Simple Prediction Model for Colorectal Serrated Polyps: Development and External Validation Study in U.S. Prospective Cohorts

Abstract

Abstract Serrated polyps (SP) are precursors for colorectal cancer and contribute disproportionately to postcolonoscopy cancers. Leveraging three U.S. cohorts (43,974 women and 5,322 men), we developed prediction models for high-risk SPs (sized ≥10 mm or ≥3) among individuals undergoing their first colonoscopy screening. We then validated the model in the Partners Colonoscopy Cohort (51,203 women and 39,077 men). We evaluated discrimination and calibration using the C-statistic and Hosmer–Lemeshow test, respectively. The age and family history model generated a C-statistic [95% confidence interval (CI)] of 0.57 (0.56–0.58) in women and 0.58 (0.55–0.61) in men. Further inclusion of smoking, alcohol, and body mass index (the simple model) increased the C-statistic (95% CI) to 0.68 (0.67–0.69) in women and 0.68 (0.66–0.71) in men (all P < 0.001). Adding more predictors did not provide much incremental predictivity. In the validation cohort, moderate discrimination was observed in both women (0.60, 0.58–0.61) and men (0.60, 0.59–0.62). Notably, the simple model also yielded similar C-statistics for a composite endpoint of SPs and high-risk conventional adenomas (women, 0.62, 0.62–0.63; men, 0.63, 0.61–0.64). The model was adequately calibrated in both sets of cohorts. In summary, we developed and externally validated a simple prediction model based on five major risk factors for high-risk SPs that may be useful for healthy lifestyle recommendations and tailored colorectal cancer screening. Prevention Relevance: On the basis of four prospective studies in the United States, we developed and externally validated a simple risk prediction model for high-risk SPs in the setting of colonoscopy screening. Our model showed moderate discriminatory accuracy and has potential utility for individualized risk assessment, healthy lifestyle recommendations, and tailored colorectal cancer prevention.