To build a simple predictive model as a guide to stratify average-risk population for colonoscopy examinations. We collected data from 92 923 males without a prior history of cancer enrolled in the Kailuan Cohort Study of China. Risk factors included in the evaluation of colorectal cancer (CRC) were collected by questionnaire-based interviews at the baseline. Logistic regression coefficients for incident CRC predictors were converted into risk scores by the absolute value of the smallest coefficient in the model and rounding up to the nearest integer. Receiver operating characteristic (ROC) analysis with the leave-one-out cross-validation method was applied to evaluate model performance. In the 10-year follow-up, 353 CRC patients were in the cohort. Age, alcohol consumption, waist circumference, occupational sitting time, and history of diabetes were selected for the scoring system, and the adjusted area under the ROC was 0.66. Population in the highest risk group (16-19 points) had a 33.12-fold (95% CI: 13.44-81.59) higher risk of CRC than those in the lowest risk group. When we defined 13 points as the cut-off, the sensitivity and specificity of the scoring system for CRC were 67.99% and 62.42%, respectively. A simple scoring system for CRC has been developed to identify men at an increased relative risk of CRC within 10 years using several well-established risk factors, which allows selection of asymptomatic candidates for priority of CRC screening and saving the health resource in cancer prevention and control.
Keywords: China; colorectal neoplasms; risk prediction; risk score.
© 2019 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.