Educational Objective:
1. To review the indications for standard indications for hyperbaric oxygen therapy (HBOT)
2. To understand the evidence regarding the role of HBOT for the treatment of acute severe ulcerative colitis
3. To understand the evidence regarding the role of HBOT for the treatment of ileoanal pouch complications
4. To review the practicalities and limitations of HBOT

Educational objectives: 

1. To understand why combination therapy is being considered? 
2. To review what we have learned from the past, when combination works
3. To review what we have learned from the past, when combination does not work
4. To review what we have learned from the past, when combination is bad and dangerous
5. To discuss and review combination therapy in IBD today and tomorrow 

Ustekinumab (UST) and vedolizumab (VDZ) are widely used to treat patients with Crohn’s disease (CD). However, limited data exist regarding comparative effectiveness of these agents for patients with CD who have failed anti-TNF treatment. We aimed to compare the efficacy of UST and VDZ utilizing the largest cohort of CD patients who failed anti-TNF in real world clinical practice.

We conducted a retrospective cohort analysis using data retrieved from the UK IBD BioResource, capturing 34,148 subjects. We identified patients with CD, who failed anti-TNF and were subsequently treated with UST or VDZ as second or third-line therapy. Inverse probability of treatment weighting (IPTW) was used to balance groups using a propensity score-weighting approach accounting for baseline patient or disease related characteristics. Persistence on therapy with clinician assessment of treatment success, without the need for treatment change or surgery was used to estimate the response to treatment. We compared treatment survival curves before and after IPTW and used a log rank test for differences between groups

654 CD patients received VDZ, either as second line (51%) or third line (49%) therapy. 365 patients received UST, 52% as a second line and 48% as a third line therapy. All patients received either infliximab or adalimumab as first and/or second biologic therapy. Baseline characteristics are detailed in table 1. Following IPTW, variables were well balanced. Patients receiving VDZ showed similar rates of treatment success compared to UST as second- and third-line biologic agent after anti TNF failure (before IPTW adjustment, log rank p 0.241; after IPTW, log rank p 0.154). Outcomes for UST were similar between 2nd and 3rd line usage (p 0.81), but outcomes for VDZ were significantly worse when used 3rd line compared to 2nd line (p <0.0001).Subgroup analysis of unadjusted survival data showed significantly better outcomes for patients with ileal disease distributiontreated with UST compared to VDZ (p=0.043) but no significant differences in outcomes for subgroups with colonic or ileocolonic disease.We estimate persistence on UST and VDZ to be 67%, 54%, 49% and 49% at 1, 2, 3 and 5 years respectively.

Using data from a multi-institutional cohort of patients with CD with larger number of participants and longer follow-up than previous cohorts, we demonstrate no difference between UST and VDZ used as second and/or third line biologic therapy, after anti-TNF failure. Subgroup analysis reveals some patient characteristics predictive of differential treatment response.

We compared the efficacy of adalimumab, infliximab, ustekinumab, and vedolizumab for achieving endoscopic healing (EH) in the ileum and colon after one-year of therapy in Crohn’s disease (CD).

A pooled analysis of patient-level data from 344 patients with CD from four clinical trial programs was performed. Patients who received continuous adalimumab, infliximab, ustekinumab, or vedolizumab throughout the trial and had at least one ileocolonic segment with a Simple Endoscopic Score for CD (SES-CD) ≥ 3 at enrolment were included. Proportions of patients achieving one-year endoscopic healing (EH), defined as SES-CD of 0, using each of four biologics were compared. Multivariate logistic regression was used to model the relationship between individual biologics and one-year outcomes, adjusted for potential confounders of EH, including disease duration, concomitant corticosteroid use, and prior anti-TNF failure.

Compared to vedolizumab [10/77 (13%)], both infliximab [29/79 (36.7%), aOR: 3.27 (95% CI: 1.34-8.01), p<0.001] and adalimumab [12/40 (30%), aOR: 3.01 (95% CI: 1.10-8.21), p=0.032] were superior for achieving one-year EH of the ileum among patients with ileal involvement at baseline.  No difference was observed between ustekinumab [5/22 (22.7%)] and vedolizumab [aOR: 2.75 (95% CI: 0.76-9.91), p=0.123].  In biologic-naïve patients, ustekinumab, adalimumab, and infliximab were superior to vedolizumab for achieving one-year EH of the ileum.  For colonic disease, in comparison to ustekinumab [9/31 (29.0%), adalimumab [30/48 (62.5%), aOR: 4.04 (95% CI: 1.88-8.71), p<0.001] and infliximab (55/105 (52.4%), aOR: 2.02 (95% CI: 1.03-3.99), p=0.041] were superior for one-year EH in the colon among patients with colonic involvements at baseline.  No difference was seen between vedolizumab [26/87 (29.9%)] and ustekinumab [aOR: 1.01 (95% CI: 0.39-2.59), p=0.987].  Similar differences were noted among biologic-naïve patients.

In this post-hoc analysis of pivotal clinical trials, TNFα antagonists were generally superior to vedolizumab and ustekinumab for achieving EH of the ileum and colon in patients with CD. However, among biologic-naïve patients, ustekinumab, adalimumab, and infliximab were superior to vedolizumab for attaining one-year EH of the ileum.

To gain insight into vedolizumab (VDZ) use as a first-line biologic in Crohn´s Disease (CD), this real-world study aimed to assess, within the maintenance phase, the 1-year comparative effectiveness and persistence of VDZ vs anti-TNF therapy in biologic-naïve CD-patients.

Between 2017-2020, 1200 consecutively enrolled biologic-naïve and biologic-experienced patients with ulcerative colitis (UC) and CD were prospectively included in the VEDO_IBD-Registry from 45 IBD-experienced centres across Germany. 294 biologic-naïve CD-patients starting a new therapy with VDZ or anti-TNF (adalimumab: ADA or infliximab: IFX) were included in this real-world evidence (RWE) study. The Kaplan-Meier was used to summarize the treatment persistence from the start of therapy through week-52. The primary outcome was week-52 clinical remission (HBI ≤ 4). Patients were analyzed on a modified intent-to-treat basis (mITT; switchers considered as outcome failure) and on a per-protocol (PP) basis (excluding switchers). To reduce selection bias in the estimation of treatment effects, the inverse probability of treatment weighting propensity score (PS) was implemented. A weighted logistic regression was used to evaluate the effectiveness. The results were reported as odds ratio (OR) and 95% confidence interval (CI).

71 VDZ and 223 anti-TNF (ADA: 59.6%, IFX: 40.4%) biologic-naïve CD-patients were evaluated. 52-weeks after treatment initiation approximately 94% of VDZ patients were still in continuous treatment vs 75% of ADA and 78% of IFX (Figure 1). The mITT 1-year clinical remission rate was 76.1% for VDZ vs 63.8% for anti-TNF (OR: 1.80, 95% CI: 0.86-3.76). Similar results were observed for VDZ vs IFX (Table 1). In contrast, the clinical remission was significantly higher in the VDZ group than in the ADA group (OR: 2.24, 95% CI: 1.04-4.85). The PP analysis suggested comparative effectiveness, having excluded more anti-TNF switchers. 91.7% of week-14 responders VDZ patients were in clinical remission from week 14 through 52 vs 66.1% of anti-TNF patients (OR: 5.69, 95% CI: 1.66-19.5). Similar, significant, results were observed for VDZ vs ADA and for VDZ vs IFX (Table 2).

In this real-world setting comparing VDZ and anti-TNF in biologic-naïve patients via PS weighted analysis, VDZ showed especially in week-14 responders higher clinical remission rates in comparison to anti-TNF. The higher treatment persistence observed for VDZ, perhaps due to a more favourable safety profile vs anti-TNF, may be considered the main driver for the better effectiveness of VDZ at one year. These findings may aid physicians’ decision-making on the choice of VDZ as the first-line biologic for CD.

Main aim: To evaluate the retention rate of ustekinumab compared to vedolizumab in Crohn’s disease patients who failed anti-TNF therapy in clinical practice. Secondary aims: To compare the short-term and long-term effectiveness, and the safety of both treatments.

755 patients were included (195 in the vedolizumab cohort and 560 in the ustekinumab cohort). After a median of 20 months (IQR 7.4-30) of follow-up, the survival rate for ustekinumab therapy was higher than vedolizumab (Figure 1). The propensity matching score verified the differences between both therapies. The short-term proportion of patients on clinical remission, steroid-free remission and clinical response was also superior in the ustekinumab cohort (Figure 2). In the long-term, significant differences were observed 2 years after the beginning of the treatments, although no differences in clinical response and remission rates were detected in patients who achieved clinical response at week 16 between both cohorts. Vedolizumab was discontinued in 142 patients and ustekinumab in 185, mainly due to primary non-response (52% in the vedolizumab and 58% in the ustekinumab cohort) and loss of response (34% and 25%, respectively) despite the fact that 35% of the patients required intensification. The predictive factors associated to the discontinuation of the therapy are described in table 1. Adverse events were observed, overall, in 12% of the patients, without differences between both groups (Table 2). Following the discontinuation of the treatment with vedolizumab/ustekinumab, other biologic agents were prescribed in 56% of the patients, and 27% underwent surgery.

In clinical practice, a relatively high proportion of Crohn’s disease patients who received ustekinumab or vedolizumab for anti-TNF failure, maintained these drugs in the medium-long term, although ustekinumab retention rate was higher in comparison with vedolizumab.

The modified Rutgeerts' score (mRS) differentiates i2a – lesions confined to the anastomosis – and i2b – neoterminal ileum lesions – categories. Its relevance for therapeutic management of Crohn's disease (CD) patients after ileocolonic resection is still debated. Our objective was to compare the postoperative recurrence (POR) risk in patients with a mRS i2a or i2b score using an individual patient data meta-analysis.

We conducted a systematic literature search of Medline, Embase and abstracts from international conferences (until July 2020) to identify all relevant studies reporting the risk of clinical and/or surgical POR and the i2a/i2b status in the year following ileocolonic resection. Initial datasets were obtained from the corresponding authors. Time from endoscopy to clinical and surgical POR was estimated using Kaplan-Meier method. The association between time to event and mRS was evaluated using a mixed Cox with centre as the random effect.

From the 17 studies identified, 7 published between 2008 and 2019 (cohort studies, n=4; clinical trials, n=2) corresponding to a total of 400 patients (median (InterQuartileRange) age at surgery 34 (26,47) years; 52% female) were included. In the year following ileocolonic resection, 189 (47%) patients displayed an i2a mRS and 211 (53%) an i2b. In the i2b group, we observed more male patients (56% versus 41%, p=0.01), more patients with previous ileocolonic resection (31% versus 21%, p=0.03) and temporary ileostomy (14% versus 6%, p=0.03) and an immunosuppressant or antiTNF therapy was more frequently initiated after endoscopy (42% versus 26%, p<0.01 and 36% versus 54%, p<0.01, respectively). The risk of clinical POR at 1, 3 and 5 years was 11% [6%-15%], 25% [18%-32%] and 36% [27%-43%] in the i2a group and 9% [5%-13%], 33% [26%-41%] and 47% [39%-56%] in the i2b group (p=0.63, p=0.12, et p=0.05 respectively). No significant difference was observed in terms of time to clinical POR (Hazard Ratio (HR)=1.27; Confidence Interval 95% [0.91,1.76]; p=0.16) (Figure 1) or surgical POR (HR=0.94; CI95% [0.44,2.00]; p=0.87). After exclusion of patients having initiated an immunosuppressant or a biologic in the 3 months after endoscopy (remaining cohort, n=361), no difference was observed in terms of clinical POR (HR=1.29 [0.92,1.80]; p=0.13) or surgical POR (HR=0.85 [0.39,1.84]; p=0.68).

In this individual patient data meta-analysis, no difference is observed between i2a and i2b mRS subcategories in terms of clinical, surgical or endoscopic POR. Limits of the mRS may explain this lack of predictive value.

Learning Objectives - Risks associated with anti-TNF treatments
1. Hypersensitivity reactions
2. Dermatological adverse effects with anti‐TNF therapy
3. Autoimmune‐like disorders
4. Infections and management strategies
5. Malignancy
6. Patients’ selection for anti TNFs vs other biologics

Educational objectives: 

1. To understand the importance of histological remission in IBD
2. To understand the hurdles in the assessment of histological remission in IBD
3. To understand the potential of automated histological scoring based on artificial analysis

HMGB1 is a ubiquitously expressed nucleoprotein with proinflammatory functions following cellular release. The protein is passively released during tissue necrosis, acting as a damage-associated molecular pattern, but can also be actively secreted by immune cells. Stool and serum HMGB1 levels have been suggested as markers of both inflammatory bowel disease (IBD) activity and colorectal cancer (CRC) invasiveness, and antibody-mediated HMGB1 neutralization was beneficial in animal models of IBD. We explored context-dependent functions of HMGB1 in the injured intestine using novel experimental mice with cell-specific genetic HMGB1 deficiency.

To circumvent the postnatal lethality of global HMGB1 deficiency in animals, we used the Cre-lox system to generate enterocyte-specific (Hmgb1^ΔIEC, Villin-Cre) and myeloid cell-specific (Hmgb1^ΔLysM, LysM-Cre) HMGB1-knockout mice. Animals were subjected to well-established models of acute (DSS, Citrobacter rodentium) and chronic (AOM+DSS, CD45RBhigh T cell transfer colitis, Apc^+/min) intestinal injury, followed by clinical, endoscopic, histological and molecular analysis. HMGB1 expression was assessed in human IBD and CRC specimens.

IBD and CRC biopsies exhibited high levels of HMGB1 expression in epithelia, immune cells, tumor cells and the peritumoral stroma. HMGB1 deficiency from enterocytes and myeloid cells did not alter Citrobacter- or T cell transfer-induced enterocolitis, when epithelial injury was comparably low. In contrast, Hmgb1^ΔIEC mice exhibited aggravated DSS-induced colitis, as evidenced by severe weight loss as well as exacerbated neutrophil- and monocyte-driven mucosal inflammation compared to Hmgb1^f/f. Whole tissue RNA sequencing indicated defective cellular proliferation in injured Hmgb1^ΔIEC intestines. In the AOM+DSS-model, Hmgb1^ΔIEC had a comparable tumor burden to Hmgb1^f/f, whereas Hmgb1^ΔLysM had significantly fewer and smaller tumors, potentially linked to metabolic alterations in the tumor micromilieu. In the Apc^+/min model, enterocyte HMGB1 deficiency effectuated more and larger tumors, whereas leukocyte HMGB1 did not affect tumor load.

Contrasting antibody-mediated HMGB1 neutralization in animal models of IBD, our findings from genetic HMGB1 deletion studies reveal a critical role of enterocyte HMGB1 in the maintenance of the intestinal barrier during severe colitis. Impaired epithelial regeneration or inefficient local immune cell expansion in Hmgb1^ΔIEC may account for the aggravated phenotype. HMGB1 from enterocytes and immune cells context-dependently affect maladaptive intestinal would healing, potentially mediated by cell-intrinsic and -extrinsic mechanisms that warrant further investigation.

Summary content

Some studies have shown decreased serological response to vaccination in patients on anti-tumor necrosis factor (TNF) medications. While the large majority of these patients do seroconvert after vaccination, titers have generally been lower and one study showed reduced neutralizing and inhibitory functions. One real-world population-based study compared found no increased infection rate in anti-TNF treated patients, but infection rates were low. The low event rate mandates exploration in longer-term population-based data. We used the epi-Israeli IBD Research Nucleus (IIRN) database to explore the effectiveness of COVID-19 vaccination in IBD patients in Israel.

We included all IBD patients insured in two of the four Israeli HMOs, covering 35% of the population, by validated algorithms, and selected those who received two doses of Pfizer BNT162b2 vaccine. These were matched by date of vaccination ±3 days and demographic variables to non-IBD controls. The primary outcome was incidence of positive COVID-19 PCR following vaccination between December, 2020 to June, 2021.

12,640 IBD patients received two vaccine doses; the matched cohort included 4,946 matched pairs (total 9,892 subjects). Mean age was 50.5±16.1 years and median follow-up was 22 weeks (range 4.1-24.4). Fifteen (0.3%) vaccinated IBD patients tested positive compared with 15 (0.3%) vaccinated non-IBD controls (OR=1 [95%CI 0.49-2.05], p=1.0). Patients on anti-TNF and/or corticosteroids did not have a higher incidence of positivity – neither compared to the entire group nor to IBD patients treated with vedolizumab/ustekinumab, even after precise matching for demographics, underlying diseases and IBD severity.

In a large population-based cohort of IBD patients in Israel, vaccine effectiveness was equivalent to non-IBD controls and was not influenced by treatment with anti-TNF or corticosteroids. Notwithstanding previous findings of impaired serological response in anti-TNF treated IBD patients, this real-world large-scale study shows that vaccine protection is robust in IBD patients, including those on immunosuppressive medications.

Robust COVID-19 vaccine-induced antibody (Ab) responses are important for protective anti-viral immunity. Data are urgently needed to determine whether vaccine-induced immunity is impacted by commonly used immunosuppressive drug regimens in IBD.

We prospectively recruited 447 adults (90 healthy controls and 357 IBD) at nine UK centres. The IBD study population was established (>12 weeks therapy) on either thiopurine monotherapy (n=78), infliximab (IFX) monotherapy (n=61), thiopurine & IFX combination therapy (n=70), ustekinumab (uste) monotherapy (n=56), vedolizumab (vedo) monotherapy (n=62) or tofacitinib (tofa) monotherapy (n=30). Participants had two doses of either ChAdOx1 nCoV-19, BNT162b2 or mRNA1273 vaccines. The primary outcome was anti-SARS-CoV-2 spike (S1 RBD) Ab concentrations, measured using the Elecsys anti-SARS-CoV-2 spike (S) Ab assay, 53-92 days after second vaccine dose, in participants without prior infection, adjusted by age & vaccine type. Secondary outcomes included proportions failing to generate protective Ab responses (defined cut-off anti-S concentration 15 U/mL, which correlated with 20% viral neutralization).

Geometric mean S Ab concentrations (figure 1) were lower in patients treated with IFX (153U/mL;p<0.0001), IFX and thiopurine combination (109U/mL;p<0.0001), tofa (430U/mL;p<0.0001) and uste (561U/mL;p=0.013) compared to controls (1596U/ml). No differences in S Ab concentrations were found between controls and thiopurine monotherapy-treated patients (1020U/mL;p=0.62), nor between controls and vedo-treated patients (944 U/mL;p=0.69). In multivariable modelling (figure 2), lower S Ab concentrations were independently associated with IFX (FC 0.10 [95% CI 0.07-0.14], p<0.0001), tofa (0.36 [95% CI 0.19-0.69],p=0.002) and uste (0.56 [95% CI 0.31-1.00],p=0.049), but not with thiopurine (0.77 [95% CI 0.54-1.11],p=0.17) or vedo (1.01 [95% CI 0.61-1.68],p=0.96). mRNA vaccines (3.67 [95% CI 2.72-4.96],p<0.0001) and older age (0.82 [95% CI 0.73-0.91],p=0.0003) were independently associated with higher & lower S Ab concentrations respectively. Protective Ab responses were generated by all thiopurine monotherapy, vedo, tofa and healthy control participants, but not by 11% of patients on IFX monotherapy, 13% on thiopurine & IFX combination therapy and 4% on uste.

COVID-19 vaccine-induced Ab responses are significantly reduced in patients treated with IFX, or tofa, and to a lesser extent with uste.  No significant reduction was seen in vedo or thiopurine monotherapy-treated patients. Our data suggest that 3rd primary or booster vaccine doses for IBD patients might be tailored to an individual’s immunosuppressive treatment.

Financial support was provided as a Research Grant by Pfizer Ltd.

Educational objectives:
1. To understand the mechanisms of action of hyperbaric oxygen therapy in IBD
2. To review current knowledge regarding the application of hyperbaric oxygen therapy in Crohn's Disease
3. To have an overview of potential areas of focus for future research 

Educational objectives:
1. To understand the phylosophy of treatment de-escalation or elective switch
2. To understand in which situation a treatment de-escalation or elective switch can be contemplated
3. To emphasize the clinical results with treatment de-escalation
4. To emphasize the need and methods of disease monitoring and management after de-escalation or elective switch

The phylosphy of treatment de-escalation or elective switch is, after achieving a state of deep remission,  to keep it with the lightest of the best tolerated treatment without jeopardizing disease control. The principle behind this is that the treatment needed once a deep state of remission has been achieved may be substantially different from the one to achieve and initially maintain this remission. Recentent studies have highlighted that not only a sustained steroid-free remission was necessary before such de-escalation, but also the need for enodscopic healing and also normalisation of biomarkers assosiated by biological disease activity. As the disease may relapse any time in up to 50% of the patients over one year and up to 80% over 7-8 years, a regular monitoring is necessary to restore appropriate treatment as soon as possible.

Educational objectives:
1. To give an overview of the main cutaneous manifestation of IBD 
2. To understand the importance of the clinico-pathological correlation for their diagnosis 

Crohn’s disease (CD) is a lifelong chronic illness with recurrent relapsing and remitting disease course that requires close follow ups and reassessments of disease status as well as screening for complications throughout a patient’s lifetime. Imaging plays a crucial role in the diagnosis and evaluation of CD. Currently, different imaging modalities can be used over the disease course of a patient’s lifetime, from the diagnosis of the disease, to determining the extent of intestinal involvement, monitoring for disease activity, and evaluating for CD related complications. Intestinal Ultrasound (IUS) is a non-invasive, radiation-free, and safe useful imaging tool that can be used in the diagnosis and management of crohn’s disease. 

Educational Objectives:

1. To emphasize on the increased rule of imaging in IBD management.
2. To review the rule of Intestinal ultrasound (IUS) in detecting Crohn’s disease complications:
   1. Strictures.
   2. Abdominal fistulas.
   3. Inflammatory Masses.
3. To review the latest guidelines statements for the use of intestinal Ultrasound in detecting crohn’s disease complications.
4. Quick overview on the innovated tools to improve the accuracy of intestinal ultrasound in detecting crohn’s disease complications.

Interactive cases will be presented to illustrate IUS diagnosis of UC complications, such as acute severe colitis, including monitoring of response to treatment, perforation, chronic structural damage in UC, infectious colitis, pseudopolips and lymphoma.

Background

Capsule endoscopy (CE) is the gold-standard for the evaluation of the enteric mucosa in patients with suspected or known inflammatory bowel disease, particularly Crohn’s disease. Ulcers and erosions of the small bowel are common findings and their identification in CE is paramount for an accurate disease stratification.

Several artificial intelligence (AI) algorithms have been developed to aid endoscopists to detect lesions in different endoscopic modalities. With this project we intend to develop and test an AI algorithm for the automatic identification of ulcers and erosions in the small bowel mucosa.

Methods

A total of 2565 CE exams from two different centers (1483 from São João University Hospital and 1082 from ManopH Gastroenterology Clinic) were used to develop the Convolutional Neural Network (CNN). 55320 frames of the enteric mucosa were obtained, 18396 containing enteric ulcers and erosions, and 36924 containing normal mucosa. 90% of the frames were used to develop the training dataset and 10% were used to test the network. The patients included on the training dataset were excluded from the testing dataset. This patient split brings the technology performance closer to that of a real-life setting. The output provided by the CNN was compared to the classification provided by a consensus of experts.

Results

Our model was able to automatically detect ulcers and erosions in the enteric mucosa with an accuracy of 93.2%, sensitivity of 90.4% and a specificity of 93.9%. The mean processing time for the validation dataset was 29 seconds (approximately 306 frames/second).
An example of the output obtained after the network application can be seen in Figure 1.

Conclusion

The authors developed a CNN for the automatic identification of enteric ulcers and erosions in CE videos and tested it in AI naïve patients. This represents an evolution in the technology readiness level into a real-life clinical setting, that will surely improve the diagnostic yield of CE exams, which will ultimately translate into better patient care.

Early prediction of intravenous corticosteroid (IVCS) resistance in Acute severe ulcerative colitis (ASUC) patients could reduce costs and delay in rescue therapy. However, most prediction models for ASUC were at high risk of bias with a lack of external validation. This study aims to construct and validate a model that accurately predicts IVCS resistance using various statistical methods.

A retrospective cohort of patients who were diagnosed with ASUC and had undergone IVCS treatment between March 2012 to January 2020 was established. Predictors evaluated included age, gender, race, medications before admission, infections, and laboratory data at baseline and during IVCS treatment, and endoscopic outcomes relied on blinded centralized endoscopy reading. The LASSO regression was used in feature selection for multivariate logistic regression model. Models based on machine learning methods (decision tree and random forest [RF]) were also constructed. Internal validity was confirmed and model performances were compared. External validation was conducted using data using an independent cohort from a tertiary referral centre.

A total of 129 patients were included in the derivation cohort. During index hospitalization, 102 (79.1%) responded to IVCS, and 27 (20.9%) failed; 16 patients underwent colectomy, 6 received cyclosporin, and 5 succeeded with IFX as rescue therapy. Ulcerative Colitis Endoscopic Index of Severity (UCEIS; odds ratio [OR] 5.39, 95% confidence interval [CI] 2.52-14.0, p<0.001) and C-reactive protein (CRP) level on the third day (OR 1.05, 95% CI 1.03-1.08, p<0.001) were selected by LASSO regression and identified as the only two independent predictors of IVCS resistance in logistic regression. The decision tree model identified a UCEIS higher than 6.5 points and CRP level at day 3 higher than 33.57 mg/dL as the proxy for IVCS resistance. UCEIS and CRP level at day 3 were also the most important predictors in the RF model. Areas under the curve receiver operating characteristic (AUC) of logistic model, decision tree model, and RF model were 0.64 (95% CI 0.49-0.80), 0.81 (95% CI 0.71-0.90), and 0.88 (95% CI 0.82-0.95), respectively. A validation cohort of 65 ASUC patients were established, and the AUC of the models in external validation were 0.57 (95% CI 0.45-0.70), 0.70 (95% CI 0.61-0.80), and 0.71 (95% CI 0.48-0.94), respectively.

In patients with ASUC, UCEIS and CRP level at day 3 of IVCS treatment appeared to allow the prompt prediction of likely IVCS nonresponders. Machine learning-based models outperformed the traditional statistical model in the prediction. The models may aid therapeutic decision-making in ASUC patients.