Construction and validation of sessile serrated lesions with dysplasia risk prediction model based on metabolic indicators and endoscopic features

To investigate the associations of metabolic abnormalities and endoscopic features with sessile serrated lesions with dysplasia (SSLD), and to develop and validate a combined prediction model and simplified clinical risk score for individualized SSLD risk stratification. This retrospective study enrolled consecutive patients who underwent colonoscopy at the Affiliated Wuxi People’s Hospital of Nanjing Medical University between January 1, 2021, and December 31, 2024. Participants were categorized into SSL and SSLD groups based on pathological diagnosis. LASSO regression with 10-fold cross-validation was applied to construct a weighted Metabolic Risk Score (MRS). Seven logistic regression models integrating baseline, metabolic, and endoscopic variables were developed and compared. The final parsimonious model was used to derive a simplified integer-based clinical risk score. Internal validation was performed using 1000 bootstrap resamples. A total of 594 patients were included, comprising 190 (32.0%) in the SSLD group and 404 (68.0%) in the SSL group. LASSO regression identified five metabolic markers (total cholesterol, triglycerides, systolic blood pressure, creatinine, and hemoglobin A1c) for MRS construction. The full combined model achieved an AUC of 0.873 (95% CI: 0.842–0.904). The final parsimonious model (Model 7: BMI, MRS, Lesion size, Lesion multiplicity, and Concomitant adenoma) yielded a comparable AUC of 0.870 (95% CI: 0.838–0.902; ΔAIC = –5.0, ΔBIC = –18.1), with good calibration and favorable net benefit in decision curve analysis. Bootstrap validation confirmed stable performance (mean AUC = 0.872). The simplified risk score stratified patients into low- (≤2 points, 9.1%), moderate- (3–4 points, 21.0%), and high-risk (≥5 points, 69.2%) groups. The combined model and simplified clinical risk score demonstrate excellent performance for SSLD prediction and may assist clinicians in early identification of high-risk patients to optimize colonoscopy surveillance strategies.