Construction and validation of a risk prediction model for colorectal adenoma

YAN Qi; ZHANG Yifeng; ZHANG Rujian

doi:10.3969/j.issn.1002-266X.2025.08.003

Abstract：Objective To analyze the risk factors for the occurrence of colorectal adenoma(CRA),to construct a risk prediction model,and to verify the predictive performance of the model.Methods A total of 718 subjects who un-derwent colonoscopy from May to December 2023 were selected and randomly divided into the training set(n=503)and internal validation set(n=215)at a ratio of 7∶3.Another 152 subjects who underwent colonoscopy in the same depart-ment from January to February 2024 were included as the external validation set.All subjects were confirmed by colonosco-py and pathological examination.We collected CRA-related indicators,including gender,age,body mass index(BMI),laboratory indicators[serum uric acid(SUA),total cholesterol(TC),triglyceride(TG),high-density lipoprotein choles-terol(HDL-C),low-density lipoprotein cholesterol(LDL-C),neutrophil-to-lymphocyte ratio(NLR),platelet-to-lympho-cyte ratio(PLR),systemic immune-inflammation index(SII),C-reactive protein(CRP),and carcinoembryonic antigen(CEA)],smoking history,drinking history,and comorbidities[type 2 diabetes mellitus(T2DM),hypertension,and non-alcoholic fatty liver disease(NAFLD)].Based on the training set,we performed univariate and multivariate Logistic re-gression analyses to identify independent risk factors for CRA occurrence,and developed a nomogram prediction model us-ing the screened factors.We plotted receiver operating characteristic(ROC)curves and calculated the area under the curve(AUC)to assess the discriminative ability;calibration curves were used to evaluate the calibration;decision curve analysis(DCA)was conducted to determine the clinical practical value of the model.Results In the training set of 503 subjects,175 had CRA and 328 did not.Results of univariate and multivariate Logistic regression analyses showed that male sex[odds ratio(OR)=2.79,95%confidence interval(CI):1.58-4.93],older age(OR=1.10,95%CI:1.08-1.13),smoking history(OR=7.05,95%CI:2.53-19.65),and elevated TG(OR=1.53,95%CI:1.12-2.09)were independent risk factors for CRA occurrence(all P<0.05).The AUC of the prediction model in the training set,internal validation set,and external validation set was 0.843(95%CI:0.808-0.877),0.812(95%CI:0.680-0.857),and 0.778(95%CI:0.702-0.854),respectively,indicating stable clinical applicability and excellent discriminative ability across all sets.Calibration curves demonstrated good consistency between predicted probabilities and actual observed prob-abilities.DCA results showed that the clinical benefit area enclosed by the model curve was larger than that of traditional strategies,confirming its clinical decision support value.Conclusions Gender,age,smoking history,and TG are inde-pendent risk factors for CRA occurrence.The CRA risk prediction model constructed based on these factors demonstrates good discriminative ability,calibration,and clinical practical value.