the extent of missing data.

Results

The proportion of pts achieving clinical remission by mMS at Week 14 in the afimkibart pooled dose arm was 32.5% (65/200; 90% confidence interval [CI]: 27.3 to 38.1) compared with 13.3% (6/45; 90% CI: 7.1 to 23.8) of pts in the PBO arm (treatment delta: 19.2%; 90% CI: 7.6 to 27.9). Based on calibrated predictions from the multivariate model, all of the individual TUSCANY-2 pts (both treated with PBO and afimkibart) had anywhere between a 2–76% probability of Week 14 remission at baseline. The clinical characteristics included in the model explained ~10% of the variability in the observed outcomes. There was no evidence that the clinical characteristics were associated with heterogeneous treatment effects on the relative scale (odds ratios appeared to be constant; p-value >0.5). On the clinically meaningful absolute scale, the treatment delta in the top 50% of pts with a higher baseline probability of remission (27–76%) was 28.5% (90% CI: 8 to 42; 45.2% in afimkibart pooled dose arm vs 16.7%, in PBO arm), compared with a treatment delta of 7.6% (90% CI: –7 to 18; 18.8% in afimkibart pooled dose arm vs 11.1%, in PBO arm) in pts with a lower baseline probability of remission (2–27%). Lower baseline CRP, lower RHI and higher serum albumin contributed to a greater probability of remission and absolute benefit from treatment with afimkibart.

Conclusion

Pts with inflammatory bowel disease (IBD) have diverse baseline characteristics that are associated with variability of treatment outcomes, yet clinicians rely on reported averages from clinical trials to make individual treatment decisions. This analysis demonstrates how precision medicine approaches, that account for diverse characteristics, may improve interpretation of positive IBD clinical trial results. Integrating genetic and novel biomarkers alongside clinical features could help to better predict individual outcomes for pts with IBD.

References

1. Danese S, et al. Presented at UEGW 2024 (Oral presentation OP079).
2. Kent D, et al. The Predictive Approaches to Treatment effect Heterogeneity (PATH) Statement. Ann Intern Med 2020;172:35–45.

Disclosure

This study was sponsored by F. Hoffmann-La Roche Ltd. Third-party medical writing support, under the direction of the authors, was provided by Victoria Dagwell, MSc, of Ashfield MedComms, an Inizio company, and was funded by F. Hoffmann-La Roche Ltd.
LP-B reports consulting or advisory role/board from AbbVie, Abivax, Adacyte, Alimentiv, Amgen, Applied Molecular Transport, Arena, Banook, Biogen, Bristol Myers Squibb, Celltrion, Connect Biopharm, Cytoki Pharma, Enthera, Ferring, Fresenius Kabi, Galapagos, Genentech, Inc., Gilead, Gossamer Bio, GSK, IAC Image Analysis, Index Pharmaceuticals, Inotrem, Janssen, Lilly, Medac, Mopac, Morphic, MSD, Nordic Pharma, Novartis, Oncodesign Precision Medicine, ONO Pharma, OSE Immunotherapeutics, Pandion, Par' Immune, Pfizer, Prometheus, Protagonist, F. Hoffmann-La Roche Ltd, Samsung, Sandoz, Sanofi, Satisfay, Takeda, Telavant, Theravance, Thermo Fischer, Tigenix, Tillots, Viatris, Vectivbio, Ventyx and Ysopia; speakers’ bureau from AbbVie, Alfasigma, Amgen, Arena, Biogen, Celltrion, Ferring, Galapagos, Genentech, Inc., Gilead, Janssen, Kern Pharma, Lilly, Medac, MSD, Nordic Pharma, Pfizer, Sandoz, Takeda, Tillots and Viatris. JRA reports consulting or advisory role/board from Janssen, Pfizer, AbbVie, Vedanta, Genentech, Inc., Seres Therapeutics, Ferring, GSK, Merck, Bristol Myers Squibb, Roivant, Celltrion, TRXBio and Shattuck Labs; speakers’ bureau from AbbVie and Janssen. SD reports consultancy fees from AbbVie, Alimentiv, Allergan, Amgen, Applied Molecular Transport, AstraZeneca, Athos, Biogen, Boehringer Ingelheim, Bristol Myers Squibb, Celgene, Celltrion, Dr Falk Pharma, Lilly, Enthera, Ferring Pharmaceuticals, Gilead, Hospira, Inotrem, Janssen, Johnson & Johnson, Morphic, MSD, Mundipharma, Mylan, Pfizer, F. Hoffmann-La Roche Ltd, Sandoz, Sublimity, Takeda, Teladoc Health, TiGenix, UCB, Vial and Vifor; lecture fees from AbbVie, Amgen, Ferring Pharmaceuticals, Gilead, Janssen, Mylan, Pfizer and Takeda. PSD reports employment at Northwestern University; consulting or advisory role/board from AbbVie, Adiso, Alimentiv, Celltrion, Genentech/Roche, Geneoscopy, Janssen, Shattuck and Takeda; research funding from Takeda and Bristol Myers Squibb. AB reports employment and stockholder from Pfizer. DEC reports employment and stock ownership from Pfizer. EP reports employment, stock or other ownership interests and patents, royalties, other intellectual property from Pfizer. SN reports employment and stock or other ownership interests from Pfizer. MSV reports employment and shareholder from Pfizer. KH reports no conflict of interest. LU reports employment at Genentech, Inc. CS reports employment, stock or other ownership interests from Genentech/Roche. DB reports employment at F. Hoffmann-La Roche Ltd. KL reports employment and stock or other ownership interests from Genentech, Inc. BES reports honoraria for consulting from AbbVie, Adiso, Agomab, Alimentiv, Amgen, AnaptysBio, Arena Pharmaceuticals, Artugen, AstraZeneca, Biolojic Design, Biora, Boehringer Ingelheim, Boston Pharmaceuticals, Calibr, Celgene, Celltrion, ClostraBio, Equillium, Enthera, Enveda, Evommune, Ferring, Fresenius Kabi, Fzata, Galapagos, Genentech/Roche, Gilead Sciences, GSK, Gossamer Bio, Imhotex, Index Pharmaceuticals, Innovation Pharmaceuticals, Inotrem, Kaleido, Kallyope, Merck, Microbiotica, Mitsubishi Tanabe, Mobius Care, Morphic, MRM Health, Nexus, Nimbus Discovery, Odyssey, OSE Immunotherapeutics, Palisade Bio, Progenity, Prometheus Biosciences, Prometheus Laboratories, Protagonist, Q32 Bio, Rasayana, Recludix, Reistone, Sanofi, Sorriso, Spyre, Surrozen, Target RWE, Teva, TLL Pharmaceutical, Tr1X, Union and Ventyx; consulting/speaking fees from Abivax; consulting/speaking fees/other support from Lilly; research grants/consulting/speaking fees/other support from Bristol Myers Squibb, Lilly, Janssen, Pfizer and Takeda; research grants/consulting fees from Janssen; stock/stock options from Ventyx.

Introduction

As a first-line biotherapy, vedolizumab showed superior efficacy compared to adalimumab in the VARSITY trial, which, combined with a favorable safety profile, should position it in a privileged position in bionaïve patients with ulcerative colitis (UC). However, a proportion of patients have an insufficient response and/or intolerance, resulting in the need for a 2nd line of advanced therapy. To date, there are no data on the efficacy of biotherapies and small molecules in the 2nd line treatment of UC, after vedolizumab has failed. The objective of our study was to compare the efficacy of anti-TNFs, ustekinumab and JAKi in 2nd line in UC, after prior exposure to vedolizumab.

Aims & Methods

We conducted a retrospective multicenter study in 27 French and Belgian GETAID centers. All UC patients treated with vedolizumab in the first-line setting who received a second-line of anti-TNFs, ustekinumab and JAKi were retrospectively included. The primary outcome was corticoid-free remission at induction (week 14) defined by a clinical partial Mayo score of 2 or less, with no subscore greater than 1, in patients still receiving 2nd line therapy. The response was defined as a decrease in partial Mayo score of at least 30%. We also assessed the persistence rate of second-line treatments using the Kaplan-Meier method.

Results

203 patients were included with a median follow-up of 5 months (IQR: 3.0-11.0). Of the 203 patients, 123 (60.6%), 64 (31.5%) and 16 (7.9%) were treated with anti-TNF, ustekinumab and JAKi in the second-line setting. The median duration of disease progression was 39 months prior to the introduction of the second line (IQR 15-137.0). Vedolizumab was discontinued in 109 patients for primary failure (53.7%), 81 patients for secondary failure (34.9%), and 14 patients for adverse reaction (7.4%).
At week 14, 31/95 (32.6%) patients on infliximab, 7/27 (25.9%) on SC antiTNF, 20/63 (31.7%) on ustekinumab and 7/16 (43.8%) treated with JAKi achieved corticosteroid-free remission. Remission rates were 37.9% for patients on infliximab, 33.3% on SC antiTNF, 50% on JAKi, and 38.1% on ustekinumab. Response rates were 51.6%, 40.8%, 52.4%, and 62.5%, respectively, for the infliximab, antiTNF SC, ustekinumab and JAKi groups.
At 1-year, overall treatment persistence was 41.9% [IC 95% 34.7-50.6%]. Treatment persistence was 45.2% [35.0-58.4] in patients treated with infliximab, 43.3% [27.8-67.5] for antiTNF SC, 40.4% [28.2-58.0] in the Ustekinumab group and 18.8% [5.5-64.3] for the JAKi.
Second-line therapy was discontinued in 37 patients (34.9%) for primary failure, 53 (50%) for secondary failure and 16 (15.1%) for adverse reaction.
The overall adverse reaction rate was 31/203 (15.3%), of which 16 resulted in discontinuation (15.1%). In patients on infliximab, the overall rate was 19/96 (19.8%), and infliximab was discontinued in 10 patients (10.4%) due to adverse reactions. With JAKi, the overall rate was 1/16 (6.3%) with no cases of herpes zoster or thrombo-embolic events. Finally, in the 64 patients treated with ustekinumab, only 5 side effects especially arthralgias were reported (7.8%).

Conclusion

This is the largest real-world study of the efficacy of second-line treatments after vedolizumab in UC. Corticosteroid-free remission rates at induction appeared to be similar between the different treatment groups. About forty percent of patients remained on second-line treatment at one year. Discontinuation of second-line treatment was predominantly related to primary or secondary failures. Efficacy at induction appeared to be similar between the different treatment groups. ​

Introduction

Long-term real-life data on vedolizumab (VDZ) are scarce. In this study, we aim to evaluate the long-term outcomes of patients (pts) with inflammatory bowel disease (IBD) who received VDZ.

Aims & Methods

The LONG-LIVE is the long-term extension (LTE) of the LIVE study, including IBD pts who had started VDZ from 04 to 06/2017 at 47 Italian centres affiliated with IG-IBD, prospectively followed-up to 06/2019. For this study, data on therapy withdrawal, clinical and endoscopic activity, need for surgery/hospitalizations, deaths, and serious adverse events (SAEs) were retrospectively extracted from 07/2019 up to the last available follow-up (fup).

Results

Seven hundred eighty-two (70% of the original cohort) pts were enrolled in the LTE (394 CD, 388 UC), 316 (40%) of whom had already definitively discontinued VDZ; 43% were female, mean age was 53±16 years, mean disease duration was 17±9 years, 76% of patients were bioexposed to anti-TNFα therapy. Among the remaining 466 still receiving VDZ, 204 (44%) subsequently suspended VDZ permanently and 9 (2%) temporarily: mean VDZ therapy duration was 113±92 weeks and 311±60 weeks among those who suspended and those who continued VDZ, respectively. Mean therapy duration was 177±124 weeks (172±124 weeks for CD and 182±125 weeks for UC, with no significant differences); longer mean persistence was observed in bionaive vs bioexposed CD pts (197±123 vs 182±122 weeks, p<0.05), while no such difference was observed in UC pts (193.1±126 vs 190±122. week, p=ns). Among CD pts who continued VDZ throughout the entire LTE, 74% and 54% were in clinical and endoscopic remission, respectively, at the last fup; among UC pts who continued VDZ throughout the entire LTE, 79% and 50% were in clinical and endoscopic remission, respectively, at the last fup. During the LTE, CD pts had a higher risk of IBD-related hospitalizations (Hazard Ration, HR, 1.37, 95% Confidence Interval, CI, 1.04-1.72, p=0.027). Specifically, 153 (20%) pts (83 CD, 70 UC) were hospitalized to receive IBD-related surgery after a mean time of 136±87 weeks from VDZ start (for CD vs UC comparison, HR 1.17, 95% CI 0.85-1.60, p=ns); 85 (11%, 54 CD, 31 UC) pts underwent IBD-related hospitalization for non-surgical reasons (CD vs UC, HR 1.71, 95% CI 1.21-2.62, p=0.015). New or recurrent cancers were reported in 17 pts (the most common localization being 4 breast cancers and 3 gastrointestinal cancers): mean interval between first VDZ administration and cancer diagnosis was 214 (±95) weeks; 20 non-oncological SAEs were recorded in pts receiving VDZ, the most common being infections (n=7), mostly pneumonia (n=4). Finally, 15 deaths were reported (of which 9 occurred in patients still receiving VDZ); the most common causes of death were infections (n=5), cardiovascular diseases (n=5) and cancer (n=3).

Conclusion

About one-third of pts were still receiving VDZ after a fup of more than 6 years, with the majority of them being in clinical remission and more than 50% also having quiescent endoscopic activity at the last observation; bionaive CD patients had longer persistence compared to bioexposed ones. During the LTE, CD pts had higher rates of IBD-related hospitalizations; most hospitalizations were due to the necessity of surgery. SAEs were observed in a minority of patients and predominantly consisted of infections; the rates of cancer diagnosis are seemingly in line with those of the general population. Overall, our study strengthens the notion that vedolizumab is associated with a long-term real-life durability in terms of both effectiveness and safety.

Disclosure

Daniela Pugliese received speaker fees and/or advisory board from AbbVie, Galapagos, MSD, Takeda and Janssen, Pfizer. Giuseppe Privitera received consultancy fees from Alphasigma and Janssen. Alessandro Armuzzi: consulting and/or advisory board fees from AbbVie, Amgen, Biogen, Boehringer Ingelheim, Bristol-Myers Squibb, Celgene, Celltrion, Eli-Lilly, Ferring, Galapagos, Gilead, Janssen, MSD, Mylan, Pfizer, Protagonist Therapeutics, Roche, Samsung Bioepis, Sandoz, Takeda; lecture and/or speaker bureau fees from AbbVie, Amgen, Arena, Biogen, Bristol-Myers Squibb, Eli-Lilly, Ferring, Galapagos, Gilead, Janssen, MSD, Mitsubishi-Tanabe, Novartis, Pfizer, Roche, Sandoz, Samsung Bioepis, Takeda; and research grants from MSD, Pfizer, Takeda and Biogen.