DEVELOPMENT OF AN OPTIMAL MACHINE LEARNING MODEL USING TREATMENT NAÏVE DIAGNOSTIC PATHOLOGY IMAGES TO PREDICT STEROID-FREE CLINICAL REMISSION AT ONE YEAR IN PEDIATRIC ULCERATIVE COLITIS

Background: Tumor necrosis factor inhibitors (TNFi) are a mainstay of pediatric Crohn’s disease (PCD) therapy; yet not all patients respond, and others lose response over time. Combination therapy with methotrexate may improve response; however, this has not been rigorously evaluated.

Objective: To compare the effectiveness of TNFi in combination with low dose oral methotrexate to TNFi monotherapy in children with PCD.

Methods: We performed a multi-center, randomized, double-blind, placebo-controlled pragmatic trial. Eligible PCD patients initiating infliximab or adalimumab as standard of care were randomized in 1:1 allocation to a weight-based dose of oral methotrexate or placebo and followed for a minimum of 12 months and maximum of 36 months. The primary outcome was a composite indicator of treatment failure, including toxicity. Secondary outcomes included development of anti-drug antibodies (ADA) and patient reported outcomes (PROs) of pain interference and fatigue. Adverse events (AEs) and Serious AEs (SAEs) were collected.

Results: We recruited 297 participants at 35 centers (Table 1), including 156 in the methotrexate arm (110 infliximab and 46 adalimumab initiators) and 141 in the placebo arm (102 infliximab and 39 adalimumab initiators). Overall, 26% of participants in the combination therapy group and 44% of participants in the monotherapy group experienced treatment failure [hazard ratio (HR) 0.69, 95% confidence interval (CI) 0.45-1.05]. Among infliximab initiators, there were no differences between combination and monotherapy (HR 0.93, 95% CI 0.55-1.56, p=0.78). Among adalimumab initiators, combination therapy was associated with longer time to failure (HR 0.40, 95% CI 0.19-0.81, p=0.01, Figure 1). In the monotherapy group, adalimumab-treated patients had higher failure rates than those receiving infliximab (HR 2.19, 95% CI 1.23-3.89, p=0.008). Combination therapy patients had numerically lower rates of ADA development [(infliximab odds ratio (OR) 0.72 (0.49-1.07); adalimumab OR 0.71 (0.24-2.07)]. No differences in PROs were observed. In comparison to monotherapy, combination therapy resulted in more AEs, including nausea and vomiting and elevated liver enzymes. However, the combination group had fewer SAEs and did not require treatment discontinuation for toxicity more frequently than the monotherapy group.

Conclusions: Compared to adalimumab monotherapy, PCD patients treated with adalimumab and methotrexate experienced a 2-fold reduction in the risk of treatment failure. In contrast, no differences were observed between infliximab combination and monotherapy. A trend towards reduced immunogenicity in the combination therapy group was observed among both infliximab and adalimumab initiators. Overall, these findings suggest improved effectiveness of combination therapy in adalimumab-treated patients with a tolerable safety profile.

Background: Mirikizumab is a humanized, IgG4 monoclonal antibody directed against the p19 subunit of IL-23, a key inflammatory mediator in inflammatory bowel disease. Mirikizumab was superior to placebo in inducing clinical remission as well as other symptomatic, clinical, and endoscopic endpoints and had an acceptable safety profile in adults with moderately to severely active ulcerative colitis (UC) in the Phase (Ph)3 LUCENT-1 induction and the Ph3 LUCENT-2 maintenance studies. Here we report the first PK, efficacy, and safety data in pediatric patients with moderately to severely active UC from the Ph2 SHINE-1 study (NCT04004611), following a 12-week induction period.

Methods: Children and adolescents aged 2 to <18 years old (n=26) with active UC as defined by a Modified Mayo Score (MMS) of 4-9 points and a centrally read Mayo endoscopic subscore ≥2 and who had an inadequate response, loss of response, or intolerance to corticosteroids, immunosuppressants, or biologic therapies, or those with corticosteroid-dependent colitis were enrolled in the study. Patients weighing >40 kg received an induction dose of 300 mg via intravenous (IV) infusion and patients weighing ≤40 kg received 5 or 10 mg/kg via(IV) infusion at Weeks 0, 4, and 8. The primary endpoint evaluated the pharmacokinetics (PK) of mirikizumab treatment to confirm doses for Phase 3. Patients were assessed at Week 12 for clinical response, clinical remission, and endoscopic remission and were monitored for safety throughout the trial.

Results: For pediatric patients >40kg, the observed and PK model-estimated exposures were similar to adult levels. For pediatric patients ≤40kg, the exposure at 5 mg/kg matched the adult range based on both observed and modeled concentrations. Exposure at 10 mg/kg was ~2 fold higher than the adult mean. MMS clinical response and remission, and endoscopic remission rates were consistent with those of the LUCENT adults (Figure 1); PUCAI response and remission rates were acceptable. The safety findings observed in the pediatric patients are consistent with the LUCENT-1 data (Table 1).

Conclusions: In this Ph2 pediatric UC study, mirikizumab demonstrated similar observed PK concentrations to those reported in the adult LUCENT-1 induction study, an acceptable safety profile, and clinically meaningful improvement in clinical response, clinical remission and endoscopic remission following induction treatment with mirikizumab. These results provide the first reported efficacy of a p19 inhibitor in pediatric patients with UC and provide the foundation for the pivotal Ph3 pediatric study.

Background: STRIDE-II guidelines recommend a treat-to-target (T2T) and tight control strategy utilizing non-invasive biomarkers to achieve endoscopic remission (ER) within one year after therapy initiation in Crohn’s disease (CD). Intestinal ultrasound (IUS) is an accurate tool for monitoring CD activity in adults, but accuracy compared to endoscopy and utility as an early treatment target for tight control in children is unknown. We aimed to evaluate the ability of IUS to measure early treatment response and accuracy to T2T endoscopy in children who achieve ER vs. those who did not.
Methods: In a single-center prospective longitudinal cohort study, consecutive children (<18 years) with terminal ileal (TI) CD, with or without colonic disease, and no prior surgery initiating biologics from 09/2020 to 09/2021 were included. IUS, clinical (Pediatric Crohn’s Disease Activity Index (PCDAI)), and biochemical (C-reactive protein (CRP), albumin) assessments were performed at baseline, 8 weeks (post-induction), 6 months, and T2T endoscopy. T2T ER was defined by a Simple Endoscopic Score for Crohn’s Disease (SES-CD) < 3. Primary outcome was measures of TI bowel wall thickness (BWT) at 8 weeks associated with T2T ER. Secondary outcomes included association of week 8 PCDAI, CRP, and albumin with ER, accuracy of T2T BWT, PCDAI, CRP, and albumin to detect ER, and correlation of T2T BWT, PCDAI, CRP, and albumin with T2T SES-CD. Descriptive statistics were reported as frequencies or median [interquartile range [IQR]]. Univariate analyses tested associations. Area under the receiver operating curve (AUROC) determined cut-off values for T2T ER. Correlation was analyzed with a Spearman’s correlation coefficient.
Results: 44 children (17 (39%) female, age 13 [12-17] years, disease duration 19.6 [4.3-115.9] weeks, 29 (66%) biologic naïve) were included (Table 1). 29 (66%) children achieved ER. Post-induction TI BWT was significantly lower in children who achieved ER vs. no ER (6.0 [3.0-8.2] mm vs. 2.9 [1.0-5.2] mm, p<0.0001). Post-induction % change in TI BWT was greater in children who achieved ER vs. no ER (-37.5% [-72.9 to -18.2] vs. 0.0% [-22.7 to 32.3], p<0.0001). Post-induction PCDAI was associated with ER (10 [0-30] with vs. 20 [0-50] without, p=0.005) but CRP (p=0.058) and albumin (p=0.153) were not. T2T TI BWT correlated more strongly with SES-CD (ρ=0.91, p<0.0001) vs. PCDAI (ρ=0.69, p <0.0001), CRP (ρ=0.38, p=0.01), or albumin (ρ=-0.42, p=0.004). T2T TI BWT < 2.5 mm (AUROC, 0.97; 95% confidence interval [CI], 0.86-0.99; p=<0.0001) was the most accurate to detect ER (Table 2).
Conclusions: Post-induction decrease in TI BWT represents a novel, non-invasive, early treatment target for ER in pediatric CD. Future studies are needed to determine if treatment escalation based on early change in TI BWT alone improves rates of ER in pediatric CD.

Background:
The majority of children with ulcerative colitis present with extensive colitis at diagnosis, but the response to therapy is heterogenous. Identifying the optimal window for biologic treatment and risk stratification of patients remains an unmet need.

Aim:
To develop a pathology based histomic model to predict corticosteroid free clinical remission (CSF) with mesalamine alone at one year.

Methods:
292 hematoxylin and eosin diagnostic treatment naïve rectal mucosal biopsies from the multi-center PROTECT study were digitized. Whole slide images (WSIs) underwent two-step pre-processing a) stain normalization and b)informative patch selection (size 512x512). We trained 23 machine learning (ML) models using 250 histomic features (texture, color, histogram and nuclei features) with 5-fold cross-validation, for patch-level classification. Feature importance was determined by the Gini index. We re-trained the classifier using the top features. Slide-level prediction was defined by threshold voting. Performance metrics at the patch and WSI level was evaluated.

A total of 161 patients underwent high-throughput RNA sequencing to define rectal gene expression. We undertook unsupervised weighted gene co-expression network analysis (WGCNA) to discover networks of co-expressed genes with shared biologic functions correlated with histomic features, histological traits, and outcomes.

Results:
187571 informative patches from 292 patients (Male:55%; Age:12.7y (IQR:11-15); CSF remission:41%) were trained on 23 ML classifiers. The best model trained on 250 features at the patch-level was random forest (RF). At a remission ratio threshold of 0.48, WSI area under the receiver operator curve (AUROC) was 0.90 (95%CI:0.70, 1.00), accuracy 90.4%, precision 90.9% and recall 84.7%. 18 top features were identified and trained, and the corresponding WSI AUROC was 0.87 (95%CI:0.72, 1.00), accuracy 90.1%, precision 89.4%, and recall 83.9% (Fig. 1). We re-trained the 18 features on an independent real-world dataset of 131 UC patients and the model WSI AUROC was 0.85 (95%CI:0.74, 1.00) and an accuracy of 88.5%.

Of the 13 modules identified by WGCNA analysis, six modules significantly correlated with clinical and/or histomic features (Fig. 2). Two of the gene co-expression modules were negatively associated with baseline clinical, endoscopic, and histologic measures of severity, and positively associated with nuclei features (Otsu area, perimeter, and equivalent diameter) and the outcome measure of CSF remission. Intersection of genes with adult single cell RNA seq data demonstrated enrichment for enterocytes (SLC26A3) and extracellular matrix (IHH)

Conclusion:
We developed a predictive model for UC disease course using histomic features from standard of care pre-treatment pathology images. Characterization of the underlying molecular basis of the histomic features is ongoing.