ChinJGastroenterol,2022,Vol.27，No.5结直肠息肉发生风险预测模型的构建和验证*张学云1#季大年2项平2郑松柏1&复旦大学附属华东医院消化内科1（200040）内镜中心2背景：结直肠息肉为常见下消化道疾病，研究其危险因素对于临床防治具有重要意义。目的：构建结直肠息肉发生风险预测模型并进行验证。方法：根据纳入和排除标准，254例于2019年1月—2021年6月在上海华东医院特需内科病房住院体检者纳入研究，根据结肠镜检查结果归入结直肠息肉组和非结直肠息肉组。收集结直肠息肉相关危险因素，包括性别、年龄、吸烟史、饮酒史、高血压、糖尿病、高脂血症、高尿酸血症、胆囊息肉/结石、脂肪肝等信息，经LASSO回归筛选后纳入多因素Logistic回归分析构建预测模型并绘制列线图。以ROC曲线、C指数、校准曲线、决策曲线评价模型并进行内部验证。结果：254例研究对象中结直肠息肉组116例，非结直肠息肉组138例。经统计分析构建风险预测模型，发现性别（OR=2.11,95%CI:1.06~4.27）、年龄（OR=2.76,95%CI:1.17~6.73）、高血压（OR=3.23,95%CI:1.52~7.12）、糖尿病（OR=4.37,95%CI:1.52~14.64）、高脂血症（OR=3.20,95%CI:1.74~5.95）和脂肪肝（OR=2.21,95%CI:1.13~4.35）是结直肠息肉发生的独立危险因素。模型具有良好的ROC曲线下面积（0.807）和C指数（0.807）。决策曲线表明，如结直肠息肉发生阈值概率>12%，则模型具有临床意义。随机抽取内部样本进行验证，C指数为0.793。结论：结合性别、年龄、高血压、糖尿病、高脂血症、脂肪肝六项危险因素构建的预测模型和产生的列线图对结直肠息肉风险预测有一定参考价值。关键词结直肠息肉；危险因素；预测模型；列线图ConstructionandValidationofRiskPredictionModelforColorectalPolypsZHANGXueyun1,JIDanian2,XIANGPing2,ZHENGSongbai1.1DepartmentofGastroenterology,2EndoscopyCenter,HuadongHospitalAffiliatedtoFudanUniversity,Shanghai(200040)Correspondenceto:ZHENGSongbai,Email:*******************Background:Colorectalpolypisacommonlowergastrointestinaldisease.Studyofitsriskfactorsisofgreatsignificanceforpreventionandtreatmentofcolorectalpolypsinclinicalpractice.Aims:Toconstructandverifyapredictionmodelforriskofcolorectalpolyps.Methods:Accordingtotheinclusionandexclusioncriteria,254subjectswhowerehospitalizedforhealthexaminationintheSpecialInternalMedicineWardofShanghaiHuadongHospitalfromJanuary2019toJune2021wereenrolledinthestudy.Theywereallocatedintocolorectalpolypsgroupandnon⁃polypgroupbasedontheresultsofcolonoscopy.Therelevantriskfactorsofcolorectalpolypwerecollected,includinggender,age,cigarettesmoking,alcoholdrinking,hypertension,diabetes,hyperlipidemia,hyperuricemia,polyps/stonesofgallbladder,fattyliver,etc.AfterscreenedbyLASSOregressionmodel,theselectedfactorswereanalyzedbymultivariateLogisticregressiontobuildthepredictionmodelandnomogram.Furthermore,thepredictionmodelwasevaluatedbyROCcurve,Cindex,calibrationcurveanddecisioncurve,andvalidatedbyinternalsamples.Results:Ofthe254subjectsenrolledinthestudy,116caseswereincolorectalpolypsgroupand138innon⁃polypgroup.Theriskpredictionmodelidentifiedthatgender(OR=2.11,95%CI:1.06⁃4.27),age(OR=2.76,95%CI:1.17⁃6.73),hypertension(OR=3.23,95%CI:1.52⁃7.12),diabetes(OR=4.37,95%CI:1.52⁃14.64),hyperlipidemia(OR=3.20,95%CI:1.74⁃5.95)andfattyliver(OR=2.21,95%CI:1.13⁃4.35)wereindependentriskfactorsforcolorectalpolyps.ThemodelshowedgoodareaundertheROCcurve(0.807)andCindex(0.807).Thedecisioncurvedemonstratedthatifthethresholdprobabilityofcolorectalpolypswasmorethan12%,themodelwouldbeofclinicalsignificance.Internalsampleswererandomlyselectedforvalidation,andtheCindexwas0.793.Conclusions:Thepredictionmodelandnomogramconstructedbycombi...
