Exabis Library

Educational objectives
1) To understand the limitation of current endoscopic scoring systems and why artificial intelligence may help 
2) To understand the basic principles of developing artificial intelligence with its potential and limitations
3) To understand how artificial intelligence may influence patient management in the future

Educational objectives:
1/ To understand the limitation of human interpretation of endoscopic images
2/ To review the rational and goals for implementing artificial intellegence in image processing
3/ To understand the different types of artificial intelligence applications in image processing
4/ To have an overview of the future potential application of artificial intelligence in IBD endoscopy

There is a scarcity of data regarding the role of histopathology evaluation in Crohn’s disease (CD) and how it relates to endoscopy. Mirikizumab (miri) has shown efficacy in Ph2 studies in CD and ulcerative colitis (UC). Here we explore the agreement of histologic response and remission with endoscopic outcomes after 12 and 52 weeks of treatment with miri in a phase 2 randomized clinical trial (SERENTIY; NCT02891226) in patients with moderate-to-severe CD.

Patients were randomized 2:1:1:2 across 4 treatment arms (PBO, 200, 600, 1000mg miri) administered intravenously (IV) every four weeks (Q4W) at Weeks 0, 4, and 8. Those who received miri and achieved ≥1 point improvement from baseline (BL) at W12 in Simple Endoscopic Score for Crohn’s Disease (SES-CD) were re-randomized 1:1 into double-blind maintenance to continue their IV dose assignment Q4W (IV/IV) or to 300mg miri SC Q4W (IV/SC); due to small numbers, maintenance IV arms were pooled for analysis. W12 non-improvers and all induction PBO patients received 1000mg miri Q4W from W12 through W52. Biopsies were obtained during endoscopy at W0, 12 and 52 from the edge of the ulcers and the most inflamed mucosa in terminal ileum and 4 colonic segments (ascending, transverse, descending, rectum; N=10/patient, 2 per location with the more severe score used), and scored by central readers blind to study treatment, timepoint, and response status. Histologic endpoints were defined post-hoc but prior to performing the analyses (see Table footnotes for definitions). Cohen’s Kappa Coefficient was used to evaluate the degree of agreement of two measures.

At Week 12, there was an overall 69.6% agreement in histologic and endoscopic response, and 84.2% agreement between histologic remission and endoscopic remission (Table 1). After 52 weeks of miri treatment, the agreement between histologic and endoscopic response was 61.9%, while the agreement between histologic and endoscopic remission was 70.6% (Table 2). Interestingly, in every comparison, the percentage of patients achieving the histologic outcome but not the endoscopic outcome was greater than that of patients achieving the endoscopic outcome but not the histologic outcome.

Histologic and endoscopic outcomes demonstrated fair association after both 12 and 52 weeks of miri treatment. Histologic endpoints, despite the stringent criteria, appear to be more sensitive to change than endoscopic endpoints. However, variability of biopsy sampling may be an additional factor contributing to these differences.

The human gut is colonized by diverse microbes including fungi such as Candida albicans, which has been implicated in the pathogenesis of Crohn’s disease (CD). Intestinal C. albicans typically exists as commensal yeast, but can also form tissue-invasive filaments that secrete Candidalysin toxin to disrupt epithelial barriers. Commensal gut bacteria produce molecules that induce host immunity, including phosphoantigen HMB-PP which stimulates Vδ2+ T-cell homing to intestine and mucosal expression of IL-22 to promote barrier integrity. We aimed to identify mechanisms of gut barrier defence against fungal invasion in health and CD.

Colonic biopsies / resected tissue were stimulated or not with HMB-PP in the presence or absence of anti-IL-22 blocking antibody and analysed by flow-cytometry / microscopy to determine extent of fungal outgrowth. Fungal isolates were identified using standard morphological and biochemical criteria before assessment of filamentation, cell wall composition, and NETosis induction.

Mucosal Vδ2+ T-cell activation with bacterial phosphoantigen HMB-PP, in the presence of epithelial damage-associated cytokine IL-15, potently suppressed growth of endogenous fungi in human colonic organ cultures. Vδ2-mediated fungal suppression required IL-22 and was sufficient to restrict the growth of multiple fungal species, including a range of emerging pathogenic moulds as well as archetypal pathobiont C. albicans. In Vδ2-deficient CD patients, mucosal biopsy stimulation with HMB-PP failed to control fungal growth. CD-derived C. albicans strains also displayed rapid filamentation ex vivo and higher levels of NOD2 ligand chitin than were observed in isolates from healthy controls. Comparing hypoxic with oxygenated cultures that mimic inflamed intestine, pSILAC proteomic analysis of a representative CD-derived C. albicans strain confirmed features of invasive growth (e.g. DAO1, IHD1), whereas a healthy control isolate exhibited metabolic hallmarks of symbiosis (Jen1p, SCH9). These divergent characteristics directly impacted on host anti-fungal immune responses, since the CD-derived strain rapidly induced neutrophil extracellular traps in vitro, whereas the healthy control isolate did not.

These data suggest that commensal bacteria can activate host Vδ2 T-cells to suppress fungal invasion of the gut epithelium via an IL22-dependent mechanism that is deficient in CD patients.

Conflicting evidence exists regarding whether low baseline Triggering Receptor Expressed on Myeloid cells (TREM)1 whole-blood gene expression predicts response or non-response to anti-tumour necrosis factor agents in patients with Ulcerative Colitis (UC) or Crohn’s Disease (CD). This study investigated whether baseline TREM1 expression predicted outcomes following adalimumab treatment in patients with UC or CD in the SERENE Phase 3 studies.

The SERENE studies (SERENE-UC, NCT02065622; and SERENE-CD, NCT02065570) compared a higher adalimumab induction dosing regimen versus the standard regimen in patients with UC or CD. Whole-blood TREM1 expression was analysed by RNA sequencing in patients who received standard-dose adalimumab and had clinical and endoscopic outcomes at Week 8 or 52 in SERENE-UC (n=95 for Week 8 and n=70 for Week 52 outcomes), or Week 12 or 56 in SERENE-CD (n=106 for all Week 12 outcomes; n=48 for clinical and n=50 for endoscopic outcomes at Week 56). Outcome definitions are summarised in Table 1. A 2-sample t-test was used to compare baseline TREM1 gene expression (log2 counts per million) in clinical or endoscopic remitters or responders versus non-remitters or non-responders at Weeks 8 and 52 (SERENE-UC) or Weeks 12 and 56 (SERENE-CD). 

In SERENE-UC, baseline TREM1 expression did not predict clinical remission at Week 8 (Figure 1a; p=0.7942) and was only weakly predictive of clinical remission at Week 52 (Figure 1b; p=0.04997). Further, it was not predictive of clinical response at Week 8 or Week 52, nor of endoscopic remission or endoscopic response (Figure 1c) at Week 8 (p>0.05 in all cases), although a weak correlation was observed with both endoscopic remission (p=0.0484) and endoscopic response (p=0.01162; Figure 1d) at Week 52. In SERENE-CD, baseline TREM1 expression was not linked to endoscopic or clinical outcomes at either Week 12 or Week 56 (p>0.05 in all cases; Figure 2a–d). Stratification by adalimumab quartiles did not increase the ability of baseline TREM1 expression to predict clinical outcomes in either study. Of note, a highly positive correlation was observed between baseline TREM1 expression and neutrophils in endoscopic responders and non-responders in both studies. Regardless of clinical or endoscopic response or non-response, TREM1 expression decreased over time following adalimumab treatment. All results obtained with standard-dose adalimumab were replicated with the higher dose (data not shown). 

The findings of this study suggest that baseline whole-blood TREM1 gene expression is not a robust predictor of clinical or endoscopic outcomes following adalimumab treatment in patients with UC or CD. 

To understand the pros and cons of evidenced based diet therapy for IBD. 
To understand the potential nutritional and psychological dangers of restrictive diets.
To highlight the importance of MDT working to support patients with diet therapy.