Abstract 2213: A simple risk prediction model for high-risk adenomatous polyps at the time of colonoscopy

Abstract

Abstract Introduction: The prediction of high risk adenoma polyp (HRAP) may help to prioritize the urgency and guide the performance of colonoscopy procedures. Our objective was to develop and internally validate a simple, scalable clinical prediction model. Methods: The study population consisted of 2,364 individuals aged 50 to 74 with no prior history of cancer who had a screening colonoscopy at the Forzani and MacPhail Colon Cancer Screening Centre in Calgary, Canada. A total of 190 HRAPs were identified (8.0%). A multivariable logistic regression model was created using colorectal cancer risk factors identified from prior research. Predictor variables were collected from a baseline health questionnaire and included patient demographic (age, sex, and ethnicity), lifestyle (body mass index, alcohol (daily vs. no), smoking (never vs. ever), physical activity (high vs. moderate to low), and non-steroidal anti-inflammatory drug (NSAID) use (yes vs. no)), medical (family history of colorectal cancer, personal history of diabetes, or fecal occult blood test within the past two years), and female-specific characteristics (menopausal status and hormone replacement therapy (yes vs. no)). The demographic variables were first used to create a baseline model and the benefits of adding the groups of lifestyle, medical, and female-specific variables into the model in various combinations was assessed. Five-fold internal cross validation was conducted. Performance was assessed using the C-statistic and Hosmer-Lemeshow goodness-of-fit test. Results: The average age of the participants was 58 years of whom 54.9% were male and 85.24% were Caucasian. The clinical prediction model included demographic and lifestyle variables. On average, the predicted probability of having a HRAP was 8.0% (IQR: 4.1% to 10.6%). The bias-adjusted C-statistic was 0.66 (95% CI: 0.62 to 0.70) and there was no evidence of a lack of calibration according to the Hosmer-Lemeshow goodness-of-fit test. The addition of the medical history variables (Δ AUC = +0.0001; p = 0.98) or female-specific variables (Δ AUC = + 0.006; p = 0.32) or both medical and female-specific groups of variables (Δ AUC = +0.004; p = 0.45) did not significantly improve predictive performance. Conclusions: A model based on demographic and lifestyle variables showed a modest predictive ability for having an HRAP at the time of colonoscopy in a population undergoing screening-related colonoscopies. Consideration of these factors may assist in guiding prioritization of limited screening resources. Next steps include external validation and testing the incremental predictive ability of circulating biomarkers. Citation Format: Devon J. Boyne, Lisa M. Lix, Susanna Town, Steven J. Heitman, Robert J. Hilsden, Darren R. Brenner. A simple risk prediction model for high-risk adenomatous polyps at the time of colonoscopy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2213.