Multicentre derivation and validation of a colitis-associated colorectal cancer risk prediction web tool

Abstract

Objective: Patients with ulcerative colitis (UC) diagnosed with low-grade dysplasia (LGD) have increased risk of developing advanced neoplasia (AN: high-grade dysplasia or colorectal cancer). We aimed to develop and validate a predictor of AN risk in patients with UC with LGD and create a visual web tool to effectively communicate the risk.

Design: In our retrospective multicentre validated cohort study, adult patients with UC with an index diagnosis of LGD, identified from four UK centres between 2001 and 2019, were followed until progression to AN. In the discovery cohort (n=246), a multivariate risk prediction model was derived from clinicopathological features using Cox regression. Validation used data from three external centres (n=198). The validated model was embedded in a web tool to calculate patient-specific risk.

Results: Four clinicopathological variables were significantly associated with AN progression in the discovery cohort: endoscopically visible LGD >1 cm (HR 2.7; 95% CI 1.2 to 5.9), unresectable or incomplete endoscopic resection (HR 3.4; 95% CI 1.6 to 7.4), moderate/severe histological inflammation within 5 years of LGD diagnosis (HR 3.1; 95% CI 1.5 to 6.7) and multifocality (HR 2.9; 95% CI 1.3 to 6.2). In the validation cohort, this four-variable model accurately predicted future AN cases with overall calibration Observed/Expected=1.01 (95% CI 0.64 to 1.52), and achieved 100% specificity for the lowest risk group over 13 years of available follow-up.

Conclusion: Multicohort validation confirms that patients with large, unresected, multifocal LGD and recent moderate/severe inflammation are at highest risk of developing AN. Personalised risk prediction provided via the Ulcerative Colitis-Cancer Risk Estimator ( www.UC-CaRE.uk ) can support treatment decision-making.

Keywords: clinical decision making; colorectal cancer; dysplasia; ulcerative colitis.

Figure 1

Flow chart of low-grade dysplasia (LGD) cases included and excluded (centre details in online supplemental table S1).

Figure 2

Kaplan-Meier plots for probability of remaining free of high-grade dysplasia (HGD) or colorectal cancer (CRC) to assess risk stratification and predictive power of multivariate model. (A) Discovery (n=246) and (B) validation (n=198) cohorts stratified by risk score (0 to 3+) defined by final multivariate model at index low-grade dysplasia (LGD) diagnosis to year 5 follow-up (see online supplemental figure S3 for similar results over total years of follow-up).

Figure 3

Ulcerative Colitis-Cancer Risk Estimator (UC-CaRE) clinical decision support web tool user pipeline. (A) Clinician records clinicopathological variables for patient at low-grade dysplasia (LGD) diagnosis (postresection, if performed) for shared decision-making consultation. (B) Simple interface in web tool to input patient characteristics. (C) Plot is created for predicted patient risk of progression to advanced neoplasia, as determined by the multivariate model, at each year of future follow-up up to 10 years, with percentages also provided for consideration.

Figure 4

Ulcerative Colitis-Cancer Risk Estimator (UC-CaRE) online risk report. Paling charts provide a user-friendly display of the predicted cumulative risk of advanced neoplasia at 1, 5 and 10 years since low-grade dysplasia diagnosis/resection given patient characteristics. HGD, high-grade dysplasia.