两种风险评估模型在结直肠癌筛查中应用效果的比较研究

Comparative study on application effect of two risk assessment models in screening colorectal cancer

目的比较两种风险评估模型在结直肠癌早期筛查中的应用效果,探讨更适用于我国居民结直肠癌早期筛查的评估模型。方法以2017—2019年参加深圳市南山区城市癌症早诊早治项目的40~74岁常住居民为研究对象,采用两种风险评估模型对同一人群进行结直肠癌风险评估,比较不同模型的筛查效果和预测价值。结果共纳入4141例研究对象,平均年龄(56.4±9.0)岁。模型一、模型二的初筛阳性率分别为15.2%和21.3%,总体一致率为93.50%(Kappa值=0.784,P<0.001)。702例肠镜检查者中,肠炎、息肉、腺瘤、肠癌、其他肠道病变的检出率分别为12.5%、12.0%、15.8%、0.7%、30.7%,两种模型初筛阳性者的肠镜检查结果分布无统计学差异(χ^(2)=8.679,P=0.123)。模型一的灵敏度为45.7%,低于模型二(61.2%);而模型一的特异度(64.8%)、阳性预测值(76.7%)、Kappa值(0.081)、约登指数(0.103)均高于模型二(41.7%、72.6%、0.026、0.029)。两种模型的ROC曲线下面积分别为0.660(95%CI0.618~0.702)和0.675(95%CI0.634~0.715),两者之间无统计学差异(P=0.584)。结论两种风险评估模型对结直肠癌早期诊断均具有一定的预测能力和优势,但在筛查准确度和筛查效益方面,模型一稍优于模型二,在大规模人群筛查中,建议两种模型取长补短,综合应用。

Objective The aim of this study was to compare the application effects of two risk assessment models in the early screening of colorectal cancer,and to explore an evaluation model that was more suitable for the early screening of colorectal cancer in Chinese residents.Methods The 40-74 years old permanent residents who participated in the urban cancer early diagnosis and early treatment project in Nanshan District,Shenzhen from 2017 to 2019 were selected as the research subjects.Two risk assessment models were used to assess the risk of colorectal cancer in the same population and compare the screening performance and predicted value.Results A total of 4141 participants were included,with an average age of 56.4±9.0 years old.The positive rates of model 1 and model 2 were 15.2%and 21.3%,respectively,and the overall agreement rate was 93.50%(Kappa value=0.784,P<0.001).Among 702 colonoscopy examiners,the detection rates of enteritis,polyps,adenomas,CRC,and other intestinal lesions were 12.5%,12.0%,15.8%,0.7%,and 30.7%,respectively.There was no statistically significant difference in the distribution of colonoscopy results(χ^(2)=8.679,P=0.123).The sensitivity of model 1 was 45.7%,which was lower than that of model 2(61.2%);while the specificity(64.8%),positive predictive value(76.7%),Kappa value(0.081)and Youden index(0.103)were higher than those in model 2(41.7%,72.6%,0.026 and 0.029).The areas under the ROC curves for the two models were 0.660(95%CI0.618-0.702)and 0.675(95%CI0.634-0.715),respectively,and there was no significant difference(P=0.584).Conclusion Both RAMs have certain predictive power and advantages for early diagnosis of CRC,but model 1 is slightly better than model 2 in terms of screening accuracy and screening benefit.In large-scale population screening,it is recommended the two models learn from each other′s strengths and complement their weaknesses,and are comprehensively applied.