鉴别克罗恩病与溃疡性结肠炎的列线图预测模型构建

Construction of a nomogram prediction model for distinguishing Crohn’s disease from ulcerative colitis

目的探讨鉴别克罗恩病(Crohn’s disease,CD)和溃疡性结肠炎(ulcerative colitis,UC)的独立危险因素并构建列线图预测模型。方法回顾性分析429例炎症性肠病(inflammatory bowel disease,IBD)的临床病理资料,通过单因素和多因素分析鉴别CD和UC的危险因素并构建列线图模型。使用受试者工作特征曲线(receiver operating characteristic curve,ROC)、Hosmer-Lemeshow拟合优度及校准曲线检验评估列线图预测模型。结果429例IBD中CD 180例、UC 249例;单因素分析显示:年龄27.50(21.00,42.00)、PLT 302.50(229.00,364.25)、男性(70.00%)、中腹(52.77%)、大便次数1~3次/天(55.55%)、十二指肠病变(5.00%)、空肠病变(11.11%)、回肠病变(68.88%)、回盲部病变(56.11%)、右半结肠病变(49.44%)、内镜溃疡(77.77%)、瘘管(4.44%)、病理溃疡(45.55%)、管腔狭窄(21.11%)、肉芽肿(72.22%)、隆起性节段性病变(33.88%)、透壁坏死(9.44%)的患者患CD的风险显著增高;而年龄44.00(29.50,54.00)、PLT 281.00(209.00,337.00)、便血(49.79%)、黏液便(64.25%)、左半结肠病变(79.51%)、直肠病变(74.29%)、糜烂(55.82%)、出血(76.70%)、内镜下连续性病变(51.00%)、腺体萎缩(37.75%)、杯状细胞减少(33.33%)、隐窝脓肿(69.47%)的患者患UC的风险显著增高。多因素分析显示:患者性别、年龄、大便次数、便血、内镜下出血、回盲部病变、回肠病变、管腔狭窄、隆起性节段性病变、肉芽肿和隐窝脓肿是鉴别CD和UC的独立危险因素;构建列线图预测模型并计算ROC的曲线下面积为0.987(95%CI:0.979~0.995);Hosmer-Lemeshow拟合优度检验χ^(2)=3.449,P=0.903。结论基于独立危险因素建立的鉴别CD、UC的列线图预测模型具有良好的准确度,可为临床诊断治疗提供帮助。

Purpose To analysis the independent risk factors for distinguishing Crohn’s disease(CD)from ulcerative colitis(UC)and to construct nomogram prediction model.Methods The clinical and pathological data of 429 patients with inflammatory bowel disease(IBD)were analyzed retrospectively.The risk factors of distinguishing CD from UC were predicted by univariate and multivariate analysis and the nomogram was constructed.The receiver operating characteristic curve(ROC),Hosmer-Lemeshow goodness-of-fit test and the calibration curve were used to evaluate the nomogram.Results Among the 429 cases of IBD,there were 180 cases of CD and 249 cases of UC.Univariate analysis showed that age 27.50(21.00,42.00),PLT 302.50(229.00,364.25),male(70.00%),middle abdomen(52.77%),stool frequency 1-3 times per day(55.55%),duodenal lesion(5.00%),jejunal lesion(11.11%),ileal lesion(68.88%),ileocecal lesion(56.11%),right colon lesion(49.44%),endoscopic ulcer(77.77%),Fistula(4.44%),pathological ulcer(45.55%),lumen stenosis(21.11%),granuloma(72.22%),protuberant segmental lesions(33.88%)and transmural necrosis(9.44%)were significantly increased the risk of CD in patients.Patients with age 44.00(29.50,54.00),PLT 281.00(209.00,337.00),hematochezia(49.79%),mucous stool(64.25%),left colon lesion(79.51%),rectal lesion(74.29%),erosion(55.82%),bleeding(76.70%),endoscopic continuous lesion(51.00%),glandular atrophy(37.75%),goblet cell decrease(33.33%)and crypt abscess(69.47%)had a significantly higher risk of UC.Multivariate analysis showed that sex,age,stool frequency,hematochezia,endoscopic bleeding,ileocecal lesion,ileal lesion,lumen stenosis,protuberant segmental lesions,granuloma and crypt abscess were independent risk factors for distinguishing CD from UC.The nomogram prediction model was constructed and the area under the ROC area under curve of the nomogram predicton model was 0.987(95%CI:0.979-0.995).Hosmer-Lemeshow goodness-of-fit test wasχ^(2)=3.449,P=0.903.Conclusion The nomogram prediction model for distinguishing CD from UC based on independent risk factors has good accuracy and provides support for clinical diagnosis and treatment.