← Issue №6/ week of Aug 9, 2026/IBD

Multi-modal ensemble learning prediction of anti-TNF treatment response in pediatric patients with Crohn's disease using clinical, radiomic, and deep learning MR enterography features.

From GI Signals issue №6: what this paper found, what it changes, and where it sits against the current standard of care, reviewed by Simon Mathews, MD.

IBD retrospective · n=92 · Aug 7, 2026 · Inflamm Bowel Dis · IF 4.5

Multi-modal ensemble learning prediction of anti-TNF treatment response in pediatric patients with Crohn's disease using clinical, radiomic, and deep learning MR enterography features.

New evidenceCrohn's diseaseanti-TNFartificial intelligencepediatric
Clinical takeawayRetrospective model without prospective validation or clinical implementation pathway. While retrospectively accurate (AUROC 0.817), clinical use requires prospective validation, definition of actionable thresholds, and workflow integration. No bedside action supported currently.
What it foundA multi-modal ensemble model combining MRE radiologist assessment, radiomic features, deep learning features, and clinical variables predicted anti-TNF treatment response (mucosal healing at 36 months) with AUROC 0.817 (95% CI 0.793-0.842) in 92 pediatric CD patients (37 responders, 55 non-responders), outperforming individual feature types (radiologist 0.706, radiomic 0.710, deep learning 0.724).
ContextAnti-TNF therapy is a major advancement for CD, but many patients fail to respond. Currently, no validated pre-treatment predictor of anti-TNF response exists in routine practice. This study attempts to use multi-modal imaging and machine learning to stratify risk in pediatric CD, extending established principles (imaging predicts outcomes) to a novel technical approach.
Measured against the standard of care· ACG 2025 (Lichtenstein et al., Am J Gastroenterol) / AGA Living Guideline 2025 (Scott et al., Gastroenterology) / AGA Clinical Practice Update on Therapeutic Endoscopy in IBD 2026 (CGH)

Decision at stakeWhether to initiate anti-TNF therapy in pediatric patients with Crohn's disease

No single passage of this standard matched the paper closely enough to quote, so none is shown. The standard is cited above.

Our full summary of this standard

Two current US guidelines govern Crohn's disease and should be read together. ACG 2025 (Lichtenstein et al.) remains the reference for diagnosis, mild-to-moderate disease, perianal, postoperative and abscess management. DIAGNOSIS: fecal calprotectin (>50-100 μg/g) separates inflammatory from noninflammatory disease of the colon. Workup is ileocolonoscopy with biopsies, including uninvolved mucosa, plus small-bowel imaging. CTE and MRE are comparable; prefer MRE in patients <35 y and when serial exams are likely. Intestinal ultrasound is an ungraded adjunct. Do routine CRC surveillance in Crohn's colitis. MILD-TO-MODERATE: recommend against oral mesalamine; sulfasalazine only for symptomatic mild colonic CD; ileal-release budesonide 9 mg for induction in ileocecal disease but not for maintenance. MODERATE-TO-SEVERE (AGA Living Guideline, Scott et al., Gastroenterology 2025;169(7):1397-1448, 16 recommendations: 1 strong, 9 conditional, 6 knowledge gaps; updated quarterly, adult outpatients only): RECOMMEND infliximab, adalimumab, ustekinumab, risankizumab, mirikizumab, guselkumab or upadacitinib over no treatment; SUGGEST certolizumab pegol or vedolizumab over no treatment. Advanced-therapy-naive: suggest higher-efficacy agents (infliximab, adalimumab, vedolizumab, ustekinumab, risankizumab, mirikizumab, guselkumab) over lower-efficacy ones (certolizumab pegol, upadacitinib). Prior advanced-therapy (especially anti-TNF) exposure: suggest higher- or intermediate-efficacy agents (adalimumab, risankizumab, guselkumab, upadacitinib; ustekinumab, mirikizumab) over lower-efficacy ones (vedolizumab, certolizumab pegol). Suggest against thiopurine monotherapy for induction; suggest it for maintenance of remission. Suggest SC/IM methotrexate monotherapy; suggest against oral methotrexate. In thiopurine-naive patients starting infliximab, suggest infliximab + thiopurine over infliximab alone. No recommendation on IFX + MTX, adalimumab + immunomodulator, or non-TNF biologic + immunomodulator. After ≥6 months of steroid-free remission on anti-TNF + immunomodulator, suggest withdrawing the immunomodulator and suggest against withdrawing the anti-TNF. SUGGEST upfront advanced therapy over step-up therapy (corticosteroids and/or immunomodulator first). No recommendation on treating to endoscopic vs symptomatic remission. The ACG 2025 moderate-to-severe recommendations (anti-TNF, IFX + thiopurine combination and IL-12/23/IL-23 agents, each strong/moderate; upadacitinib after anti-TNF exposure; steroids for short-term induction only; TPMT/NUDT15 testing before thiopurines) still stand alongside AGA. ACG perianal (infliximab first; drain abscesses and place setons before escalating), postoperative (endoscopy at 6-12 months; anti-TNF strong, vedolizumab conditional in high-risk patients) and intra-abdominal abscess (>2 cm: antibiotics plus drainage, hold immunosuppression) positions are unchanged. THERAPEUTIC ENDOSCOPY (AGA Clinical Practice Update, CGH 2026; details below come from AGA news coverage, not the full text, and are ungraded advice): get cross-sectional imaging before the procedure. Endoscopic balloon dilation is first-line for short strictures (about ≤4-5 cm) with few risk factors (fistula, prestenotic dilation >5 cm, high CRP, prior anti-TNF). Longer or multiple strictures, or ≥3 risk factors, favor surgery.

ACG 2025 (Lichtenstein et al., Am J Gastroenterol) / AGA Living Guideline 2025 (Scott et al., Gastroenterology) / AGA Clinical Practice Update on Therapeutic Endoscopy in IBD 2026 (CGH) · reviewed 2026-10-11 ↗
Li H … Dillman JR · Inflammatory Bowel Diseases · IF 4.5 · PubMed ↗Permalink
← Read the whole of issue №6 Every paper GI Signals surfaces gets a page like this one. All issues