A Score to Predict Advanced Colorectal Neoplasia in Adults Younger than Age 45

Abstract

Background: Early-onset colorectal cancer (CRC) requires identifying adults at heightened risk of advanced colorectal neoplasia (AN) who may warrant colonoscopy initiation < age 45 years. Aims: We aim to develop and validate a model estimating the likelihood of AN in adults age < 45 years. Methods: We performed a cross-sectional analysis of adults’ ages 18–44 years who underwent a colonoscopy between 2011 and 2021 at a tertiary center. Subjects with AN constituted the case group while those with a normal colonoscopy or non-advanced neoplasia (NAN) formed the control group. We used backward elimination multivariable logistic regression methods to construct a model based on significant associations (p < 0.05) between risk factors and the presence of AN in a randomly selected training set and confirmed the associations in a validation set. Results: AN was detected in 346 (3.7%) of the 9,446 participants included. The reduced logistic regression model based on the training set identified BMI (p = 0.0157), family history of CRC (first-degree relative < 60, p < 0.0001; other family history of CRC p = 0.0117), and tobacco use (current vs. never, p = 0.0015, former vs. never, p = 0.0009) as risk factors for AN. In the validation set, the model exhibited moderate discriminatory power (c-statistic 0.645). The prediction score estimated the likelihood of detecting AN in the complete dataset, from 1.8% for individuals scoring 1 to > 14% for individuals scoring ≥ 9. Conclusion: We developed and internally validated a simple score using clinical factors which successfully predicts the likelihood of AN in adults < 45 years undergoing colonoscopy. Once externally validated, the proposed risk score may be useful for individualized CRC screening strategies.