Multivariable Model Outperforms Guideline-Based Model for Predicting CRC

A multivariable prediction model for the presence of colorectal cancer (CRC) at surveillance colonoscopy outperformed a guideline-based model that focused solely on polyp characteristics, according to study results published online ahead of print in Gastro Hep Advances.

“Demand for surveillance colonoscopy can sometimes exceed capacity, such as during and following the coronavirus disease 2019 pandemic, yet no tools exist to prioritize the patients most likely to be diagnosed with CRC among those awaiting surveillance colonoscopy,” wrote corresponding author Theodore R. Levin, MD, of Kaiser Permanente Northern California, Oakland, California, and coauthors in the study background.

The model was developed using data from more than 52,000 patients who received a surveillance colonoscopy after polypectomy between 2014 and 2019. Candidate predictors for the model included index colonoscopy indication, findings, and endoscopist adenoma detection rate (ADR), as well as patient and clinical characteristics at surveillance.

Polyp size of 10 mm or greater, an ADR of less than 32.5% or missing, patient age, and having ever smoked tobacco were significant predictors of CRC, the study found.

The multivariable prediction model performed better than a guideline-based model that designated patients as low- or high-risk based solely on polyp characteristics, according to researchers. In an internal validation cohort consisting of 15,854 patients, the model’s area under the receiver-operating characteristic curve was 0.73.

In an external validation cohort of 30,015 patients, the model’s performance declined. However, the area under the receiver-operating characteristic curve recovered to 0.72 after model updating.

“When surveillance colonoscopy demand exceeds capacity, a prediction model featuring common clinical predictors may help prioritize patients at highest risk for CRC among those awaiting surveillance,” researchers wrote. “Also, regular model updates can address model performance drift.”

Reference

Levin TR, Jensen CD, Marks AR, et al. Development and external validation of a prediction model for colorectal cancer among patients awaiting surveillance colonoscopy following polypectomy. Gastro Hep Advances. Published online March 15, 2024. doi:10.1016/j.gastha.2024.03.008