Y-ECCO Literature Review: Gaurav Nigam

Empirical evidence on the varying risk levels during the natural history of the disease are lacking. Nevertheless, surveillance intervals in current guidelines are predominantly informed by expert opinion using a top risk approach, which may result in unnecessary procedures for low-risk patients.[4–6] Wijnands and colleagues sought to develop and validate a dynamic prediction model for advanced colorectal neoplasia (aCRN), including high-grade dysplasia and CRC, in IBD patients.[8] This model sought to improve risk stratification, enabling targeted surveillance for high-risk patients while reducing unnecessary procedures for those at lower risk.

Methods

The study pooled data from six cohort studies across Canada, the Netherlands, the United Kingdom, and the United States of America, comprising IBD patients with CRC surveillance indication and at least one follow-up procedure. Patients with prior aCRN, colectomy, or unclear surveillance indication were excluded.

A dynamic prediction model was developed using a landmarking approach based on Cox proportional hazard modelling.[9] This method allows risk assessment at multiple follow-up time points, accounting for changes in patient characteristics and risk factors over time. This approach is particularly useful in situations where the risk of an event (in this case, aCRN) may change as the duration of follow-up increases.

Predictor variables were selected based on existing literature and included extensive disease, PSC, prior dysplasia, male sex, IBD type, post-inflammatory polyps (PIPs), age at IBD diagnosis, and highest grade of endoscopic inflammation.[7] Model performance was assessed using Harrell's concordance statistic (c-statistic) for discrimination and calibration curves. Generalisability was evaluated through internal-external cross-validation. Predictions were made for 5 and 10 years after landmark times. Sensitivity analyses were conducted to evaluate the model's performance with recent data (2010 onwards) and the impact of time since dysplasia. Patients were stratified according to the British Society of Gastroenterology (BSG) and European Crohn's and Colitis Organisation (ECCO) guidelines and presented with predicted risks.[5, 6] This stratification aimed to align the model's predictions with established clinical guidelines and enhance its practical applicability.

Key findings

The study included 3,731 patients across six surveillance cohorts, with a median follow-up of 5.7 years, totalling 26,336 patient-years. During this period, 146 individuals were diagnosed with aCRN. The final model, which incorporated eight selected predictors, demonstrated good discrimination ability. The cross-validation median c-statistics were 0.74 for the 5-year prediction window and 0.75 for the 10-year prediction window. Calibration plots showed good alignment between predicted probabilities and observed event rates. Internal-external cross-validation indicated medium discrimination and reasonable to good calibration across different cohorts, suggesting some variability in the model's performance across diverse patient populations.

Discussion

This study represents a significant advancement in CRC surveillance for IBD patients by addressing limitations of current guidelines.[4–6] Unlike existing approaches that rely heavily on expert opinion and static risk factors, this model accounts for the dynamic nature of IBD and its associated risk factors. By incorporating changes in risk factors over time, the model enables more personalised and adaptive surveillance strategies. If validated and implemented, this model could lead to more targeted surveillance strategies, potentially reducing unnecessary procedures for low-risk patients while ensuring adequate monitoring for those at higher risk.

The study's strengths include the use of a large, diverse cohort and a robust statistical approach. The landmarking method allows for dynamic predictions that can be updated as new information becomes available, making the model more adaptable to individual patient trajectories. However, the variability in model performance across different cohorts highlights the need for further external validation before widespread clinical implementation.

Importantly, the study stratified patients according to the BSG and ECCO guidelines and presented them with predicted risks. This approach aligns the model's predictions with established clinical guidelines, potentially enhancing its practical applicability. For instance, a patient classified as high-risk by BSG/ECCO guidelines but with a low predicted aCRN risk might benefit from less frequent surveillance, while a patient with a high predicted aCRN risk within the same group requires more intensive monitoring.

Future research should focus on formal external validation of the model, exploring its integration into clinical practice, and investigating how predicted aCRN risks can be related to surveillance intervals to create a decision support tool for clinicians. Prospective studies evaluating the impact of implementing this model on patient outcomes and healthcare resource utilisation would provide evidence for its effectiveness in real-world clinical settings.

Additionally, future research should explore the potential of leveraging large datasets from electronic patient records with detailed clinical information. These comprehensive datasets could provide the foundation for developing more robust and generalisable dynamic risk prediction models, ultimately leading to more accurate and personalised clinical decision support tools for CRC surveillance in IBD patients.

Conclusions

While further validation and refinement are necessary, this study represents a significant advance towards a more data-driven, personalised approach to CRC surveillance in IBD patients. The model's ability to provide predicted aCRN risks in relation to established BSG and ECCO surveillance risk groups offers a potential bridge between current guidelines and more personalised risk assessment, potentially optimising resource utilisation while maintaining patient safety.

References

1. Olén O, Erichsen R, Sachs MC, et al. Colorectal cancer in Crohn’s disease: a Scandinavian population-based cohort study. Lancet Gastroenterol Hepatol 2020; 5: 475–484.
2. Olén O, Erichsen R, Sachs MC, et al. Colorectal cancer in ulcerative colitis: a Scandinavian population-based cohort study. The Lancet 2020; 395: 123–131.
3. Bye WA, Nguyen TM, Parker CE, et al. Strategies for detecting colon cancer in patients with inflammatory bowel disease. Cochrane Database of Systematic Reviews.
4. Murthy SK, Feuerstein JD, Nguyen GC, et al. AGA Clinical Practice Update on Endoscopic Surveillance and Management of Colorectal Dysplasia in Inflammatory Bowel Diseases: Expert Review. Gastroenterology 2021; 161: 1043-1051.e4.
5. Lamb CA, Kennedy NA, Raine T, et al. British Society of Gastroenterology consensus guidelines on the management of inflammatory bowel disease in adults. Gut 2019; 68: s1–s106.
6. Gordon H, Biancone L, Fiorino G, et al. ECCO Guidelines on Inflammatory Bowel Disease and Malignancies. J Crohns Colitis 2023; 17: 827–854.
7. Wijnands AM, de Jong ME, Lutgens MWMD, et al. Prognostic factors for advanced colorectal neoplasia in inflammatory bowel disease: systematic review and meta-analysis. Gastroenterology 2021; 160: 1584–1598.
8. Wijnands AM, Penning de Vries BBL, Lutgens MWMD, et al. Dynamic Prediction of Advanced Colorectal Neoplasia in Inflammatory Bowel Disease. Clin Gastroenterol Hepatol; 22. Epub ahead of print 1 August 2024. DOI: 10.1016/J.CGH.2024.02.014.
9. Van Houwelingen HC. Dynamic prediction by landmarking in event history analysis. Scandinavian Journal of Statistics 2007; 34: 70–85.

Profile

Dr Gaurav Nigam is a gastroenterology specialist registrar in the Thames Valley (Oxford) deanery, in the United Kingdom. He is currently pursuing an NIHR-funded doctoral research fellowship focused on machine learning-based risk prediction tools at the University of Oxford. His research interests include gastrointestinal bleeding, endoscopy in inflammatory bowel disease (IBD), and colorectal cancer screening.