
Abstract Intestinal ultrasound (IUS) is increasingly used as a point-of-care tool during routine follow-up care for the assessment and longitudinal monitoring of inflammatory bowel disease (IBD). Although its clinical advantages are well established, implementation across Canada remains uneven. Implementation in Canada has been shaped by variable training capacity, infrastructure, reimbursement, workflow, and provincial health-system organization. A virtual national roundtable of experts was convened to explore the practical role of IUS in contemporary Canadian IBD care, identify barriers to wider adoption, and define priorities for future implementation. The panel included adult and pediatric gastroenterologists from multiple Canadian provinces from diverse practice environments. The discussion was guided by 3 domains: clinical utility, barriers to implementation, and strategies for national expansion. IUS is positioned to become an increasingly important component of treat-to-target IBD care in Canada, but its equitable and sustainable adoption will require coordinated attention to training, quality, infrastructure, and health-system value.
Abstract Background Accurate colorectal polyp measurement is essential as it guides surveillance intervals and colorectal cancer risk assessment. Emerging vision-language models (VLMs) can estimate polyp size from endoscopic images, but their performance remains unclear. This study compared accuracy of VLMs for colorectal polyp size estimation. Methods In a single-center prospective study, patients ≥18 years undergoing routine colonoscopy were enrolled in an endoscopic video databank. Polyps were resected and measured ex vivo using on-site microscopy, serving as reference standard. Expert endoscopists provided real-time visual estimation. Still images for each polyp were analyzed by 4 VLMs for size estimation. Identical images were then presented to blinded expert endoscopists for size estimation. Primary outcome was correct classification across clinically relevant size categories (diminutive ≤5 mm, small 5.001-9.999 mm, large ≥10 mm) and secondary outcomes included continuous size estimation accuracy. Results For 132 polyps from 85 patients (78.0% diminutive, 21.2% small, 0.8% large), real-time expert visual estimation achieved highest correct size classification (86.0%; 95% CI 78.8-91.1) followed by non-real-time visual estimation (75.4%; 95% CI 70.0-80.1). Performance among the VLMs was variable (ranging from 70.6%; 95% CI 61.5-78.3 to 23.6% 95% CI 16.7-32.2). Three models (Llama4, Qwen3-VL, Mistral-small3.2) demonstrated categorical classification accuracy not statistically different from non-real-time expert assessment. Conclusion Expert endoscopist assessment achieved highest accuracy for clinically relevant polyp size categorization. While current off-the-shelf VLMs are not ready for routine polyp sizing, several models performed similarly to non-real-time expert estimation, suggesting targeted domain-specific training and calibration may enable future AI systems to assist endoscopists with polyp sizing.
Abstract Background The impact of bowel preparation on outcomes of computer-aided detection (CADe) and mucosal exposure devices (MEDs) is unknown. We examined the association between Boston Bowel Preparation Scale (BBPS) score, CADe use, MED use, and colonoscopic detection metrics in a large real-world cohort. Methods This was a retrospective secondary analysis of a prospective observational cohort at a single academic center (2021-2025). Patients aged ≥45 undergoing elective colonoscopy were included. Adenoma detection rate (ADR), advanced ADR (AADR), sessile serrated lesion detection rate (SSLDR), and adenomas per colonoscopy (APC) were modeled using generalized linear models with piecewise BBPS (threshold 5), withdrawal time, CADe use, MED use, and clinical covariates. Results Three thousand six hundred and sixty-one (3661) patients were included (mean age 65, 51.2% female; 38.4% CADe-assisted, 17.5% MED-assisted). From BBPS 0 to 5, detection improved numerically with increasing preparation quality. Above BBPS 5, no further improvement was observed; AADR and APC declined significantly (odds ratio [OR] 0.63, 95% CI 0.19-0.89; risk ratio [RR] 0.66, 0.46-0.93, respectively). ADR and SSLDR declines were not significant. CADe and MED were independently associated with higher detection across all BBPS levels (eg, ADR at BBPS 5: CADe-assisted 38.4% vs unassisted 28.6%; MED-assisted 38.7% vs 28.6%). MED nearly doubled SSLDR (OR 1.97, 1.38-2.81); CADe did not significantly improve SSLDR (OR 1.21, 0.87-1.67). The benefit of CADe and MED appeared consistent across all preparation categories. Conclusions Adequate, rather than excellent, preparation, is critical to achieve detection thresholds. CADe and MED consistently improved lesion detection across all BBPS scores, supporting their routine use. Clinical Trial Registration ClinicalTrials.gov NCT06822816.
Abstract Background and Aims Bowel urgency is a burdensome symptom of inflammatory bowel disease (IBD) that greatly impacts quality of life. This study evaluated vedolizumab effectiveness on bowel urgency in Canadian patients with ulcerative colitis (UC) and Crohn’s disease (CD). Methods Adults with moderately to severely active UC or CD with baseline Urgency Numeric Rating Scale (UNRS) scores enrolled in the Canadian Patient Support Program (July 2024–March 2025) were included. Bowel urgency and clinical endpoints were evaluated at 2, 6, and 14 weeks. Clinically meaningful improvement (CMI) was defined as UNRS change from baseline ≥3 among patients with baseline UNRS ≥3. UC clinical remission was defined as pM (partial Mayo) score <3. CD clinical response was defined as Harvey-Bradshaw Index (HBI) decrease ≥3 and ≥25% from baseline; remission as HBI <5. Results Of 412 vedolizumab-treated patients (319 UC, 93 CD), most were bio-naïve (UC: 89%, CD: 63%) with median disease durations of 4 years (UC) and 7 years (CD). At baseline, 268 (84%) UC and 74 (80%) CD patients had UNRS ≥3. Median time to CMI was 6 weeks in both cohorts. By week 14, UC patients experienced a 5-point median UNRS reduction (80%) and CD patients a 4-point reduction (62%). Patients achieving CMI at weeks 6 and 14 were significantly more likely to achieve clinical remission. Week 2 urgency outcomes were significant predictors of later outcomes. Conclusions Most UC and CD patients achieved CMI by 6 weeks. CMI was significantly associated with validated clinical outcomes, demonstrating real-world effectiveness of vedolizumab in reducing bowel urgency.
Abstract Objectives To describe Ontario’s colonoscopy quality strategy (OCQS), to review outcomes to date, to assess it against the Health Quality Ontario framework, and to identify areas for improvement. Methods We provide a descriptive analysis of the OCQS, tracing its development from the launch of the ColonCancerCheck program in 2007 through subsequent initiatives, including funding mechanisms, evidence-based standards, systematic data collection, provider-level performance reporting and feedback. The impact of the OCQS is assessed via temporal trends in colonoscopy quality indicators. Results The OCQS has evolved into a multipronged, system-wide strategy emphasizing evidence generation, collaboration with clinical leaders, and embedding quality improvement within regional cancer programs. Key achievements include standardized reporting, provincial and regional accountability structures, and integration of quality metrics into practice. Over time, there has been improvement in colonoscopy and screening indicators, reflecting the impact of the OCQS. Challenges include addressing endoscopists with persistent performance issues, optimizing accountability mechanisms, and sustaining engagement in a resource-constrained system. Conclusions Ontario’s experience demonstrates that implementing colonoscopy quality initiatives within a complex health system requires pragmatism, iterative refinement, and continuous evaluation. Lessons learned may inform other jurisdictions in designing and scaling endoscopy quality strategies.