Patients with IBD (chronic inflammatory bowel disease) are at increased risk of cancer due to intestinal inflammation and the use of immunosuppressants to which they are exposed at an increasingly early age, whether it is colorectal adenocarcinoma ( CRC) or small intestine, cholangiocarcinoma, or anal cancer.
Thus, patients who have extensive old colitis (ulcerative colitis, UC, or Crohn’s disease, CD, colic) have a 5 to 8 times higher risk of developing CRC. Over a lifetime, this risk can reach 20 to 30% in patients who have pancolitis from adolescence. Prevention and detection of CRC is based on endoscopic surveillance of all patients, and on chemoprevention with 5-aminosalicylates in high-risk patients.
The precise modalities of this colonoscopic surveillance are not the subject of consensus. Dutch authors wanted to develop and validate a dynamic prediction model for the appearance of advanced colorectal neoplasia or NCRA (adenoma measuring more than 10 mm, containing a significant villous component or presenting high-grade dysplasia and cancer) in IBD.
6 international surveillance cohorts, 8 predictors
They pooled data from 6 international surveillance cohort studies (Canada, Netherlands, United Kingdom, United States). Patients were included if they met the following criteria: (i) unclassified UC, CD or IBD involving at least 30% of the colonic surface area, (ii) disease duration >7.5 years or any duration of disease. disease in case of concomitant primary sclerosing cholangitis (PSC), (iii) at least one colonoscopic procedure or colectomy (surgical specimen evaluated by a pathologist).
Exclusion criteria were neoplasia diagnosed before or up to 3 months after study initiation, colectomy before study initiation, or unclear indication for CRC surveillance.
Predictor variables were selected based on the literature. Eight predictors were thus selected: extensive disease, PSC, anterior dysplasia (often low grade), male sex, type of IBD, post-inflammatory polyps, age at diagnosis and highest grade of endoscopic inflammation. A dynamic prediction model was developed using Cox proportional hazards modeling.
The surveillance cohorts included 3,731 patients, recruited and followed during the period from 1973 to 2021, with a median follow-up period of 5.7 years (26,336 patient-years of follow-up assessment); 146 people were diagnosed with CRC. The model contained 8 predictors, with a median cross-validation agreement statistic of 0.74 and 0.75 for a prediction window of 5 and 10 years, respectively. Internal-external cross-validation results showed average discrimination and correct calibration.
A dynamic model for individualized prediction
This study has several strengths: the results are based on a large combined international database on colonoscopic surveillance of IBD allowing the development and internal-external validation of this surveillance model. The pre-selection of predictors is based on an in-depth meta-analysis and literature review.
A dynamic modeling approach enabled repeated use of this model to derive updated risk predictions throughout a patient’s journey. Unlike studies prior to the 2010s, this modeling approach only includes recent data and also made it possible to include time-varying predictors.
Weak points concern the retrospective nature, heterogeneity of colonoscopic surveillance intervals and endoscopic techniques between cohorts/time periods. In this model, anterior dysplasia significantly increased the risk of CRC. For the highest degree of endoscopic inflammation, it should be understood that this time-varying predictor might not adequately describe the severity of inflammation because the course of inflammation is not taken into account.
In addition to the morphology and time of appearance of these lesions, the authors did not have information on colonic stenoses, the presence of a tubular colon, family history of CRC, the use of chromoendoscopy or the repeated negative colonoscopies in patients without real risk factors.
This retrospective study of 6 international cohorts matches the results of Lutgens et Coll. (2015) and demonstrates the benefit of individualized predictions using a multivariable model instead of current classification schemes. A substantial number of patients classified as high risk were found to have low predicted risks of CRC.
A less intensive monitoring protocol may be warranted in these cases. In contrast, patients assigned low risk with relatively high CRC predictive risks with this model may ultimately benefit from intensified surveillance to reduce interval CRCs.
Pending formal external validation
Future research should focus on formal external validation and evaluation of model performance in non-academic contexts. Next, to translate the model into a clinical decision support tool, risk scores should be linked to monitoring intervals based on expert and patient opinions, as evidence on this topic is lacking.
In the meantime, the interval between two screening colonoscopies must be 6 months in the event of suspected dysplasia left in place, one year in the case of associated PSC, 1 to 5 years in other cases depending on the findings of index colonoscopy.
Unlike CD, the chemopreventive effect of 5-aminosalicylated derivatives (5-ASA) in UC is now confirmed in high-quality meta-analyses. Exposure to a daily dose of at least one gram is associated with a 50% reduction in the risk of colorectal neoplasia. This chemoprevention is recommended from the diagnosis except in patients who have pure proctitis in the event of long-standing colonic disease.
There are no data available regarding the possible chemopreventive effect of anti-TNF antibodies and new molecules. Regardless of the IBD phenotype, it is useful to remind all patients of the lifestyle elements that contribute to the primary prevention of sporadic CRC: stopping smoking, moderate consumption of alcohol and red meat, fighting against sedentary lifestyle and overweight.
In conclusion, this new predictive model showed good discrimination and calibration to predict the cancerization of IBD without being able to produce formal generalization criteria. Future research should focus on external validation of this model and link predicted risks of advanced colorectal neoplasia to personalized monitoring intervals to enable the creation of a clinical decision support tool.
Prevention and detection of CRC is based on endoscopic surveillance with chromoendoscopy in all patients from the index colonoscopy and chemoprevention with 5-aminosalicylates in high-risk patients.
